1 2018-05-18 Martin Liska <mliska@suse.cz>
3 * dbgcnt.c (limit_low): Renamed from limit.
4 (limit_high): New variable.
5 (dbg_cnt_is_enabled): Check for upper limit.
6 (dbg_cnt): Adjust dumping.
7 (dbg_cnt_set_limit_by_index): Add new argument for high
9 (dbg_cnt_set_limit_by_name): Likewise.
10 (dbg_cnt_process_single_pair): Parse new format.
11 (dbg_cnt_process_opt): Use strtok.
12 (dbg_cnt_list_all_counters): Remove 'value' and add
14 * doc/invoke.texi: Document changes.
16 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
18 * doc/sourcebuild.texi (scalar_all_fma): Document.
19 * tree.def (FMA_EXPR): Delete.
20 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
21 * internal-fn.c (ternary_direct): New macro.
22 (expand_ternary_optab_fn): Likewise.
23 (direct_ternary_optab_supported_p): Likewise.
24 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
25 * builtins.c (fold_builtin_fma): Delete.
26 (fold_builtin_3): Don't call it.
27 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
28 * expr.c (expand_expr_real_2): Likewise.
29 * fold-const.c (operand_equal_p): Likewise.
30 (fold_ternary_loc): Likewise.
31 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
32 * gimple.c (DEFTREECODE): Likewise.
33 * gimplify.c (gimplify_expr): Likewise.
34 * optabs-tree.c (optab_for_tree_code): Likewise.
35 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
36 * tree-eh.c (operation_could_trap_p): Likewise.
37 (stmt_could_throw_1_p): Likewise.
38 * tree-inline.c (estimate_operator_cost): Likewise.
39 * tree-pretty-print.c (dump_generic_node): Likewise.
40 (op_code_prio): Likewise.
41 * tree-ssa-loop-im.c (stmt_cost): Likewise.
42 * tree-ssa-operands.c (get_expr_operands): Likewise.
43 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
44 * fold-const-call.h (fold_fma): Delete.
45 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
46 CFN_FNMA and CFN_FNMS.
48 * genmatch.c (combined_fn): New enum.
49 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
50 (commutative_op): New function.
51 (commutate): Use it. Handle more than 2 operands.
52 (dt_operand::gen_gimple_expr): Use commutative_op.
53 (parser::parse_expr): Allow :c to be used with non-binary
54 operators if the commutative operand is known.
55 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
56 CFN_FMS, CFN_FNMA and CFN_FNMS.
57 (backprop::process_assign_use): Remove FMA_EXPR handling.
58 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
59 (gen_hsa_fma): New function.
60 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
61 IFN_FNMA and IFN_FNMS.
62 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
63 * gimple-fold.h (follow_all_ssa_edges): Declare.
64 * gimple-fold.c (follow_all_ssa_edges): New function.
65 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
66 gimple_build interface and use follow_all_ssa_edges to fold the result.
67 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
68 instead of checking for optabs directly.
69 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
70 rather than FMA_EXPRs.
71 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
72 call to IFN_FMA instead of an FMA_EXPR.
74 2018-05-17 Jim Wilson <jimw@sifive.com>
76 * expr.c (do_tablejump): When converting index to Pmode, if we have a
77 sign extended promoted subreg, and the range does not have the sign bit
78 set, then do a sign extend.
80 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
81 test, check for sign extended subreg and/or constant operands, and
82 do a sign extend in that case.
84 2018-05-17 Steve Ellcey <sellcey@cavium.com>
86 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
87 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
89 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
90 Change logics_shift_reg to logics_shift_imm.
91 (thunderx2t99_fp_loadpair_basic): Delete.
92 (thunderx2t99_fp_storepair_basic): Delete.
93 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
94 (thunderx2t99_asimd_polynomial): Delete.
95 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
96 and neon_fp_mul_d_scalar_q.
97 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
98 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
99 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
100 (thunderx2t99_asimd_lut): Add missing tbl types.
101 (thunderx2t99_asimd_ext): Delete.
102 (thunderx2t99_asimd_load1_1_mult): Delete.
103 (thunderx2t99_asimd_load1_2_mult): Delete.
104 (thunderx2t99_asimd_load1_ldp): New.
105 (thunderx2t99_asimd_load1): New.
106 (thunderx2t99_asimd_load2): Add missing *load2* types.
107 (thunderx2t99_asimd_load3): New.
108 (thunderx2t99_asimd_load4): New.
109 (thunderx2t99_asimd_store1_1_mult): Delete.
110 (thunderx2t99_asimd_store1_2_mult): Delete.
111 (thunderx2t99_asimd_store2_mult): Delete.
112 (thunderx2t99_asimd_store2_onelane): Delete.
113 (thunderx2t99_asimd_store_stp): New.
114 (thunderx2t99_asimd_store1): New.
115 (thunderx2t99_asimd_store2): New.
116 (thunderx2t99_asimd_store3): New.
117 (thunderx2t99_asimd_store4): New.
119 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
121 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
122 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
124 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
125 Segher Boessenkool <segher@kernel.crashing.org>
128 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
131 2018-05-17 Richard Biener <rguenther@suse.de>
133 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
134 for pruning loop and prune defs feeding only already visited PHIs.
136 2018-05-17 Richard Biener <rguenther@suse.de>
138 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
140 2018-05-17 Bin Cheng <bin.cheng@arm.com>
141 Richard Biener <rguenther@suse.de>
143 PR tree-optimization/85793
144 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
145 for VMAT_ELEMENTWISE.
147 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
149 * internal-fn.h (lookup_internal_fn): Declare
150 * internal-fn.c (lookup_internal_fn): New function.
151 * gimple.c (gimple_build_call_from_tree): Handle calls to
153 * gimple-pretty-print.c (dump_gimple_call): Print "." before
154 internal function names.
155 * tree-pretty-print.c (dump_generic_node): Likewise.
156 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
158 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
160 * gimple-fold.h (gimple_build): Make the function forms take
161 combined_fn rather than built_in_function.
162 (gimple_simplify): Likewise.
163 * gimple-match-head.c (gimple_simplify): Likewise.
164 * gimple-fold.c (gimple_build): Likewise.
165 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
166 rather than gimple_build_call_internal.
167 (get_initial_defs_for_reduction): Likewise.
168 (vect_create_epilog_for_reduction): Likewise.
169 (vectorizable_live_operation): Likewise.
171 2018-05-17 Martin Liska <mliska@suse.cz>
173 * gimple-ssa-sprintf.c (format_directive): Do not use
174 space in between 'G_' and '('.
176 2018-05-17 Jakub Jelinek <jakub@redhat.com>
179 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
180 even if the mask is not all ones.
183 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
185 (ix86_gimple_fold_builtin): Likewise.
188 * config/i386/i386.c: Include tree-vector-builder.h.
189 (ix86_vector_shift_count): New function.
190 (ix86_fold_builtin): Fold shift builtins by scalar count.
191 (ix86_gimple_fold_builtin): Likewise.
193 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
194 _mm512_setzero): New intrinsics.
196 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
197 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
199 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
200 code generation for cases where splatting a value is not useful.
201 * simplify-rtx.c (simplify_ternary_operation): Simplify
202 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
203 mode to a vec_concat.
205 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
207 * config.gcc: Support "goldmont-plus".
208 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
210 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
211 PROCESSOR_GOLDMONT_PLUS.
212 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
213 (processor_target_table): Add "goldmont-plus".
214 (PTA_GOLDMONT_PLUS): Define.
215 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
216 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
217 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
218 (fold_builtin_cpu): Add "goldmont-plus".
219 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
220 (ix86_option_override_internal): Add "goldmont-plus".
221 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
222 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
223 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
224 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
226 2018-05-17 Richard Biener <rguenther@suse.de>
228 PR tree-optimization/85757
229 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
230 remove defs that only feed that PHI from further processing.
232 2018-05-16 Jim Wilson <jimw@sifive.com>
234 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
236 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
238 2018-05-16 Mark Wielaard <mark@klomp.org>
240 * dwarf2out.c (count_index_strings): New function.
241 (output_indirect_strings): Call count_index_strings and generate
242 header for dwarf_version >= 5.
244 2018-05-16 Mark Wielaard <mark@klomp.org>
246 * dwarf2out.c (dwarf_FORM): New function.
247 (set_indirect_string): Use dwarf_FORM.
248 (reset_indirect_string): Likewise.
249 (size_of_die): Likewise.
250 (value_format): Likewise.
251 (output_die): Likewise.
252 (add_skeleton_AT_string): Likewise.
253 (output_macinfo_op): Likewise.
254 (index_string): Likewise.
255 (output_index_string_offset): Likewise.
256 (output_index_string): Likewise.
257 (count_index_strings): Likewise.
259 2018-05-16 Carl Love <cel@us.ibm.com>
261 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
262 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
264 2018-05-16 Martin Jambor <mjambor@suse.cz>
266 * ipa-prop.c (ipa_free_all_edge_args): Remove.
267 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
269 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
271 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
272 (fnma<mode>4): Likewise.
273 (fms<mode>4): Likewise.
274 (fnms<mode>4): Likewise.
275 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
276 (aarch64_fnma<mode>4): Likewise.
277 (aarch64_fms<mode>4): Likewise.
278 (aarch64_fnms<mode>4): Likewise.
279 (aarch64_fnmadd<mode>4): Likewise.
281 2018-05-16 Jason Merrill <jason@redhat.com>
283 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
285 2018-05-16 Richard Biener <rguenther@suse.de>
287 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
288 (dump_stmt_cost): Declare.
289 (add_stmt_cost): Dump cost we add.
290 (add_stmt_costs): New function.
291 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
293 (vect_analyze_stmt): Adjust prototype.
294 (vectorizable_condition): Likewise.
295 (vectorizable_live_operation): Likewise.
296 (vectorizable_reduction): Likewise.
297 (vectorizable_induction): Likewise.
298 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
299 cost vector to pass to vectorizable_ and record afterwards.
300 (vect_model_reduction_cost): Take cost vector argument and adjust.
301 (vect_model_induction_cost): Likewise.
302 (vectorizable_reduction): Likewise.
303 (vectorizable_induction): Likewise.
304 (vectorizable_live_operation): Likewise.
305 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
306 SLP_TREE_NUMBER_OF_VEC_STMTS.
307 (vect_analyze_slp_cost_1): Remove.
308 (vect_analyze_slp_cost): Likewise.
309 (vect_slp_analyze_node_operations): Take visited args and
310 a target cost vector. Avoid processing already visited stmt sets.
311 (vect_slp_analyze_operations): Use a local cost vector to gather
312 costs and register those of non-discarded instances.
313 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
314 (vect_schedule_slp_instance): Remove copying of
315 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
317 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
318 adding cost. Record cost entry location.
319 (vect_prologue_cost_for_slp_op): Function to compute cost of
320 a constant or invariant generated for SLP vect in the prologue,
321 split out from vect_analyze_slp_cost_1.
322 (vect_model_simple_cost): Make static. Adjust for SLP costing.
323 (vect_model_promotion_demotion_cost): Likewise.
324 (vect_model_store_cost): Likewise, make static.
325 (vect_model_load_cost): Likewise.
326 (vectorizable_bswap): Add cost vector arg and adjust.
327 (vectorizable_call): Likewise.
328 (vectorizable_simd_clone_call): Likewise.
329 (vectorizable_conversion): Likewise.
330 (vectorizable_assignment): Likewise.
331 (vectorizable_shift): Likewise.
332 (vectorizable_operation): Likewise.
333 (vectorizable_store): Likewise.
334 (vectorizable_load): Likewise.
335 (vectorizable_condition): Likewise.
336 (vectorizable_comparison): Likewise.
337 (can_vectorize_live_stmts): Likewise.
338 (vect_analyze_stmt): Likewise.
339 (vect_transform_stmt): Adjust calls to vectorizable_*.
340 * tree-vectorizer.c: Include gimple-pretty-print.h.
341 (dump_stmt_cost): New function.
343 2018-05-16 Richard Biener <rguenther@suse.de>
345 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
346 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
347 * tree-ssa-dse.c: Include tree-ssa-loop.h.
348 (check_name): New callback.
349 (dse_classify_store): Track cycles via a visited bitmap of PHI
350 defs and simplify handling of in-loop and across loop dead stores
351 and properly fail for loop-variant refs. Handle byte-tracking with
352 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
355 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
357 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
358 (vect_get_mask_type_for_stmt): Likewise.
359 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
361 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
362 to determine the statement's vector type and the vector type that
363 should be used for calculating nunits. Deal with cases in which
364 the type has to be deferred.
365 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
366 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
367 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
368 (vect_determine_vf_for_stmt): New functions, split out from...
369 (vect_determine_vectorization_factor): ...here.
370 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
371 (vect_get_mask_type_for_stmt): New functions, split out from
372 vect_determine_vectorization_factor.
374 2018-05-16 Richard Biener <rguenther@suse.de>
376 * tree-cfg.c (verify_gimple_assign_ternary): Properly
377 verify the [VEC_]COND_EXPR embedded comparison.
379 2018-05-15 Martin Sebor <msebor@redhat.com>
381 PR tree-optimization/85753
382 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
383 RECORD_TYPE in addition to ARRAY_TYPE.
385 2018-05-15 Martin Sebor <msebor@redhat.com>
388 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
390 2018-05-15 Richard Biener <rguenther@suse.de>
392 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
393 add by_clobber_p one. Change algorithm to collect all defs
394 representing uses we need to walk and try reducing them to
395 a single one before failing.
396 (dse_dom_walker::dse_optimize_stmt): Adjust.
398 2018-05-13 Mark Wielaard <mark@klomp.org>
400 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
401 (size_of_loc_descr): Likewise.
402 (output_loc_operands): Likewise.
403 (output_loc_operands_raw): Likewise.
404 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
405 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
406 (hash_loc_operands): Likewise.
407 (compare_loc_operands): Likewise.
409 2018-05-14 Mark Wielaard <mark@klomp.org>
411 * dwarf2out.c (count_index_addrs): New function.
412 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
414 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
416 PR tree-optimization/83648
417 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
418 return value as malloc candidate.
420 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
423 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
424 as true in call to suggest_attribute.
426 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
428 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
431 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
433 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
435 (aarch64_crypto_aesd_fused): Likewise.
437 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
439 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
440 (movsi_aarch64): Likewise.
441 (load_pairsi): Likewise.
442 (load_pairdi): Likewise.
443 (store_pairsi): Likewise.
444 (store_pairdi): Likewise.
445 (load_pairsf): Likewise.
446 (load_pairdf): Likewise.
447 (store_pairsf): Likewise.
448 (store_pairdf): Likewise.
449 (zero_extend): Likewise.
450 (trunc): Swap alternatives.
451 (fcvt_target): Add '?' to prefer w over r.
453 2018-05-14 Jakub Jelinek <jakub@redhat.com>
456 * config/i386/i386.md: Disallow non-commutative arithmetics in
457 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
458 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
459 in the peephole2 before it.
461 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
463 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
464 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
465 (ix86_handle_option): Handle -mcldemote.
466 * config.gcc: New header.
467 * config/i386/cldemoteintrin.h: New file.
468 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
469 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
471 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
472 OPTION_MASK_ISA_CLDEMOTE.
473 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
474 (ix86_valid_target_attribute_inner_p): Ditto.
475 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
476 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
477 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
478 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
479 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
481 * config/i386/i386.opt: Add -mcldemote.
482 * config/i386/x86intrin.h: New header.
483 * doc/invoke.texi: Add -mcldemote.
485 2018-05-14 Richard Biener <rguenther@suse.de>
487 * doc/match-and-simplify.texi: Adjust :s documentation.
489 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
491 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
492 intended memcpy size.
493 (REORDER_45): Likewise.
495 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
498 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
499 * vec.c (qsort_chk): Use gcc_qsort.
500 * Makefile.in (OBJS-libcommon): Add sort.o.
501 (build/sort.o): New target. Use it...
502 (BUILD_RTL): ... here, and...
503 (build/gencfn-macros): ... here, and...
504 (build/genmatch): ... here.
506 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
507 Chung-Ju Wu <jasonwucj@gmail.com>
509 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
510 * config/nds32/nds32-graywolf.md: New file.
511 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
512 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
514 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
515 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
516 * config/nds32/nds32.md (pipeline_model): Add graywolf.
517 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
518 * config/nds32/pipelines.md: Include n15 settings.
520 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
521 Chung-Ju Wu <jasonwucj@gmail.com>
523 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
524 * config/nds32/nds32-n13.md: New file.
525 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
526 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
528 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
529 * config/nds32/nds32.md (pipeline_model): Add n13.
530 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
531 * config/nds32/pipelines.md: Include n13 settings.
533 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
534 Chung-Ju Wu <jasonwucj@gmail.com>
536 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
537 * config/nds32/nds32-n10.md: New file.
538 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
539 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
541 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
542 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
543 * config/nds32/nds32.md (pipeline_model): Add n10.
544 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
545 * config/nds32/pipelines.md: Include n10 settings.
547 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
548 Kito Cheng <kito.cheng@gmail.com>
549 Chung-Ju Wu <jasonwucj@gmail.com>
551 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
552 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
553 Add enum values for DSP extension instructions.
554 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
556 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
557 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
559 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
560 * config/nds32/nds32-dspext.md: New file for DSP implementation.
561 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
562 * config/nds32/nds32-intrinsic.md: Likewise.
563 * config/nds32/nds32_intrinsic.h: Likewise.
564 * config/nds32/nds32-md-auxiliary.c: Likewise.
565 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
566 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
567 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
568 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
569 * config/nds32/nds32-protos.h: New declarations for DSP extension.
570 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
571 TYPE_DMAC in switch statement.
572 * config/nds32/nds32.c: New checking and implementation for DSP
573 extension instructions.
574 * config/nds32/nds32.h: Likewise.
575 * config/nds32/nds32.md: Likewise.
576 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
577 * config/nds32/predicates.md: Implement new predicates for DSP
580 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
582 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
583 Reformat alternatives and attributes so it is easier to identify
584 which constraints/attributes go with which instruction.
585 (mov<mode>_hardfloat32, FMOVE64): Likewise.
586 (mov<mode>_softfloat32, FMOVE64): Likewise.
587 (mov<mode>_hardfloat64, FMOVE64): Likewise.
588 (mov<mode>_softfloat64, FMOVE64): Likewise.
590 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
592 * doc/extend.texi (PowerPC Built-in Functions): Rename this
594 (Basic PowerPC Built-in Functions): The new name of the
595 subsection previously known as "PowerPC Built-in Functions".
596 (Basic PowerPC Built-in Functions Available on all Configurations):
598 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
599 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
600 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
601 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
603 2018-05-11 Martin Jambor <mjambor@suse.cz>
606 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
609 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
612 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
614 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
616 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
617 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
618 (ix86_handle_option): Handle -mwaitpkg.
619 * config.gcc: New header.
620 * config/i386/cpuid.h (bit_WAITPKG): New bit.
621 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
622 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
624 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
625 OPTION_MASK_ISA_WAITPKG.
626 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
627 (ix86_option_override_internal): Add PTA_WAITPKG.
628 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
629 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
631 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
632 __builtin_ia32_umwait and __builtin_ia32_tpause.
633 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
634 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
635 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
636 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
637 UNSPECV_TPAUSE): New.
638 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
639 * config/i386/i386.opt: Add -mwaitpkg.
640 * config/i386/waitpkgintrin.h: New file.
641 * config/i386/x86intrin.h: New header.
642 * doc/invoke.texi: Add -mwaitpkg.
644 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
647 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
649 (cortex-m0): Use armv6s-m isa.
650 (cortex-m0plus): Likewise.
651 (cortex-m1): Likewise.
652 (cortex-m0.small-multiply): Likewise.
653 (cortex-m0plus.small-multiply): Likewise.
654 (cortex-m1.small-multiply): Likewise.
656 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
657 Jakub Jelinek <jakub@redhat.com>
659 PR tree-optimization/85692
660 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
661 source permute as well.
663 2018-05-11 Martin Liska <mliska@suse.cz>
666 * doc/extend.texi: Document LLVM style format for no_sanitize
669 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
671 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
672 mode_supports_vsx_dform_quad to mode_supports_dq_form.
673 (mode_supports_vsx_dform_quad): Likewise.
674 (mode_supports_vmx_dform): Move these functions to be next to the
675 other mode_supports functions.
676 (mode_supports_dq_form): Likewise.
677 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
678 mode_supports_dq_form.
679 (reg_offset_addressing_ok_p): Likewise.
680 (offsettable_ok_by_alignment): Likewise.
681 (rs6000_legitimate_offset_address_p): Likewise.
682 (legitimate_lo_sum_address_p): Likewise.
683 (rs6000_legitimize_address): Likewise.
684 (rs6000_legitimize_reload_address): Likewise.
685 (rs6000_secondary_reload_inner): Likewise.
686 (rs6000_preferred_reload_class): Likewise.
687 (rs6000_output_move_128bit): Likewise.
689 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
691 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
692 Generate SImode target register for null target.
693 <case IX86_BUILTIN_XGETBV>: Ditto.
694 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
695 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
697 2018-05-10 Carl Love <cel@us.ibm.com>
699 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
700 dcbtt and dcbtstt if operands[2] is 0.
702 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
705 * config/i386/sse.md (usadv64qi): New expander.
707 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
709 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
710 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
711 -maltivec=be support.
712 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
713 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
714 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
715 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
716 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
717 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
718 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
719 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
720 altivec_vsumsws): Adjust.
721 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
722 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
723 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
725 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
726 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
727 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
728 (altivec_lve<VI_char>x): Delete expand.
729 (*altivec_lve<VI_char>x_internal): Rename to...
730 (altivec_lve<VI_char>x): ... this.
731 (altivec_lvxl_<mode>): Delete expand.
732 (*altivec_lvxl_<mode>_internal): Rename to ...
733 (altivec_lvxl_<mode>): ... this.
734 (altivec_stvxl_<mode>): Delete expand.
735 (*altivec_stvxl_<mode>_internal): Rename to ...
736 (altivec_stvxl_<mode>): ... this.
737 (altivec_stve<VI_char>x): Delete expand.
738 (*altivec_stve<VI_char>x_internal): Rename to ...
739 (altivec_stve<VI_char>x): ... this.
740 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
741 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
742 reduc_plus_scal_<mode>): Adjust.
743 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
745 (rs6000_cpu_cpp_builtins): Adjust.
746 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
747 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
748 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
749 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
750 -maltivec=be support.
751 (rs6000_split_vec_extract_var): Adjust.
752 (rs6000_split_v4si_init): Adjust.
753 (swap_selector_for_mode): Delete.
754 (altivec_expand_lvx_be, altivec_expand_stvx_be,
755 altivec_expand_stvex_be): Delete.
756 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
757 -maltivec=be support.
758 (rs6000_gimple_fold_builtin): Ditto.
759 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
761 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
762 (TARGET_DIRECT_MOVE_64BIT): Adjust.
763 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
764 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
765 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
766 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
767 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
768 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
769 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
770 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
771 anonymous split): Adjust.
772 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
773 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
775 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
777 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
778 when --with-gxx-include-dir is also specified.
779 * configure: Regenerate.
781 2018-05-09 Jim Wilson <jimw@sifive.com>
784 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
785 * config/riscv/t-withmultilib: New.
786 * config/riscv/withmultilib.h: New.
787 * doc/install.texi: Document RISC-V --with-multilib-list support.
789 2018-05-09 Richard Biener <rguenther@suse.de>
791 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
793 (vect_bb_vectorization_profitable_p): Adjust. Compute
794 actual scalar cost using the cost vector and the add_stmt_cost
797 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
799 PR rtl-optimization/85645
800 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
801 in the REG_CFA_REGISTER note for LR, don't leave it empty.
803 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
805 PR rtl-optimization/85645
806 * shrink-wrap.c (spread_components): Return a boolean saying if
807 anything was changed.
808 (try_shrink_wrapping_separate): Iterate spread_components until
809 nothing changes anymore.
811 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
813 PR rtl-optimization/85645
814 * regrename.c (build_def_use): Also kill the chains that include the
815 destination of a REG_CFA_REGISTER note.
817 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
819 PR rtl-optimization/85645
820 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
821 insn that has a REG_CFA_REGISTER note.
823 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
825 * cfgexpand.c (expand_clobber): New function.
826 (expand_gimple_stmt_1): Use it.
827 * tree-vect-stmts.c (vect_clobber_variable): New function,
829 (vectorizable_simd_clone_call): ...here.
830 (vectorizable_store): Emit a clobber either side of an
831 IFN_STORE_LANES sequence.
832 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
834 2018-05-09 Tom de Vries <tom@codesourcery.com>
837 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
838 (define_insn "trap_if_false"): Add exit after trap.
840 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
842 PR rtl-optimization/85638
843 * bb-reorder.c: Include common/common-target.h.
844 (create_forwarder_block): New function extracted from...
845 (fix_up_crossing_landing_pad): ...here. Rename into...
846 (dw2_fix_up_crossing_landing_pad): ...this.
847 (sjlj_fix_up_crossing_landing_pad): New function.
848 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
849 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
850 from both partitions and exit the loop after one iteration.
852 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
855 * doc/extend.texi (PowerPC Built-in Functions): Rename this
857 (Basic PowerPC Built-in Functions): The new name of the
858 subsection previously known as "PowerPC Built-in Functions".
859 (Basic PowerPC Built-in Functions Available on all Configurations):
861 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
863 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
865 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
867 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
870 2018-05-08 Jim Wilson <jimw@sifive.com>
872 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
873 (LD_EMUL_SUFFIX): New.
876 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
878 * doc/extend.texi (PowerPC Built-in Functions): Rename this
880 (Basic PowerPC Built-in Functions): The new name of the
881 subsection previously known as "PowerPC Built-in Functions".
882 (Basic PowerPC Built-in Functions Available on all Configurations):
884 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
886 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
888 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
890 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
893 2018-05-08 Jakub Jelinek <jakub@redhat.com>
896 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
897 after cmpelim optimization.
899 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
901 * config.gcc: Support "goldmont".
902 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
903 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
905 * config/i386/i386.c (m_GOLDMONT): Define.
906 (processor_target_table): Add "goldmont".
907 (PTA_GOLDMONT): Define.
908 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
909 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
910 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
911 (fold_builtin_cpu): Add "goldmont".
912 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
913 (ix86_option_override_internal): Add "goldmont".
914 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
915 (processor_type): Add PROCESSOR_GOLDMONT.
916 * config/i386/i386.md: Add CPU "glm".
917 * config/i386/glm.md: New file.
918 * config/i386/x86-tune.def: Add m_GOLDMONT.
919 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
921 2018-05-08 Jakub Jelinek <jakub@redhat.com>
924 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
926 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
927 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
928 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
931 * config/i386/i386.c (ix86_fold_builtin): Handle
932 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
935 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
936 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
938 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
941 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
942 (check_arch): Likewise.
943 (check_fpu): Return the result rather than printing it.
944 (end arch): Fix operator precedence.
946 (END): Print the result from check_fpu.
948 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
949 Alan Hayward <alan.hayward@arm.com>
950 David Sherwood <david.sherwood@arm.com>
952 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
953 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
954 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
955 (*fcmuo<mode>_and): New patterns.
957 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
959 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
960 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
961 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
962 (cmp_op, sve_imm_con): New code attributes.
963 (SVE_COND_INT_CMP, imm_con): Delete.
964 (cmp_op): Remove above unspecs from int attribute.
965 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
967 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
968 comparison-specific unspecs.
969 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
970 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
971 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
972 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
973 (*vec_fcm<cmp_op><mode>): Rename to...
974 (*fcm<cmp_op><mode>): ...this and adjust likewise.
975 (*vec_fcmuo<mode>): Rename to...
976 (*fcmuo<mode>): ...this and adjust likewise.
977 (*pred_fcm<cmp_op><mode>): New pattern.
978 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
979 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
981 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
983 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
984 (aarch64_emit_sve_predicated_cond): New function.
985 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
986 (aarch64_emit_unspec_cond_or): Replace with...
987 (aarch64_emit_sve_or_conds): ...this new function. Use
988 aarch64_emit_sve_ptrue_op for the individual comparisons and
989 aarch64_emit_binop to OR them together.
990 (aarch64_emit_inverted_unspec_cond): Replace with...
991 (aarch64_emit_sve_inverted_cond): ...this new function. Use
992 aarch64_emit_sve_ptrue_op for the comparison and
993 aarch64_emit_unop to invert the result.
994 (aarch64_expand_sve_vec_cmp_float): Update after the above
995 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
997 2018-05-07 Nathan Sidwell <nathan@acm.org>
999 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
1000 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
1001 (Backwards Compatibility): Likewise.
1003 2018-05-07 Luis Machado <luis.machado@linaro.org>
1007 2018-05-07 Luis Machado <luis.machado@linaro.org>
1009 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1010 <prefetch_dynamic_strides>: New const bool field.
1011 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1012 prefetch_dynamic_strides.
1013 (exynosm1_prefetch_tune): Likewise.
1014 (thunderxt88_prefetch_tune): Likewise.
1015 (thunderx_prefetch_tune): Likewise.
1016 (thunderx2t99_prefetch_tune): Likewise.
1017 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1018 (aarch64_override_options_internal): Update to set
1019 PARAM_PREFETCH_DYNAMIC_STRIDES.
1020 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1021 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1022 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1023 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1024 prefetch-dynamic-strides setting.
1026 2018-05-07 Luis Machado <luis.machado@linaro.org>
1028 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1029 <minimum_stride>: New const int field.
1030 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1031 minimum_stride field.
1032 (exynosm1_prefetch_tune): Likewise.
1033 (thunderxt88_prefetch_tune): Likewise.
1034 (thunderx_prefetch_tune): Likewise.
1035 (thunderx2t99_prefetch_tune): Likewise.
1036 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1037 (aarch64_override_options_internal): Update to set
1038 PARAM_PREFETCH_MINIMUM_STRIDE.
1039 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1040 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1041 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1042 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1043 stride is constant and is below the minimum stride threshold.
1045 2018-05-07 Luis Machado <luis.machado@linaro.org>
1047 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
1050 2018-05-07 Luis Machado <luis.machado@linaro.org>
1052 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1053 <prefetch_dynamic_strides>: New const bool field.
1054 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1055 prefetch_dynamic_strides.
1056 (exynosm1_prefetch_tune): Likewise.
1057 (thunderxt88_prefetch_tune): Likewise.
1058 (thunderx_prefetch_tune): Likewise.
1059 (thunderx2t99_prefetch_tune): Likewise.
1060 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1061 (aarch64_override_options_internal): Update to set
1062 PARAM_PREFETCH_DYNAMIC_STRIDES.
1063 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1064 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1065 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1066 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1067 prefetch-dynamic-strides setting.
1069 2018-05-07 Luis Machado <luis.machado@linaro.org>
1071 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1072 <minimum_stride>: New const int field.
1073 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1074 minimum_stride field.
1075 (exynosm1_prefetch_tune): Likewise.
1076 (thunderxt88_prefetch_tune): Likewise.
1077 (thunderx_prefetch_tune): Likewise.
1078 (thunderx2t99_prefetch_tune): Likewise.
1079 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1080 (aarch64_override_options_internal): Update to set
1081 PARAM_PREFETCH_MINIMUM_STRIDE.
1082 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1083 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1084 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1085 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1086 stride is constant and is below the minimum stride threshold.
1088 2018-05-06 Jakub Jelinek <jakub@redhat.com>
1091 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
1092 the type is addressable. Don't force op into register if it has
1095 2018-05-05 Roland McGrath <mcgrathr@google.com>
1098 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
1099 any section for which we don't know a specific type it should have,
1100 regardless of name. Previously this was done only for the exact
1101 names ".init_array", ".fini_array", and ".preinit_array".
1102 (default_elf_asm_named_section): Add comment about
1103 relationship with default_section_type_flags and SECTION_NOTYPE.
1104 (get_section): Don't consider it a type conflict if one side has
1105 SECTION_NOTYPE and the other doesn't, as long as neither has the
1106 SECTION_BSS et al used in the default_section_type_flags logic.
1108 2018-05-05 Tom de Vries <tom@codesourcery.com>
1111 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
1112 (workaround_barsyncs): New function.
1113 (nvptx_reorg): Use workaround_barsyncs.
1114 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1115 (define_expand "nvptx_membar_cta"): New define_expand.
1116 (define_insn "*nvptx_membar_cta"): New insn.
1118 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
1120 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
1121 To improve optimization opportunities.
1122 * builtin-types.def: The new needed builtin types for the above.
1124 2018-05-04 Richard Biener <rguenther@suse.de>
1126 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
1127 * gimple-ssa-store-merging.c
1128 (imm_store_chain_info::output_merged_store): Remove redundant create,
1129 release split_store vector contents on failure.
1130 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
1131 scalar stmt vector on cache hit.
1133 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
1135 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
1137 * config.gcc (powerpc-xilinx-eabi*): Remove.
1138 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
1140 (fusion_addis_mem_combo_load): Ditto.
1141 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
1143 (rs6000_cpu_cpp_builtins): Ditto.
1144 * config/rs6000/rs6000-linux.c
1145 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
1146 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
1147 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
1149 (rs6000_setup_reg_addr_masks): Ditto.
1150 (rs6000_init_hard_regno_mode_ok): Ditto.
1151 (rs6000_option_override_internal): Ditto.
1152 (legitimate_lo_sum_address_p): Ditto.
1153 (rs6000_legitimize_address): Ditto.
1154 (rs6000_legitimize_reload_address): Ditto.
1155 (rs6000_legitimate_address_p): Ditto.
1156 (abi_v4_pass_in_fpr): Ditto.
1157 (setup_incoming_varargs): Ditto.
1158 (rs6000_gimplify_va_arg): Ditto.
1159 (rs6000_split_multireg_move): Ditto.
1160 (rs6000_savres_strategy): Ditto.
1161 (rs6000_emit_prologue_components): Ditto.
1162 (rs6000_emit_epilogue_components): Ditto.
1163 (rs6000_emit_prologue): Ditto.
1164 (rs6000_emit_epilogue): Ditto.
1165 (rs6000_elf_file_end): Ditto.
1166 (rs6000_function_value): Ditto.
1167 (rs6000_libcall_value): Ditto.
1168 * config/rs6000/rs6000.h: Ditto.
1169 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
1170 (TARGET_MINMAX): ... this. New.
1171 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
1172 * config/rs6000/rs6000.md: Remove Xilinx FP support.
1173 (*movsi_internal1_single): Delete.
1174 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
1175 mfpu=, mxilinx-fpu): Delete.
1176 * config/rs6000/singlefp.h: Delete.
1177 * config/rs6000/sysv4.h: Remove Xilinx FP support.
1178 * config/rs6000/t-rs6000: Ditto.
1179 * config/rs6000/t-xilinx: Delete.
1180 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
1181 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
1182 (VStype_simple): Delete.
1183 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
1184 * config/rs6000/xfpu.h: Delete.
1185 * config/rs6000/xfpu.md: Delete.
1186 * config/rs6000/xilinx.h: Delete.
1187 * config/rs6000/xilinx.opt: Delete.
1188 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
1189 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
1191 2018-05-04 Tom de Vries <tom@codesourcery.com>
1194 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
1197 2018-05-04 Richard Biener <rguenther@suse.de>
1200 * tree-complex.c (update_complex_assignment): We are always in SSA form.
1201 (expand_complex_div_wide): Likewise.
1202 (expand_complex_operations_1): Likewise.
1203 (expand_complex_libcall): Preserve EH info of the original stmt.
1204 (tree_lower_complex): Handle removed blocks.
1205 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
1206 on complex multiplication and division libcall builtins.
1208 2018-05-04 Richard Biener <rguenther@suse.de>
1211 * fold-const.c (negate_expr_p): Restrict negation of operand
1212 zero of a division to when we know that can happen without
1214 (fold_negate_expr_1): Likewise.
1216 2018-05-04 Jakub Jelinek <jakub@redhat.com>
1219 * real.h (real_nextafter): Declare.
1220 * real.c (real_nextafter): New function.
1221 * fold-const-call.c (fold_const_nextafter): New function.
1222 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
1223 CASE_CFN_NEXTTOWARD.
1224 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
1225 even when arg1_mode is different from arg0_mode.
1227 2018-05-03 Nathan Sidwell <nathan@acm.org>
1229 * doc/extend.texi (Deprecated Features): Remove
1231 (Backwards Compatibility): Likewise.
1232 * doc/invoke.texi (C++ Language Options): Likewise.
1233 (C++ Dialect Options): Likewise.
1235 2018-05-03 Jakub Jelinek <jakub@redhat.com>
1238 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1239 _mm512_mask_mullox_epi64): New intrinsics.
1241 2018-05-03 Tom de Vries <tom@codesourcery.com>
1244 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1245 dump files): Add offload-tree.
1247 2018-05-03 Richard Biener <rguenther@suse.de>
1249 PR tree-optimization/85615
1250 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1251 to loops not nested in BBs loop father to avoid creating multi-entry
1254 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1256 PR tree-optimization/70291
1257 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1258 arguments. Change return type to tree. Emit libcall as a new
1259 statement rather than replacing existing one when inplace_p is true.
1260 (expand_complex_multiplication_components): New function.
1261 (expand_complex_multiplication): Expand floating-point complex
1262 multiplication using the above.
1263 (expand_complex_division): Rename inner_type parameter to type.
1264 Update expand_complex_libcall call-site.
1265 (expand_complex_operations_1): Update expand_complex_multiplication
1266 and expand_complex_division call-sites.
1268 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1271 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1272 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1273 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1274 the highest significant bit of the shift count mask is clear. In
1275 check whether and[sq]i3 is needed verify that all significant bits
1276 of the shift count other than the highest are set.
1278 2018-05-02 Tom de Vries <tom@codesourcery.com>
1281 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1282 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1283 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1284 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1285 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1286 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1287 __builtin_goacc_parlevel_size.
1289 2018-05-02 Richard Biener <rguenther@suse.de>
1291 PR tree-optimization/85597
1292 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1293 do not use split vect_get_vec_defs call but call vect_get_slp_defs
1296 2018-05-02 Tom de Vries <tom@codesourcery.com>
1299 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1300 dump files): Add ltrans-tree.
1302 2018-05-02 Tom de Vries <tom@codesourcery.com>
1305 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1306 dump files): Add wpa-ipa.
1308 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
1310 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
1311 powerpc*-*-linux*paired* target.
1312 * config/rs6000/750cl.h: Delete.
1313 * config/rs6000/paired.h: Delete.
1314 * config/rs6000/paired.md: Delete.
1315 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1317 * config/rs6000/rs6000-builtin.def: Remove paired float support.
1318 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1319 comment. Remove paired float support.
1320 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1321 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1323 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1324 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1326 * config/rs6000/rs6000.c: Remove paired float support.
1327 (paired_expand_vector_init, paired_expand_vector_move,
1328 paired_emit_vector_compare, paired_emit_vector_cond_expr,
1329 (paired_expand_lv_builtin, paired_expand_stv_builtin,
1330 paired_expand_builtin, paired_expand_predicate_builtin,
1331 paired_init_builtins): Delete.
1332 * config/rs6000/rs6000.h: Remove paired float support.
1333 * config/rs6000/rs6000.md: Remove paired float support.
1334 (move_from_CR_ov_bit): Delete.
1335 * config/rs6000/rs6000.opt (mpaired): Delete.
1336 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1337 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1339 2018-05-02 Richard Biener <rguenther@suse.de>
1342 * gimplify.c (gimplify_save_expr): When in SSA form allow
1343 SAVE_EXPRs to compute to SSA vars.
1345 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1348 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1349 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1350 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1351 clobber operands[2], instead use a new pseudo. Formatting fixes.
1353 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1355 PR tree-optimization/85586
1356 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1357 exit early for statements in the same group if the accesses are
1360 2018-05-02 Tom de Vries <tom@codesourcery.com>
1363 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1366 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
1368 PR tree-optimization/85143
1369 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1371 2018-05-01 Tom de Vries <tom@codesourcery.com>
1374 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1375 not found" error message.
1377 2018-05-01 Tom de Vries <tom@codesourcery.com>
1380 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1381 * vec.c (test_ordered_remove_if): New function.
1382 (vec_c_tests): Call test_ordered_remove_if.
1383 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1384 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1385 * tree-vect-patterns.c (vect_pattern_recog_1): Use
1386 VEC_ORDERED_REMOVE_IF.
1388 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1390 PR tree-optimization/82665
1391 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1392 pointer subtraction where arguments come from a memchr call.
1394 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1396 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1397 --push-state --as-needed and --pop-state instead of --as-needed and
1398 --no-as-needed if ld supports it.
1399 * configure: Regenerated.
1402 * doc/install.texi2html: Replace _002d with - and _002a with * in
1403 generated html files using sed.
1405 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1408 * gcc-rich-location.c (blank_line_before_p): New function.
1409 (use_new_line): New function.
1410 (gcc_rich_location::add_fixit_insert_formatted): New function.
1411 * gcc-rich-location.h
1412 (gcc_rich_location::add_fixit_insert_formatted): New function.
1414 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1416 * selftest.c (assert_streq): Rename "expected" and "actual" to
1417 "val1" and "val2". Extend NULL-handling to cover both inputs
1418 symmetrically, while still requiring both to be non-NULL for a pass.
1419 * selftest.h (assert_streq): Rename "expected" and "actual" to
1421 (ASSERT_EQ): Likewise.
1422 (ASSERT_EQ_AT): Likewise.
1423 (ASSERT_KNOWN_EQ): Likewise.
1424 (ASSERT_KNOWN_EQ_AT): Likewise.
1425 (ASSERT_NE): Likewise.
1426 (ASSERT_MAYBE_NE): Likewise.
1427 (ASSERT_MAYBE_NE_AT): Likewise.
1428 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1429 the assertion to pass.
1430 (ASSERT_STREQ_AT): Likewise.
1432 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1434 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1435 interaction with -pie.
1437 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1439 * selftest.h: Fix alphabetization of per-source-file selftest
1442 2018-04-30 Jason Merrill <jason@redhat.com>
1444 PR c++/61982 - dead stores to destroyed objects.
1445 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1448 2018-04-30 Jason Merrill <jason@redhat.com>
1450 * tree.c (build_clobber): New.
1451 * tree.h: Declare it.
1452 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1454 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1456 * diagnostic-show-locus.c (layout::layout): Update for
1457 location_get_source_line returning a char_span.
1458 (struct char_span): Move to input.h.
1459 (struct correction): Update for fields in char_span becoming
1461 (struct source_line): Update for location_get_source_line
1462 returning a char_span.
1463 (layout::print_line): Likewise.
1464 * edit-context.c (edited_file::print_content): Likewise.
1465 (edited_file::print_diff_hunk): Likewise.
1466 (edited_file::print_run_of_changed_lines): Likewise.
1467 (edited_file::get_num_lines): Likewise.
1468 (edited_line::edited_line): Likewise.
1469 * final.c (asm_show_source): Likewise.
1470 * input.c (location_get_source_line): Convert return type
1471 from const char * to char_span, losing the final "line_len"
1473 (dump_location_info): Update for the above.
1474 (get_substring_ranges_for_loc): Likewise. Use a char_span
1475 when handling the literal within the line.
1476 (test_reading_source_line): Update for location_get_source_line
1477 returning a char_span.
1478 * input.h (class char_span): Move here from
1479 diagnostic-show-locus.c, converting from a struct to a class.
1480 Make data members private.
1481 (char_span::operator bool): New.
1482 (char_span::length): New.
1483 (char_span::get_buffer): New.
1484 (char_span::operator[]): New.
1485 (char_span::subspan): Make const.
1486 (char_span::xstrdup): New.
1487 (location_get_source_line): Convert return type from const char *
1488 to char_span, losing the final "line_size" param.
1490 2018-04-30 Jan Hubicka <jh@suse.cz>
1492 * lto-wrapper.c (ltrans_priorities): New static var.
1493 (cmp_priority): New.
1494 (run_gcc): Read priorities and if doing parallel build order
1495 the Makefile by them.
1497 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1499 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
1501 2018-04-30 Richard Biener <rguenther@suse.de>
1503 * tree-cfg.c (verify_address): Remove base argument, add
1504 flag whether to check TREE_ADDRESSABLE and do that.
1505 (verify_expr): Remove.
1506 (verify_types_in_gimple_reference): Add pieces from verify_expr.
1507 (verify_gimple_assign_single): Likewise.
1508 (verify_gimple_switch): Likewise.
1509 (verify_expr_location_1): Dereference tp once. Add (disabled)
1510 piece from verify_expr.
1511 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1513 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1515 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1517 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1519 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1520 (small_data_pattern): Likewise.
1521 (arc_rewrite_small_data): Likewise.
1522 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1523 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1524 (get_symbol_alignment): New function.
1525 (legitimate_small_data_address_p): Likewise.
1526 (legitimate_scaled_address): Update, call
1527 legitimate_small_data_address_p.
1528 (output_sdata): New static variable.
1529 (arc_print_operand): Update how we handle small data operands.
1530 (arc_print_operand_address): Likewise.
1531 (arc_legitimate_address_p): Update, use
1532 legitimate_small_data_address_p.
1533 (arc_rewrite_small_data_p): Remove.
1534 (arc_rewrite_small_data_1): Likewise.
1535 (arc_rewrite_small_data): Likewise.
1536 (small_data_pattern): Likewise.
1537 (compact_sda_memory_operand): Update to use
1538 legitimate_small_data_address_p and get_symbol_alignment.
1539 (prepare_move_operands): Don't rewite sdata pattern.
1540 (prepare_extend_operands): Remove.
1541 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
1543 (zero_extendqisi2): Likewise.
1544 (zero_extendhisi2): Likewise.
1545 (extendqihi2): Likewise.
1546 (extendqisi2): Likewise.
1547 (extendhisi2): Likewise.
1551 * config/arc/constraints.md (Usd): Change it to memory constraint.
1553 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1555 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
1556 as source of std instructions.
1557 * config/arc/arc.md (movsi_insn): Update pattern predicate to
1558 allow 6-bit constants as source for store instructions.
1559 (movdi_insn): Update instruction pattern to allow 6-bit constants
1560 as source for store instructions.
1562 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1564 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
1566 2018-04-30 Nathan Sidwell <nathan@acm.org>
1567 Sandra Loosemore <sandra@codesourcery.com>
1569 * dumpfile.c (dump_open): Allow '-' for stdout.
1570 * doc/invoke.texi (Developer Options): Document dump filename
1571 determination early. Document stdin/stdout selection.
1573 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
1575 Microblaze Target: PIC data text relative
1577 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
1578 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
1580 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
1581 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
1582 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
1583 New addressing mode for data-text relative position indepenedent code.
1584 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
1585 'ADDRESS_SYMBOLIC_TXT_REL'.
1586 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
1587 (microblaze_legitimate_pic_operand): Exclude function calls from
1588 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
1589 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
1591 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1592 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1593 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
1594 for 'address + offset'.
1595 (microblaze_expand_prologue): Add new function prologue call for
1597 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
1598 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
1599 table in case of TARGET_PIC_DATA_TEXT_REL.
1600 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
1601 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
1602 Add new macros 'UNSPEC_TEXT',
1603 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
1604 + exclude function calls from 'UNSPEC_PLT' in case of data text
1606 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1607 new target hook for generating address diff vector tables in case of
1609 * doc/tm.texi : Regenerate.
1610 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
1611 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
1612 of addr diff vector generation.
1613 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1614 target hook definition.
1615 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
1616 Add default function for generate_pic_addr_diff_vec -> flag_pic.
1617 * doc/invoke.texi (Add new pic option): Add new microblaze pic
1618 option for data text relative.
1620 2018-04-30 Richard Biener <rguenther@suse.de>
1622 * tree-chrec.h (evolution_function_is_constant_p): Remove
1624 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
1626 2018-04-30 Richard Biener <rguenther@suse.de>
1629 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
1631 2018-04-30 Richard Biener <rguenther@suse.de>
1633 PR tree-optimization/28364
1634 PR tree-optimization/85275
1635 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
1636 copying first exit test.
1638 2018-04-28 Mark Wielaard <mark@klomp.org>
1640 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
1642 (dwarf_AT): Handle DW_AT_addr_base.
1643 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
1645 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
1648 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
1649 (*ashl<dwi>3_doubleword_mask_1): Ditto.
1650 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
1651 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
1653 2018-04-28 Richard Biener <rguenther@suse.de>
1655 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
1656 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
1657 to reflect use. Only add interesting stmts.
1659 2018-04-27 Martin Jambor <mjambor@suse.cz>
1662 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
1663 the jump function allows for passing through aggregate values.
1665 2018-04-27 David Malcolm <dmalcolm@redhat.com>
1667 * input.h (in_system_header_at): Convert from macro to inline
1669 (from_macro_expansion_at): Likewise.
1670 (from_macro_definition_at): Likewise.
1672 2018-04-27 Jeff Law <law@redhat.com>
1674 * config.gcc: Mark tile* targets as deprecated/obsolete.
1676 2018-04-27 Richard Biener <rguenther@suse.de>
1678 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
1681 2018-04-27 Richard Biener <rguenther@suse.de>
1683 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
1685 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
1687 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
1688 with Yd constraint. Set "preferred_for_speed" attribute from
1689 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
1691 (*movdi_internal): Ditto.
1692 (movti_interunit splitters): Remove
1693 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
1694 (movdi_interunit splitters): Ditto.
1695 * config/i386/constraints.md (Ye): Remove.
1696 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
1698 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1701 * config/aarch64/constraints.md (Usg): Limit to 31.
1704 2018-04-27 Jakub Jelinek <jakub@redhat.com>
1706 PR tree-optimization/85529
1707 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
1708 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
1709 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
1710 zero extension or masking of the MSB bit.
1711 (optimize_range_tests): Add FIRST_BB argument, pass it through
1712 to optimize_range_tests_var_bound.
1713 (maybe_optimize_range_tests, reassociate_bb): Adjust
1714 optimize_range_tests callers.
1716 2018-04-26 Richard Biener <rguenther@suse.de>
1717 Jakub Jelinek <jakub@redhat.com>
1719 * cgraph.h (symbol_table): Just declare debug method here.
1720 * symtab.c (symbol_table::debug): Define.
1722 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
1724 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
1726 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
1728 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
1729 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
1730 (*movdi_internal): Substitute Yi and Yj constraint with x
1731 and Ym and Yn constraint with y constraint. Update "isa"
1732 attribute and set "preferred_for_speed" attribute from
1733 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
1734 (*movsi_internal): Ditto.
1735 (*movdf_internal): Ditto.
1736 (*movsf_internal): Ditto.
1737 (*zero_extendsidi2): Ditto.
1738 * config/i386/sse.md (vec_set<mode>_0): Ditto.
1739 (sse2_loadld): Ditto.
1740 (*vec_extract<ssevecmodelower>_0): Ditto.
1741 (*vec_extractv4si_0_zext_sse4): Ditto.
1742 (vec_concatv2di): Ditto.
1743 (*vec_dup<mode>): Ditto.
1744 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
1745 * config/i386/constraints.md (Yi): Remove.
1750 2018-04-26 Nathan Sidwell <nathan@acm.org>
1752 * dumpfile.c (dump_open): New.
1753 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
1754 (dump_finish): Detect stdio/stderr by value not name.
1756 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
1758 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
1760 2018-04-26 Tom de Vries <tom@codesourcery.com>
1763 * config/nvptx/nvptx.c (verify_neutering_jumps)
1764 (verify_neutering_labels): New function
1765 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
1767 2018-04-26 Tom de Vries <tom@codesourcery.com>
1770 * config/nvptx/nvptx.c (needs_neutering_p): New function.
1771 (nvptx_single): Use needs_neutering_p to skip over insns that do not
1774 2018-04-26 Richard Biener <rguenther@suse.de>
1775 Tom de Vries <tom@codesourcery.com>
1778 * lto-streamer-out.c (output_function): Fixup loops if required to match
1779 discovery done in the reader.
1781 2018-04-26 Richard Biener <rguenther@suse.de>
1783 PR tree-optimization/85116
1784 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
1785 have a loop exit from the single latch predecessor. Remove
1786 case of header with just condition.
1787 (ch_base::copy_headers): Exclude infinite loops from any
1789 (pass_ch::execute): Record exits.
1791 2018-04-26 Richard Biener <rguenther@suse.de>
1793 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
1794 prologue cost vector and pass it to vect_get_load_cost.
1795 (vect_get_peeling_costs_all_drs): Likewise.
1796 (vect_peeling_hash_get_lowest_cost): Likewise.
1797 (vect_enhance_data_refs_alignment): Likewise.
1799 2018-04-26 Richard Biener <rguenther@suse.de>
1802 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
1803 checking of integer<->pointer conversions.
1804 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
1805 sign-/zero-extending pointer types.
1806 (expand_omp_for_static_chunk): Likewise.
1808 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
1809 Jean Lee <xiaoyur347@gmail.com>
1811 * config/mips/mips.c (mips_asan_shadow_offset): New function.
1812 (TARGET_ASAN_SHADOW_OFFSET): Define.
1813 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
1814 true for -fsanitize=address.
1816 2018-04-25 Mark Wielaard <mark@klomp.org>
1818 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
1821 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1823 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
1824 than "alu", remove explicit "memory" and "imm_disp" attributes.
1825 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
1828 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
1829 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
1830 gen_lowpart_no_emit.
1832 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
1835 * config/i386/i386.c (ix86_expand_builtin): Change memory
1836 operand to XI, extend p0 to Pmode.
1837 * config/i386/i386.md: Change unspec volatile and operand
1838 1 mode to XI, change operand 0 mode to P.
1840 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1842 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
1843 GET_MODE_MASK before any checking.
1844 (nds32_can_use_bset_p): Likewise.
1845 (nds32_can_use_btgl_p): Likewise.
1847 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1849 * config/nds32/nds32-doubleword.md: New define_split pattern for
1850 illegal register number.
1852 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1854 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
1856 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1858 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
1860 2018-04-25 Richard Biener <rguenther@suse.de>
1862 * lto-streamer.h (LTO_major_version): Bump to 8.
1864 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1866 * BASE-VER: Set to 9.0.0.
1868 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
1870 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
1871 in __abskf2 and __powikf2.
1873 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1876 * config/aarch64/constraints.md (Usg, Usj): New constraints.
1877 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
1878 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
1879 Use the above on operand 2. Reindent.
1880 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
1882 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1885 * common/config/i386/i386-common.c (ix86_handle_option): Don't
1887 * config/i386/i386.opt (mcet): Removed.
1888 * doc/install.texi: Remove -mcet documentation.
1889 * doc/invoke.texi: Likewise.
1891 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1894 * doc/install.texi: Remove -mcet from bootstrap-cet.
1896 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1899 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
1900 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
1904 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
1905 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
1906 containing a CONST_VECTOR.
1908 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
1910 * doc/install.texi: Update newlib dependency for nvptx.
1912 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1915 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
1916 instead of INTVAL when shifting x left.
1918 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
1920 PR tree-optimization/85478
1921 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
1922 vect_grouped_store_supported for single element vectors.
1924 2018-04-24 Richard Biener <rguenther@suse.de>
1927 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
1928 load cost increase to the case of non-constant step.
1930 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1933 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
1934 destination if any_malformed_asm.
1936 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
1939 * expr.c (store_field): In the bitfield case, if the value comes from
1940 a function call and is returned in registers by means of a PARALLEL,
1941 do not change the mode of the temporary unless BLKmode and VOIDmode.
1943 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
1945 PR rtl-optimization/85423
1946 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
1947 dependencies to debug insns when the previous insn is non-debug.
1949 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1951 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
1952 enums into a single definition.
1953 (fls): Fix predicates and printing.
1956 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1958 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
1959 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
1960 and short u6 immediate.
1961 (check_if_valid_sleep_operand): Remove.
1962 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
1964 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1966 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
1967 flag_always_save_lp condition.
1968 * config/nds32/nds32.opt (malways-save-lp): New option.
1970 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1972 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
1973 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
1974 * config/nds32/nds32.h
1975 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
1976 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
1978 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1980 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
1981 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
1983 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1984 Chung-Ju Wu <jasonwucj@gmail.com>
1986 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
1988 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
1989 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
1991 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1993 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
1995 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1997 * config/nds32/nds32-protos.h (nds32_data_alignment,
1998 nds32_local_alignment): Declare.
1999 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
2000 nds32_local_alignment): New functions.
2001 (TARGET_CONSTANT_ALIGNMENT): Define.
2002 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
2004 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2006 * config/nds32/nds32.c
2007 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
2008 (TARGET_MODES_TIEABLE_P): Likewise.
2010 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2012 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
2015 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
2016 Chung-Ju Wu <jasonwucj@gmail.com>
2018 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2019 for unaligned access.
2020 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2022 * config/nds32/nds32-intrinsic.md: Likewise.
2023 * config/nds32/nds32_intrinsic.h: Likewise.
2024 * config/nds32/nds32.h (nds32_builtins): Likewise.
2025 * config/nds32/nds32.opt (munaligned-access): New option.
2026 * config/nds32/nds32.c (nds32_asm_file_start): Display
2027 flag_unaligned_access status.
2029 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2031 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
2032 -mno-relax is present.
2033 * config/riscv/linux.h (LINK_SPEC): Ditto.
2035 2018-04-20 Martin Sebor <msebor@redhat.com>
2038 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
2040 (gimple_fold_builtin_stxcpy_chk): Same.
2041 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
2043 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
2046 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
2047 __powikf2 when long double is IEEE 128-bit.
2049 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2051 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
2052 step to make sure stack always aligned.
2054 2018-04-20 Carl Love <cel@us.ibm.com>
2057 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
2058 size check for arg0.
2060 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
2061 Tom de Vries <tom@codesourcery.com>
2064 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
2065 Emit insns for calls too.
2066 (nvptx_find_par): Always look for worker-level predecessor insn.
2067 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
2069 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
2070 (nvptx_process_pars): Propagate frames for calls.
2072 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
2075 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
2077 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
2078 (ix86_handle_option): Don't handle OPT_mibt.
2079 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
2081 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
2083 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
2084 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
2085 (ix86_target_macros): Define __CET__ with flag_cf_protection
2086 for -fcf-protection.
2087 * config/i386/i386.c (isa2_opts): Remove -mibt.
2088 * config/i386/i386.h (TARGET_IBT): Removed.
2089 (TARGET_IBT_P): Likewise.
2090 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
2091 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
2092 * config/i386/i386.opt (mcet): Update help message.
2095 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
2096 -mcet as an alias for -mshstk.
2098 2018-04-20 Richard Biener <rguenther@suse.de>
2101 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
2102 complexity by forcing a single use of the multiply operand.
2104 2018-04-20 Martin Jambor <mjambor@suse.cz>
2107 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
2108 recursion dependency to only apply to non-clones.
2110 2018-04-20 Martin Jambor <mjambor@suse.cz>
2113 * ipa-cp.c (create_specialized_node): Check that clones of
2114 self-recursive edges exist during IPA-CP.
2116 2018-04-19 Toon Moene <toon@moene.org>
2118 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
2121 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2123 PR tree-optimization/85467
2124 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
2125 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
2126 VECTOR_CST element to type.
2128 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2131 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
2132 * config/i386/i386.md (builtin_setjmp_setup): Removed.
2133 (builtin_longjmp): Likewise.
2134 (save_stack_nonlocal): New pattern.
2135 (restore_stack_nonlocal): Likewise.
2137 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2140 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2141 Replace ASM_OUTPUT_LABEL with fprintf.
2143 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2146 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2147 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
2148 * config/i386/i386-c.c (ix86_target_macros_internal): Also
2149 define __IBT__ and __SHSTK__ for -fcf-protection.
2150 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
2152 (ix86_trampoline_init): Likewise.
2153 (x86_output_mi_thunk): Likewise.
2154 (ix86_notrack_prefixed_insn_p): Likewise.
2155 (ix86_option_override_internal): Don't disallow -fcf-protection.
2156 * config/i386/i386.md (rdssp<mode>): Also enable for
2158 (incssp<mode>): Likewise.
2159 (nop_endbr): Likewise.
2160 * config/i386/i386.opt (mcet): Change help message to built-in
2164 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
2165 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
2166 enable CET built-in functions.
2168 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
2170 * common/config/i386/i386-common.c
2171 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
2172 OPTION_MASK_ISA_MOVDIRI_UNSET,
2173 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
2174 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
2175 * config.gcc (movdirintrin.h): New header.
2176 * config/i386/cpuid.h (bit_MOVDIRI,
2177 bit_MOVDIR64B): New bits.
2178 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
2180 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
2181 (VOID, PVOID, PCVOID)): New function types.
2182 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
2183 __builtin_ia32_directstoreu_u64,
2184 __builtin_ia32_movdir64b): New builtins.
2185 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
2186 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
2188 (ix86_valid_target_attribute_inner_p): Ditto.
2189 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
2190 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
2191 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
2192 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
2193 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
2194 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
2195 (movdiri<mode>, movdir64b_<mode>): New.
2196 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
2197 * config/i386/immintrin.h: Include movdirintrin.h.
2198 * config/i386/movdirintrin.h: New file.
2199 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
2201 2018-04-19 Richard Biener <rguenther@suse.de>
2204 * cfg.c (clear_bb_flags): When loop state says we have
2205 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
2207 2018-04-19 Richard Biener <rguenther@suse.de>
2209 PR tree-optimization/84737
2210 * tree-vect-data-refs.c (vect_copy_ref_info): New function
2211 copying restrict info.
2212 (vect_setup_realignment): Use it.
2213 * tree-vectorizer.h (vect_copy_ref_info): Declare.
2214 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
2215 the first DR to all generated stores.
2216 (vectorizable_load): Likewise for loads.
2218 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2220 PR tree-optimization/85446
2221 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
2222 the integral and pointer types to have the same precision.
2224 * doc/install.texi: Document --disable-cet being the default and
2227 2018-04-18 Martin Liska <mliska@suse.cz>
2229 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
2232 2018-04-18 Martin Liska <mliska@suse.cz>
2235 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2238 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2239 arguments if they are comparable.
2241 2018-04-18 Martin Liska <mliska@suse.cz>
2244 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2247 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2250 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
2253 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2254 ENDBR after calling __morestack.
2256 2018-04-18 David Malcolm <dmalcolm@redhat.com>
2259 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2260 by using gcc_base_ver to generate a gcc_driver_version, and use
2261 it when generating GCC_DRIVER_NAME.
2262 * configure: Regenerate.
2264 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2267 * config.gcc: Obsolete powerpc*-*-*spe*.
2269 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2272 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2273 (stabstr_D): Change type of unum from unsigned int to
2274 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
2277 2018-04-17 Jim Wilson <jimw@sifive.com>
2280 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2281 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2282 Set arg_pointer_offset after using pretend_args_size.
2284 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2286 PR rtl-optimization/85431
2287 * dse.c (record_store): Ignore zero width stores.
2290 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
2291 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2292 __builtin_stack_restore rather than after it.
2293 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2294 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2295 argument instead of virtual_dynamic_stack_rtx.
2297 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
2299 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2301 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2302 Add note to error message to explain internal mapping of overloaded
2303 built-in function name to non-overloaded built-in function name.
2304 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2307 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2310 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2311 where the inputs overlap with the output.
2313 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2315 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2316 (=v, v) alternative and explicit "memory" attribute.
2317 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
2318 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2320 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2321 "sselog1" type instead of "sselog".
2322 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2323 "sselog". Remove explicit "memory" attribute.
2324 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2325 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2327 (vec_extract_hi_v32hi): Merge all alternatives into one, use
2328 "sselog1" type instead of "sselog". Remove explicit "memory"
2330 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2331 use "sselog1" type instead of "sselog". Remove explicit "memory"
2333 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2334 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2336 (vec_extract_hi_v64qi): Merge all alternatives into one, use
2337 "sselog1" type instead of "sselog". Remove explicit "memory"
2339 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2340 use "sselog1" type instead of "sselog". Remove explicit "memory"
2344 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2347 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2350 2018-04-17 Martin Jambor <mjambor@suse.cz>
2353 * ipa-cp.c (create_specialized_node): Call
2354 expand_all_artificial_thunks if necessary.
2356 2018-04-17 Martin Liska <mliska@suse.cz>
2359 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2360 in message, remote space in between '_G' and '('.
2362 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2365 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2366 avx512f_vmcmp<mode>3<round_saeonly_name>,
2367 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2368 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2369 avx512f_rndscale<mode><round_saeonly_name>,
2370 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2371 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2372 Use %<iptr>2 instead of %2 for -masm=intel.
2373 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2374 avx512f_vcvttss2usi<round_saeonly_name>,
2375 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2377 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2378 avx512f_vcvttsd2usi<round_saeonly_name>,
2379 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2380 Use %q1 instead of %1 for -masm=intel.
2381 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2382 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2383 of %3 for -masm=intel.
2384 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2386 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2388 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2390 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2391 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2393 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2395 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2396 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2397 %g1 and one with %0 and %1.
2398 (avx512er_vmrcp28<mode><round_saeonly_name>,
2399 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2401 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2402 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2403 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2404 of %0 and %{%4%} for -masm=intel.
2405 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2406 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2407 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2408 order of %0 and %{%5%}%{z%} for -masm=intel.
2410 2018-04-17 Jan Hubicka <jh@suse.cz>
2413 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2415 2018-04-17 Martin Liska <mliska@suse.cz>
2418 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2419 for target_clone error message. Make default implementation
2420 clone to be a local declaration.
2421 (separate_attrs): Add new argument and check for an empty
2423 (expand_target_clones): Handle it.
2424 (ipa_target_clone): Make redirection just for target_clones
2427 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2428 Tom de Vries <tom@codesourcery.com>
2431 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2432 tiled basic blocks without omp continue statements.
2434 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2437 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2438 vec_extract expression as having side effects to make sure it gets
2441 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2444 * config/i386/i386.c (get_builtin_code_for_version): Check
2447 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2450 * gcc/config.gcc: Support "skylake".
2451 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2453 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2454 (processor_target_table): Add "skylake".
2455 (ix86_option_override_internal): Add "skylake".
2456 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2457 PROCESSOR_CANNONLAKE.
2458 (get_builtin_code_for_version): Fix priority for
2459 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2460 PROCESSOR_SKYLAKE-AVX512.
2461 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2462 (processor_type): Add PROCESSOR_SKYLAKE.
2464 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
2465 Jason Merrill <jason@redhat.com>
2468 * convert.c (convert_to_integer_1): Use direct recursion for
2469 enumeral types and types with a precision less than the number
2470 of bits in their mode.
2472 2018-04-16 Julia Koval <julia.koval@intel.com>
2475 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2476 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2478 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
2481 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2482 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2483 and -mno-direct-move.
2485 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
2488 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
2489 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
2490 Ensure negative shifts result in {0}.
2492 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
2494 PR rtl-optimization/79916
2495 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
2496 regs (if any) to define how to gnerate SD moves when LRA is in
2499 2018-04-13 Jakub Jelinek <jakub@redhat.com>
2501 PR rtl-optimization/85393
2502 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
2503 * except.c (expand_dw2_landing_pad_for_region): Make static.
2504 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2505 a label and unconditional jump to old_bb, rather than
2506 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2509 PR rtl-optimization/85376
2510 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2511 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2512 instead of a specific value.
2514 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2515 Bin Cheng <bin.cheng@arm.com>
2517 PR tree-optimization/82965
2518 PR tree-optimization/83991
2519 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2520 by_profile_only parameter.
2521 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2522 information if the loop was predicted to iterate too many times.
2523 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2525 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2528 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
2531 2018-04-13 Martin Liska <mliska@suse.cz>
2532 Jakub Jelinek <jakub@redhat.com>
2535 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
2536 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
2537 * builtins.c (expand_builtin_memory_copy_args): Use
2538 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
2539 handle dest_addr == pc_rtx.
2541 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
2544 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
2545 asked to not generate direct moves.
2546 (fix_trunc<mode>si2_stfiwx): Similar.
2547 (fix_trunc<mode>si2_internal): Similar.
2549 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2552 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
2553 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
2554 lookup if dest in some wider mode is known to be const0_rtx and
2555 if so, record permanent equivalence for it to be ZERO_EXTEND of
2556 the narrower mode destination.
2558 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2560 * lto-streamer-out.c (output_function): Revert 259346.
2561 * omp-expand.c (expand_oacc_for): Likewise.
2563 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
2565 PR rtl-optimization/85354
2566 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
2567 * sel-sched.c (sel_global_init): ... here.
2569 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
2572 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
2573 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2574 mode for PE-COFF targets.
2575 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
2576 (i386_pe_asm_lto_end): Likewise.
2577 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
2578 (TARGET_ASM_LTO_END): Likewise.
2579 * config/i386/winnt.c (saved_debug_info_level): New static variable.
2580 (i386_pe_asm_lto_start): New function.
2581 (i386_pe_asm_lto_end): Likewise.
2583 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2584 Richard Biener <rguenther@suse.de>
2587 * lto-streamer-out.c (output_function): Fix CFG loop state before
2589 * omp-expand.c (expand_oacc_for): Handle calls to internal
2590 functions like regular functions.
2592 2018-04-12 Richard Biener <rguenther@suse.de>
2595 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
2596 for the early LTO debug to properly generate references to it
2597 during DIE emission. Do not re-use that for the skeleton for
2599 (dwarf2out_early_finish): Likewise.
2601 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2604 * config/i386/sse.md
2605 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
2606 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
2607 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
2608 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
2609 and output is a reg, avoid creating invalid lowpart subreg, but
2610 instead split into a 512-bit move. Don't split if not AVX512VL,
2611 input is xmm16+ reg and output is a mem.
2612 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
2613 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
2614 xmm16+ reg and output is a mem.
2616 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2618 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
2619 also for flag_dwarf2_cfi_asm.
2621 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2623 PR rtl-optimization/85342
2624 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
2625 a bool scalar var inside of the loop instead. Don't try to update
2626 recog_data.operand after failed apply_change_group.
2628 2018-04-12 Tom de Vries <tom@codesourcery.com>
2631 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
2632 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
2633 array with flexible array member as array without given dimension.
2634 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
2635 argument for undefined param to true.
2637 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
2640 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2641 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
2642 from PowerPC section.
2643 * config/rs6000/sysv4.opt (mcall-): Improve help text.
2644 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
2645 help text that is too long.
2646 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
2647 help text that is too long.
2648 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
2649 help text that is too long.
2651 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
2653 * config/alpha/alpha.md (stack_probe_internal): Rename
2654 from "probe_stack". Update all callers.
2656 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2658 PR rtl-optimization/84566
2659 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
2660 sched_macro_fuse_insns.
2662 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2665 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
2666 (compute_block_dependences): ... from here.
2668 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2670 PR tree-optimization/85331
2671 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
2672 from int to HOST_WIDE_INT.
2674 2018-04-11 Martin Jambor <mjambor@suse.cz>
2677 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
2678 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
2679 not the same as the source val.
2680 (cgraph_edge_brings_value_p): New parameter.
2681 (gather_edges_for_value): Pass destination value to
2682 cgraph_edge_brings_value_p.
2683 (perhaps_add_new_callers): Likewise.
2684 (get_info_about_necessary_edges): Likewise and exclude values brought
2685 only by self-recursive edges.
2686 (create_specialized_node): Redirect only clones of self-calling edges.
2687 (+self_recursive_pass_through_p): New function.
2688 (find_more_scalar_values_for_callers_subset): Use it.
2689 (find_aggregate_values_for_callers_subset): Likewise.
2690 (known_aggs_to_agg_replacement_list): Removed.
2691 (decide_whether_version_node): Re-calculate known constants for all
2692 remaining context clones.
2694 2018-04-11 Richard Biener <rguenther@suse.de>
2697 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
2698 from early DWARF output.
2699 (dwarf2out_early_finish): Output line info unconditionally into
2700 early DWARF and add reference to it.
2702 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2705 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
2706 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
2707 other than V2DFmode using iptr mode attribute.
2708 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
2710 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2712 PR rtl-optimization/84659
2713 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
2715 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2718 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
2720 (output_loc_list): Pass address of a dummy size variable even in the
2721 locview handling loop.
2722 (index_location_lists): Add comment on why skip_loc_list_entry can't
2725 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
2728 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
2731 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
2734 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2735 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
2736 and -mstring-compare-inline-limit.
2738 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2741 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
2742 for stack clash protection in a register whenever we need it to be in
2745 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2747 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
2748 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
2750 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2753 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
2755 (mlong-double-): Ditto.
2756 * config/rs6000/sysv4.opt (msdata=): Ditto.
2757 (mtls-size=): Ditto.
2759 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
2761 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2762 erroneous entries for
2763 "vector int vec_ldl (int, long int *)", and
2764 "vector unsigned int vec_ldl (int, unsigned long int *)".
2765 Add comments and entries for
2766 "vector bool char vec_ldl (int, bool char *)",
2767 "vector bool short vec_ldl (int, bool short *)",
2768 "vector bool int vec_ldl (int, bool int *)",
2769 "vector bool long long vec_ldl (int, bool long long *)",
2770 "vector pixel vec_ldl (int, pixel *)",
2771 "vector long long vec_ldl (int, long long *)",
2772 "vector unsigned long long vec_ldl (int, unsigned long long *)".
2773 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
2774 type tree bool_long_long_type_node and correct definition of
2775 bool_V2DI_type_node to make reference to this new type tree.
2776 (rs6000_mangle_type): Replace erroneous reference to
2777 bool_long_type_node with bool_long_long_type_node.
2778 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
2779 comments to emphasize sign distinctions for char and int types and
2780 replace RS6000_BTI_bool_long constant with
2781 RS6000_BTI_bool_long_long constant. Also add comment to restrict
2782 use of RS6000_BTI_pixel.
2783 (bool_long_type_node): Remove this macro definition.
2784 (bool_long_long_type_node): New macro definition
2786 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2788 PR rtl-optimization/85300
2789 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
2790 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
2791 simplify_unary_operation fails.
2793 2018-04-10 Martin Liska <mliska@suse.cz>
2795 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
2796 cgraph_edge and ipa_ref.
2798 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2802 * config/i386/sse.md
2803 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
2804 computation of the VEC_MERGE selector from mask.
2805 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
2806 Fix decoding of the VEC_MERGE selector into mask.
2808 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2810 PR tree-optimization/85286
2811 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
2813 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2815 * final.c (final_1): Set insn_last_address as well as
2816 insn_current_address.
2818 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2821 * explow.c (emit_stack_probe): Call validize_mem on memory location
2822 before passing it to gen_probe_stack. Create address operand and
2823 legitimize it for the probe_stack_address case.
2825 2018-04-09 Jan Hubicka <jh@suse.cz>
2828 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
2829 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
2830 * tree.c (free_lang_data_in_type): Fix handling of binfos;
2832 (free_lang_data): Rebuild type inheritance graph.
2834 2018-04-09 Martin Sebor <msebor@redhat.com>
2836 * invoke.texi (-finline-small-functions): Mention other optimization
2838 (-findirect-inlining, -fpartial-inlining): Same.
2839 (-finline-functions-called-once): Same.
2840 (-freorder-blocks-and-partition): Same.
2842 2018-04-09 Jan Hubicka <jh@suse.cz>
2845 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
2846 jumps; choose last target that matches the criteria (i.e.
2847 no partition changes for non-crossing jumps).
2848 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
2849 support for redirecting crossing jumps to non-crossing.
2851 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
2853 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
2854 also for naked functions.
2856 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
2858 * config/arc/arc.md (add_shift): New pattern.
2859 (add_shift2): Likewise.
2860 (sub_shift): Likewise.
2861 (sub_shift_cmp0_noout): Likewise.
2862 (compare_si_ashiftsi): Likewise.
2863 (xbfu_cmp0_noout): New combine pattern.
2864 (xbfu_cmp0"): Likewise.
2865 (movsi_set_cc_insn): Place the predicable variant first.
2866 (commutative_binary_cmp0_noout): Remove clobber.
2867 (commutative_binary_cmp0): New pattern.
2868 (noncommutative_binary_cmp0): Likewise.
2869 (noncommutative_binary_cmp0_noout): Likewise.
2870 (noncommutative_binary_comparison_result_used): Removed.
2871 (rsub_cmp0): New pattern.
2872 (rsub_cmp0_noout): Likewise.
2873 (extzvsi): Changed, keep only meaningful variants.
2874 (SQH, SEZ): New iterators.
2875 (SQH_postfix): New mode attribute.
2876 (SEZ_prefix): New code attribute.
2877 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
2878 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
2879 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
2881 (noncommutative_operator): Check the availability of barrel
2884 2018-04-09 Richard Biener <rguenther@suse.de>
2886 PR tree-optimization/85284
2887 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
2888 Only use the niter constraining form of simple_iv when the exit
2891 2018-04-09 Tom de Vries <tom@codesourcery.com>
2894 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2895 (define_expand "*memory_barrier"): New define_expand.
2896 (define_insn "memory_barrier"): New insn.
2898 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2900 PR rtl-optimization/80463
2901 PR rtl-optimization/83972
2902 PR rtl-optimization/83480
2904 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
2905 correct producer for the insn.
2906 (tidy_control_flow): Fixup seqnos in case of debug insns.
2908 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2910 PR rtl-optimization/83913
2912 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
2913 different sched-times when merging exprs.
2915 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2917 PR rtl-optimization/83962
2919 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
2920 tidy_fallthru_edge and tidy_control_flow.
2922 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2924 PR rtl-optimization/83530
2926 * sel-sched.c (force_next_insn): New global variable.
2927 (remove_insn_for_debug): When force_next_insn is true, also leave only
2928 next insn in the ready list.
2929 (sel_sched_region): When the region wasn't scheduled, make another pass
2930 over it with force_next_insn set to 1.
2932 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2934 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
2936 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2937 for interrupt control.
2938 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2939 functions for interrupt control.
2940 * config/nds32/nds32-intrinsic.md: Likewise.
2941 * config/nds32/nds32_intrinsic.h: Likewise.
2942 * config/nds32/nds32.h (nds32_builtins): Likewise.
2944 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2946 * config/nds32/nds32.c (nds32_init_machine_status,
2947 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
2948 strict_aligned_p field.
2949 (nds32_expand_to_rtl_hook): New function.
2950 (TARGET_EXPAND_TO_RTL_HOOK): Define.
2951 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
2953 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2954 Chung-Ju Wu <jasonwucj@gmail.com>
2956 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
2957 * config/nds32/nds32-n7.md: New file.
2958 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
2959 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
2961 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
2962 * config/nds32/nds32.md (pipeline_model): Add n7.
2963 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
2964 * config/nds32/pipelines.md: Include n7 settings.
2966 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2967 Chung-Ju Wu <jasonwucj@gmail.com>
2969 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
2970 * config/nds32/nds32-e8.md: New file.
2971 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
2972 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
2974 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
2975 * config/nds32/nds32.md (pipeline_model): Add e8.
2976 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
2977 * config/nds32/pipelines.md: Include e8 settings.
2979 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2980 Chung-Ju Wu <jasonwucj@gmail.com>
2982 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
2983 * config/nds32/nds32-n8.md: New file.
2984 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
2985 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
2987 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
2988 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
2989 * config/nds32/nds32.md (pipeline_model): Add n8.
2990 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
2991 * config/nds32/pipelines.md: Include n8 settings.
2993 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2994 Chung-Ju Wu <jasonwucj@gmail.com>
2996 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
2997 * config/nds32/nds32-n9-2r1w.md: New file.
2998 * config/nds32/nds32-n9-3r2w.md: New file.
2999 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
3000 nds32_register_ports): New or modify for cpu n9.
3001 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
3003 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
3004 * config/nds32/nds32-utils.c: New file.
3005 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
3006 TARGET_MUL_SLOW): Define.
3007 * config/nds32/nds32.md (pipeline_model): New attribute.
3008 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
3009 New options that support cpu n9.
3010 * config/nds32/pipelines.md: Include n9 settings.
3011 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
3013 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3015 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
3016 information if necessary.
3017 (output_cond_branch_compare_zero): Likewise.
3018 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
3019 (nds32_target_alignment): Refine for alignment.
3020 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
3021 (FUNCTION_BOUNDARY): Modify.
3022 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
3024 * config/nds32/nds32.opt (malways-align, malign-functions): New.
3026 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3028 * config/nds32/constants.md (unspec_volatile_element): Add values for
3029 TLB operation and data prefetch.
3030 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3031 functions for TLB operation and data prefetch.
3032 * config/nds32/nds32-intrinsic.md: Likewise.
3033 * config/nds32/nds32_intrinsic.h: Likewise.
3034 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
3035 (nds32_print_operand): Likewise.
3036 * config/nds32/nds32.h (nds32_builtins): Likewise.
3038 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
3039 Andrew Pinski <pinsika@gcc.gnu.org>
3042 * match.pd: Use constant_boolean_node of correct type instead of
3043 boolean_true_node or boolean_false_node for simplifying
3044 pointer comparisons to zero.
3046 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3048 PR tree-optimization/80021
3049 * tree.c (verify_type_variant): Make error call in verify_variant_match
3050 translatable and remove final full stop.
3052 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3054 * config/nds32/constants.md (unspec_volatile_element): Add
3055 UNSPEC_VOLATILE_EH_RETURN.
3056 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
3057 nds32_output_stack_pop): Support dwarf exception handling process.
3058 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
3059 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
3060 exception handling process.
3061 (nds32_compute_stack_frame): Likewise.
3062 (nds32_return_addr_rtx): Likewise.
3063 (nds32_initial_elimination_offset): Likewise.
3064 (nds32_expand_prologue): Likewise.
3065 (nds32_expand_epilogue): Likewise.
3066 (nds32_dynamic_chain_address): New function.
3067 * config/nds32/nds32.h (machine_function): Add fields for dwarf
3069 (DYNAMIC_CHAIN_ADDRESS): Define.
3070 (EH_RETURN_DATA_REGNO): Define.
3071 (EH_RETURN_STACKADJ_RTX): Define.
3072 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
3073 patterns for dwarf exception handling.
3075 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3077 * config/nds32/nds32.h: Clean up obsolete macros.
3079 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3081 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3082 Add enum values for particular instructions.
3083 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3084 particular intrinsic functions.
3085 * config/nds32/nds32-intrinsic.md: Likewise.
3086 * config/nds32/nds32_intrinsic.h: Likewise.
3087 * config/nds32/nds32.h (nds32_builtins): Likewise.
3088 * config/nds32/nds32.md (type): Add pbsad and pbsada.
3089 (btst, ave): New patterns for particular instructions.
3091 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3093 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3094 Add enum values for atomic load/store and memory sync.
3095 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
3097 * config/nds32/nds32-intrinsic.md: Likewise.
3098 * config/nds32/nds32_intrinsic.h: Likewise.
3099 * config/nds32/nds32.h (nds32_builtins): Likewise.
3101 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3103 PR tree-optimization/85257
3104 * fold-const.c (native_encode_vector): If not all elts could fit
3105 and off is -1, return 0 rather than offset.
3106 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
3107 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
3108 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
3109 adjust buffer in native_interpret_expr call.
3111 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3113 * config/nds32/constants.md (unspec_volatile_element): Add cache
3114 control enum values.
3115 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
3116 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
3117 * config/nds32/nds32.c (nds32_cctl_names): New.
3118 (nds32_print_operand): Handle cache control register names.
3119 * config/nds32/nds32.h (nds32_builtins): New enum values.
3120 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
3122 * config/nds32/nds32.md (type): Add mmu.
3123 * config/nds32/pipelines.md (simple_insn): Add mmu.
3125 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3127 * config/nds32/nds32.md (type): Remove call.
3128 * config/nds32/pipelines.md (simple_insn): Likewise.
3130 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3132 * config/nds32/constants.md (unspec_volatile_element): Add
3133 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
3134 UNSPEC_VOLATILE_FMFCFG.
3135 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
3136 description for fmfcfg and fmfcsr.
3137 (bdesc_1arg): Add fmtcsr.
3138 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
3139 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
3140 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
3141 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
3142 unspec_fmfcfg): New patterns.
3143 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
3144 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
3145 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
3146 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
3147 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
3148 __nds32__fmfcfg): Define.
3150 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3152 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
3153 intrinsic register names.
3154 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
3155 intrinsic register enum values and macros.
3157 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3159 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
3160 for load/store addressing form.
3161 (nds32_print_operand_address): Likewise.
3163 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
3166 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
3167 based on LABEL_REF. Remove useless assertion.
3168 (pic_address_needs_scratch): Fix formatting.
3169 (sparc_legitimize_pic_address): Minor tweaks.
3170 (sparc_delegitimize_address): Adjust assertion accordingly.
3171 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
3172 into symbolic_operand.
3173 (movsi_high_pic_label_ref): Likewise.
3174 (movsi_lo_sum_pic_label_ref): Likewise.
3175 (movdi_pic_label_ref): Likewise.
3176 (movdi_high_pic_label_ref): Likewise.
3177 (movdi_lo_sum_pic_label_ref): Likewise.
3179 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
3181 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
3182 custom LIB_SPEC setup.
3184 2018-04-06 Ruslan Bukin <br@bsdpad.com>
3185 Kito Cheng <kito.cheng@gmail.com>
3187 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
3188 * config/riscv/freebsd.h: New.
3190 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3192 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
3193 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
3196 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3197 Kito Cheng <kito.cheng@gmail.com>
3199 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
3200 nds32_output_call, nds32_symbol_binds_local_p): New functions.
3201 * config/nds32/nds32-protos.h (nds32_output_call,
3202 nds32_output_return): Declare.
3203 * config/nds32/nds32.md: Refine all the call and return patterns.
3205 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3208 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
3209 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
3211 PR rtl-optimization/84872
3212 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
3213 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
3216 2018-04-06 Tamar Christina <tamar.christina@arm.com>
3218 * expr.c (copy_blkmode_to_reg): Revert 254862.
3219 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
3221 2018-04-06 Richard Biener <rguenther@suse.de>
3224 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
3225 after seeing a component reference with an adjacent field. Treat
3226 refs to arrays at struct end of external decls similar to
3227 refs to unconstrained commons.
3229 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3232 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
3233 look through SAVE_EXPRs with non-side-effects argument. Adjust
3235 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3238 2018-04-06 Richard Biener <rguenther@suse.de>
3241 * alias.c (find_base_term): New wrapper around find_base_term
3242 unwinding CSELIB_VAL_PTR changes.
3243 (find_base_term): Do not restore CSELIB_VAL_PTR during the
3246 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3248 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3250 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3251 constant definitions.
3252 ("nop"): lr 0,0 -> nopr r0
3253 ("nop_lr0", "nop_lr1"): New insn definitions.
3255 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3257 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3258 NDS32_V3PUSH_AVAILABLE_P macro.
3260 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
3261 Chung-Ju Wu <jasonwucj@gmail.com>
3263 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3264 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3265 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3266 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3267 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3268 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3269 * config/nds32/constraints.md: New constraints and checking for hard
3270 float configuration.
3271 * config/nds32/iterators.md: New mode iterator and attribute for hard
3272 float configuration.
3273 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3275 * config/nds32/nds32-fpu.md: New file.
3276 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3277 deal with hard float code generation.
3278 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3280 (abi_type, float_reg_number): New enum type.
3281 * config/nds32/nds32-predicates.c: New predicates for hard float.
3282 * config/nds32/nds32-protos.h: Declare functions for hard float.
3283 * config/nds32/nds32.c: Implementation for hard float configuration.
3284 * config/nds32/nds32.h: Definitions for hard float configuration.
3285 * config/nds32/nds32.md: Include hard float machine description and
3286 modify patterns for hard float configuration.
3287 * config/nds32/nds32.opt: New options for hard float configuration.
3288 * config/nds32/predicates.md: New predicates for hard float
3291 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3293 * common/config/nds32/nds32-common.c
3294 (nds32_option_optimization_table): Enable -mreleax-hint by default.
3296 2018-04-05 Jakub Jelinek <jakub@redhat.com>
3299 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3300 CONSTRUCTOR_ELT (ctor, ...)->value.
3302 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
3305 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3307 2018-04-05 Tom de Vries <tom@codesourcery.com>
3310 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3313 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3314 Kito Cheng <kito.cheng@gmail.com>
3316 * config/nds32/constraints.md (U33): Fine-tune checking condition.
3317 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3318 * config/nds32/nds32.h (nds32_16bit_address_type): Add
3319 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3321 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3322 Kito Cheng <kito.cheng@gmail.com>
3324 * config/nds32/constraints.md (Ufe): New memory constraint.
3325 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3326 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3327 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3329 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3330 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3332 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3334 * config/nds32/nds32.md: Use optimize_size in the condition for
3335 alu-shift instructions.
3337 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3339 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3341 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3343 * config/nds32/nds32.md (negsi2): Refine pattern.
3345 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3346 Chung-Ju Wu <jasonwucj@gmail.com>
3348 * config/nds32/iterators.md (shift_rotate): New code iterator.
3349 (shift): New code attribute.
3350 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3351 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3352 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3353 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3354 bit-wise operations.
3355 (andsi3, *andsi3): Ditto.
3356 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3357 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3358 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3359 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3360 nds32_ior_operand, nds32_xor_operand): New predicates.
3362 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3364 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3365 (addsi3, subsi3): ... this.
3367 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3369 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3371 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3373 * config/nds32/nds32.md: Adjust indention.
3375 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3377 * config/nds32/nds32.md (feature): New attribute.
3379 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3381 * config/nds32/nds32.md (subtype): New attribute.
3383 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
3386 * config/arm/arm-builtins.c (arm_expand_builtin): Change
3387 expansion to perform a bitwise AND of the argument followed by a
3388 boolean negation of the result.
3390 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3392 PR rtl-optimization/84878
3393 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3394 the basic block. Assert the use reference is not artificial and that
3395 it has an associated insn.
3397 2018-04-04 Michael Matz <matz@suse.de>
3399 * builtins.c (compute_objsize): Pass correct operand
3400 to array_at_struct_end_p.
3402 2018-04-04 Richard Biener <rguenther@suse.de>
3405 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3406 from contexts for DINFO_LEVEL_TERSE and below.
3408 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3410 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3411 resiter_operand condition.
3412 * config/nds32/nds32.md (*move<mode>): Ditto.
3414 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3415 Monk Chiang <sh.chiang04@gmail.com>
3417 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3419 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3421 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3423 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3424 Kito Cheng <kito.cheng@gmail.com>
3426 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3427 nds32_cond_code_str, output_cond_branch,
3428 output_cond_branch_compare_zero, nds32_expand_cbranch,
3429 nds32_expand_cstore, nds32_expand_movcc,
3430 nds32_output_cbranchsi4_equality_zero,
3431 nds32_output_cbranchsi4_equality_reg,
3432 nds32_output_cbranchsi4_equality_reg_or_const_int,
3433 nds32_output_cbranchsi4_greater_less_zero: New functions.
3434 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3435 nds32_expand_cstore, nds32_expand_movcc,
3436 nds32_output_cbranchsi4_equality_zero,
3437 nds32_output_cbranchsi4_equality_reg,
3438 nds32_output_cbranchsi4_equality_reg_or_const_int,
3439 nds32_output_cbranchsi4_greater_less_zero): Declare.
3440 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3441 nds32_rimm11s_operand): New predicates.
3442 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3443 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3446 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3448 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3449 * config/nds32/nds32.md: Ditto.
3450 * config/nds32/pipelines.md: Ditto.
3452 2018-04-04 Richard Biener <rguenther@suse.de>
3454 PR tree-optimization/85168
3455 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3456 propagating abnormals.
3458 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3460 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3462 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3463 Kito Cheng <kito.cheng@gmail.com>
3465 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3466 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3467 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3468 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3469 * config/nds32/nds32.md (sibcall_internal): New.
3470 (sibcall_register): Remove.
3471 (sibcall_immediate): Remove.
3472 (sibcall_value_internal): New.
3473 (sibcall_value_register): Remove.
3474 (sibcall_value_immediate): Remove.
3475 * config/nds32/predicates.md (nds32_general_register_operand): New.
3476 (nds32_call_address_operand): New.
3478 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3480 PR rtl-optimization/85167
3481 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3482 bb_defs if *split_p, instead preinitialize it to NULL.
3484 PR tree-optimization/85156
3485 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3486 evaluating the argument multiple times.
3488 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
3490 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
3492 (_mm_cvtpd_ps): Likewise.
3493 (_mm_cvttpd_epi32): Likewise.
3494 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
3495 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
3496 vector, pixel, and bool following altivec.h include.
3498 2018-04-03 Martin Sebor <msebor@redhat.com>
3500 * doc/extend.texi (Common Function Attributes): Clarify.
3501 (const attribute): Likewise.
3502 (pure attribute): Likewise.
3504 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3507 * config/i386/i386.c (ix86_expand_vector_set): Use
3508 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
3510 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
3512 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3513 instructions when changing rounding bits to preserve precision bits
3514 in the x87 control word.
3516 2018-04-03 Martin Liska <mliska@suse.cz>
3518 PR tree-optimization/82491
3519 * rtl.h (strip_offset_and_add): Replace += suboffset with
3520 poly_uint64 () + suboffset.
3522 2018-03-29 Martin Liska <mliska@suse.cz>
3523 Martin Jambor <mjambor@suse.cz>
3526 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
3527 param_type is not an integral or pointer type.
3529 2018-04-03 Richard Biener <rguenther@suse.de>
3531 * sese.h (recompute_all_dominators): Remove.
3533 2018-04-02 Martin Sebor <msebor@redhat.com>
3535 * doc/invoke.texi (-Wrestrict): Fix typos.
3537 2018-04-02 Jim Wilson <jimw@sifive.com>
3539 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
3540 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
3541 (<optab>di3, <optab>si3_extend): Likewise.
3542 (<optab>si3_mask, <optab>si3_mask_1): New.
3543 (<optab>di3_mask, <optab>di3_mask_1): New.
3544 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
3545 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
3546 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
3548 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
3550 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
3553 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
3555 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
3556 (nds32_canonicalize_comparison): New function.
3558 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3559 Kito Cheng <kito.cheng@gmail.com>
3560 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3562 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
3563 * config/nds32/constants.md (unspec_volatile_element): Add
3564 UNSPEC_VOLATILE_RELAX_GROUP.
3565 * config/nds32/nds32-relax-opt.c: New file.
3566 * config/nds32/nds32-predicates.c
3567 (nds32_symbol_load_store_p): New function.
3568 * config/nds32/nds32-protos.h
3569 (nds32_symbol_load_store_p): Declare function.
3570 (make_pass_nds32_relax_opt): Declare new rtl pass function.
3571 * config/nds32/nds32.c
3572 (nds32_register_pass): New function to register pass.
3573 (nds32_register_passes): New function to register passes.
3574 * config/nds32/nds32.md (relax_group): New pattern.
3575 * config/nds32/nds32.opt (mrelax-hint): New option.
3576 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
3578 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
3580 * config/nds32/t-nds32: Modify files dependency.
3582 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3584 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
3585 (PROFILE_HOOK): Define its implementation.
3587 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3589 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
3590 type and 32-bit size.
3592 2018-04-01 Jakub Jelinek <jakub@redhat.com>
3595 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
3596 (V_128_256): New mode iterator.
3597 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
3598 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
3599 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
3601 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
3602 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
3604 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
3607 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
3608 NaN inputs correctly.
3610 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
3613 * config/rs6000/vsx.md (??r): New mode attribute.
3614 (*vsx_mov<mode>_64bit): Use it.
3615 (*vsx_mov<mode>_32bit): Likewise.
3617 2018-03-30 Martin Sebor <msebor@redhat.com>
3619 PR tree-optimization/84818
3620 * builtins.c (check_access): Use warning_n.
3622 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3625 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
3627 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
3630 2018-03-30 Julia Koval <julia.koval@intel.com>
3633 * x86-tune.def (movx, partial_reg_dependency): Enable for
3636 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
3639 * lra-constraints.c (process_alt_operands): Move setting
3640 this_alternative_matches below.
3642 2018-03-29 Martin Liska <mliska@suse.cz>
3645 * doc/invoke.texi: Document how LTO works with debug info.
3646 Describe auto-load support of binutils. Mention 'x86-64'
3647 as valid option value of -march option.
3649 2018-03-29 Jakub Jelinek <jakub@redhat.com>
3651 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
3654 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
3655 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
3656 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
3659 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3662 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
3663 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
3664 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
3665 for RS6000_BTM_POWERPC64.
3666 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
3667 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
3668 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
3673 2018-03-28 Carl Love <cel@us.ibm.com>
3676 2017-09-27 Carl Love <cel@us.ibm.com>
3678 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
3679 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
3680 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
3683 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3685 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
3686 instead of __vector bool.
3687 (_mm_max_pu8): Likewise.
3688 (_mm_min_pi16): Likewise.
3690 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3693 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
3694 (DIVWEUO): Likewise.
3696 (DIVDEUO): Likewise.
3697 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
3698 DIVWEUO and DIVDEUO.
3699 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
3700 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
3701 (div_extend): Likewise.
3702 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
3704 (__builtin_divweuo): Likewise.
3705 (__builtin_divdeo): Likewise.
3706 (__builtin_divdeuo): Likewise.
3708 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3711 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
3712 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
3714 PR tree-optimization/82004
3715 * gimple-match-head.c (optimize_pow_to_exp): New function.
3716 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
3717 Don't fold to exp if optimize_pow_to_exp is false.
3719 2018-03-28 Martin Liska <mliska@suse.cz>
3722 * calls.c (initialize_argument_information): Fix trailing space.
3723 * common.opt: Fix typo and provide better explanation for
3724 -fsanitize-coverage option.
3725 * config/i386/i386.opt: Fix typo.
3727 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3728 Martin Liska <mliska@suse.cz>
3731 * gimplify.c (asan_poison_variable): Don't do the check for
3732 gimplify_omp_ctxp here.
3733 (gimplify_decl_expr): Do it here.
3734 (gimplify_target_expr): Likewise.
3736 2018-03-28 Martin Liska <mliska@suse.cz>
3739 * config/i386/i386.c (ix86_function_arg_advance): Do not call
3740 chkp_type_bounds_count if MPX is not enabled.
3742 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
3744 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
3746 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
3749 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
3750 function to create the function decl for complex long double
3751 multiply and divide for -mabi=ieeelongdouble.
3752 (init_float128_ieee): Call it.
3754 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
3757 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
3758 -fcf-protection=branch -mibt.
3759 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
3761 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3764 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
3766 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
3769 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
3770 extern array declarations.
3772 2018-03-27 Richard Biener <rguenther@suse.de>
3775 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
3776 explicit single_use checks.
3778 2018-03-27 Richard Biener <rguenther@suse.de>
3780 PR tree-optimization/85082
3781 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
3784 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3786 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
3787 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
3788 Turn on fasynchronous-unwind-tables and funwind-tables.
3790 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
3793 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
3794 (*bmi_blsr_<mode>_ccz): Ditto.
3796 2018-03-26 Tom de Vries <tom@codesourcery.com>
3798 PR tree-optimization/85063
3799 * omp-general.c (offloading_function_p): New function. Factor out
3801 * omp-offload.c (pass_omp_target_link::gate): ... here.
3802 * omp-general.h (offloading_function_p): Declare.
3803 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
3804 with attribute omp declare target for offloading functions.
3806 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
3808 PR tree-optimization/84005
3809 * tree-data-ref.h (get_base_for_alignment): Declare.
3810 * tree-data-ref.c (get_base_for_alignment_1): New function.
3811 (get_base_for_alignment): Likewise.
3812 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
3813 get_base_for_alignment to find a suitable base object, instead
3814 of always using drb->base_address.
3816 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3819 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
3820 known size by default.
3822 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
3825 * lra-constraints.c (process_alt_operands): Don't match BLKmode
3826 and non BLKmode operands.
3828 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3831 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
3832 Clean up attributes.
3834 2018-03-23 Richard Biener <rguenther@suse.de>
3837 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
3838 we are going to emit early debug for LTO.
3840 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3843 * function.c (match_asm_constraints_1): Don't optimize if input
3844 doesn't satisfy general_operand predicate for output's mode.
3847 * alias.c (write_dependence_p): Don't require for x_canonicalized
3848 non-VOIDmode if x has VOIDmode.
3851 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
3852 just don't try to optimize it rather than assert it never happens.
3854 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3856 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
3857 macro expansions for definition of ST_INTERNAL_<mode> and
3858 LD_INTERNAL_<mode> builtins.
3859 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
3861 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
3863 (altivec_expand_st_builtin): Likewise.
3864 (altivec_expand_builtin): Remove calls to deleted functions.
3865 (rs6000_address_for_altivec): Delete this function.
3866 * config/rs6000/vector.md: Remove expands for
3867 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
3869 2018-03-22 Sudakshina Das <sudi.das@arm.com>
3872 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
3873 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
3874 re-computing once computed.
3875 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
3876 (arm_init_machine_status): Initialize
3877 machine->static_chain_stack_bytes.
3879 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3882 * doc/extend.texi: Add four new prototypes for vec_ld.
3883 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
3884 definitions for more logical presentation.
3885 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
3886 entries for V1TI variants of __builtin_altivec_ld builtin.
3887 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
3888 handling of V1TI variant of LVX icode pattern.
3889 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
3890 (rs6000_gimple_fold_builtin): Likewise.
3891 (altivec_init_builtins): Add code to define
3892 __builtin_altivec_lvx_v1ti function.
3894 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3897 * function.c (match_asm_constraints_1): Don't do the optimization
3898 if input isn't a REG, SUBREG, MEM or constant.
3900 2018-03-22 Tom de Vries <tom@codesourcery.com>
3902 PR tree-optimization/84956
3903 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
3904 bb_has_abnormal_pred.
3906 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3909 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
3910 DECL_INITIAL (decl) to decl at the end.
3911 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
3914 2018-03-21 Joseph Myers <joseph@codesourcery.com>
3916 * doc/extend.texi (__builtin_tgmath): Document when complex
3917 integer types are treated as _Complex _Float64.
3919 2018-03-21 Tom de Vries <tom@codesourcery.com>
3921 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
3923 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3925 PR tree-optimization/84960
3926 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
3927 if it is ENTRY block, move them into single succ of ENTRY in that case.
3929 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
3931 PR tree-optimization/84811
3932 * poly-int.h (poly_span_traits): Remove the T3 parameter and
3933 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
3934 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
3935 (known_subrange_p): Update accordingly. Cast each value involved
3936 in the size comparison, rather than casting the result of the
3939 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3941 PR tree-optimization/84982
3942 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
3943 by flipping the least significant bit rather than all bits from
3944 bitpos to bitpos + bitsize - 1.
3946 2018-03-21 Nathan Sidwell <nathan@acm.org>
3948 * doc/extend.texi (Deprecated Features): Remove mention of
3949 long-deleted deprecations.
3951 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3954 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
3955 * configure: Regenerate.
3957 2018-03-21 Tom de Vries <tom@codesourcery.com>
3959 PR tree-optimization/83126
3960 * tree-parloops.c (num_phis): New function.
3961 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
3963 2018-03-21 Nathan Sidwell <nathan@acm.org>
3965 * doc/extend.texi (Deprecated Features): Update deprecated flags,
3966 mention anon-struct/union members and trailing attributes.
3968 2018-03-21 Bin Cheng <bin.cheng@arm.com>
3970 PR tree-optimization/84969
3971 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
3972 builtin memset partitions if they set different rhs values.
3974 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3976 PR rtl-optimization/84989
3977 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
3978 VEC_DUPLICATE with scalar result mode.
3980 2018-03-21 Martin Liska <mliska@suse.cz>
3983 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
3984 not intended return statement.
3986 2018-03-21 Martin Liska <mliska@suse.cz>
3989 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
3990 (chkp_find_bound_slots_1): Limit number of iterations.
3992 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
3995 * Minor grammar fixes for x86 options.
3997 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4000 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
4001 holding REG_CFA_RESTORE notes, instead turn them into a USE.
4003 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
4006 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
4007 (altivec_lvx_<mode>_1op): Likewise.
4008 (altivec_stvx_<mode>_2op): Likewise.
4009 (altivec_stvx_<mode>_1op): Likewise.
4010 (altivec_lvx_<VM2:mode>): New define_expand.
4011 (altivec_stvx_<VM2:mode>): Likewise.
4012 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
4013 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4014 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
4015 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4016 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
4017 (rs6000_gen_lvx): Likewise.
4018 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
4019 (altivec_expand_stv_builtin): Likewise.
4020 (altivec_expand_builtin): Likewise.
4021 * config/rs6000/vector.md: Likewise.
4023 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4026 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
4029 2018-03-20 Richard Biener <rguenther@suse.de>
4032 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
4033 sign-conversions as zero, fall back to standard scalar_stmt
4036 2018-03-20 Martin Liska <mliska@suse.cz>
4039 * predict.c (rebuild_frequencies): Handle case when we have
4040 PROFILE_ABSENT, but flag_guess_branch_prob is false.
4042 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4045 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
4046 flag_section_anchors.
4047 * varasm.c (use_blocks_for_decl_p): Remove hack for
4048 dw2_force_const_mem.
4051 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
4053 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
4054 be created, use lowpart_subreg of operands[0] rather than operands[0]
4056 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
4057 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
4058 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
4059 and n constraint instead of aarch64_shift_imm_di and Usd.
4060 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
4061 (*aarch64_<optab>_reg_minus<mode>3): ... this.
4063 2018-03-20 Sudakshina Das <sudi.das@arm.com>
4066 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
4067 to favor GPR over NEON registers.
4068 (<shift>di3_neon): Likewise.
4070 2018-03-20 Tom de Vries <tom@codesourcery.com>
4073 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
4074 (nvptx_process_pars): Emit bar.sync asap and alap.
4076 2018-03-20 Tom de Vries <tom@codesourcery.com>
4079 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
4080 seen_label if seen_label is already set.
4082 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4085 * config/i386/i386.c (fold_builtin_cpu): For features above 31
4086 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
4087 Use 1U instead of 1. Formatting fixes.
4090 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
4091 instead of TREE_TYPE (s1) for the return value.
4093 2018-03-19 Jakub Jelinek <jakub@redhat.com>
4095 PR tree-optimization/84946
4096 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
4097 bitsize + bitsize in poly_uint64 rather than poly_int64.
4100 * dwarf2asm.c: Include fold-const.c.
4101 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
4102 of decl rather than decl itself.
4104 PR rtl-optimization/84643
4105 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
4107 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
4110 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
4111 calling assemble_variable.
4113 2018-03-19 Sudakshina Das <sudi.das@arm.com>
4116 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
4117 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
4119 2018-03-19 Jim Wilson <jimw@sifive.com>
4122 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
4123 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
4124 (riscv_first_stack_step): Likewise.
4125 (riscv_option_override): Use STACK_BOUNDARY instead of
4127 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
4129 (BIGGEST_ALIGNMENT): Set to 128.
4130 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
4131 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
4134 2018-03-19 Richard Biener <rguenther@suse.de>
4136 PR tree-optimization/84933
4137 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
4138 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
4140 2018-03-19 Richard Biener <rguenther@suse.de>
4142 PR tree-optimization/84859
4143 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
4144 (cond_if_else_store_replacement): Perform sinking operation on
4145 single-store BBs regardless of MAX_STORES_TO_SINK setting.
4146 Generalize what a BB with a single eligible store is.
4148 2018-03-19 Richard Biener <rguenther@suse.de>
4150 PR tree-optimization/84929
4151 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
4152 chrec_is_positive against non-chrec arg.
4154 2018-03-19 Tamar Christina <tamar.christina@arm.com>
4157 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
4159 2018-03-18 Martin Liska <mliska@suse.cz>
4161 PR rtl-optimization/84635
4162 * regrename.c (build_def_use): Use matches_mode only when
4165 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
4167 PR tree-optimization/84913
4168 * tree-vect-loop.c (vectorizable_reduction): Don't try to
4169 vectorize chains of COND_EXPRs.
4171 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4173 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
4175 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4177 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
4179 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4181 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
4183 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4184 Kito Cheng <kito.cheng@gmail.com>
4186 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
4187 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
4188 (nds32_adjust_reg_alloc_order): New function.
4189 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
4191 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
4193 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
4194 nds32_print_operand, nds32_print_operand_address): Use
4195 HOST_WIDE_INT_PRINT_DEC instead.
4197 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4199 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
4201 2018-03-17 Jakub Jelinek <jakub@redhat.com>
4204 * config/i386/i386.c (initial_ix86_tune_features,
4205 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
4207 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
4208 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
4209 rather than 1u << ix86_tune. Formatting fix.
4210 (ix86_option_override_internal): Change ix86_arch_mask from
4211 unsigned int to unsigned HOST_WIDE_INT, initialize to
4212 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
4213 (ix86_function_specific_restore): Likewise.
4215 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4218 * postreload.c (reload_combine_recognize_pattern): Perform
4219 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
4220 truncate_int_for_mode the result for the destination's mode.
4223 * hsa-gen.c (mem_type_for_type): Fix comment typo.
4224 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
4226 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4229 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
4232 * lra-assigns.c (lra_split_hard_reg_for): Don't use
4233 regno_allocno_class_array and sorted_pseudos.
4234 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
4235 insns where regno is used.
4237 2018-03-16 Martin Liska <mliska@suse.cz>
4240 * multiple_target.c (create_dispatcher_calls): Redirect
4241 reference in the symbol table.
4243 2018-03-16 Martin Liska <mliska@suse.cz>
4246 * multiple_target.c (create_dispatcher_calls): Redirect also
4249 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4253 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4254 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4255 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4257 2018-03-16 Julia Koval <julia.koval@intel.com>
4259 * doc/invoke.texi (Skylake Server): Add CLWB.
4260 Cannonlake): Remove CLWB.
4262 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4264 PR tree-optimization/84841
4265 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4267 (FLOAT_ONE_CONST_TYPE): Define.
4268 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4269 (sort_by_operand_rank): Put entries with higher constant_type last
4270 rather than first to match comments.
4272 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
4274 * config/nios2/nios2.md (movsi_internal): Fix thinko in
4277 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4280 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4281 check and use address of target always.
4283 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
4286 * config/i386/i386.c (indirect_thunk_needed): Update comments.
4287 (indirect_thunk_bnd_needed): Likewise.
4288 (indirect_thunks_used): Likewise.
4289 (indirect_thunks_bnd_used): Likewise.
4290 (indirect_return_needed): New.
4291 (indirect_return_bnd_needed): Likewise.
4292 (output_indirect_thunk_function): Add a bool argument for
4294 (output_indirect_thunk_function): Don't generate alias for
4295 function return thunk.
4296 (ix86_code_end): Call output_indirect_thunk_function to generate
4297 function return thunks.
4298 (ix86_output_function_return): Set indirect_return_bnd_needed
4299 and indirect_return_needed instead of indirect_thunk_bnd_needed
4300 and indirect_thunk_needed.
4302 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
4304 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4305 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4306 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4308 2018-03-15 David Malcolm <dmalcolm@redhat.com>
4309 Paul Hua <paul.hua.gm@gmail.com>
4312 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4314 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
4316 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4317 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4320 2018-03-15 Tamar Christina <tamar.christina@arm.com>
4323 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4324 instead of GET_MODE_SIZE when comparing Units.
4326 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4329 * varasm.c (hash_section): Return an unchangeble hash value
4330 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4331 Return !aarch64_can_use_per_function_literal_pools_p ().
4333 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4336 * optabs.c (emit_conditional_move): Pass address of cmode's copy
4337 rather than address of cmode as last argument to prepare_cmp_insn.
4339 2018-03-15 Julia Koval <julia.koval@intel.com>
4341 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4342 F_AVX512VNNI, F_AVX512BITALG): New.
4344 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4347 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
4348 insn for floating-point loads and stores.
4350 2018-03-14 Carl Love <cel@us.ibm.com>
4352 * config/rs6000/rs6000-c.c: Add macro definitions for
4353 ALTIVEC_BUILTIN_VEC_PERMXOR.
4354 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4355 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4356 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4357 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4359 * config/doc/extend.texi: Add prototypes for vec_permxor.
4361 2018-03-14 David Malcolm <dmalcolm@redhat.com>
4364 * diagnostic-show-locus.c (class layout_point): Convert m_line
4365 from int to linenum_type.
4366 (line_span::comparator): Use linenum "compare" function when
4367 comparing line numbers.
4368 (test_line_span): New function.
4369 (layout_range::contains_point): Convert param "row" from int to
4371 (layout_range::intersects_line_p): Likewise.
4372 (layout::will_show_line_p): Likewise.
4373 (layout::print_source_line): Likewise.
4374 (layout::should_print_annotation_line_p): Likewise.
4375 (layout::print_annotation_line): Likewise.
4376 (layout::print_leading_fixits): Likewise.
4377 (layout::annotation_line_showed_range_p): Likewise.
4378 (struct line_corrections): Likewise for field m_row.
4379 (line_corrections::line_corrections): Likewise for param "row".
4380 (layout::print_trailing_fixits): Likewise.
4381 (layout::get_state_at_point): Likewise.
4382 (layout::get_x_bound_for_row): Likewise.
4383 (layout::print_line): Likewise.
4384 (diagnostic_show_locus): Likewise for locals "last_line" and
4386 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4387 * input.c (selftest::test_linenum_comparisons): New function.
4388 (selftest::input_c_tests): Call it.
4389 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4390 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4391 * selftest.h (ASSERT_GT): New macro.
4392 (ASSERT_GT_AT): New macro.
4393 (ASSERT_LT): New macro.
4394 (ASSERT_LT_AT): New macro.
4396 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4398 PR rtl-optimization/84780
4399 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4401 2018-03-14 Martin Liska <mliska@suse.cz>
4403 * tree.c (record_node_allocation_statistics): Use
4404 get_stats_node_kind.
4405 (get_stats_node_kind): New function extracted from
4406 record_node_allocation_statistics.
4407 (free_node): Use get_stats_node_kind.
4409 2018-03-14 Richard Biener <rguenther@suse.de>
4411 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4412 that the value-set of ANTIC_IN doesn't grow.
4415 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4417 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4418 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4420 2018-03-14 Julia Koval <julia.koval@intel.com>
4422 * config.gcc (icelake-client, icelake-server): New.
4424 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4425 (initial_ix86_arch_features): Ditto.
4426 (PTA_SKYLAKE): Add SGX.
4427 (PTA_ICELAKE): Remove.
4428 (PTA_ICELAKE_CLIENT): New.
4429 (PTA_ICELAKE_SERVER): New.
4430 (ix86_option_override_internal): Split up icelake on icelake client and
4432 (get_builtin_code_for_version): Ditto.
4433 (fold_builtin_cpu): Ditto.
4434 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4435 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4436 * config/i386/i386.h (processor_type): Ditto.
4437 * doc/invoke.texi: Ditto.
4439 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4442 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4443 INTEGER_CST offset, add it together with bitpos / 8 and
4444 sign extend based on POINTER_SIZE.
4448 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4451 * config/i386/constraints.md (Yc): New register constraint.
4452 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4453 Use Yc constraint for alternative 2 of operand 0. Remove
4454 preferred_for_speed attribute.
4456 2018-03-14 Richard Biener <rguenther@suse.de>
4458 PR tree-optimization/84830
4459 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4460 with the old one to avoid oscillations.
4462 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
4465 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4467 (assign_by_spills): Return a flag of reload assignment failure.
4468 Do not process the reload assignment failures. Do not spill other
4469 reload pseudos if they has the same reg class. Update n if
4471 (lra_assign): Add a return arg. Set up from the result of
4472 assign_by_spills call.
4473 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4474 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4475 usage_insns if it is not NULL.
4476 (spill_hard_reg_in_range): New function.
4477 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4478 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4479 function prototypes.
4480 (lra_assign): Change prototype.
4481 * lra.c (lra): Add code to deal with fails by splitting hard reg
4484 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
4486 * config/riscv/riscv.opt (mrelax): New option.
4487 * config/riscv/riscv.c (riscv_file_start): Emit ".option
4488 "norelax" when riscv_mrelax is disabled.
4489 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
4491 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4494 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
4495 reassociation for int modes.
4497 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4499 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
4500 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
4502 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
4503 * config/aarch64/aarch64-sve.md
4504 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4505 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4506 (*extend<mode><Vwide>2): Rename to...
4507 (aarch64_sve_extend<mode><Vwide>2): ...this.
4508 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4509 renaming the old pattern to...
4510 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
4512 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4513 define_expand, renaming the old pattern to...
4514 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4515 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4516 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4517 account when deciding which SVE instruction the optab should use.
4518 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4520 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4522 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4523 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4524 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4525 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4526 (tlsdesc_small_<mode>): Turn a define_expand and use
4527 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
4528 (tlsdesc_small_advsimd_<mode>): ...this.
4529 (tlsdesc_small_sve_<mode>): New pattern.
4531 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4533 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
4534 (UNSPEC_UMUL_HIGHPART): New constants.
4535 (MUL_HIGHPART): New int iteraor.
4536 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
4537 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
4539 (*<su>mul<mode>3_highpart): New define_insn.
4541 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4544 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4547 2018-03-13 Martin Liska <mliska@suse.cz>
4550 * (sem_item_optimizer::sem_item_optimizer): Initialize new
4552 (sem_item_optimizer::~sem_item_optimizer): Release it.
4553 (sem_item_optimizer::merge_classes): Register variable aliases.
4554 (sem_item_optimizer::fixup_pt_set): New function.
4555 (sem_item_optimizer::fixup_points_to_sets): Likewise.
4556 * ipa-icf.h: Declare new variables and functions.
4558 2018-03-13 Jakub Jelinek <jakub@redhat.com>
4561 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
4562 integer_pow2p@2 and test integer_pow2p in condition.
4563 (A < 0 ? C : 0): Similarly for @1.
4566 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
4567 characters starting at p contain '\0' character, don't look beyond
4571 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
4572 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
4575 * reg-stack.c (change_stack): Change update_end var from int to
4576 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
4577 also call set_block_for_insn on the newly added insns and rescan.
4580 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
4581 on the last operand.
4584 * tree.c (stabilize_reference_1): Return save_expr (e) for
4585 STATEMENT_LIST even if it doesn't have side-effects.
4587 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
4589 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
4591 2018-03-12 Renlin Li <renlin.li@arm.com>
4593 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
4594 aarch64_output_scalar_simd_mov_immediate.
4596 2018-03-12 Martin Sebor <msebor@redhat.com>
4598 PR tree-optimization/83456
4599 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
4600 for perfectly overlapping calls to memcpy.
4601 (gimple_fold_builtin_memory_chk): Same.
4602 (gimple_fold_builtin_strcpy): Handle no-warning.
4603 (gimple_fold_builtin_stxcpy_chk): Same.
4604 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
4606 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4608 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
4609 parameter. Use it for SFmode.
4610 (rs6000_function_arg_advance_1): Adjust.
4611 (rs6000_function_arg): Adjust.
4612 (rs6000_gimplify_va_arg): Pass false for that new parameter.
4614 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4616 PR rtl-optimization/84169
4617 PR rtl-optimization/84780
4618 * combine.c (can_combine_p): Check for a 2-insn combination whether
4619 the destination register is used between the two insns, too.
4621 2018-03-12 Richard Biener <rguenther@suse.de>
4623 PR tree-optimization/84803
4624 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
4625 for refs DR analysis didn't process.
4627 2018-03-12 Richard Biener <rguenther@suse.de>
4629 PR tree-optimization/84777
4630 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
4631 force-vectorize loops ignore whether we are optimizing for size.
4633 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
4635 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
4636 (TARGET_MD_ASM_ADJUST): Define.
4638 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
4639 Kito Cheng <kito.cheng@gmail.com>
4640 Chung-Ju Wu <jasonwucj@gmail.com>
4642 * config/nds32/nds32.c (nds32_compute_stack_frame,
4643 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
4644 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
4645 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
4646 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
4647 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
4648 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
4649 * config/nds32/nds32.md (prologue, epilogue): Use macro
4650 NDS32_V3PUSH_AVAILABLE_P to do checking.
4652 2018-03-11 Jakub Jelinek <jakub@redhat.com>
4655 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
4656 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
4657 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
4658 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
4659 addition of most attributes on !orig_type_die or the attribute not
4660 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
4662 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4663 Chung-Ju Wu <jasonwucj@gmail.com>
4665 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
4667 * config/nds32/nds32.opt (mvh): New option.
4669 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4670 Chung-Ju Wu <jasonwucj@gmail.com>
4672 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
4674 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
4675 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
4678 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4679 Chung-Ju Wu <jasonwucj@gmail.com>
4681 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
4683 * config/nds32/nds32-multiple.md (strlensi): New pattern.
4684 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
4686 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
4687 Kito Cheng <kito.cheng@gmail.com>
4688 Chung-Ju Wu <jasonwucj@gmail.com>
4690 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
4691 UNSPEC_FFMISM and UNSPEC_FLMISM.
4692 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
4693 for ffb, ffmism and flmism.
4694 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
4695 (unspec_ffmism): Ditto.
4696 (unspec_flmism): Ditto.
4697 (nds32_expand_builtin_impl): Check if string extension is available.
4698 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
4699 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
4701 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
4704 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4707 * lra-assigns.c (assign_by_spills): Return a flag of reload
4708 assignment failure. Do not process the reload assignment
4709 failures. Do not spill other reload pseudos if they has the same
4711 (lra_assign): Add a return arg. Set up from the result of
4712 assign_by_spills call.
4713 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4714 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4715 usage_insns if it is not NULL.
4716 (spill_hard_reg_in_range): New function.
4717 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4718 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4719 function prototypes.
4720 (lra_assign): Change prototype.
4721 * lra.c (lra): Add code to deal with fails by splitting hard reg
4724 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
4727 * config/i386/i386.opt: Replace Enforcment with Enforcement.
4729 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
4732 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
4733 (dw_val_node): Add val_symbolic_view.
4734 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
4735 (symview_upper_bound): New.
4736 (new_line_info_table): Initialize symviews_since_reset.
4737 (dwarf2out_source_line): Count symviews_since_reset and set
4738 symview_upper_bound.
4739 (dw_val_equal_p): Handle symview.
4740 (add_AT_symview): New.
4741 (print_dw_val): Handle symview.
4742 (attr_checksum, attr_checksum_ordered): Likewise.
4743 (same_dw_val_p, size_of_die): Likewise.
4744 (value_format, output_die): Likewise.
4745 (add_high_low_attributes): Use add_AT_symview for entry_view.
4746 (dwarf2out_finish): Reset symview_upper_bound, clear
4749 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
4752 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
4753 Add strict argument and use it.
4754 (rs6000_split_multireg_move): Update for new strict argument.
4755 (mem_operand_gpr): Disallow all non-offsettable addresses.
4756 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
4758 2018-03-09 Jakub Jelinek <jakub@redhat.com>
4761 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
4762 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
4763 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
4766 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
4767 decl, use remap_type if we want to use the type.
4769 2018-03-09 Martin Sebor <msebor@redhat.com>
4771 PR tree-optimization/84526
4772 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4774 (builtin_access::generic_overlap): Be prepared to handle non-array
4777 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
4779 PR rtl-optimization/84682
4780 * lra-constraints.c (process_address_1): Check is_address flag
4781 for address constraints.
4782 (process_alt_operands): Likewise.
4783 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
4784 preprocess_constraints.
4785 * recog.h (preprocess_constraints): Add oploc parameter.
4787 * recog.c (preprocess_constraints): Test address_operand for
4788 CT_ADDRESS constraints.
4790 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4793 * lra-assigns.c (assign_by_spills): Return a flag of reload
4794 assignment failure. Do not process the reload assignment
4795 failures. Do not spill other reload pseudos if they has the same
4797 (lra_assign): Add a return arg. Set up from the result of
4798 assign_by_spills call.
4799 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4800 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4801 usage_insns if it is not NULL.
4802 (spill_hard_reg_in_range): New function.
4803 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4804 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4805 function prototypes.
4806 (lra_assign): Change prototype.
4807 * lra.c (lra): Add code to deal with fails by splitting hard reg
4810 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4813 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
4814 Accept complain bool parameter. Only emit errors if it is true.
4815 (arm_parse_cpu_option_name): Likewise.
4816 (arm_target_thumb_only): Adjust callers of the above.
4817 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
4818 prototype to take a default true bool parameter.
4819 (arm_parse_arch_option_name): Likewise.
4821 2018-03-09 David Malcolm <dmalcolm@redhat.com>
4822 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
4826 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
4827 * configure: Regenerate.
4828 * configure.ac ("linker --version-script option"): New.
4829 ("linker soname option"): New.
4831 2018-03-09 Richard Biener <rguenther@suse.de>
4833 PR tree-optimization/84775
4834 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
4835 immediate uses of predicate stmts and mark them modified.
4838 PR tree-optimization/84178
4839 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4841 (version_loop_for_if_conversion): Delay update_ssa call.
4842 (tree_if_conversion): Delay update_ssa until after predicate
4845 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
4848 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
4849 when the function accesses prior frames.
4851 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4854 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
4855 gen_llsym, otherwise call maybe_gen_llsym.
4858 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
4859 has ',' character inside of it.
4861 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4864 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
4865 as clobbering CC_REGNUM.
4867 2018-03-08 Richard Biener <rguenther@suse.de>
4870 * tree-scalar-evolution.c: Include tree-into-ssa.h.
4871 (follow_copies_to_constant): Do not follow SSA names registered
4874 2018-03-08 Richard Biener <rguenther@suse.de>
4876 PR tree-optimization/84178
4877 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4879 (version_loop_for_if_conversion): Delay update_ssa call.
4880 (tree_if_conversion): Delay update_ssa until after predicate
4883 2018-03-08 David Malcolm <dmalcolm@redhat.com>
4885 PR tree-optimization/84178
4886 * tree-if-conv.c (release_bb_predicate): Remove the
4887 the assertion that the stmts have NULL use_ops.
4888 Discard the statements, asserting that they haven't
4889 yet been added to a BB.
4891 2018-03-08 Richard Biener <rguenther@suse.de>
4893 PR tree-optimization/84746
4894 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
4895 (phi_translate): Pass in destination ANTIC_OUT set.
4896 (phi_translate_1): Likewise. For a simplified result lookup
4897 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
4898 (phi_translate_set): Adjust.
4899 (do_pre_regular_insertion): Likewise.
4900 (do_pre_partial_partial_insertion): Likewise.
4902 2018-03-08 Martin Liska <mliska@suse.cz>
4904 PR gcov-profile/84735
4905 * doc/gcov.texi: Document usage of profile files.
4906 * gcov-io.h: Document changes in the format.
4908 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
4912 * dwarf2out.c (struct dw_line_info_table): Update comments for
4914 (FORCE_RESET_NEXT_VIEW): New.
4915 (FORCE_RESETTING_VIEW_P): New.
4916 (RESETTING_VIEW_P): Check for -1 too.
4917 (ZERO_VIEW_P): Likewise.
4918 (new_line_info_table): Force-reset next view.
4919 (dwarf2out_begin_function): Likewise.
4920 (dwarf2out_source_line): Simplify zero_view_p initialization.
4921 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
4922 view directly. Omit view when omitting .loc at line 0.
4924 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4926 PR tree-optimization/84740
4927 * tree-switch-conversion.c (process_switch): Call build_constructors
4928 only if info.phi_count is non-zero.
4930 PR tree-optimization/84739
4931 * tree-tailcall.c (find_tail_calls): Check call arguments against
4932 DECL_ARGUMENTS (current_function_decl) rather than
4933 DECL_ARGUMENTS (func) when checking for tail recursion.
4935 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4937 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
4938 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
4939 Volker Reichelt's entry and add entries for people that perform
4940 GCC fuzzy testing and report numerous bugs.
4942 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
4945 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
4946 readonly data in sdata, if that is disabled.
4947 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
4948 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
4949 -mreadonly-in-sdata option.
4951 2018-03-07 Martin Sebor <msebor@redhat.com>
4953 PR tree-optimization/84468
4954 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
4955 basic block when looking for nul assignment.
4957 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
4960 * except.h (output_function_exception_table): Adjust prototype.
4961 * except.c (output_function_exception_table): Remove FNNAME parameter
4962 and add SECTION parameter. Ouput one part of the table at a time.
4963 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
4964 the first part of the exception table and emit unwind directives.
4965 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
4966 (i386_pe_seh_cold_init): Likewise.
4967 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
4968 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
4969 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
4970 (ix86_output_call_insn): Emit a nop in one more case for SEH.
4971 * config/i386/winnt.c: Include except.h.
4972 (struct seh_frame_state): Add reg_offset, after_prologue and
4973 in_cold_section fields.
4974 (i386_pe_seh_end_prologue): Set seh->after_prologue.
4975 (i386_pe_seh_cold_init): New function.
4976 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
4977 to seh->in_cold_section.
4978 (seh_emit_push): Record the offset of the push.
4979 (seh_emit_save): Record the offet of the save.
4980 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
4981 Test seh->after_prologue to disregard the epilogue.
4982 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
4983 (i386_pe_end_cold_function): New function.
4985 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4988 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
4989 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
4992 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
4994 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
4995 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
4998 * multiple_target.c: Include tree-inline.h and intl.h.
4999 (expand_target_clones): Diagnose and fail if node->definition and
5000 !tree_versionable_function_p (node->decl).
5002 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
5004 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
5006 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
5007 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
5008 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
5010 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5013 * combine.c (try_combine): Use reg_or_subregno instead of handling
5014 just paradoxical SUBREGs and REGs.
5016 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
5018 * config/arc/arc.c (arc_finalize_pic): Remove function.
5019 (arc_must_save_register): We use single base PIC register, remove
5020 checks to save/restore the PIC register.
5021 (arc_expand_prologue): Likewise.
5022 * config/arc/arc-protos.h (arc_set_default_type_attributes):
5024 (arc_verify_short): Likewise.
5025 (arc_attr_type): Likewise.
5026 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
5027 (walk_stores): Likewise.
5028 (arc_address_cost): Make it static.
5029 (arc_verify_short): Likewise.
5030 (branch_dest): Likewise.
5031 (arc_attr_type): Likewise.
5032 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
5033 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
5034 (arc_final_prescan_insn): Remove inserting the nops due to
5035 hardware hazards. It is done in reorg step.
5036 (insn_length_variant_t): Remove.
5037 (insn_length_parameters_t): Likewise.
5038 (arc_insn_length_parameters): Likewise.
5039 (arc_get_insn_variants): Likewise.
5040 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
5042 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5045 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
5048 PR tree-optimization/84687
5049 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
5051 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
5053 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5055 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
5056 Rename to ppc_speculation_barrier.
5057 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
5058 __builtin_ppc_speculation_barrier.
5060 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5063 * combine.c (combine_simplify_rtx): Don't try to simplify if
5064 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
5067 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
5069 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
5070 to 32 bytes when compiling for POWER9.
5072 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5075 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
5076 regparm >= 3 with no arg reg available also for calls with
5077 flag_force_indirect_call. Pass decl to ix86_function_regparm.
5080 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
5082 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
5084 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
5087 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
5089 2018-03-05 Richard Biener <rguenther@suse.de>
5091 PR tree-optimization/84486
5092 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
5093 When inserting a __builtin_assume_aligned call set the LHS
5094 SSA name alignment info accordingly.
5096 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
5098 PR tree-optimization/84114
5099 * config/aarch64/aarch64.c (aarch64_reassociation_width)
5100 Avoid reassociation of FLOAT_MODE addition.
5102 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
5104 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
5105 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
5106 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
5107 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
5108 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
5109 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
5110 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
5112 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
5113 __builtin_ia32_wbinvd): New builtins.
5114 (SPECIAL_ARGS2): New.
5115 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
5116 (SPECIAL_ARGS2): New.
5117 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
5118 (ix86_valid_target_attribute_inner_p): Ditto.
5119 (ix86_init_mmx_sse_builtins): Add special_args2.
5120 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
5121 TARGET_WBNOINVD_P): New.
5122 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
5123 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
5124 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
5125 * config/i386/immintrin.h (_wbinvd): New intrinsic.
5126 * config/i386/pconfigintrin.h: New file.
5127 * config/i386/wbnoinvdintrin.h: Ditto.
5128 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
5129 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
5131 2018-03-05 Richard Biener <rguenther@suse.de>
5133 PR tree-optimization/84670
5134 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5136 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5137 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5138 (compute_antic_aux): Only assert the number of values in ANTIC_IN
5139 doesn't grow if all successors (recursively) were visited at least
5142 2018-03-05 Richard Biener <rguenther@suse.de>
5144 PR tree-optimization/84650
5145 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
5146 if executed in the loop pipeline.
5148 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
5150 * doc/configfiles.texi (Configuration Files): Move info about
5151 conditionalizing $target-protos.h to...
5152 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
5153 differs from $target-protos.h.
5155 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
5156 Chung-Ju Wu <jasonwucj@gmail.com>
5158 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
5159 * config/nds32/nds32-multiple.md (setmemsi): Define.
5160 * config/nds32/nds32-memory-manipulation.c
5161 (nds32_gen_dup_4_byte_to_word_value): New.
5162 (emit_setmem_word_loop): New.
5163 (emit_setmem_byte_loop): New.
5164 (nds32_expand_setmem_loop): New.
5165 (nds32_expand_setmem_loop_v3m): New.
5166 (nds32_expand_setmem_unroll): New.
5167 (nds32_expand_setmem): New.
5169 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5170 Chung-Ju Wu <jasonwucj@gmail.com>
5172 * config/nds32/nds32-memory-manipulation.c
5173 (nds32_emit_load_store): New.
5174 (nds32_emit_post_inc_load_store): New.
5175 (nds32_emit_mem_move): New.
5176 (nds32_emit_mem_move_block): New.
5177 (nds32_expand_movmemsi_loop_unknown_size): New.
5178 (nds32_expand_movmemsi_loop_known_size): New.
5179 (nds32_expand_movmemsi_loop): New.
5180 (nds32_expand_movmemsi_unroll): New.
5181 (nds32_expand_movmemqi): Rename ...
5182 (nds32_expand_movmemsi): ... to this.
5183 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
5184 (movmemsi): ... to this.
5185 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
5186 (nds32_expand_movmemsi): ... to this.
5188 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5189 Monk Chiang <sh.chiang04@gmail.com>
5190 Chung-Ju Wu <jasonwucj@gmail.com>
5192 * config/nds32/nds32-protos.h
5193 (nds32_expand_load_multiple): New arguments.
5194 (nds32_expand_store_multiple): Ditto.
5195 (nds32_valid_multiple_load_store): Rename ...
5196 (nds32_valid_multiple_load_store_p): ... to this.
5197 * config/nds32/nds32-memory-manipulation.c
5198 (nds32_expand_load_multiple): Refine implementation.
5199 (nds32_expand_store_multiple): Ditto.
5200 * config/nds32/nds32-multiple.md
5201 (load_multiple): Update nds32_expand_load_multiple interface.
5202 (store_multiple): Update nds32_expand_store_multiple interface.
5203 * config/nds32/nds32-predicates.c
5204 (nds32_valid_multiple_load_store): Rename ...
5205 (nds32_valid_multiple_load_store_p): ... to this and refine
5207 * config/nds32/predicates.md
5208 (nds32_load_multiple_and_update_address_operation): New predicate.
5209 (nds32_store_multiple_and_update_address_operation): New predicate.
5211 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5212 Chung-Ju Wu <jasonwucj@gmail.com>
5214 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
5215 (combo): New attribute.
5216 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
5218 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
5220 * config/nds32/nds32.opt: Change -mcmodel= default value.
5222 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
5223 Monk Chiang <sh.chiang04@gmail.com>
5224 Chung-Ju Wu <jasonwucj@gmail.com>
5226 * config/nds32/constants.md (unspec_element): New enum.
5227 * config/nds32/constraints.md (Umw): New constraint.
5228 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
5229 * config/nds32/nds32-intrinsic.md: Likewise.
5230 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
5231 (nds32_valid_smw_lwm_base_p): New.
5232 (nds32_output_smw_single_word): New.
5233 (nds32_output_lmw_single_word): New.
5234 (nds32_expand_unaligned_load): New.
5235 (nds32_expand_unaligned_store): New.
5236 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5237 (nds32_output_smw_single_word): Declare.
5238 (nds32_output_lmw_single_word): Declare.
5239 (nds32_expand_unaligned_load): Declare.
5240 (nds32_expand_unaligned_store): Declare.
5241 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5242 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5243 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5244 NDS32_BUILTIN_UASTORE_DW.
5245 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5248 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
5249 Kito Cheng <kito.cheng@gmail.com>
5250 Chung-Ju Wu <jasonwucj@gmail.com>
5252 * config/nds32/nds32-intrinsic.c
5253 (nds32_expand_builtin_null_ftype_reg): Delete.
5254 (nds32_expand_builtin_reg_ftype_imm): Ditto.
5255 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5256 (nds32_read_argument): New.
5257 (nds32_legitimize_target): Ditto.
5258 (nds32_legitimize_argument): Ditto.
5259 (nds32_check_constant_argument): Ditto.
5260 (nds32_expand_unop_builtin): Ditto.
5261 (nds32_expand_unopimm_builtin): Ditto.
5262 (nds32_expand_binop_builtin): Ditto.
5263 (nds32_builtin_decl_impl): Ditto.
5264 (builtin_description): Ditto.
5265 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5266 (nds32_init_builtins_impl): Ditto.
5267 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5268 (nds32_builtin_decl): New.
5269 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5270 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5272 2018-03-02 Jeff Law <law@redhat.com>
5274 * reorg.c (stop_search_p): Handle DEBUG_INSN.
5275 (redundant_insn, fill_simple_delay_slots): Likewise.
5276 (fill_slots_from_thread): Likewise.
5277 * resource.c (mark_referenced_resources): Likewise.
5278 (mark_set_resources, find_dead_or_set_registers): Likewise.
5280 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5282 * substring-locations.h (format_warning_va): Formatting fix for
5284 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5286 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5287 * substring-locations.c: Include intl.h.
5288 (format_warning_va): Turned into small wrapper around
5289 format_warning_n_va, renamed to ...
5290 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5291 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5293 (format_warning_at_substring_n): New function.
5294 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5295 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
5296 format_warning_at_substring with just a shorter name instead of
5297 const function pointer.
5298 (fmtwarn_n): New function.
5299 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5300 appropriate, get rid of all the fmtstr temporaries, move conditionals
5301 with G_() wrapped string literals directly into fmtwarn arguments,
5302 cast dir.len to (int), formatting fixes.
5304 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
5306 * doc/invoke.texi: Remove "Cilk Plus" references.
5308 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5309 Richard Biener <rguenther@suse.de>
5312 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5313 for error or warning attributes if CALL_FROM_THUNK_P is set.
5316 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5319 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5320 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5323 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5324 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5326 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5327 instead of -1U in last predictors element's probability member.
5329 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5332 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5333 arguments if they are comparable.
5335 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5337 PR tree-optimization/84634
5338 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5339 masks and masked_loop_p with a single loop_masks, making sure it's
5340 null for bb vectorization.
5342 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5344 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5345 (vect_analyze_data_ref_access): Use loop->safe_len rather than
5346 loop->force_vectorize to check whether there is no alias.
5348 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5351 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5353 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5355 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5356 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5357 instead of a loop around prev_real_insn.
5358 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5362 * config/i386/i386.c (ix86_print_operand): Use conditional
5363 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5366 2018-03-02 Richard Biener <rguenther@suse.de>
5368 PR tree-optimization/84427
5369 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5370 (bitmap_set_subtract_values): Rewrite to handle multiple
5373 (prune_clobbered_mems): Likewise.
5374 (phi_translate): Take edge instead of pred/phiblock.
5375 (phi_translate_1): Likewise.
5376 (phi_translate_set): Likewise. Insert all translated
5377 exprs for a value into the set, keeping possibly multiple
5378 expressions per value.
5379 (compute_antic_aux): Adjust for phi_translate changes.
5380 When intersecting union the expressions and prune those
5381 not in the final value set, keeping possibly multiple
5382 expressions per value. Do not use value-insertion
5383 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5384 all expressions. Add verification that the value-sets
5385 only shrink during iteration.
5386 (compute_partial_antic_aux): Adjust for the phi_translate changes.
5387 (do_pre_regular_insertion): Likewise.
5388 (do_pre_partial_partial_insertion): Likewise.
5390 2018-03-02 Richard Biener <rguenther@suse.de>
5393 * config/darwin.c (saved_debug_info_level): New static global.
5394 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5395 (darwin_asm_lto_end): Restore debug info generation settings.
5397 2018-03-01 Martin Liska <mliska@suse.cz>
5400 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5403 2018-03-01 Richard Biener <rguenther@suse.de>
5406 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5407 type annotation with LTO when debug was disabled at compile-time.
5409 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5411 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5413 (mips_final_postscan_insn): Likewise.
5415 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5417 PR rtl-optimization/84528
5418 * alias.c (init_alias_target): Add commentary.
5419 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5420 a unique base value if the frame pointer is not eliminated
5421 to the stack pointer.
5423 2018-03-01 Tom de Vries <tom@codesourcery.com>
5425 PR rtl-optimization/83327
5426 * lra-int.h (hard_regs_spilled_into): Declare.
5427 * lra.c (hard_regs_spilled_into): Define.
5428 (init_reg_info): Init hard_regs_spilled_into.
5429 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5430 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5431 (process_bb_lives): Handle hard_regs_spilled_into.
5432 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5433 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5435 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5437 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5438 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5439 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5440 * config/rs6000/aix72.h: New file.
5442 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5444 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5445 instead of warning_at with conditional singular and plural messages
5449 * stor-layout.c (update_alignment_for_field): For
5450 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5451 && !DECL_PACKED (field), do the alignment update, just use
5452 only desired_align instead of MAX (type_align, desired_align)
5454 (place_field): Don't do known_align < desired_align handling
5455 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5456 is non-NULL, instead do it after rli->prev_field handling and
5457 only if not within a bitfield word. For DECL_PACKED (field)
5458 use type_align of BITS_PER_UNIT.
5460 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5462 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5463 superfluous parentheses and trailing spaces.
5465 2018-02-28 Richard Biener <rguenther@suse.de>
5467 PR tree-optimization/84584
5468 * graphite-scop-detection.c (scop_detection::add_scop): Discard
5469 SCoPs with fake exit edge.
5471 2018-02-28 Martin Liska <mliska@suse.cz>
5474 * timevar.c (timer::print): Fix format to properly print 100%
5477 2018-02-28 Richard Biener <rguenther@suse.de>
5480 * genmatch.c (capture_info::walk_match): Do not mark
5481 captured expressions without operands as expr_p given
5482 they act more like predicates and should be subject to
5483 "lost tail" side-effect preserving.
5485 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
5487 PR rtl-optimization/81611
5488 * auto-inc-dec.c (attempt_change): Move dead note from
5489 mem_insn if it's the next use of regno
5490 (find_address): Take address use of reg holding
5491 non-incremented value. Add parm to limit search to the named
5493 (merge_in_block): Attempt to use a mem insn that is the next
5494 use of the original regno.
5496 2018-02-27 Martin Sebor <msebor@redhat.com>
5499 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
5500 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
5502 2018-02-27 Martin Sebor <msebor@redhat.com>
5504 PR translation/84207
5505 * diagnostic-core.h (warning_n, error_n, inform_n): Change
5506 n argument to unsigned HOST_WIDE_INT.
5507 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5508 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
5509 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5510 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5512 2018-02-27 Richard Biener <rguenther@suse.de>
5514 PR tree-optimization/84512
5515 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5516 Do not use the estimate returned from record_stmt_cost for
5517 the scalar iteration cost but sum properly using add_stmt_cost.
5519 2018-02-27 Richard Biener <rguenther@suse.de>
5521 PR tree-optimization/84466
5522 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5523 Adjust last change to less strictly validate use operands.
5525 2018-02-27 Martin Liska <mliska@suse.cz>
5527 PR gcov-profile/84548
5528 * gcov.c (process_file): Allow partial overlap and consider it
5529 also as group functions.
5530 (output_lines): Properly calculate range of lines for a group.
5532 2018-02-27 Martin Liska <mliska@suse.cz>
5534 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
5535 'ggc' suffixes. Change first column width.
5536 (timer::print): Fix formatting of the column.
5538 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
5540 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
5541 preserve inline entry blocks for the sake of debug inline
5542 entry point markers alone.
5543 (remove_unused_locals): Suggest in comments a better place to
5544 force the preservation of inline entry blocks that are
5545 otherwise unused, but do not preserve them.
5547 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5549 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
5551 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5554 * config/i386/constraints.md (Bs): Replace
5555 ix86_indirect_branch_register with
5556 TARGET_INDIRECT_BRANCH_REGISTER.
5558 * config/i386/i386.md (indirect_jump): Likewise.
5559 (tablejump): Likewise.
5560 (*sibcall_memory): Likewise.
5561 (*sibcall_value_memory): Likewise.
5562 Peepholes of indirect call and jump via memory: Likewise.
5563 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
5564 (*sibcall_value_GOT_32): Likewise.
5565 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5566 (GOT_memory_operand): Likewise.
5567 (call_insn_operand): Likewise.
5568 (sibcall_insn_operand): Likewise.
5569 (GOT32_symbol_operand): Likewise.
5570 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
5572 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5574 PR rtl-optimization/83496
5575 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
5576 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
5577 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
5578 redundant insn, if any.
5579 (relax_delay_slots): Likewise.
5580 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
5582 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
5584 PR tree-optimization/83965
5585 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
5586 that grouped statements are part of a reduction chain. Return
5587 true if the statement is not marked as a reduction itself but
5589 (vect_recog_dot_prod_pattern): Don't check whether the statement
5590 is part of a group here.
5591 (vect_recog_sad_pattern): Likewise.
5592 (vect_recog_widen_sum_pattern): Likewise.
5594 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5597 * final.c (rest_of_clean_state): Also look for calls inside sequences.
5599 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5602 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
5604 (ix86_output_indirect_function_return): New prototype.
5605 (ix86_split_simple_return_pop_internal): Likewise.
5606 * config/i386/i386.c (indirect_return_via_cx): New.
5607 (indirect_return_via_cx_bnd): Likewise.
5608 (indirect_thunk_name): Handle return va CX_REG.
5609 (output_indirect_thunk_function): Create alias for
5610 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
5611 (ix86_output_indirect_jmp): Remove the bool argument.
5612 (ix86_output_indirect_function_return): New function.
5613 (ix86_split_simple_return_pop_internal): Likewise.
5614 * config/i386/i386.md (*indirect_jump): Don't pass false
5615 to ix86_output_indirect_jmp.
5616 (*tablejump_1): Likewise.
5617 (simple_return_pop_internal): Change it to define_insn_and_split.
5618 Call ix86_split_simple_return_pop_internal to split it for
5620 (simple_return_indirect_internal): Call
5621 ix86_output_indirect_function_return instead of
5622 ix86_output_indirect_jmp.
5624 2018-02-26 Jakub Jelinek <jakub@redhat.com>
5627 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
5628 memset and value initialization afterwards.
5630 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
5632 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
5634 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5637 * common/config/aarch64/aarch64-common.c
5638 (aarch_option_optimization_table[]): Switch
5639 off fomit-frame-pointer
5641 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5642 Chung-Ju Wu <jasonwucj@gmail.com>
5644 * config/nds32/nds32-multiple.md (load_multiple): Disallow
5646 (store_multiple): Ditto.
5648 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5650 * config.gcc: Add --with-cpu support for nds32 target.
5651 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
5652 * config/nds32/nds32.opt: Add -mcpu= option.
5654 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
5656 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
5657 isel=yes): Warn for these deprecated options.
5659 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
5661 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
5662 ISA_2_5_MASKS_EMBEDDED.
5664 2018-02-23 Jakub Jelinek <jakub@redhat.com>
5666 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
5667 p->max as pointers rather than using iterative_hash_expr.
5669 2018-02-23 Carl Love <cel@us.ibm.com>
5671 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
5672 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
5674 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
5675 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
5676 P8V_BUILTIN_VEC_VUNSIGNED2.
5678 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
5681 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
5682 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
5684 * lra-constraints.c (curr_insn_transform): Set up
5685 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
5687 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
5688 * lra-eliminations.c (spill_pseudos): Ditto.
5689 (process_insn_for_elimination): Ditto.
5690 * lra-lives.c (reg_early_clobber_p): Use the new macros.
5691 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
5692 LRA_NON_CLOBBERED_ALT.
5694 2018-02-22 Martin Sebor <msebor@redhat.com>
5696 PR tree-optimization/84480
5697 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
5698 to maybe_diag_stxncpy_trunc. Call it.
5699 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
5700 from gimple_fold_builtin_strcpy. Print inlining stack.
5701 (handle_builtin_stxncpy): Print inlining stack.
5702 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
5704 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
5707 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
5708 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
5709 and -fcheck-pointer-bounds are used together.
5710 (indirect_thunk_prefix): New enum.
5711 (indirect_thunk_need_prefix): New function.
5712 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
5713 "_nt" instead of "_bnd" for NOTRACK prefix.
5714 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
5715 (output_indirect_thunk_function): Likewise.
5717 (ix86_code_end): Update output_indirect_thunk_function calls.
5718 (ix86_output_indirect_branch_via_reg): Replace
5719 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
5720 (ix86_output_indirect_branch_via_push): Likewise.
5721 (ix86_output_function_return): Likewise.
5722 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
5723 incompatible with -fcf-protection=branch and
5724 -fcheck-pointer-bounds.
5726 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5729 * config/aarch64/aarch64.c (aarch64_print_address_internal):
5730 Change gcc_assert call to output_operand_lossage.
5732 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5734 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
5736 2018-02-22 DJ Delorie <dj@redhat.com>
5737 Sebastian Perta <sebastian.perta@renesas.com>
5738 Oleg Endo <olegendo@gcc.gnu.org>
5740 * config/rx/rx.c (rx_rtx_costs): New function.
5741 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
5743 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
5745 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
5747 2018-02-22 Martin Liska <mliska@suse.cz>
5750 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
5751 Add "native" as a possible value.
5753 2018-02-22 Martin Liska <mliska@suse.cz>
5756 * config/i386/i386.c (ix86_option_override_internal):
5757 Add "native" as a possible value for -march and -mtune.
5759 2018-02-22 Jakub Jelinek <jakub@redhat.com>
5762 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
5763 to all type variants.
5765 PR tree-optimization/84503
5766 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
5767 width as info->bitpos + info->bitsize - start.
5768 (merged_store_group::merge_overlapping): Simplify width computation.
5769 (check_no_overlap): New function.
5770 (imm_store_chain_info::try_coalesce_bswap): Compute expected
5771 start + width and last_order of the group, fail if check_no_overlap
5773 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
5774 to group if check_no_overlap fails.
5776 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5778 * config/rs6000/altivec.md: Delete contraint arguments to
5779 define_expand, define_split, and define_peephole2, and in
5780 define_insn_and_split if always unused.
5781 * config/rs6000/darwin.md: Ditto.
5782 * config/rs6000/dfp.md: Ditto.
5783 * config/rs6000/rs6000.md: Ditto.
5784 * config/rs6000/sync.md: Ditto.
5785 * config/rs6000/vector.md: Ditto.
5786 * config/rs6000/vsx.md: Ditto.
5788 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5790 * config/rs6000/altivec.md: Write output control strings as braced
5791 blocks instead of double-quoted strings.
5792 * config/rs6000/darwin.md: Ditto.
5793 * config/rs6000/rs6000.md: Ditto.
5794 * config/rs6000/vector.md: Ditto.
5795 * config/rs6000/vsx.md: Ditto.
5797 2018-02-21 Jason Merrill <jason@redhat.com>
5799 PR c++/84314 - ICE with templates and fastcall attribute.
5800 * attribs.c (build_type_attribute_qual_variant): Remove assert.
5802 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5804 * ipa-cp.c (determine_versionability): Fix comment typos.
5806 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5809 * ipa-cp.c (determine_versionability): Do not version functions caling
5812 2018-02-21 Martin Liska <mliska@suse.cz>
5815 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
5816 Add "native" as a possible value.
5817 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
5818 the macro when native cpu detection is available.
5820 2018-02-21 Martin Liska <mliska@suse.cz>
5823 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
5824 Add "native" as a possible value.
5825 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
5826 when native cpu detection is available.
5828 2018-02-21 Jakub Jelinek <jakub@redhat.com>
5829 Martin Sebor <msebor@redhat.com>
5831 PR tree-optimization/84478
5832 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
5834 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
5835 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
5836 support which is conservatively correct, for 2 only stay conservative
5837 for maxlen. Formatting and comment capitalization fixes. Add STRICT
5838 argument to the 2 argument get_range_strlen, adjust 6 arg
5839 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
5841 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
5842 (gimple_fold_builtin_strlen): Pass true as last argument to
5845 2018-02-20 Martin Sebor <msebor@redhat.com>
5848 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
5849 (builtin_memref::set_base_and_offset): Same. Handle inner references.
5850 (builtin_memref::builtin_memref): Factor out parts into
5851 set_base_and_offset and call it.
5853 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5856 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
5857 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
5858 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
5859 search at the associated MODE_INT.
5861 2018-02-20 Jeff Law <law@redhat.com>
5864 PR tree-optimization/81592
5866 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
5867 for range data rather than using global data.
5868 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
5869 range data rather than using global data.
5870 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
5871 pass it to children as needed.
5872 (struct directive::fmtresult): Similarly.
5873 (struct directive::set_width): Similarly.
5874 (struct directive::set_precision): Similarly.
5875 (format_integer, format_directive, parse_directive): Similarly.
5876 (format_none): Accept unnamed vr_values parameter.
5877 (format_percent, format_floating, format_character): Similarly.
5878 (format_string, format_plain): Similarly.
5879 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
5880 the EVRP range analyzer for range data rather than using global data.
5881 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
5882 gimple-ssa-evrp-analyze.h
5883 (class sprintf_dom_walker): Add after_dom_children member function.
5884 Add evrp_range_analyzer member.
5885 (sprintf_dom_walker::before_dom_children): Call into the EVRP
5886 range analyzer as needed.
5887 (sprintf_dom_walker::after_dom_children): New member function.
5888 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
5890 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5891 (evrp_range_analyzer::pop_to_marker): Likewise.
5893 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5895 PR tree-optimization/84419
5896 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
5897 with the required type if its current type is compatible but
5900 2018-02-20 Jakub Jelinek <jakub@redhat.com>
5903 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
5904 after vectorization.
5906 2018-02-20 Martin Liska <mliska@suse.cz>
5909 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
5910 possible values if we don't have a hint.
5912 2018-02-20 Martin Liska <mliska@suse.cz>
5916 * final.c (shorten_branches): Build align_tab array with one
5918 * opts.c (finish_options): Add alignment option limit check.
5919 (MAX_CODE_ALIGN): Likewise.
5920 (MAX_CODE_ALIGN_VALUE): Likewise.
5921 * doc/invoke.texi: Document maximum allowed option value for
5922 all -falign-* options.
5924 2018-02-19 Jakub Jelinek <jakub@redhat.com>
5927 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
5928 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
5929 * var-tracking.c (emit_note_insn_var_location): Remove all references
5930 to NOTE_INSN_CALL_ARG_LOCATION.
5931 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
5932 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
5933 Use copy_rtx_if_shared.
5934 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
5935 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
5936 (dwarf2out_var_location): Remove handling of
5937 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
5939 * final.c (final_scan_insn): Remove all references to
5940 NOTE_INSN_CALL_ARG_LOCATION.
5941 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
5942 before dumping final insns.
5943 * except.c (emit_note_eh_region_end): Remove all references to
5944 NOTE_INSN_CALL_ARG_LOCATION.
5945 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
5946 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
5947 * config/arc/arc.c (hwloop_optimize): Likewise.
5948 * config/arm/arm.c (create_fix_barrier): Likewise.
5949 * config/s390/s390.c (s390_chunkify_start): Likewise.
5950 * config/sh/sh.c (find_barrier): Likewise.
5951 * config/i386/i386.c (rest_of_insert_endbranch,
5952 ix86_seh_fixup_eh_fallthru): Likewise.
5953 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
5954 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
5955 * config/frv/frv.c (frv_function_prologue): Likewise.
5956 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
5958 (note_outside_basic_block_p): Remove all references to
5959 NOTE_INSN_CALL_ARG_LOCATION.
5960 * gengtype.c (adjust_field_rtx_def): Likewise.
5961 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
5963 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
5964 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
5967 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
5970 PR tree-optimization/84452
5971 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
5972 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
5975 2018-02-19 Martin Liska <mliska@suse.cz>
5978 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
5980 2018-02-19 Martin Liska <mliska@suse.cz>
5981 Richard Sandiford <richard.sandiford@linaro.org>
5983 PR tree-optimization/82491
5984 * gimple-fold.c (get_base_constructor): Make earlier bail out
5987 2018-02-19 Carl Love <cel@us.ibm.com>
5989 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
5990 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
5992 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
5993 P8V_BUILTIN_VEC_NEG.
5995 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
5997 * config/rl78/rl78.md (movdf): New define expand.
5999 2018-02-19 Martin Liska <mliska@suse.cz>
6002 * doc/invoke.texi: Fix typo.
6003 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
6005 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
6007 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
6008 handle rs6000_single_float and rs6000_double_float specially for
6011 2018-02-16 Jeff Law <law@redhat.com>
6013 * config/rx/rx.c (add_pop_cfi_notes): New function.;
6016 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6019 * ipa-inline.c (inline_small_functions): Fix a typo.
6021 2018-02-16 Nathan Sidwell <nathan@acm.org>
6023 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
6025 2018-02-16 Carl Love <cel@us.ibm.com>
6027 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
6028 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
6029 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
6030 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
6031 expansion to P8V_BUILTIN_VEC_FLOAT2.
6033 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
6035 PR rtl-optimization/70023
6036 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
6037 src_regno into account.
6039 2018-02-16 Carl Love <cel@us.ibm.com>
6041 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
6042 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
6043 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
6044 * config/rs6000/rs6000.c: Remove case statements for
6045 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
6046 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
6047 and P9V_BUILTIN_VEC_VINSERT4B.
6048 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
6049 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
6050 * config/rs6000/vsx.md:
6051 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
6054 2018-02-16 Carl Love <cel@us.ibm.com>
6056 * config/rs6000/altivec.h: Add builtin names vec_extract4b
6058 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
6060 * config/rs6000/rs6000-c.c: Add the definitions for
6061 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
6062 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
6063 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
6064 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
6065 definition for insert4b and define insn *insert3b_internal.
6066 * doc/extend.texi: Add documentation for vec_extract4b.
6068 2018-02-16 Nathan Sidwell <nathan@acm.org>
6070 * doc/extend.texi (Backwards Compatibility): Mention friend
6071 injection. Note for-scope is deprecated.
6072 * doc/invoke.texi (-ffriend-injection): Deprecate.
6074 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
6076 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
6077 that moved to I2, also allow destinations that are a paradoxical
6078 subreg (instead of a normal reg).
6080 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
6083 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
6086 2018-02-16 Richard Biener <rguenther@suse.de>
6088 PR tree-optimization/84037
6089 PR tree-optimization/84016
6091 * config/i386/i386.c (ix86_builtin_vectorization_cost):
6092 Adjust vec_construct for the fact we need additional higher latency
6093 128bit inserts for AVX256 and AVX512 vector builds.
6094 (ix86_add_stmt_cost): Scale vector construction cost for
6097 2018-02-16 Richard Biener <rguenther@suse.de>
6099 PR tree-optimization/84417
6100 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
6101 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
6102 (non_rewritable_lvalue_p): Likewise, use poly-ints.
6104 2018-02-16 Martin Liska <mliska@suse.cz>
6107 * internal-fn.def (ASAN_CHECK): Set proper flags.
6108 (ASAN_MARK): Likewise.
6110 2018-02-16 Julia Koval <julia.koval@intel.com>
6112 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
6113 from PTA_CANNONLAKE.
6115 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6118 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
6119 Use ++iter rather than iter++ for std::list iterators.
6120 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
6121 defer deleting them until all nodes in the forest are processed. Do
6122 free even leaf nodes. Change to_process into auto_vec.
6125 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
6126 * vec.h (vec_default_construct): Use memset instead of placement new
6127 if BROKEN_VALUE_INITIALIZATION is defined.
6128 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
6129 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
6132 PR rtl-optimization/83723
6133 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
6134 * lra.c (lra_substitute_pseudo): Likewise. If true, use
6135 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
6137 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
6139 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
6141 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
6143 PR rtl-optimization/81443
6144 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
6145 from inner REGs to paradoxical SUBREGs.
6147 2018-02-16 Richard Biener <rguenther@suse.de>
6149 PR tree-optimization/84399
6150 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6151 For operands we can analyze at their definition make sure we can
6152 analyze them at each use as well.
6154 2018-02-16 Richard Biener <rguenther@suse.de>
6156 PR tree-optimization/84190
6157 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
6158 volatile accesses if the decl isn't volatile.
6160 2018-02-15 Jason Merrill <jason@redhat.com>
6162 PR c++/84314 - ICE with templates and fastcall attribute.
6163 * attribs.c (build_type_attribute_qual_variant): Don't clobber
6164 TYPE_CANONICAL on an existing type.
6166 2018-02-15 Jakub Jelinek <jakub@redhat.com>
6168 PR tree-optimization/84383
6169 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
6170 dstoff nor call operand_equal_p if dstbase is NULL.
6172 PR tree-optimization/84334
6173 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
6174 also a CONSTANT_CLASS_P, punt.
6176 2018-02-14 Jim Wilson <jimw@sifive.com>
6178 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
6179 first SMALL_OPERAND check. New local min_second_step. Move assert
6180 to where locals are set. Add TARGET_RVC support.
6181 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
6183 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
6185 * doc/invoke.texi: Correct -Wformat-overflow code sample.
6187 2018-02-14 Martin Sebor <msebor@redhat.com>
6189 PR tree-optimization/83698
6190 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
6191 arrays constrain the offset range to their bounds.
6192 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
6193 (builtin_access::overlap): Avoid setting the size of overlap if it's
6195 (maybe_diag_overlap): Also consider arrays when deciding what values
6196 of offsets to include in diagnostics.
6198 2018-02-14 Martin Sebor <msebor@redhat.com>
6201 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
6202 that correspond to the kind of a declaration.
6204 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
6207 * config/pa/pa.md: Load address of PIC label using the linkage table
6208 if the label is nonlocal.
6210 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
6212 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
6213 warning message if user requests -maltivec=be.
6214 * doc/invoke.texi: Document deprecation of -maltivec=be.
6216 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
6219 * config/rs6000/rs6000-c.c: Update definitions for
6220 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
6221 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
6223 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6226 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
6227 add _get_ssp intrinsics. Remove argument from
6228 __builtin_ia32_rdssp[d|q].
6229 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
6230 * config/i386/i386-builtin.def: Remove argument from
6231 __builtin_ia32_rdssp[d|q].
6232 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
6233 ix86_expand_special_args_builtin for _rdssp[d|q].
6234 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
6235 Clear register before usage.
6236 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6237 Add documentation for new _get_ssp and _inc_ssp intrinsics.
6239 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
6241 PR tree-optimization/84357
6242 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6243 operand 1 of an ARRAY_REF too.
6245 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
6248 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6249 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6251 (set_of_reg): New struct.
6252 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6253 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6254 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6256 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6257 Split into bitclr, bitset, bitinvert patterns if appropriate.
6258 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6259 use rx_fuse_in_memory_bitop.
6260 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6261 to named insn, correct maximum insn length.
6263 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6266 * machmode.def: Define a complex mode for PARTIAL_INT.
6267 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6269 * doc/rtl.texi: Document CSPImode.
6270 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6272 (msp430_hard_regno_nregs_with_padding): Likewise.
6274 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
6277 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6279 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
6281 PR rtl-optimization/84169
6282 * combine.c (try_combine): New variable split_i2i3. Set it to true if
6283 we generated a parallel as new i3 and we split that to new i2 and i3
6284 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
6285 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
6286 those to i2, not i1. Partially rewrite this scan code.
6288 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6291 * stor-layout.c (place_field): For variable length fields, adjust
6292 offset_align afterwards not just based on the field's alignment,
6293 but also on the size.
6296 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6297 of exps and logs in the use_exp2 case.
6299 2018-02-13 Jeff Law <law@redhat.com>
6301 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6304 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6307 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
6311 * common.opt (gas-loc-support, gas-locview-support): New.
6312 (ginline-points, ginternal-reset-location-views): New.
6313 * doc/invoke.texi: Document them. Use @itemx where intended.
6314 (gvariable-location-views): Adjust.
6315 * target.def (reset_location_view): New.
6316 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6317 (TARGET_RESET_LOCATION_VIEW): New.
6318 * doc/tm.texi: Rebuilt.
6319 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6320 (dwarf2out_default_as_locview_support): New.
6321 (output_asm_line_debug_info): Use option variables.
6322 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6323 (output_loc_list): Likewise.
6324 (add_high_low_attributes): Check option variables.
6325 Don't output entry view attribute in strict mode.
6326 (gen_inlined_subroutine_die): Check option variables.
6327 (dwarf2out_inline_entry): Likewise.
6328 (init_sections_and_labels): Likewise.
6329 (dwarf2out_early_finish): Likewise.
6330 (maybe_reset_location_view): New, from...
6331 (dwarf2out_var_location): ... here. Call it.
6332 * debug.h (dwarf2out_default_as_loc_support): Declare.
6333 (dwarf2out_default_as_locview_support): Declare.
6334 * hooks.c (hook_int_rtx_insn_0): New.
6335 * hooks.h (hook_int_rtx_insn_0): Declare.
6336 * toplev.c (process_options): Take -gas-loc-support and
6337 -gas-locview-support from dwarf2out. Enable
6338 -gvariable-location-views by default only with locview
6339 assembler support. Enable -ginternal-reset-location-views by
6340 default only if the target defines the corresponding hook.
6341 Enable -ginline-points by default if location views are
6342 enabled; force it disabled if statement frontiers are
6344 * tree-inline.c (expand_call_inline): Check option variables.
6345 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6347 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
6349 PR tree-optimization/84321
6350 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6351 handling. Also check whether the anti-range contains any values
6352 that satisfy the mask; switch to a VR_RANGE if not.
6354 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
6357 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6359 2018-02-13 Martin Jambor <mjambor@suse.cz>
6362 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6363 of call statements, also set location of a load to a temporary.
6365 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
6367 * config/rl78/rl78.c (add_vector_labels): New function.
6368 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6369 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6370 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6371 which checks that no arguments are passed.
6372 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6373 * doc/extend.texi: Documentation for the new attribute.
6375 2018-02-13 Andreas Schwab <schwab@suse.de>
6377 * config/riscv/linux.h (CPP_SPEC): Define.
6379 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6382 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6383 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6384 OPTION_MASK_ISA_AES as first argument to def_builtin_const
6385 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6386 instead of OPTION_MASK_ISA_PCLMUL as first argument to
6387 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6388 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6389 temporarily for AES and PCLMUL builtins.
6391 PR tree-optimization/84339
6392 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6393 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6397 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6398 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6399 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6401 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6403 * omp-simd-clone.h: New file.
6404 * omp-simd-clone.c: Include omp-simd-clone.h.
6405 (expand_simd_clones): No longer static.
6406 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6407 cgraph.h and omp-simd-clone.h.
6408 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6409 (vect_recog_widen_shift_pattern): Formatting fix.
6410 (vect_pattern_recog_1): Don't check optab for calls.
6413 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6414 operands[2] into a REG before using gen_lowpart on it.
6416 2018-02-12 Jeff Law <law@redhat.com>
6419 * config/sh/sh.c (find_barrier): Consider a sibling call
6422 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6423 successfully back substituting a reg.
6425 2018-02-12 Richard Biener <rguenther@suse.de>
6427 PR tree-optimization/84037
6428 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6429 parameter, move visited init to caller.
6430 (vect_slp_analyze_operations): Separate cost from validity
6431 check, initialize visited once for all instances.
6432 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6434 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6436 (vect_model_promotion_demotion_cost): Likewise.
6437 (vectorizable_bswap): Guard cost modeling with !slp_node
6438 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6440 (vectorizable_call): Likewise.
6441 (vectorizable_conversion): Likewise.
6442 (vectorizable_assignment): Likewise.
6443 (vectorizable_shift): Likewise.
6444 (vectorizable_operation): Likewise.
6445 (vectorizable_store): Likewise.
6446 (vectorizable_load): Likewise.
6447 (vectorizable_condition): Likewise.
6448 (vectorizable_comparison): Likewise.
6450 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6453 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6454 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6455 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6457 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6460 * params.def (inline-min-speedup): Increase from 8 to 15.
6461 (max-inline-insns-auto): Decrease from 40 to 30.
6462 * ipa-split.c (consider_split): Add some buffer for function to
6463 be considered inlining candidate.
6464 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6467 2018-02-12 Richard Biener <rguenther@suse.de>
6469 PR tree-optimization/84037
6470 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6471 matched stmts if we cannot swap the non-matched ones.
6473 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6475 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6476 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6477 _mm_maskz_scalef_round_ss): New intrinsics.
6478 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6479 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6480 __builtin_ia32_scalefss_round): Remove.
6481 (__builtin_ia32_scalefsd_mask_round,
6482 __builtin_ia32_scalefss_mask_round): New intrinsics.
6483 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6484 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6485 ((match_operand:VF_128 2 "<round_nimm_predicate>"
6486 "<round_constraint>")): Changed to ...
6487 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
6488 "<round_scalar_constraint>")): ... this.
6489 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
6490 %0, %1, %2<round_op3>}"): Changed to ...
6491 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
6492 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
6493 %2<round_scalar_mask_op3>}"): ... this.
6494 * config/i386/subst.md (round_scalar_nimm_predicate): New.
6496 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6498 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
6499 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
6500 (_mm_maskz_sqrt_round_ss): New intrinsics.
6501 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
6502 (__builtin_ia32_sqrtsd_mask_round)
6503 (__builtin_ia32_sqrtss_mask_round): New builtins.
6504 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6505 (__builtin_ia32_sqrtss_round): Remove.
6506 (__builtin_ia32_sqrtsd_mask_round)
6507 (__builtin_ia32_sqrtss_mask_round): New builtins.
6508 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6509 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6510 ((match_operand:VF_128 1 "vector_operand"
6511 "xBm,<round_constraint>")): Changed to ...
6512 ((match_operand:VF_128 1 "vector_operand"
6513 "xBm,<round_scalar_constraint>")): ... this.
6514 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6515 %0, %2, %<iptr>1<round_op3>}): Changed to ...
6516 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6517 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6518 %<iptr>1<round_scalar_mask_op3>}): ... this.
6519 ((set_attr "prefix" "<round_prefix>")): Changed to ...
6520 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6522 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
6525 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6526 Cast vec_cmpeq result to correct type.
6527 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
6528 Cast vec_cmpgt result to correct type.
6530 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
6532 * final.c (final_scan_insn_1): Renamed from...
6533 (final_scan_insn): ... this. New wrapper, to recover
6534 seen from the outermost call in recursive ones.
6535 * config/sparc/sparc.c (output_return): Drop seen from call.
6536 (output_sibcall): Likewise.
6537 * config/visium/visium.c (output_branch): Likewise.
6539 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
6541 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
6544 2018-02-10 Alan Modra <amodra@gmail.com>
6547 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
6548 Specify LR as an input.
6550 2018-02-10 Jakub Jelinek <jakub@redhat.com>
6553 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
6554 remove_member_access_dummy_vars): New functions.
6555 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
6556 lower_omp_1, execute_lower_omp): Use them.
6558 PR rtl-optimization/84308
6559 * shrink-wrap.c (spread_components): Release todo vector.
6561 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
6563 PR rtl-optimization/57193
6564 * ira-color.c (struct allocno_color_data): Add member
6565 conflict_allocno_hard_prefs.
6566 (update_conflict_allocno_hard_prefs): New.
6567 (bucket_allocno_compare_func): Add a preference based on
6568 conflict_allocno_hard_prefs.
6569 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
6570 (color_allocnos): Remove a dead code. Initiate
6571 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
6573 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6576 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
6577 constraint from =wa to wa. Avoid a subreg on the output operand,
6578 instead use a pseudo and subreg it in a move.
6579 (p9_xxbrd_<mode>): Changed to ...
6580 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
6581 (p9_xxbrd_v2df): New expander.
6582 (p9_xxbrw_<mode>): Changed to ...
6583 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
6584 (p9_xxbrw_v4sf): New expander.
6586 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6588 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
6590 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
6593 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
6594 multiply in 32-bit mode.
6595 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
6596 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
6599 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6601 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
6602 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
6603 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
6604 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
6606 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
6609 * dwarf2out.c (is_trivial_indirect_ref): New function.
6610 (dwarf2out_late_global_decl): Do not generate a location
6611 attribute for variables that have a non-trivial DECL_VALUE_EXPR
6612 and that are not defined in the current unit.
6614 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6616 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
6617 instead of a libcall for UNORDERED.
6619 2018-02-09 Tamar Christina <tamar.christina@arm.com>
6622 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
6623 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
6625 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6628 * config/s390/s390.c (s390_set_current_function): Invoke
6629 s390_indirect_branch_settings also if fndecl didn't change.
6631 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6633 * config/rs6000/rs6000.md (blockage): Set length to zero.
6635 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6637 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
6639 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6642 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
6643 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
6647 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
6648 PARALLEL incoming that failed vt_get_decl_and_offset check.
6651 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
6652 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
6654 (get_variable_section): For decls in named .bss* sections pass true as
6655 second argument to bss_initializer_p.
6657 2018-02-09 Marek Polacek <polacek@redhat.com>
6658 Jakub Jelinek <jakub@redhat.com>
6661 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
6662 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
6663 Sync some changes from cxx_fold_indirect_ref.
6665 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6667 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
6669 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
6670 (BLOCK_INLINE_ENTRY_LABEL): New.
6671 (dwarf2out_var_location): Disregard inline entry markers.
6672 (inline_entry_data): New struct.
6673 (inline_entry_data_hasher): New hashtable type.
6674 (inline_entry_data_hasher::hash): New.
6675 (inline_entry_data_hasher::equal): New.
6676 (inline_entry_data_table): New variable.
6677 (add_high_low_attributes): Add DW_AT_entry_pc and
6678 DW_AT_GNU_entry_view attributes if a pending entry is found
6679 in inline_entry_data_table. Add old entry_pc attribute only
6680 if debug nonbinding markers are disabled.
6681 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
6682 markers are enabled.
6683 (block_within_block_p, dwarf2out_inline_entry): New.
6684 (dwarf2out_finish): Check that no entries remained in
6685 inline_entry_data_table.
6686 * final.c (reemit_insn_block_notes): Handle inline entry notes.
6687 (final_scan_insn, notice_source_line): Likewise.
6688 (rest_of_clean_state): Skip inline entry markers.
6689 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
6691 * gimple.c (gimple_build_debug_inline_entry): New.
6692 * gimple.h (enum gimple_debug_subcode): Add
6693 GIMPLE_DEBUG_INLINE_ENTRY.
6694 (gimple_build_debug_inline_entry): Declare.
6695 (gimple_debug_inline_entry_p): New.
6696 (gimple_debug_nonbind_marker_p): Adjust.
6697 * insn-notes.def (INLINE_ENTRY): New.
6698 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
6699 inline entry marker notes.
6700 (print_insn): Likewise.
6701 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
6702 (INSN_DEBUG_MARKER_KIND): Likewise.
6703 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
6704 * tree-inline.c (expand_call_inline): Build and insert
6705 debug_inline_entry stmt.
6706 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
6707 inline entry blocks early, if nonbind markers are enabled.
6708 (dump_scope_block): Dump fragment info.
6709 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
6710 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
6711 (gimple_build_debug_inline_entry): New.
6712 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
6713 Enable/disable inline entry points too.
6714 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
6715 (DEBUG_INSN): Describe inline entry markers.
6717 * common.opt (gvariable-location-views): New.
6718 (gvariable-location-views=incompat5): New.
6719 * config.in: Rebuilt.
6720 * configure: Rebuilt.
6721 * configure.ac: Test assembler for view support.
6722 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
6723 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
6724 * dwarf2out.c (var_loc_view): New typedef.
6725 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
6726 (dwarf2out_locviews_in_attribute): New.
6727 (dwarf2out_locviews_in_loclist): New.
6728 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
6729 (enum dw_line_info_opcode): Add LI_adv_address.
6730 (struct dw_line_info_table): Add view.
6731 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
6732 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
6733 (zero_view_p): New variable.
6734 (ZERO_VIEW_P): New macro.
6735 (output_asm_line_debug_info): New.
6736 (struct var_loc_node): Add view.
6737 (add_AT_view_list, AT_loc_list): New.
6738 (add_var_loc_to_decl): Add view param. Test it against last.
6739 (new_loc_list): Add view params. Record them.
6740 (AT_loc_list_ptr): Handle loc and view lists.
6741 (view_list_to_loc_list_val_node): New.
6742 (print_dw_val): Handle dw_val_class_view_list.
6743 (size_of_die): Likewise.
6744 (value_format): Likewise.
6745 (loc_list_has_views): New.
6746 (gen_llsym): Set vl_symbol too.
6747 (maybe_gen_llsym, skip_loc_list_entry): New.
6748 (dwarf2out_maybe_output_loclist_view_pair): New.
6749 (output_loc_list): Output view list or entries too.
6750 (output_view_list_offset): New.
6751 (output_die): Handle dw_val_class_view_list.
6752 (output_dwarf_version): New.
6753 (output_compilation_unit_header): Use it.
6754 (output_skeleton_debug_sections): Likewise.
6755 (output_rnglists, output_line_info): Likewise.
6756 (output_pubnames, output_aranges): Update version comments.
6757 (output_one_line_info_table): Output view numbers in asm comments.
6758 (dw_loc_list): Determine current endview, pass it to new_loc_list.
6759 Call maybe_gen_llsym.
6760 (loc_list_from_tree_1): Adjust.
6761 (add_AT_location_description): Create view list attribute if
6762 needed, check it's absent otherwise.
6763 (convert_cfa_to_fb_loc_list): Adjust.
6764 (maybe_emit_file): Call output_asm_line_debug_info for test.
6765 (dwarf2out_var_location): Reset views as needed. Precompute
6766 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
6767 attribute. Set view.
6768 (new_line_info_table): Reset next view.
6769 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
6770 (dwarf2out_source_line): Likewise. Output view resets and labels to
6771 the assembler, or select appropriate line info opcodes.
6772 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
6773 (optimize_string_length): Catch it. Adjust.
6774 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
6775 dw_val_class_view_list, and remove it if no longer needed.
6776 (hash_loc_list): Hash view numbers.
6777 (loc_list_hasher::equal): Compare them.
6778 (optimize_location_lists): Check whether a view list symbol is
6779 needed, and whether the locview attribute is present, and
6780 whether they match. Remove the locview attribute if no longer
6782 (index_location_lists): Call skip_loc_list_entry for test.
6783 (dwarf2out_finish): Call output_asm_line_debug_info for test.
6784 Use output_dwarf_version.
6785 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
6786 (struct dw_val_node): Add val_view_list.
6787 * final.c (SEEN_NEXT_VIEW): New.
6788 (set_next_view_needed): New.
6789 (clear_next_view_needed): New.
6790 (maybe_output_next_view): New.
6791 (final_start_function): Rename to...
6792 (final_start_function_1): ... this. Take pointer to FIRST,
6793 add SEEN parameter. Emit param bindings in the initial view.
6794 (final_start_function): Reintroduce SEEN-less interface.
6795 (final): Rename to...
6796 (final_1): ... this. Take SEEN parameter. Output final pending
6797 next view at the end.
6798 (final): Reintroduce seen-less interface.
6799 (final_scan_insn): Output pending next view before switching
6800 sections or ending a block. Mark the next view as needed when
6801 outputting variable locations. Notify debug backend of section
6802 changes, and of location view changes.
6803 (rest_of_handle_final): Adjust.
6804 * toplev.c (process_options): Autodetect value for debug variable
6805 location views option. Warn on incompat5 without -gdwarf-5.
6806 * doc/invoke.texi (gvariable-location-views): New.
6807 (gvariable-location-views=incompat5): New.
6808 (gno-variable-location-views): New.
6810 2018-02-08 David Malcolm <dmalcolm@redhat.com>
6812 PR tree-optimization/84136
6813 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
6814 that the result of find_edge is non-NULL.
6816 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6819 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
6820 storing integer register in SImode. Fix cost of 256 and 512
6821 byte aligned SSE register store.
6823 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6825 * config/i386/i386.c (ix86_multiplication_cost): Fix
6826 multiplication cost for TARGET_AVX512DQ.
6828 2018-02-08 Marek Polacek <polacek@redhat.com>
6830 PR tree-optimization/84238
6831 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
6834 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6836 PR tree-optimization/84265
6837 * tree-vect-stmts.c (vectorizable_store): Don't treat
6838 VMAT_CONTIGUOUS accesses as grouped.
6839 (vectorizable_load): Likewise.
6841 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6843 PR tree-optimization/81635
6844 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
6845 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
6846 (test_round_for_mask): New functions.
6847 (wide_int_cc_tests): Call test_round_for_mask.
6848 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
6849 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
6850 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
6851 range returned by get_range_info.
6853 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6856 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
6857 * symtab.c: Include builtins.h
6858 (symtab_node::output_to_lto_symbol_table_p): Move here
6859 from lto-streamer-out.c:output_symbol_p.
6860 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
6861 (output_symbol_p): Move all logic to symtab.c
6862 (produce_symtab): Update.
6864 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6866 * config/s390/s390-opts.h (enum indirect_branch): Define.
6867 * config/s390/s390-protos.h (s390_return_addr_from_memory)
6868 (s390_indirect_branch_via_thunk)
6869 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
6870 (enum s390_indirect_branch_type): Define.
6871 * config/s390/s390.c (struct s390_frame_layout, struct
6872 machine_function): Remove.
6873 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
6874 (indirect_branch_table_label_no, indirect_branch_table_name):
6876 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
6877 (enum s390_indirect_branch_option): Define.
6878 (s390_return_addr_from_memory): New function.
6879 (s390_handle_string_attribute): New function.
6880 (s390_attribute_table): Add new attribute handler.
6881 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
6882 (s390_indirect_branch_via_thunk): New function.
6883 (s390_indirect_branch_via_inline_thunk): New function.
6884 (s390_function_ok_for_sibcall): When jumping via thunk disallow
6885 sibling call optimization for non z10 compiles.
6886 (s390_emit_call): Force indirect branch target to be a single
6887 register. Add r1 clobber for non-z10 compiles.
6888 (s390_emit_epilogue): Emit return jump via return_use expander.
6889 (s390_reorg): Handle JUMP_INSNs as execute targets.
6890 (s390_option_override_internal): Perform validity checks for the
6891 new command line options.
6892 (s390_indirect_branch_attrvalue): New function.
6893 (s390_indirect_branch_settings): New function.
6894 (s390_set_current_function): Invoke s390_indirect_branch_settings.
6895 (s390_output_indirect_thunk_function): New function.
6896 (s390_code_end): Implement target hook.
6897 (s390_case_values_threshold): Implement target hook.
6898 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
6900 * config/s390/s390.h (struct s390_frame_layout)
6901 (struct machine_function): Move here from s390.c.
6902 (TARGET_INDIRECT_BRANCH_NOBP_RET)
6903 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
6904 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
6905 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
6906 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
6907 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
6908 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
6909 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
6910 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
6911 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
6912 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
6913 (mnemonic attribute): Add values which aren't recognized
6915 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
6916 pattern for branch conversion. Fix mnemonic attribute.
6917 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
6918 indirect branch via thunk if requested.
6919 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
6920 ("*indirect_jump"): Disable for branch conversion using out of
6922 ("indirect_jump_via_thunk<mode>_z10")
6923 ("indirect_jump_via_thunk<mode>")
6924 ("indirect_jump_via_inlinethunk<mode>_z10")
6925 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
6926 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
6927 ("casesi_jump_via_inlinethunk<mode>_z10")
6928 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
6929 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
6930 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
6931 ("*indirect2_jump"): Disable for branch conversion.
6932 ("casesi_jump"): Turn into expander and expand patterns for branch
6934 ("return_use"): New expander.
6935 ("*return"): Emit return via thunk and rename it to ...
6936 ("*return<mode>"): ... this one.
6937 * config/s390/s390.opt: Add new options and and enum for the
6940 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6942 * lra-constraints.c (match_reload): Unconditionally use
6943 gen_lowpart_SUBREG, rather than selecting between that
6944 and equivalent gen_rtx_SUBREG code.
6946 2018-02-08 Richard Biener <rguenther@suse.de>
6948 PR tree-optimization/84233
6949 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
6950 changed flag instead of boguously re-using phi_inserted.
6952 2018-02-08 Martin Jambor <mjambor@suse.cz>
6954 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
6955 static local variables.
6957 2018-02-08 Richard Biener <rguenther@suse.de>
6959 PR tree-optimization/84278
6960 * tree-vect-stmts.c (vectorizable_store): When looking for
6961 smaller vector types to perform grouped strided loads/stores
6962 make sure the mode is supported by the target.
6963 (vectorizable_load): Likewise.
6965 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6967 * config/aarch64/aarch64.c (aarch64_components_for_bb):
6968 Increase LDP/STP opportunities by adding adjacent callee-saves.
6970 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6972 PR rtl-optimization/84068
6973 PR rtl-optimization/83459
6974 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
6976 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
6978 PR tree-optimization/84224
6979 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
6980 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
6983 2018-02-07 Iain Sandoe <iain@codesourcery.com>
6986 * config/rs6000/altivec.md (*restore_world): Remove LR use.
6987 * config/rs6000/predicates.md (restore_world_operation): Adjust op
6988 count, remove one USE.
6990 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
6992 * doc/install.texi (Configuration): Document the
6993 --with-long-double-format={ibm,ieee} PowerPC configuration
6997 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
6998 Convert from define_expand to be define_insn_and_split. Rework
6999 float/double/_Float128 conversions to QI/HI/SImode to work with
7000 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
7001 conversions to QI/HImode types did a store and then a load to
7002 truncate the value. For conversions to VSX registers, don't split
7003 the insn, instead emit the code directly. Use the code iterator
7004 any_fix to combine signed and unsigned conversions.
7005 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
7006 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
7007 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7008 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
7009 (fix_<mode>di2_hw): Likewise.
7010 (fixuns_<mode>di2_hw): Likewise.
7011 (fix_<mode>si2_hw): Likewise.
7012 (fixuns_<mode>si2_hw): Likewise.
7013 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
7014 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7015 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
7016 fix<uns>_trunc<SFDF:mode>si2_p8.
7017 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
7019 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
7020 (fix<uns>_<mode>_mem): Likewise.
7021 (fctiw<u>z_<mode>_mem): Likewise.
7022 (fix<uns>_<mode>_mem): Likewise.
7023 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
7024 the register allocator from doing a direct move to the GPRs to do
7025 a store, and instead use the ISA 3.0 store byte/half-word from
7026 vector register instruction. For IEEE 128-bit floating point,
7027 also optimize stores of 32-bit ints.
7028 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
7030 2018-02-07 Alan Hayward <alan.hayward@arm.com>
7032 * genextract.c (push_pathstr_operand): New function to support
7034 (walk_rtx): Call push_pathstr_operand.
7035 (print_path): Support [a-zA-Z].
7037 2018-02-07 Richard Biener <rguenther@suse.de>
7039 PR tree-optimization/84037
7040 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
7041 (cse_and_gimplify_to_preheader): Declare.
7042 (vect_get_place_in_interleaving_chain): Likewise.
7043 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7045 (_loop_vec_info::~_loop_vec_info): Delete it.
7046 (cse_and_gimplify_to_preheader): New function.
7047 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
7048 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
7049 (vectorizable_load): Likewise. For grouped stores always base
7050 the IV on the first element.
7051 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
7052 condition before gimplifying.
7054 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7056 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
7057 *DIV_EXPR and *MOD_EXPR.
7059 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
7062 * config/i386/i386.c (ix86_option_override_internal): Mask out
7063 the CF_SET bit when checking -fcf-protection.
7065 2018-02-07 Tom de Vries <tom@codesourcery.com>
7068 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
7071 2018-02-07 Richard Biener <rguenther@suse.de>
7073 PR tree-optimization/84204
7074 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
7077 PR tree-optimization/84205
7078 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
7079 special-case isl_ast_op_zdiv_r.
7081 PR tree-optimization/84223
7082 * graphite-scop-detection.c (gather_bbs::before_dom_children):
7083 Only add conditions from within the region.
7084 (gather_bbs::after_dom_children): Adjust.
7086 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
7089 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
7090 * config/avr/avr.md: Only post-reload split REG-REG moves if
7091 either register is GENERAL_REG_P.
7093 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7095 PR tree-optimization/84235
7096 * tree-ssa-scopedtables.c
7097 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
7098 if the subtraction is performed in floating point type where NaNs are
7099 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
7100 build 1. Formatting fix.
7102 2018-02-06 Jakub Jelinek <jakub@redhat.com>
7105 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
7106 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
7107 and skip it regardless of bb boundaries. Use CALL_P macro,
7108 don't test INSN_P (insn) together with CALL_P or JUMP_P check
7109 unnecessarily, formatting fix.
7111 2018-02-06 Michael Collison <michael.collison@arm.com>
7113 * config/arm/thumb2.md:
7114 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
7115 (*thumb_mov_notscc): Ditto.
7117 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
7120 * config/rs6000/rs6000.md (su code attribute): Use "u" for
7121 unsigned_fix, not "s".
7123 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7125 * configure.ac (gcc_fn_eh_frame_ro): New function.
7126 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
7127 correct .eh_frame permissions.
7128 * configure: Regenerate.
7130 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
7132 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
7135 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7137 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7138 Display warning message for -mno-speculate-indirect-jumps.
7140 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
7142 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
7144 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
7146 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
7148 PR tree-optimization/84225
7149 * tree-eh.c (find_trapping_overflow): Only call
7150 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
7152 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7155 * config/i386/i386.c: Reimplement the check of possible options
7156 -mibt/-mshstk conbination. Change error messages.
7157 * doc/invoke.texi: Fix a typo: remove extra '='.
7159 2018-02-06 Marek Polacek <polacek@redhat.com>
7161 PR tree-optimization/84228
7162 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
7164 2018-02-06 Tamar Christina <tamar.christina@arm.com>
7167 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
7168 emitted arch directives.
7169 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
7170 __ARM_FEATURE_COPROC before changing architectures.
7172 2018-02-06 Richard Biener <rguenther@suse.de>
7174 * config/i386/i386.c (print_reg): Fix typo.
7175 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
7177 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
7179 * configure: Regenerate.
7181 2018-02-05 Martin Sebor <msebor@redhat.com>
7183 PR tree-optimization/83369
7184 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
7187 2018-02-05 Martin Liska <mliska@suse.cz>
7189 * doc/invoke.texi: Cherry-pick upstream r323995.
7191 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
7193 * ira.c (ira_init_register_move_cost): Adjust comment.
7195 2018-02-05 Martin Liska <mliska@suse.cz>
7197 PR gcov-profile/84137
7198 * doc/gcov.texi: Fix typo in documentation.
7200 2018-02-05 Martin Liska <mliska@suse.cz>
7202 PR gcov-profile/83879
7203 * doc/gcov.texi: Document necessity of --dynamic-list-data when
7204 using dlopen functionality.
7206 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
7208 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
7209 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
7210 _mm_maskz_range_ss, _mm_mask_range_round_ss,
7211 _mm_maskz_range_round_ss): New intrinsics.
7212 (__builtin_ia32_rangesd128_round)
7213 (__builtin_ia32_rangess128_round): Remove.
7214 (__builtin_ia32_rangesd128_mask_round,
7215 __builtin_ia32_rangess128_mask_round): New builtins.
7216 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
7217 __builtin_ia32_rangess128_round): Remove.
7218 (__builtin_ia32_rangesd128_mask_round,
7219 __builtin_ia32_rangess128_mask_round): New builtins.
7220 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
7221 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
7222 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
7223 "<round_saeonly_constraint>")): Changed to ...
7224 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
7225 "<round_saeonly_scalar_constraint>")): ... this.
7226 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
7227 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
7228 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
7229 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
7230 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
7232 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
7234 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
7236 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7237 Remove all values except native, 8540 and 8548.
7239 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
7241 * config/i386/i386.c (ix86_output_function_return): Pass
7242 INVALID_REGNUM, instead of -1, as invalid register number to
7243 indirect_thunk_name and output_indirect_thunk.
7245 2018-02-02 Julia Koval <julia.koval@intel.com>
7247 * config.gcc: Add -march=icelake.
7248 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7249 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7250 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7251 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7252 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7253 (processor_target_table): Add icelake.
7254 (ix86_option_override_internal): Handle new PTAs.
7255 (get_builtin_code_for_version): Handle icelake.
7256 (M_INTEL_COREI7_ICELAKE): New.
7257 (fold_builtin_cpu): Handle icelake.
7258 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7259 * doc/invoke.texi: Add -march=icelake.
7261 2018-02-02 Julia Koval <julia.koval@intel.com>
7263 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7264 to wide_int_bitmask.
7265 * wide-int-bitmask.h: New.
7267 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7270 * config/i386/i386.md: Replace Pmode with word_mode in
7271 builtin_setjmp_setup and builtin_longjmp to support x32.
7273 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
7277 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7279 (rs6000_supported_cpu_names): New static variable.
7280 (linux_cpu_translation_table): Likewise.
7281 (elf_platform) <cpu>: Define new static variable and use it.
7282 Translate kernel AT_PLATFORM name to canonical name if needed.
7283 Error if platform name is unknown.
7285 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
7288 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7290 2018-02-01 Jeff Law <law@redhat.com>
7293 * config/i386/i386.c (release_scratch_register_on_entry): Add new
7294 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
7295 the scratch if RELEASE_VIA_POP is false.
7296 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7297 If we have to save a temporary register, decrement SIZE appropriately.
7298 Pass new arguments to release_scratch_register_on_entry.
7299 (ix86_adjust_stack_and_probe): Likewise.
7300 (ix86_emit_probe_stack_range): Pass new arguments to
7301 release_scratch_register_on_entry.
7303 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
7305 PR rtl-optimization/84157
7306 * combine.c (change_zero_ext): Use REG_P predicate in
7307 front of HARD_REGISTER_P predicate.
7309 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
7311 * config/avr/avr.c (avr_option_override): Move disabling of
7312 -fdelete-null-pointer-checks to...
7313 * common/config/avr/avr-common.c (avr_option_optimization_table):
7316 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7318 PR tree-optimization/81635
7319 * tree-data-ref.c (split_constant_offset_1): For types that
7320 wrap on overflow, try to use range info to prove that wrapping
7323 2018-02-01 Renlin Li <renlin.li@arm.com>
7326 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7328 (aarch64_register_move_cost): Likewise.
7329 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7331 (REG_CLASS_NAMES): Likewise.
7332 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7333 TAILCALL_ADDR_REGS. Remove IP registers.
7334 * config/aarch64/aarch64.md (Ucs): Update register constraint.
7336 2018-02-01 Richard Biener <rguenther@suse.de>
7338 * domwalk.h (dom_walker::dom_walker): Add additional constructor
7339 for specifying RPO order and allow NULL for that.
7340 * domwalk.c (dom_walker::dom_walker): Likewise.
7341 (dom_walker::walk): Handle NULL RPO order.
7342 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7344 (rewrite_update_dom_walker): Likewise.
7345 (mark_def_dom_walker): Likewise.
7347 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7349 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7350 (aarch64_maybe_expand_sve_subreg_move): Declare.
7351 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7352 * config/aarch64/predicates.md (aarch64_any_register_operand): New
7354 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7355 that are semantically a reverse operation.
7356 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7357 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7358 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7360 (aarch64_can_change_mode_class): For big-endian, forbid changes
7361 between two SVE modes if they have different element sizes.
7363 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7365 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7366 the TImode handling for big-endian targets.
7368 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7370 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7371 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
7373 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7374 Remove BSWAP handing for big-endian targets and use the form of
7375 LD1RQ appropariate for the mode.
7377 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7379 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7380 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7383 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7386 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7387 check for operands that need to go through aarch64_sve_reload_be.
7389 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7391 PR tree-optimization/81661
7392 PR tree-optimization/84117
7393 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7394 * tree-eh.c: Include gimplify.h.
7395 (find_trapping_overflow, replace_trapping_overflow,
7396 rewrite_to_non_trapping_overflow): New functions.
7397 * tree-vect-loop.c: Include tree-eh.h.
7398 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7399 * tree-data-ref.c: Include tree-eh.h.
7400 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7402 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7404 PR rtl-optimization/84123
7405 * combine.c (change_zero_ext): Check if hard register satisfies
7406 can_change_dest_mode before calling gen_lowpart_SUBREG.
7408 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7411 * ira.c (ira_init_register_move_cost): Remove assert.
7413 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7415 PR rtl-optimization/84071
7416 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7417 * doc/tm.texi: Regenerate.
7419 2018-01-31 Richard Biener <rguenther@suse.de>
7421 PR tree-optimization/84132
7422 * tree-data-ref.c (analyze_miv_subscript): Properly
7423 check whether evolution_function_is_affine_multivariate_p
7424 before calling gcd_of_steps_may_divide_p.
7426 2018-01-31 Julia Koval <julia.koval@intel.com>
7429 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7430 * config/i386/i386.md (rdpid_rex64) New.
7431 (rdpid): Make 32bit only.
7433 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7436 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7437 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7439 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7442 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7444 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7446 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7448 PR rtl-optimization/84071
7449 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7450 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7452 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7454 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7455 (arc_attribute_table): Add 'aux' attribute.
7456 (arc_in_small_data_p): Consider aux like variables.
7457 (arc_is_aux_reg_p): New function.
7458 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7459 (arc_get_aux_arg): New function.
7460 (prepare_move_operands): Handle aux-register access.
7461 (arc_handle_aux_attribute): New function.
7462 * doc/extend.texi (ARC Variable attributes): Add subsection.
7464 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7466 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7467 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7468 (arc_attribute_table): Add 'uncached' attribute.
7469 (arc_print_operand): Print '.di' flag for uncached memory
7471 (arc_in_small_data_p): Do not consider for small data the uncached
7473 (arc_is_uncached_mem_p): New function.
7474 * config/arc/predicates.md (compact_store_memory_operand): Check
7475 for uncached memory accesses.
7476 (nonvol_nonimm_operand): Likewise.
7477 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7479 2018-01-31 Jakub Jelinek <jakub@redhat.com>
7482 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7483 falign-loops=): Add Optimization flag.
7485 2018-01-30 Jeff Law <law@redhat.com>
7488 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
7489 INT_REGISTERS_SAVED. Check it prior to calling
7490 get_scratch_register_on_entry.
7491 (ix86_adjust_stack_and_probe): Similarly.
7492 (ix86_emit_probe_stack_range): Similarly.
7493 (ix86_expand_prologue): Corresponding changes.
7495 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7498 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
7499 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
7501 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
7504 * lra-constraints.c (curr_insn_transform): Process AND in the
7507 2018-01-30 Jakub Jelinek <jakub@redhat.com>
7509 PR rtl-optimization/83986
7510 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7511 dependence against last_pending_memory_flush in addition to
7514 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
7516 PR tree-optimization/81611
7517 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7520 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7523 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7526 2018-01-30 Richard Biener <rguenther@suse.de>
7527 Jakub Jelinek <jakub@redhat.com>
7529 PR tree-optimization/84111
7530 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
7531 inner loops added during recursion, as they don't have up-to-date
7534 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7537 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
7538 (can_inline_edge_by_limits_p): ... here.
7539 (can_early_inline_edge_p, check_callers,
7540 update_caller_keys, update_callee_keys, recursive_inlining,
7541 add_new_edges_to_heap, speculation_useful_p,
7542 inline_small_functions,
7543 inline_small_functions, flatten_function,
7544 inline_to_all_callers_1): Update.
7546 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7548 * profile-count.c (profile_count::combine_with_ipa_count): Handle
7551 2018-01-30 Richard Biener <rguenther@suse.de>
7553 PR tree-optimization/83008
7554 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
7555 invariant and constant vector uses in stmts when they need
7558 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7561 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
7562 * configure: Regenerate.
7564 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7566 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
7568 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
7569 Use gen_rtx_REG rather than gen_lowpart.
7571 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7573 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
7574 rather than 0 when creating partial subregs.
7576 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7578 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
7581 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
7584 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
7585 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
7586 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
7587 flags. This restores the settings used before the 2017-07-24.
7588 Turning off pre increment/decrement/modify allows IVOPTS to
7589 optimize DF/SF loops where the index is an int.
7591 2018-01-29 Richard Biener <rguenther@suse.de>
7592 Kelvin Nilsen <kelvin@gcc.gnu.org>
7595 * tree-vect-stmts.c (vectorizable_call): Don't call
7596 targetm.vectorize_builtin_md_vectorized_function if callee is
7599 2018-01-22 Carl Love <cel@us.ibm.com>
7601 * doc/extend.tex: Fix typo in second arg in
7602 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
7604 2018-01-29 Richard Biener <rguenther@suse.de>
7606 PR tree-optimization/84086
7607 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
7608 (flush_ssaname_freelist): When SSA names were released reset
7609 the SCEV hash table.
7611 2018-01-29 Richard Biener <rguenther@suse.de>
7613 PR tree-optimization/84057
7614 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
7615 removed paths when removing edges.
7617 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
7619 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
7620 -mfunction-return=@var{choice}.
7622 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
7625 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
7627 (layout::print_source_line): Likewise.
7628 (test_get_line_width_without_trailing_whitespace): Add test cases.
7630 2018-01-27 Jakub Jelinek <jakub@redhat.com>
7633 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
7636 2018-01-26 Jim Wilson <jimw@sifive.com>
7638 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
7640 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
7643 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7645 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
7646 and CMP + SUB-immediate -> SUBS.
7648 2018-01-26 Martin Sebor <msebor@redhat.com>
7650 PR tree-optimization/83896
7651 * tree-ssa-strlen.c (get_string_len): Rename...
7652 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
7653 Avoid assuming length is constant.
7654 (handle_char_store): Use HOST_WIDE_INT for string length.
7656 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
7659 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
7660 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
7662 2018-01-26 Richard Biener <rguenther@suse.de>
7664 PR rtl-optimization/84003
7665 * dse.c (record_store): Only record redundant stores when
7666 the earlier store aliases at least all accesses the later one does.
7668 2018-01-26 Jakub Jelinek <jakub@redhat.com>
7670 PR rtl-optimization/83985
7671 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
7672 REG_CFA_RESTORE insns.
7673 (delete_unmarked_insns): Don't ignore separate shrink wrapping
7674 REG_CFA_RESTORE insns here.
7677 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
7678 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
7680 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7682 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
7683 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
7684 (arc_init): Likewise.
7685 (arc_override_options): Likewise.
7686 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
7688 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
7690 * config/arc/arc.h (TARGET_DBNZ): Define.
7691 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
7692 properly set the tune attribute.
7693 (dbnz): Use TARGET_DBNZ guard.
7694 * config/arc/arc.opt (mtune): Add core3 option.
7696 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7698 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
7699 recognize new pic like addresses.
7700 (arc_delegitimize_address): Clean up.
7702 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7704 * config/arc/arc-arches.def: Option mrf16 valid for all
7706 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
7707 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
7708 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
7709 * config/arc/arc-tables.opt: Regenerate.
7710 * config/arc/arc.c (arc_conditional_register_usage): Handle
7711 reduced register file case.
7712 (arc_file_start): Set must have build attributes.
7713 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
7715 * config/arc/arc.opt (mrf16): Add new option.
7716 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
7717 * config/arc/genmultilib.awk: Handle new mrf16 option.
7718 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
7719 * config/arc/t-multilib: Regenerate.
7720 * doc/invoke.texi (ARC Options): Document mrf16 option.
7722 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7724 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
7725 * config/arc/arc.c (arc_handle_secure_attribute): New function.
7726 (arc_attribute_table): Add 'secure_call' attribute.
7727 (arc_print_operand): Print secure call operand.
7728 (arc_function_ok_for_sibcall): Don't optimize tail calls when
7730 (arc_is_secure_call_p): New function. * config/arc/arc.md
7731 (call_i): Add support for sjli instruction.
7732 (call_value_i): Likewise.
7733 * config/arc/constraints.md (Csc): New constraint.
7735 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7736 John Eric Martin <John.Martin@emmicro-us.com>
7738 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
7739 * config/arc/arc.c (_arc_jli_section): New struct.
7740 (arc_jli_section): New type.
7741 (rc_jli_sections): New static variable.
7742 (arc_handle_jli_attribute): New function.
7743 (arc_attribute_table): Add jli_always and jli_fixed attribute.
7744 (arc_file_end): New function.
7745 (TARGET_ASM_FILE_END): Define.
7746 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
7747 (arc_add_jli_section): New function.
7748 (jli_call_scan): Likewise.
7749 (arc_reorg): Call jli_call_scan.
7750 (arc_output_addsi): Remove 'S' from printing asm operand.
7751 (arc_is_jli_call_p): New function.
7752 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
7754 (movhi_insn): Likewise.
7755 (movsi_insn): Likewise.
7756 (movsi_set_cc_insn): Likewise.
7757 (loadqi_update): Likewise.
7758 (load_zeroextendqisi_update): Likewise.
7759 (load_signextendqisi_update): Likewise.
7760 (loadhi_update): Likewise.
7761 (load_zeroextendhisi_update): Likewise.
7762 (load_signextendhisi_update): Likewise.
7763 (loadsi_update): Likewise.
7764 (loadsf_update): Likewise.
7765 (movsicc_insn): Likewise.
7766 (bset_insn): Likewise.
7767 (bxor_insn): Likewise.
7768 (bclr_insn): Likewise.
7769 (bmsk_insn): Likewise.
7770 (bicsi3_insn): Likewise.
7771 (cmpsi_cc_c_insn): Likewise.
7772 (movsi_ne): Likewise.
7773 (movsi_cond_exec): Likewise.
7774 (clrsbsi2): Likewise.
7785 (call_i): Remove 'S' asm letter, add jli instruction.
7786 (call_value_i): Likewise.
7787 * config/arc/arc.op (mjli-always): New option.
7788 * config/arc/constraints.md (Cji): New constraint.
7789 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
7791 (subsf3_fpx): Likewise.
7792 (mulsf3_fpx): Likewise.
7793 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
7795 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
7797 * doc/invoke.texi (ARC): Document mjli-always option.
7799 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
7801 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
7802 avoid addition with 0 and use incw and decw where possible.
7804 2018-01-26 Richard Biener <rguenther@suse.de>
7806 PR tree-optimization/81082
7807 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
7808 association if it requires casting to unsigned.
7809 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
7810 from fold_plusminus_mult_expr to catch important cases late when
7811 range info is available.
7813 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7815 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
7816 * configure.ac (hidden_linkonce): New test.
7817 * configure: Regenerate.
7818 * config.in: Regenerate.
7820 2018-01-26 Julia Koval <julia.koval@intel.com>
7822 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
7823 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
7824 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
7825 _mm_mask_bitshuffle_epi64_mask): Fix type.
7826 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
7827 USI_FTYPE_V4DI_V4DI_USI): Remove.
7828 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
7829 __builtin_ia32_vpshufbitqmb256_mask,
7830 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
7831 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
7832 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
7834 2018-01-26 Alan Modra <amodra@gmail.com>
7837 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
7838 UNSPEC_VBPERMQ. Sort other unspecs.
7840 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
7842 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
7844 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
7847 * predict.c (drop_profile): Do not push/pop cfun; update also
7849 (handle_missing_profiles): Fix logic looking for zero profiles.
7851 2018-01-25 Jakub Jelinek <jakub@redhat.com>
7854 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
7855 on functions with #pragma omp declare simd or functions with simd
7857 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
7858 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
7859 Remove trailing \n from warning_at calls.
7861 2018-01-25 Tom de Vries <tom@codesourcery.com>
7864 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7865 for neutered workers.
7867 2018-01-24 Joseph Myers <joseph@codesourcery.com>
7870 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
7871 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
7873 2018-01-24 Jeff Law <law@redhat.com>
7876 * i386.c (get_probe_interval): Move to earlier point.
7877 (ix86_compute_frame_layout): If -fstack-clash-protection and
7878 the frame is larger than the probe interval, then use pushes
7879 to save registers rather than reg->mem moves.
7880 (ix86_expand_prologue): Remove conditional for int_registers_saved
7883 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
7886 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
7887 min/max for never referenced object.
7889 2018-01-24 Jakub Jelinek <jakub@redhat.com>
7892 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
7894 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
7895 attributes from DECL_ATTRIBUTES (decl) without affecting
7896 DECL_ATTRIBUTES (current_function_decl).
7897 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
7898 functions with non-NULL DECL_ABSTRACT_ORIGIN.
7900 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
7902 PR tree-optimization/83979
7903 * fold-const.c (fold_comparison): Use constant_boolean_node
7904 instead of boolean_{true,false}_node.
7906 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
7908 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
7911 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7913 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7914 Simplify the clause that sets the length attribute.
7915 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7916 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
7917 clause that sets the length attribute.
7918 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7920 2018-01-24 Tom de Vries <tom@codesourcery.com>
7923 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
7924 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
7925 Add strict parameter.
7926 (prevent_branch_around_nothing): Insert dummy insn between branch to
7927 label and label with no ptx insn inbetween.
7928 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
7930 2018-01-24 Tom de Vries <tom@codesourcery.com>
7933 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7934 for neutered threads in warp.
7935 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
7937 2018-01-24 Richard Biener <rguenther@suse.de>
7939 PR tree-optimization/83176
7940 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
7943 2018-01-24 Richard Biener <rguenther@suse.de>
7945 PR tree-optimization/82819
7946 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
7947 code generating pluses that are no-ops in the target precision.
7949 2018-01-24 Richard Biener <rguenther@suse.de>
7952 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
7954 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7956 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
7957 to merge probabilities.
7958 * predict.c (probably_never_executed): Also mark as cold functions
7959 with global 0 profile and guessed local profile.
7960 * profile-count.c (profile_probability::combine_with_count): New
7962 * profile-count.h (profile_probability::operator*,
7963 profile_probability::operator*=, profile_probability::operator/,
7964 profile_probability::operator/=): Reduce precision to adjusted
7965 and set value to guessed on contradictory divisions.
7966 (profile_probability::combine_with_freq): Remove.
7967 (profile_probability::combine_wiht_count): Declare.
7968 (profile_count::force_nonzero):: Set to adjusted.
7969 (profile_count::probability_in):: Set quality to adjusted.
7970 * tree-ssa-tail-merge.c (replace_block_by): Use
7973 2018-01-23 Andrew Waterman <andrew@sifive.com>
7974 Jim Wilson <jimw@sifive.com>
7976 * config/riscv/riscv.c (riscv_stack_boundary): New.
7977 (riscv_option_override): Set riscv_stack_boundary. Handle
7978 riscv_preferred_stack_boundary_arg.
7979 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
7980 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
7981 (STACK_BOUNDARY): Set to riscv_stack_boundary.
7982 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
7983 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
7984 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
7986 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
7989 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
7990 of struct ix86_frame.
7991 (ix86_expand_epilogue): Likewise. Add a local variable for
7992 the reg_save_offset field in struct ix86_frame.
7994 2018-01-23 Bin Cheng <bin.cheng@arm.com>
7996 PR tree-optimization/82604
7997 * tree-loop-distribution.c (enum partition_kind): New enum item
7998 PKIND_PARTIAL_MEMSET.
7999 (partition_builtin_p): Support above new enum item.
8000 (generate_code_for_partition): Ditto.
8001 (compute_access_range): Differentiate cases that equality can be
8002 proven at all loops, the innermost loops or no loops.
8003 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
8004 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
8005 (finalize_partitions, distribute_loop): Don't fuse partition of
8006 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
8007 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
8010 2018-01-23 Martin Liska <mliska@suse.cz>
8012 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
8013 order to ignore the predictor.
8014 (PRED_POLYMORPHIC_CALL): Likewise.
8015 (PRED_RECURSIVE_CALL): Likewise.
8017 2018-01-23 Martin Liska <mliska@suse.cz>
8019 * tree-profile.c (tree_profiling): Print function header to
8020 aware reader which function we are working on.
8021 * value-prof.c (gimple_find_values_to_profile): Do not print
8022 not interesting value histograms.
8024 2018-01-23 Martin Liska <mliska@suse.cz>
8026 * profile-count.h (enum profile_quality): Add
8027 profile_uninitialized as the first value. Do not number values
8028 as they are zero based.
8029 (profile_count::verify): Update sanity check.
8030 (profile_probability::verify): Likewise.
8032 2018-01-23 Nathan Sidwell <nathan@acm.org>
8034 * doc/invoke.texi (ffor-scope): Deprecate.
8036 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8038 PR tree-optimization/83510
8039 * domwalk.c (set_all_edges_as_executable): New function.
8040 (dom_walker::dom_walker): Convert bool param
8041 "skip_unreachable_blocks" to enum reachability. Move setup of
8042 edge flags to set_all_edges_as_executable and only do it when
8043 reachability is REACHABLE_BLOCKS.
8044 * domwalk.h (enum dom_walker::reachability): New enum.
8045 (dom_walker::dom_walker): Convert bool param
8046 "skip_unreachable_blocks" to enum reachability.
8047 (set_all_edges_as_executable): New decl.
8048 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
8049 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
8051 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
8052 but converting true to REACHABLE_BLOCKS.
8053 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
8055 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
8056 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
8057 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
8059 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
8060 if check_all_array_refs will be called.
8062 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8064 * tree.c (selftest::test_location_wrappers): Add more test
8067 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8069 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
8070 (selftest::test_bit_in_range): Likewise.
8072 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8075 * doc/sourcebuild.texi (vect_float): Say that the selector
8076 only describes the situation when -funsafe-math-optimizations is on.
8077 (vect_float_strict): Document.
8079 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8081 PR tree-optimization/83965
8082 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
8083 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
8084 instead of checking only for a reduction.
8085 (vect_recog_widen_sum_pattern): Likewise.
8087 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8089 * predict.c (probably_never_executed): Only use precise profile info.
8090 (compute_function_frequency): Skip after inlining hack since we now
8091 have quality checking.
8093 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8095 * profile-count.h (profile_probability::very_unlikely,
8096 profile_probability::unlikely, profile_probability::even): Set
8097 precision to guessed.
8099 2018-01-23 Richard Biener <rguenther@suse.de>
8101 PR tree-optimization/83963
8102 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
8103 Properly terminate dominator walk when crossing the exit edge not
8104 when visiting its source block.
8106 2018-01-23 Jakub Jelinek <jakub@redhat.com>
8109 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
8110 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
8112 2018-01-22 Jakub Jelinek <jakub@redhat.com>
8114 PR tree-optimization/83957
8115 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
8116 semicolon after for body surrounded by braces.
8118 PR tree-optimization/83081
8119 * profile-count.h (profile_probability::split): New method.
8120 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
8121 Use profile_probability::split.
8122 (do_compare_rtx_and_jump): Fix adjustment of probabilities
8123 when splitting a single conditional jump into 2.
8125 2018-01-22 David Malcolm <dmalcolm@redhat.com>
8127 PR tree-optimization/69452
8128 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
8131 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8133 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
8134 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
8135 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
8137 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8139 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
8140 * config/rl78/rl78.md (movdi): New define_expand.
8141 * config/rl78/rl78.c (rl78_split_movdi): New function.
8143 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
8146 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
8148 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
8149 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
8150 128-bit to produce an UNSPEC move to get the double word with the
8151 signbit and then a shift directly to do signbit.
8152 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
8153 implementation with a new version that just does either a direct
8154 move or a regular move. Move memory interface to separate insns.
8155 Move insns so they are next to the expander.
8156 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
8157 with signbit move. Split big and little endian case.
8158 (signbit<mode>2_dm_mem_le): Likewise.
8159 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
8160 (signbit<mode>2_dm2): Likewise.
8162 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8164 * config/rl78/rl78.md (anddi3): New define_expand.
8166 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8168 * config/rl78/rl78.md (umindi3): New define_expand.
8170 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8172 * config/rl78/rl78.md (smindi3): New define_expand.
8174 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8176 * config/rl78/rl78.md (smaxdi3): New define_expand.
8178 2018-01-22 Carl Love <cel@us.ibm.com>
8180 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
8181 LVX_V1TI): Add macro expansion.
8182 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
8183 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
8184 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
8185 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
8186 Change check to determine if the instruction is a byte reversing
8187 entry. Fix typo in comment.
8188 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
8189 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
8190 Add def_builtin calls for new builtins.
8191 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
8192 Add define_insn expansion.
8194 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8196 * config/rl78/rl78.md (umaxdi3): New define_expand.
8198 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8200 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
8201 for non-QImode registers.
8203 2018-01-22 Richard Biener <rguenther@suse.de>
8205 PR tree-optimization/83963
8206 * graphite-scop-detection.c (scop_detection::get_sese): Delay
8207 including the loop exit block.
8208 (scop_detection::merge_sese): Likewise.
8209 (scop_detection::add_scop): Do it here instead.
8211 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8213 * doc/sourcebuild.texi (arm_softfloat): Document.
8215 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
8218 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
8219 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
8220 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
8222 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8223 David Edelsohn <dje.gcc@gmail.com>
8226 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8227 Change "crset eq" to "crset 2".
8228 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8229 (*call_indirect_aix<mode>_nospec): Likewise.
8230 (*call_value_indirect_aix<mode>_nospec): Likewise.
8231 (*call_indirect_elfv2<mode>_nospec): Likewise.
8232 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
8233 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
8234 change assembly output from . to $.
8235 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8236 (indirect_jump<mode>_nospec): Change assembly output from . to $.
8237 (*tablejump<mode>_internal1_nospec): Likewise.
8239 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
8242 * config/sh/sh_optimize_sett_clrt.cc:
8243 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8245 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
8247 PR tree-optimization/83940
8248 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8249 offset_dt to vect_constant_def rather than vect_unknown_def_type.
8250 (vect_check_load_store_mask): Add a mask_dt_out parameter and
8251 use it to pass back the definition type.
8252 (vect_check_store_rhs): Likewise rhs_dt_out.
8253 (vect_build_gather_load_calls): Add a mask_dt argument and use
8254 it instead of a call to vect_is_simple_use.
8255 (vectorizable_store): Update calls to vect_check_load_store_mask
8256 and vect_check_store_rhs. Use the dt returned by the latter instead
8257 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
8258 instead of calls to vect_is_simple_use. Pass the scalar rather
8259 than the vector operand to vect_is_simple_use when handling
8260 second and subsequent copies of an rhs value.
8261 (vectorizable_load): Update calls to vect_check_load_store_mask
8262 and vect_build_gather_load_calls. Use the cached mask_dt and
8263 gs_info.offset_dt instead of calls to vect_is_simple_use.
8265 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8268 * tree-emutls.c: Include gimplify.h.
8269 (lower_emutls_2): New function.
8270 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8271 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8272 it before further processing.
8275 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8276 UINTVAL (trueop1) instead of INTVAL (op1).
8278 2018-01-19 Jakub Jelinek <jakub@redhat.com>
8282 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8283 INCOMING_FRAME_SP_OFFSET if not defined.
8284 (scan_trace): Add ENTRY argument. If true and
8285 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8286 emit a note to adjust the CFA offset.
8287 (create_cfi_notes): Adjust scan_trace callers.
8288 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8289 INCOMING_FRAME_SP_OFFSET in the CIE.
8290 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8291 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8293 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8294 * doc/tm.texi: Regenerated.
8296 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8298 PR rtl-optimization/83147
8299 * lra-constraints.c (remove_inheritance_pseudos): Use
8300 lra_substitute_pseudo_within_insn.
8302 2018-01-19 Tom de Vries <tom@codesourcery.com>
8303 Cesar Philippidis <cesar@codesourcery.com>
8306 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8308 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
8311 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8312 spaces for function labels.
8314 2018-01-19 Martin Liska <mliska@suse.cz>
8316 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8317 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8318 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8319 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8320 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8321 (PRED_CONST_RETURN): Change from 69 to 65.
8322 (PRED_NULL_RETURN): Change from 91 to 71.
8323 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8324 (PRED_LOOP_GUARD): Change from 66 to 73.
8326 2018-01-19 Martin Liska <mliska@suse.cz>
8328 * predict.c (predict_insn_def): Add new assert.
8329 (struct branch_predictor): Change type to signed integer.
8330 (test_prediction_value_range): Amend test to cover
8332 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8333 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8334 (PRED_LOOP_ITERATIONS_MAX): Likewise.
8335 (PRED_LOOP_IV_COMPARE): Likewise.
8336 * predict.h (PROB_UNINITIALIZED): Define new constant.
8338 2018-01-19 Martin Liska <mliska@suse.cz>
8340 * predict.c (dump_prediction): Add new format for
8341 analyze_brprob.py script which is enabled with -details
8343 * profile-count.h (precise_p): New function.
8345 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8347 PR tree-optimization/83922
8348 * tree-vect-loop.c (vect_verify_full_masking): Return false if
8349 there are no statements that need masking.
8350 (vect_active_double_reduction_p): New function.
8351 (vect_analyze_loop_operations): Use it when handling phis that
8352 are not in the loop header.
8354 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8356 PR tree-optimization/83914
8357 * tree-vect-loop.c (vectorizable_induction): Don't convert
8358 init_expr or apply the peeling adjustment for inductions
8359 that are nested within the vectorized loop.
8361 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8363 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8366 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8370 * function.h (gimplify_parameters): Add gimple_seq * argument.
8371 * function.c: Include gimple.h and options.h.
8372 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8373 for the added local temporaries if needed.
8374 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8375 if there are any parameter cleanups, wrap whole body into a
8376 try/finally with the cleanups.
8378 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8381 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8382 Use GET_MODE_CLASS for scalar floating point.
8384 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8388 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8389 Fix call of call_cgraph_insertion_hooks.
8391 2018-01-18 Martin Sebor <msebor@redhat.com>
8393 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8395 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8398 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8401 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8404 * common.opt: (-ffile-prefix-map): New option.
8405 * opts.c (common_handle_option): Defer it.
8406 * opts-global.c (handle_common_deferred_options): Handle it.
8407 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8408 * file-prefix-map.h: New file.
8409 (remap_debug_filename, add_debug_prefix_map): ...here.
8410 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8411 * final.c (debug_prefix_map, add_debug_prefix_map
8412 remap_debug_filename): Move to...
8413 * file-prefix-map.c: New file.
8414 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8415 generalize, get rid of alloca(), use strrchr() instead of strchr().
8416 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8417 Implement in terms of add_prefix_map().
8418 (remap_macro_filename, remap_debug_filename): Implement in term of
8420 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8421 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8422 * dbxout.c: Include file-prefix-map.h.
8423 * varasm.c: Likewise.
8424 * vmsdbgout.c: Likewise.
8425 * xcoffout.c: Likewise.
8426 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8427 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8428 * doc/invoke.texi (-ffile-prefix-map): Document.
8429 (-fdebug-prefix-map): Update description.
8431 2018-01-18 Martin Liska <mliska@suse.cz>
8433 * config/i386/i386.c (indirect_thunk_name): Document that also
8435 (output_indirect_thunk): Document why both instructions
8436 (pause and lfence) are generated.
8438 2018-01-18 Richard Biener <rguenther@suse.de>
8440 PR tree-optimization/83887
8441 * graphite-scop-detection.c
8442 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8443 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8444 (scop_detection::merge_sese): Re-implement with a flood-fill
8445 algorithm that properly finds a SESE region if it exists.
8447 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8450 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8451 pointer_diff optimizations use view_convert instead of convert.
8453 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8455 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8456 Generate different code for -mno-speculate-indirect-jumps.
8457 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8458 (*call_indirect_aix<mode>): Disable for
8459 -mno-speculate-indirect-jumps.
8460 (*call_indirect_aix<mode>_nospec): New define_insn.
8461 (*call_value_indirect_aix<mode>): Disable for
8462 -mno-speculate-indirect-jumps.
8463 (*call_value_indirect_aix<mode>_nospec): New define_insn.
8464 (*sibcall_nonlocal_sysv<mode>): Generate different code for
8465 -mno-speculate-indirect-jumps.
8466 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8468 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8470 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8471 long double type, set the flags for noting the default long double
8472 type, even if we don't pass or return a long double type.
8474 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
8477 * ipa-inline.c (flatten_function): Do not overwrite final inlining
8480 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
8482 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8483 support for merge[hl].
8484 (fold_mergehl_helper): New helper function.
8485 (tree-vector-builder.h): New #include for tree_vector_builder usage.
8486 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
8487 (altivec_vmrglw_direct): Add xxmrglw insn.
8489 2018-01-17 Andrew Waterman <andrew@sifive.com>
8491 * config/riscv/riscv.c (riscv_conditional_register_usage): If
8492 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
8494 2018-01-17 David Malcolm <dmalcolm@redhat.com>
8497 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
8498 call the lto_location_cache before reading the
8499 DECL_SOURCE_LOCATION of the types.
8501 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
8502 Richard Sandiford <richard.sandiford@linaro.org>
8504 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8505 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8506 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8507 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8508 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8510 * config/aarch64/constraints.md (aarch64_movti_operand):
8512 * config/aarch64/predicates.md (Uti): Add new constraint.
8514 2018-01-17 Carl Love <cel@us.ibm.com>
8516 * config/rs6000/vsx.md (define_expand xl_len_r,
8517 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8518 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8520 (define_expand, define_insn): Move the shift left from the
8521 define_insn to the define_expand for lxvl and stxvl instructions.
8522 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8523 and XL_LEN_R definitions to PURE.
8525 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
8527 * config/i386/i386.c (indirect_thunk_name): Declare regno
8528 as unsigned int. Compare regno with INVALID_REGNUM.
8529 (output_indirect_thunk): Ditto.
8530 (output_indirect_thunk_function): Ditto.
8531 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
8532 in the call to output_indirect_thunk_function.
8534 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
8537 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
8538 rather than the size of inner_type to determine the stack slot size
8539 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
8541 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
8544 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
8547 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
8549 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
8550 endian Linux systems to optionally enable multilibs for selecting
8551 the long double type if the user configured an explicit type.
8552 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
8553 have no long double multilibs if not defined.
8554 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
8555 warn if the user used -mabi={ieee,ibm}longdouble and we built
8556 multilibs for long double.
8557 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
8558 appropriate multilib option.
8559 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
8561 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
8562 for building long double multilibs.
8563 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
8565 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
8567 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
8570 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
8572 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
8575 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
8578 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
8581 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8583 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
8584 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
8586 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8588 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
8589 different rtl trees depending on TARGET_64BIT.
8590 (rs6000_gen_lvx): Likewise.
8592 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8594 * config/visium/visium.md (nop): Tweak comment.
8595 (hazard_nop): Likewise.
8597 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8599 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
8600 -mspeculate-indirect-jumps.
8601 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
8602 for -mno-speculate-indirect-jumps.
8603 (*call_indirect_elfv2<mode>_nospec): New define_insn.
8604 (*call_value_indirect_elfv2<mode>): Disable for
8605 -mno-speculate-indirect-jumps.
8606 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
8607 (indirect_jump): Emit different RTL for
8608 -mno-speculate-indirect-jumps.
8609 (*indirect_jump<mode>): Disable for
8610 -mno-speculate-indirect-jumps.
8611 (*indirect_jump<mode>_nospec): New define_insn.
8612 (tablejump): Emit different RTL for
8613 -mno-speculate-indirect-jumps.
8614 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
8615 (tablejumpsi_nospec): New define_expand.
8616 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
8617 (tablejumpdi_nospec): New define_expand.
8618 (*tablejump<mode>_internal1): Disable for
8619 -mno-speculate-indirect-jumps.
8620 (*tablejump<mode>_internal1_nospec): New define_insn.
8621 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
8624 2018-01-16 Artyom Skrobov tyomitch@gmail.com
8626 * caller-save.c (insert_save): Drop unnecessary parameter. All
8629 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8630 Richard Biener <rguenth@suse.de>
8633 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
8634 return early, inline manually is_gimple_sizepos. Make sure if we
8635 call gimplify_expr we don't end up with a gimple constant.
8636 * tree.c (variably_modified_type_p): Don't return true for
8637 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
8638 * gimplify.h (is_gimple_sizepos): Remove.
8640 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8642 PR tree-optimization/83857
8643 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
8644 vectorizable_live_operation for pure SLP statements.
8645 (vectorizable_live_operation): Handle PHIs.
8647 2018-01-16 Richard Biener <rguenther@suse.de>
8649 PR tree-optimization/83867
8650 * tree-vect-stmts.c (vect_transform_stmt): Precompute
8651 nested_in_vect_loop_p since the scalar stmt may get invalidated.
8653 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8656 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
8657 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
8658 If off is not INTEGER_CST, issue a may not be aligned warning
8659 rather than isn't aligned. Use isn%'t rather than isn't.
8660 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
8662 <case MULT_EXPR>: Improve the case when bottom and one of the
8663 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
8664 operand, in that case check if the other operand is multiple of
8665 bottom divided by the INTEGER_CST operand.
8667 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8670 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
8671 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
8672 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
8673 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
8674 * config/pa/pa.c (pa_function_arg_advance): Likewise.
8675 (pa_function_arg, pa_arg_partial_bytes): Likewise.
8676 (pa_function_arg_size): New function.
8678 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8680 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
8681 in a separate statement.
8683 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8685 PR tree-optimization/83847
8686 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
8687 group gathers and scatters.
8689 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8691 PR rtl-optimization/86620
8692 * params.def (max-sched-ready-insns): Bump minimum value to 1.
8694 PR rtl-optimization/83213
8695 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
8696 to last if both are JUMP_INSNs.
8698 PR tree-optimization/83843
8699 * gimple-ssa-store-merging.c
8700 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
8701 store_immediate_info for bswap/nop orig_stores.
8703 2018-01-15 Andrew Waterman <andrew@sifive.com>
8705 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
8707 <UDIV>: Increase cost if !TARGET_DIV.
8709 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
8711 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
8712 (define_attr "cr_logical_3op"): New.
8713 (cceq_ior_compare): Adjust.
8714 (cceq_ior_compare_complement): Adjust.
8715 (*cceq_rev_compare): Adjust.
8716 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
8717 (is_cracked_insn): Adjust.
8718 (insn_must_be_first_in_group): Adjust.
8719 * config/rs6000/40x.md: Adjust.
8720 * config/rs6000/440.md: Adjust.
8721 * config/rs6000/476.md: Adjust.
8722 * config/rs6000/601.md: Adjust.
8723 * config/rs6000/603.md: Adjust.
8724 * config/rs6000/6xx.md: Adjust.
8725 * config/rs6000/7450.md: Adjust.
8726 * config/rs6000/7xx.md: Adjust.
8727 * config/rs6000/8540.md: Adjust.
8728 * config/rs6000/cell.md: Adjust.
8729 * config/rs6000/e300c2c3.md: Adjust.
8730 * config/rs6000/e500mc.md: Adjust.
8731 * config/rs6000/e500mc64.md: Adjust.
8732 * config/rs6000/e5500.md: Adjust.
8733 * config/rs6000/e6500.md: Adjust.
8734 * config/rs6000/mpc.md: Adjust.
8735 * config/rs6000/power4.md: Adjust.
8736 * config/rs6000/power5.md: Adjust.
8737 * config/rs6000/power6.md: Adjust.
8738 * config/rs6000/power7.md: Adjust.
8739 * config/rs6000/power8.md: Adjust.
8740 * config/rs6000/power9.md: Adjust.
8741 * config/rs6000/rs64.md: Adjust.
8742 * config/rs6000/titan.md: Adjust.
8744 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8746 * config/i386/predicates.md (indirect_branch_operand): Rewrite
8747 ix86_indirect_branch_register logic.
8749 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8751 * config/i386/constraints.md (Bs): Update
8752 ix86_indirect_branch_register check. Don't check
8753 ix86_indirect_branch_register with GOT_memory_operand.
8755 * config/i386/predicates.md (GOT_memory_operand): Don't check
8756 ix86_indirect_branch_register here.
8757 (GOT32_symbol_operand): Likewise.
8759 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8761 * config/i386/predicates.md (constant_call_address_operand):
8762 Rewrite ix86_indirect_branch_register logic.
8763 (sibcall_insn_operand): Likewise.
8765 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8767 * config/i386/constraints.md (Bs): Replace
8768 ix86_indirect_branch_thunk_register with
8769 ix86_indirect_branch_register.
8771 * config/i386/i386.md (indirect_jump): Likewise.
8772 (tablejump): Likewise.
8773 (*sibcall_memory): Likewise.
8774 (*sibcall_value_memory): Likewise.
8775 Peepholes of indirect call and jump via memory: Likewise.
8776 * config/i386/i386.opt: Likewise.
8777 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8778 (GOT_memory_operand): Likewise.
8779 (call_insn_operand): Likewise.
8780 (sibcall_insn_operand): Likewise.
8781 (GOT32_symbol_operand): Likewise.
8783 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8786 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
8787 type rather than type addr's type points to.
8788 (expand_omp_atomic_mutex): Likewise.
8789 (expand_omp_atomic): Likewise.
8791 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8794 * config/i386/i386.c (output_indirect_thunk_function): Use
8795 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
8796 for __x86_return_thunk.
8798 2018-01-15 Richard Biener <rguenther@suse.de>
8801 * expmed.c (extract_bit_field_1): Fix typo.
8803 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8806 * config/arm/iterators.md (VF): New mode iterator.
8807 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
8808 Remove integer-related logic from pattern.
8809 (neon_vabd<mode>_3): Likewise.
8811 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8814 * common.opt (fstrict-overflow): No longer an alias.
8815 (fwrapv-pointer): New option.
8816 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
8817 also for pointer types based on flag_wrapv_pointer.
8818 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
8819 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
8820 opts->x_flag_wrapv got set.
8821 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
8822 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
8823 POINTER_TYPE_OVERFLOW_UNDEFINED.
8824 * match.pd: Likewise in address comparison pattern.
8825 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
8827 2018-01-15 Richard Biener <rguenther@suse.de>
8830 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
8831 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
8832 Reset type names to their identifier if their TYPE_DECL doesn't
8833 have linkage (and thus is used for ODR and devirt).
8834 (save_debug_info_for_decl): Remove.
8835 (save_debug_info_for_type): Likewise.
8836 (add_tree_to_fld_list): Adjust.
8837 * tree-pretty-print.c (dump_generic_node): Make dumping of
8838 type names more robust.
8840 2018-01-15 Richard Biener <rguenther@suse.de>
8842 * BASE-VER: Bump to 8.0.1.
8844 2018-01-14 Martin Sebor <msebor@redhat.com>
8847 * builtins.c (check_access): Avoid warning when the no-warning bit
8850 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
8852 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
8853 * ira-color (allocno_hard_regs_compare): Likewise.
8855 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
8858 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
8859 Use .pushsection/.popsection.
8861 2018-01-14 Martin Sebor <msebor@redhat.com>
8864 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
8866 2018-01-14 Jakub Jelinek <jakub@redhat.com>
8868 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
8869 entry from extra_headers.
8870 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
8871 extra_headers, make the list bitwise identical to the i?86-*-* one.
8873 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8875 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
8876 -mcmodel=large with -mindirect-branch=thunk,
8877 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
8878 -mfunction-return=thunk-extern.
8879 * doc/invoke.texi: Document -mcmodel=large is incompatible with
8880 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
8881 -mfunction-return=thunk and -mfunction-return=thunk-extern.
8883 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8885 * config/i386/i386.c (print_reg): Print the name of the full
8886 integer register without '%'.
8887 (ix86_print_operand): Handle 'V'.
8888 * doc/extend.texi: Document 'V' modifier.
8890 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8892 * config/i386/constraints.md (Bs): Disallow memory operand for
8893 -mindirect-branch-register.
8895 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8896 (GOT_memory_operand): Likewise.
8897 (call_insn_operand): Likewise.
8898 (sibcall_insn_operand): Likewise.
8899 (GOT32_symbol_operand): Likewise.
8900 * config/i386/i386.md (indirect_jump): Call convert_memory_address
8901 for -mindirect-branch-register.
8902 (tablejump): Likewise.
8903 (*sibcall_memory): Likewise.
8904 (*sibcall_value_memory): Likewise.
8905 Disallow peepholes of indirect call and jump via memory for
8906 -mindirect-branch-register.
8907 (*call_pop): Replace m with Bw.
8908 (*call_value_pop): Likewise.
8909 (*sibcall_pop_memory): Replace m with Bs.
8910 * config/i386/i386.opt (mindirect-branch-register): New option.
8911 * doc/invoke.texi: Document -mindirect-branch-register option.
8913 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8915 * config/i386/i386-protos.h (ix86_output_function_return): New.
8916 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
8917 set function_return_type.
8918 (indirect_thunk_name): Add ret_p to indicate thunk for function
8920 (output_indirect_thunk_function): Pass false to
8921 indirect_thunk_name.
8922 (ix86_output_indirect_branch_via_reg): Likewise.
8923 (ix86_output_indirect_branch_via_push): Likewise.
8924 (output_indirect_thunk_function): Create alias for function
8925 return thunk if regno < 0.
8926 (ix86_output_function_return): New function.
8927 (ix86_handle_fndecl_attribute): Handle function_return.
8928 (ix86_attribute_table): Add function_return.
8929 * config/i386/i386.h (machine_function): Add
8930 function_return_type.
8931 * config/i386/i386.md (simple_return_internal): Use
8932 ix86_output_function_return.
8933 (simple_return_internal_long): Likewise.
8934 * config/i386/i386.opt (mfunction-return=): New option.
8935 (indirect_branch): Mention -mfunction-return=.
8936 * doc/extend.texi: Document function_return function attribute.
8937 * doc/invoke.texi: Document -mfunction-return= option.
8939 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8941 * config/i386/i386-opts.h (indirect_branch): New.
8942 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
8943 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
8944 with local indirect jump when converting indirect call and jump.
8945 (ix86_set_indirect_branch_type): New.
8946 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
8947 (indirectlabelno): New.
8948 (indirect_thunk_needed): Likewise.
8949 (indirect_thunk_bnd_needed): Likewise.
8950 (indirect_thunks_used): Likewise.
8951 (indirect_thunks_bnd_used): Likewise.
8952 (INDIRECT_LABEL): Likewise.
8953 (indirect_thunk_name): Likewise.
8954 (output_indirect_thunk): Likewise.
8955 (output_indirect_thunk_function): Likewise.
8956 (ix86_output_indirect_branch_via_reg): Likewise.
8957 (ix86_output_indirect_branch_via_push): Likewise.
8958 (ix86_output_indirect_branch): Likewise.
8959 (ix86_output_indirect_jmp): Likewise.
8960 (ix86_code_end): Call output_indirect_thunk_function if needed.
8961 (ix86_output_call_insn): Call ix86_output_indirect_branch if
8963 (ix86_handle_fndecl_attribute): Handle indirect_branch.
8964 (ix86_attribute_table): Add indirect_branch.
8965 * config/i386/i386.h (machine_function): Add indirect_branch_type
8966 and has_local_indirect_jump.
8967 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
8969 (tablejump): Likewise.
8970 (*indirect_jump): Use ix86_output_indirect_jmp.
8971 (*tablejump_1): Likewise.
8972 (simple_return_indirect_internal): Likewise.
8973 * config/i386/i386.opt (mindirect-branch=): New option.
8974 (indirect_branch): New.
8977 (thunk-inline): Likewise.
8978 (thunk-extern): Likewise.
8979 * doc/extend.texi: Document indirect_branch function attribute.
8980 * doc/invoke.texi: Document -mindirect-branch= option.
8982 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
8985 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
8987 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
8989 * ipa-inline.c (want_inline_small_function_p): Return false if
8990 inlining has already failed with CIF_FINAL_ERROR.
8991 (update_caller_keys): Call want_inline_small_function_p before
8993 (update_callee_keys): Likewise.
8995 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8997 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
8999 (rs6000_quadword_masked_address_p): Likewise.
9000 (quad_aligned_load_p): Likewise.
9001 (quad_aligned_store_p): Likewise.
9002 (const_load_sequence_p): Add comment to describe the outer-most loop.
9003 (mimic_memory_attributes_and_flags): New function.
9004 (rs6000_gen_stvx): Likewise.
9005 (replace_swapped_aligned_store): Likewise.
9006 (rs6000_gen_lvx): Likewise.
9007 (replace_swapped_aligned_load): Likewise.
9008 (replace_swapped_load_constant): Capitalize argument name in
9009 comment describing this function.
9010 (rs6000_analyze_swaps): Add a third pass to search for vector loads
9011 and stores that access quad-word aligned addresses and replace
9012 with stvx or lvx instructions when appropriate.
9013 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
9014 New function prototype.
9015 (rs6000_quadword_masked_address_p): Likewise.
9016 (rs6000_gen_lvx): Likewise.
9017 (rs6000_gen_stvx): Likewise.
9018 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
9019 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
9020 when memory address is aligned.
9021 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
9022 this split to select lvx instruction when memory address is aligned.
9023 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
9024 instruction when memory address is aligned.
9025 (*vsx_le_perm_load_v16qi): Likewise.
9026 (four unnamed splitters): Modify to select the stvx instruction
9027 when memory is aligned.
9029 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
9031 * predict.c (determine_unlikely_bbs): Handle correctly BBs
9032 which appears in the queue multiple times.
9034 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9035 Alan Hayward <alan.hayward@arm.com>
9036 David Sherwood <david.sherwood@arm.com>
9038 * tree-vectorizer.h (vec_lower_bound): New structure.
9039 (_loop_vec_info): Add check_nonzero and lower_bounds.
9040 (LOOP_VINFO_CHECK_NONZERO): New macro.
9041 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
9042 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
9043 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
9044 fields. Make seg_len the distance travelled, not including the
9046 (dr_direction_indicator): Declare.
9047 (dr_zero_step_indicator): Likewise.
9048 (dr_known_forward_stride_p): Likewise.
9049 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
9051 (runtime_alias_check_p): Allow runtime alias checks with
9053 (operator ==): Compare access_size and align.
9054 (prune_runtime_alias_test_list): Rework for new distinction between
9055 the access_size and seg_len.
9056 (create_intersect_range_checks_index): Likewise. Cope with polynomial
9058 (get_segment_min_max): New function.
9059 (create_intersect_range_checks): Use it.
9060 (dr_step_indicator): New function.
9061 (dr_direction_indicator): Likewise.
9062 (dr_zero_step_indicator): Likewise.
9063 (dr_known_forward_stride_p): Likewise.
9064 * tree-loop-distribution.c (data_ref_segment_size): Return
9065 DR_STEP * (niters - 1).
9066 (compute_alias_check_pairs): Update call to the dr_with_seg_len
9068 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
9069 (vect_preserves_scalar_order_p): New function, split out from...
9070 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
9071 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
9072 (vect_vfa_access_size): New function.
9073 (vect_vfa_align): Likewise.
9074 (vect_compile_time_alias): Take access_size_a and access_b arguments.
9075 (dump_lower_bound): New function.
9076 (vect_check_lower_bound): Likewise.
9077 (vect_small_gap_p): Likewise.
9078 (vectorizable_with_step_bound_p): Likewise.
9079 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
9080 depencies if the vectorization factor is 1. Convert the checks
9081 for nonzero steps into checks on the bounds of DR_STEP. Try using
9082 a bunds check for variable steps if the minimum required step is
9083 relatively small. Update calls to the dr_with_seg_len
9084 constructor and to vect_compile_time_alias.
9085 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
9087 (vect_loop_versioning): Call it.
9088 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
9090 (vect_estimate_min_profitable_iters): Account for any bounds checks.
9092 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9093 Alan Hayward <alan.hayward@arm.com>
9094 David Sherwood <david.sherwood@arm.com>
9096 * doc/sourcebuild.texi (vect_scatter_store): Document.
9097 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
9099 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
9101 * genopinit.c (main): Add supports_vec_scatter_store and
9102 supports_vec_scatter_store_cached to target_optabs.
9103 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
9104 IFN_MASK_SCATTER_STORE.
9105 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
9107 * internal-fn.h (internal_store_fn_p): Declare.
9108 (internal_fn_stored_value_index): Likewise.
9109 * internal-fn.c (scatter_store_direct): New macro.
9110 (expand_scatter_store_optab_fn): New function.
9111 (direct_scatter_store_optab_supported_p): New macro.
9112 (internal_store_fn_p): New function.
9113 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
9114 IFN_MASK_SCATTER_STORE.
9115 (internal_fn_mask_index): Likewise.
9116 (internal_fn_stored_value_index): New function.
9117 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
9119 * optabs-query.h (supports_vec_scatter_store_p): Declare.
9120 * optabs-query.c (supports_vec_scatter_store_p): New function.
9121 * tree-vectorizer.h (vect_get_store_rhs): Declare.
9122 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
9123 true for scatter stores.
9124 (vect_gather_scatter_fn_p): Handle scatter stores too.
9125 (vect_check_gather_scatter): Consider using scatter stores if
9126 supports_vec_scatter_store_p.
9127 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
9129 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9130 internal_fn_stored_value_index.
9131 (check_load_store_masking): Handle scatter stores too.
9132 (vect_get_store_rhs): Make public.
9133 (vectorizable_call): Use internal_store_fn_p.
9134 (vectorizable_store): Handle scatter store internal functions.
9135 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
9136 when deciding whether the end of the group has been reached.
9137 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
9138 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
9139 (mask_scatter_store<mode>): New insns.
9141 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9142 Alan Hayward <alan.hayward@arm.com>
9143 David Sherwood <david.sherwood@arm.com>
9145 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
9146 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
9147 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
9149 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
9150 Use vect_truncate_gather_scatter_offset if we can't treat the
9151 operation as a normal gather load or scatter store.
9152 (get_group_load_store_type): Take the gather_scatter_info
9153 as argument. Try using a gather load or scatter store for
9154 single-element groups.
9155 (get_load_store_type): Update calls to get_group_load_store_type
9156 and vect_use_strided_gather_scatters_p.
9158 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9159 Alan Hayward <alan.hayward@arm.com>
9160 David Sherwood <david.sherwood@arm.com>
9162 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
9163 optional tree argument.
9164 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
9166 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
9167 but continue to use the current value as a fallback.
9168 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
9169 to compare the updates.
9170 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
9171 (get_load_store_type): Use it when handling a strided access.
9172 (vect_get_strided_load_store_ops): New function.
9173 (vect_get_data_ptr_increment): Likewise.
9174 (vectorizable_load): Handle strided gather loads. Always pass
9175 a step to vect_create_data_ref_ptr and bump_vector_ptr.
9177 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9178 Alan Hayward <alan.hayward@arm.com>
9179 David Sherwood <david.sherwood@arm.com>
9181 * doc/md.texi (gather_load@var{m}): Document.
9182 (mask_gather_load@var{m}): Likewise.
9183 * genopinit.c (main): Add supports_vec_gather_load and
9184 supports_vec_gather_load_cached to target_optabs.
9185 * optabs-tree.c (init_tree_optimization_optabs): Use
9186 ggc_cleared_alloc to allocate target_optabs.
9187 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
9188 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
9190 * internal-fn.h (internal_load_fn_p): Declare.
9191 (internal_gather_scatter_fn_p): Likewise.
9192 (internal_fn_mask_index): Likewise.
9193 (internal_gather_scatter_fn_supported_p): Likewise.
9194 * internal-fn.c (gather_load_direct): New macro.
9195 (expand_gather_load_optab_fn): New function.
9196 (direct_gather_load_optab_supported_p): New macro.
9197 (direct_internal_fn_optab): New function.
9198 (internal_load_fn_p): Likewise.
9199 (internal_gather_scatter_fn_p): Likewise.
9200 (internal_fn_mask_index): Likewise.
9201 (internal_gather_scatter_fn_supported_p): Likewise.
9202 * optabs-query.c (supports_at_least_one_mode_p): New function.
9203 (supports_vec_gather_load_p): Likewise.
9204 * optabs-query.h (supports_vec_gather_load_p): Declare.
9205 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
9206 and memory_type field.
9207 (NUM_PATTERNS): Bump to 15.
9208 * tree-vect-data-refs.c: Include internal-fn.h.
9209 (vect_gather_scatter_fn_p): New function.
9210 (vect_describe_gather_scatter_call): Likewise.
9211 (vect_check_gather_scatter): Try using internal functions for
9212 gather loads. Recognize existing calls to a gather load function.
9213 (vect_analyze_data_refs): Consider using gather loads if
9214 supports_vec_gather_load_p.
9215 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
9216 (vect_get_gather_scatter_offset_type): Likewise.
9217 (vect_convert_mask_for_vectype): Likewise.
9218 (vect_add_conversion_to_patterm): Likewise.
9219 (vect_try_gather_scatter_pattern): Likewise.
9220 (vect_recog_gather_scatter_pattern): New pattern recognizer.
9221 (vect_vect_recog_func_ptrs): Add it.
9222 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9223 internal_fn_mask_index and internal_gather_scatter_fn_p.
9224 (check_load_store_masking): Take the gather_scatter_info as an
9225 argument and handle gather loads.
9226 (vect_get_gather_scatter_ops): New function.
9227 (vectorizable_call): Check internal_load_fn_p.
9228 (vectorizable_load): Likewise. Handle gather load internal
9230 (vectorizable_store): Update call to check_load_store_masking.
9231 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
9232 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
9233 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
9234 (aarch64_gather_scale_operand_d): New predicates.
9235 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
9236 (mask_gather_load<mode>): New insns.
9238 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9239 Alan Hayward <alan.hayward@arm.com>
9240 David Sherwood <david.sherwood@arm.com>
9242 * optabs.def (fold_left_plus_optab): New optab.
9243 * doc/md.texi (fold_left_plus_@var{m}): Document.
9244 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9245 * internal-fn.c (fold_left_direct): Define.
9246 (expand_fold_left_optab_fn): Likewise.
9247 (direct_fold_left_optab_supported_p): Likewise.
9248 * fold-const-call.c (fold_const_fold_left): New function.
9249 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9250 * tree-parloops.c (valid_reduction_p): New function.
9251 (gather_scalar_reductions): Use it.
9252 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9253 (vect_finish_replace_stmt): Declare.
9254 * tree-vect-loop.c (fold_left_reduction_fn): New function.
9255 (needs_fold_left_reduction_p): New function, split out from...
9256 (vect_is_simple_reduction): ...here. Accept reductions that
9257 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9258 (vect_force_simple_reduction): Also store the reduction type in
9259 the assignment's STMT_VINFO_REDUC_TYPE.
9260 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9261 (merge_with_identity): New function.
9262 (vect_expand_fold_left): Likewise.
9263 (vectorize_fold_left_reduction): Likewise.
9264 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9265 scalar phi in place for it. Check for target support and reject
9266 cases that would reassociate the operation. Defer the transform
9267 phase to vectorize_fold_left_reduction.
9268 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9269 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9270 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9272 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9274 * tree-if-conv.c (predicate_mem_writes): Remove redundant
9275 call to ifc_temp_var.
9277 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9278 Alan Hayward <alan.hayward@arm.com>
9279 David Sherwood <david.sherwood@arm.com>
9281 * target.def (legitimize_address_displacement): Take the original
9282 offset as a poly_int.
9283 * targhooks.h (default_legitimize_address_displacement): Update
9285 * targhooks.c (default_legitimize_address_displacement): Likewise.
9286 * doc/tm.texi: Regenerate.
9287 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9288 as an argument, moving assert of ad->disp == ad->disp_term to...
9289 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
9290 Try calling targetm.legitimize_address_displacement before expanding
9291 the address rather than afterwards, and adjust for the new interface.
9292 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9293 Match the new hook interface. Handle SVE addresses.
9294 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9297 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9299 * Makefile.in (OBJS): Add early-remat.o.
9300 * target.def (select_early_remat_modes): New hook.
9301 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9302 * doc/tm.texi: Regenerate.
9303 * targhooks.h (default_select_early_remat_modes): Declare.
9304 * targhooks.c (default_select_early_remat_modes): New function.
9305 * timevar.def (TV_EARLY_REMAT): New timevar.
9306 * passes.def (pass_early_remat): New pass.
9307 * tree-pass.h (make_pass_early_remat): Declare.
9308 * early-remat.c: New file.
9309 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9311 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
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-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9318 vfm1 with a bound_epilog parameter.
9319 (vect_do_peeling): Update calls accordingly, and move the prologue
9320 call earlier in the function. Treat the base bound_epilog as 0 for
9321 fully-masked loops and retain vf - 1 for other loops. Add 1 to
9322 this base when peeling for gaps.
9323 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9324 with fully-masked loops.
9325 (vect_estimate_min_profitable_iters): Handle the single peeled
9326 iteration in that case.
9328 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9329 Alan Hayward <alan.hayward@arm.com>
9330 David Sherwood <david.sherwood@arm.com>
9332 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9333 single-element interleaving even if the size is not a power of 2.
9334 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9335 accesses for single-element interleaving if the group size is
9338 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9339 Alan Hayward <alan.hayward@arm.com>
9340 David Sherwood <david.sherwood@arm.com>
9342 * doc/md.texi (fold_extract_last_@var{m}): Document.
9343 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9344 * optabs.def (fold_extract_last_optab): New optab.
9345 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9346 * internal-fn.c (fold_extract_direct): New macro.
9347 (expand_fold_extract_optab_fn): Likewise.
9348 (direct_fold_extract_optab_supported_p): Likewise.
9349 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9350 * tree-vect-loop.c (vect_model_reduction_cost): Handle
9351 EXTRACT_LAST_REDUCTION.
9352 (get_initial_def_for_reduction): Do not create an initial vector
9353 for EXTRACT_LAST_REDUCTION reductions.
9354 (vectorizable_reduction): Leave the scalar phi in place for
9355 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
9356 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
9357 epilogue code for EXTRACT_LAST_REDUCTION and defer the
9358 transform phase to vectorizable_condition.
9359 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9361 (vect_finish_stmt_generation): ...here.
9362 (vect_finish_replace_stmt): New function.
9363 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9364 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9366 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9368 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9369 Alan Hayward <alan.hayward@arm.com>
9370 David Sherwood <david.sherwood@arm.com>
9372 * doc/md.texi (extract_last_@var{m}): Document.
9373 * optabs.def (extract_last_optab): New optab.
9374 * internal-fn.def (EXTRACT_LAST): New internal function.
9375 * internal-fn.c (cond_unary_direct): New macro.
9376 (expand_cond_unary_optab_fn): Likewise.
9377 (direct_cond_unary_optab_supported_p): Likewise.
9378 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9379 loops using EXTRACT_LAST.
9380 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9381 (extract_last_<mode>): ...this optab.
9382 (vec_extract<mode><Vel>): Update accordingly.
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 * target.def (empty_mask_is_expensive): New hook.
9389 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9390 * doc/tm.texi: Regenerate.
9391 * targhooks.h (default_empty_mask_is_expensive): Declare.
9392 * targhooks.c (default_empty_mask_is_expensive): New function.
9393 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9394 if the target says that empty masks are expensive.
9395 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9397 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9399 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9400 Alan Hayward <alan.hayward@arm.com>
9401 David Sherwood <david.sherwood@arm.com>
9403 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9404 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9405 (vect_use_loop_mask_for_alignment_p): New function.
9406 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9407 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9408 niters_skip argument. Make sure that the first niters_skip elements
9409 of the first iteration are inactive.
9410 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9411 Update call to vect_set_loop_masks_directly.
9412 (get_misalign_in_elems): New function, split out from...
9413 (vect_gen_prolog_loop_niters): ...here.
9414 (vect_update_init_of_dr): Take a code argument that specifies whether
9415 the adjustment should be added or subtracted.
9416 (vect_update_init_of_drs): Likewise.
9417 (vect_prepare_for_masked_peels): New function.
9418 (vect_do_peeling): Skip prologue peeling if we're using a mask
9419 instead. Update call to vect_update_inits_of_drs.
9420 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9422 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9423 alignment. Do not include the number of peeled iterations in
9424 the minimum threshold in that case.
9425 (vectorizable_induction): Adjust the start value down by
9426 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9427 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9428 Take the number of skipped iterations into account when calculating
9430 * tree-vect-stmts.c (vect_gen_while_not): New function.
9432 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9433 Alan Hayward <alan.hayward@arm.com>
9434 David Sherwood <david.sherwood@arm.com>
9436 * doc/sourcebuild.texi (vect_fully_masked): Document.
9437 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9439 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9441 (vect_analyze_loop_2): ...here. Don't check the vectorization
9442 factor against the number of loop iterations if the loop is
9445 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9446 Alan Hayward <alan.hayward@arm.com>
9447 David Sherwood <david.sherwood@arm.com>
9449 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9450 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9451 (dump_groups): Update accordingly.
9452 (iv_use::mem_type): New member variable.
9453 (address_p): New function.
9454 (record_use): Add a mem_type argument and initialize the new
9456 (record_group_use): Add a mem_type argument. Use address_p.
9457 Remove obsolete null checks of base_object. Update call to record_use.
9458 (find_interesting_uses_op): Update call to record_group_use.
9459 (find_interesting_uses_cond): Likewise.
9460 (find_interesting_uses_address): Likewise.
9461 (get_mem_type_for_internal_fn): New function.
9462 (find_address_like_use): Likewise.
9463 (find_interesting_uses_stmt): Try find_address_like_use before
9464 calling find_interesting_uses_op.
9465 (addr_offset_valid_p): Use the iv mem_type field as the type
9466 of the addressed memory.
9467 (add_autoinc_candidates): Likewise.
9468 (get_address_cost): Likewise.
9469 (split_small_address_groups_p): Use address_p.
9470 (split_address_groups): Likewise.
9471 (add_iv_candidate_for_use): Likewise.
9472 (autoinc_possible_for_pair): Likewise.
9473 (rewrite_groups): Likewise.
9474 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9475 (determine_group_iv_cost): Update after split of USE_ADDRESS.
9476 (get_alias_ptr_type_for_ptr_address): New function.
9477 (rewrite_use_address): Rewrite address uses in calls that were
9478 identified by find_address_like_use.
9480 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9481 Alan Hayward <alan.hayward@arm.com>
9482 David Sherwood <david.sherwood@arm.com>
9484 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9486 * gimple-expr.h (is_gimple_addressable: Likewise.
9487 * gimple-expr.c (is_gimple_address): Likewise.
9488 * internal-fn.c (expand_call_mem_ref): New function.
9489 (expand_mask_load_optab_fn): Use it.
9490 (expand_mask_store_optab_fn): Likewise.
9492 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9493 Alan Hayward <alan.hayward@arm.com>
9494 David Sherwood <david.sherwood@arm.com>
9496 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
9497 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
9498 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
9499 (cond_umax@var{mode}): Document.
9500 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
9501 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
9502 (cond_umin_optab, cond_umax_optab): New optabs.
9503 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
9504 (COND_IOR, COND_XOR): New internal functions.
9505 * internal-fn.h (get_conditional_internal_fn): Declare.
9506 * internal-fn.c (cond_binary_direct): New macro.
9507 (expand_cond_binary_optab_fn): Likewise.
9508 (direct_cond_binary_optab_supported_p): Likewise.
9509 (get_conditional_internal_fn): New function.
9510 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9511 Cope with reduction statements that are vectorized as calls rather
9513 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9514 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9515 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9516 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9517 (UNSPEC_COND_EOR): New unspecs.
9518 (optab): Add mappings for them.
9519 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9520 (sve_int_op, sve_fp_op): New int attributes.
9522 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9523 Alan Hayward <alan.hayward@arm.com>
9524 David Sherwood <david.sherwood@arm.com>
9526 * optabs.def (while_ult_optab): New optab.
9527 * doc/md.texi (while_ult@var{m}@var{n}): Document.
9528 * internal-fn.def (WHILE_ULT): New internal function.
9529 * internal-fn.h (direct_internal_fn_supported_p): New override
9530 that takes two types as argument.
9531 * internal-fn.c (while_direct): New macro.
9532 (expand_while_optab_fn): New function.
9533 (convert_optab_supported_p): Likewise.
9534 (direct_while_optab_supported_p): New macro.
9535 * wide-int.h (wi::udiv_ceil): New function.
9536 * tree-vectorizer.h (rgroup_masks): New structure.
9537 (vec_loop_masks): New typedef.
9538 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
9540 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
9541 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
9542 (vect_max_vf): New function.
9543 (slpeel_make_loop_iterate_ntimes): Delete.
9544 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
9545 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
9546 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
9547 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
9548 internal-fn.h, stor-layout.h and optabs-query.h.
9549 (vect_set_loop_mask): New function.
9550 (add_preheader_seq): Likewise.
9551 (add_header_seq): Likewise.
9552 (interleave_supported_p): Likewise.
9553 (vect_maybe_permute_loop_masks): Likewise.
9554 (vect_set_loop_masks_directly): Likewise.
9555 (vect_set_loop_condition_masked): Likewise.
9556 (vect_set_loop_condition_unmasked): New function, split out from
9557 slpeel_make_loop_iterate_ntimes.
9558 (slpeel_make_loop_iterate_ntimes): Rename to..
9559 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
9560 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
9561 (vect_do_peeling): Update call accordingly.
9562 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
9564 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9565 mask_compare_type, can_fully_mask_p and fully_masked_p.
9566 (release_vec_loop_masks): New function.
9567 (_loop_vec_info): Use it to free the loop masks.
9568 (can_produce_all_loop_masks_p): New function.
9569 (vect_get_max_nscalars_per_iter): Likewise.
9570 (vect_verify_full_masking): Likewise.
9571 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
9572 retries, and free the mask rgroups before retrying. Check loop-wide
9573 reasons for disallowing fully-masked loops. Make the final decision
9574 about whether use a fully-masked loop or not.
9575 (vect_estimate_min_profitable_iters): Do not assume that peeling
9576 for the number of iterations will be needed for fully-masked loops.
9577 (vectorizable_reduction): Disable fully-masked loops.
9578 (vectorizable_live_operation): Likewise.
9579 (vect_halve_mask_nunits): New function.
9580 (vect_double_mask_nunits): Likewise.
9581 (vect_record_loop_mask): Likewise.
9582 (vect_get_loop_mask): Likewise.
9583 (vect_transform_loop): Handle the case in which the final loop
9584 iteration might handle a partial vector. Call vect_set_loop_condition
9585 instead of slpeel_make_loop_iterate_ntimes.
9586 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
9587 (check_load_store_masking): New function.
9588 (prepare_load_store_mask): Likewise.
9589 (vectorizable_store): Handle fully-masked loops.
9590 (vectorizable_load): Likewise.
9591 (supportable_widening_operation): Use vect_halve_mask_nunits for
9593 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
9594 (vect_gen_while): New function.
9595 * config/aarch64/aarch64.md (umax<mode>3): New expander.
9596 (aarch64_uqdec<mode>): New insn.
9598 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9599 Alan Hayward <alan.hayward@arm.com>
9600 David Sherwood <david.sherwood@arm.com>
9602 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
9603 (reduc_xor_scal_optab): New optabs.
9604 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
9605 (reduc_xor_scal_@var{m}): Document.
9606 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
9607 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
9609 * fold-const-call.c (fold_const_call): Handle them.
9610 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
9611 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
9612 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
9613 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
9614 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
9615 (UNSPEC_XORV): New unspecs.
9616 (optab): Add entries for them.
9617 (BITWISEV): New int iterator.
9618 (bit_reduc_op): New int attributes.
9620 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9621 Alan Hayward <alan.hayward@arm.com>
9622 David Sherwood <david.sherwood@arm.com>
9624 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
9625 * internal-fn.def (VEC_SHL_INSERT): New internal function.
9626 * optabs.def (vec_shl_insert_optab): New optab.
9627 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
9628 (duplicate_and_interleave): Likewise.
9629 * tree-vect-loop.c: Include internal-fn.h.
9630 (neutral_op_for_slp_reduction): New function, split out from
9631 get_initial_defs_for_reduction.
9632 (get_initial_def_for_reduction): Handle option 2 for variable-length
9633 vectors by loading the neutral value into a vector and then shifting
9634 the initial value into element 0.
9635 (get_initial_defs_for_reduction): Replace the code argument with
9636 the neutral value calculated by neutral_op_for_slp_reduction.
9637 Use gimple_build_vector for constant-length vectors.
9638 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
9639 but the first group_size elements have a neutral value.
9640 Use duplicate_and_interleave otherwise.
9641 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
9642 Update call to get_initial_defs_for_reduction. Handle SLP
9643 reductions for variable-length vectors by creating one vector
9644 result for each scalar result, with the elements associated
9645 with other scalar results stubbed out with the neutral value.
9646 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
9647 Require IFN_VEC_SHL_INSERT for double reductions on
9648 variable-length vectors, or SLP reductions that have
9649 a neutral value. Require can_duplicate_and_interleave_p
9650 support for variable-length unchained SLP reductions if there
9651 is no neutral value, such as for MIN/MAX reductions. Also require
9652 the number of vector elements to be a multiple of the number of
9653 SLP statements when doing variable-length unchained SLP reductions.
9654 Update call to vect_create_epilog_for_reduction.
9655 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
9656 and remove initial values.
9657 (duplicate_and_interleave): Make public.
9658 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
9659 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
9661 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9662 Alan Hayward <alan.hayward@arm.com>
9663 David Sherwood <david.sherwood@arm.com>
9665 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
9666 (can_duplicate_and_interleave_p): New function.
9667 (vect_get_and_check_slp_defs): Take the vector of statements
9668 rather than just the current one. Remove excess parentheses.
9669 Restriction rejectinon of vect_constant_def and vect_external_def
9670 for variable-length vectors to boolean types, or types for which
9671 can_duplicate_and_interleave_p is false.
9672 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
9673 (duplicate_and_interleave): New function.
9674 (vect_get_constant_vectors): Use gimple_build_vector for
9675 constant-length vectors and suitable variable-length constant
9676 vectors. Use duplicate_and_interleave for other variable-length
9677 vectors. Don't defer the update when inserting new statements.
9679 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9680 Alan Hayward <alan.hayward@arm.com>
9681 David Sherwood <david.sherwood@arm.com>
9683 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
9684 min_profitable_iters doesn't go negative.
9686 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9687 Alan Hayward <alan.hayward@arm.com>
9688 David Sherwood <david.sherwood@arm.com>
9690 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
9691 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
9692 * optabs.def (vec_mask_load_lanes_optab): New optab.
9693 (vec_mask_store_lanes_optab): Likewise.
9694 * internal-fn.def (MASK_LOAD_LANES): New internal function.
9695 (MASK_STORE_LANES): Likewise.
9696 * internal-fn.c (mask_load_lanes_direct): New macro.
9697 (mask_store_lanes_direct): Likewise.
9698 (expand_mask_load_optab_fn): Handle masked operations.
9699 (expand_mask_load_lanes_optab_fn): New macro.
9700 (expand_mask_store_optab_fn): Handle masked operations.
9701 (expand_mask_store_lanes_optab_fn): New macro.
9702 (direct_mask_load_lanes_optab_supported_p): Likewise.
9703 (direct_mask_store_lanes_optab_supported_p): Likewise.
9704 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
9706 (vect_load_lanes_supported): Likewise.
9707 * tree-vect-data-refs.c (strip_conversion): New function.
9708 (can_group_stmts_p): Likewise.
9709 (vect_analyze_data_ref_accesses): Use it instead of checking
9710 for a pair of assignments.
9711 (vect_store_lanes_supported): Take a masked_p parameter.
9712 (vect_load_lanes_supported): Likewise.
9713 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
9714 vect_store_lanes_supported and vect_load_lanes_supported.
9715 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9716 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
9717 parameter. Don't allow gaps for masked accesses.
9718 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
9719 and vect_load_lanes_supported.
9720 (get_load_store_type): Take a masked_p parameter and update
9721 call to get_group_load_store_type.
9722 (vectorizable_store): Update call to get_load_store_type.
9723 Handle IFN_MASK_STORE_LANES.
9724 (vectorizable_load): Update call to get_load_store_type.
9725 Handle IFN_MASK_LOAD_LANES.
9727 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9728 Alan Hayward <alan.hayward@arm.com>
9729 David Sherwood <david.sherwood@arm.com>
9731 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
9733 * config/aarch64/aarch64-protos.h
9734 (aarch64_sve_struct_memory_operand_p): Declare.
9735 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
9736 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
9737 (VPRED, vpred): Handle SVE structure modes.
9738 * config/aarch64/constraints.md (Utx): New constraint.
9739 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
9740 (aarch64_sve_struct_nonimmediate_operand): New predicates.
9741 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
9742 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
9743 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
9744 structure modes. Split into pieces after RA.
9745 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
9746 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
9748 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
9749 SVE structure modes.
9750 (aarch64_classify_address): Likewise.
9751 (sizetochar): Move earlier in file.
9752 (aarch64_print_operand): Handle SVE register lists.
9753 (aarch64_array_mode): New function.
9754 (aarch64_sve_struct_memory_operand_p): Likewise.
9755 (TARGET_ARRAY_MODE): Redefine.
9757 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9758 Alan Hayward <alan.hayward@arm.com>
9759 David Sherwood <david.sherwood@arm.com>
9761 * target.def (array_mode): New target hook.
9762 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
9763 * doc/tm.texi: Regenerate.
9764 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
9765 * hooks.c (hook_optmode_mode_uhwi_none): New function.
9766 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
9768 * stor-layout.c (mode_for_array): Likewise. Support polynomial
9771 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9772 Alan Hayward <alan.hayward@arm.com>
9773 David Sherwood <david.sherwood@arm.com>
9775 * fold-const.c (fold_binary_loc): Check the argument types
9776 rather than the result type when testing for a vector operation.
9778 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9780 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
9781 * doc/tm.texi: Regenerate.
9783 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9784 Alan Hayward <alan.hayward@arm.com>
9785 David Sherwood <david.sherwood@arm.com>
9787 * doc/invoke.texi (-msve-vector-bits=): Document new option.
9788 (sve): Document new AArch64 extension.
9789 * doc/md.texi (w): Extend the description of the AArch64
9790 constraint to include SVE vectors.
9791 (Upl, Upa): Document new AArch64 predicate constraints.
9792 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
9794 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
9795 (msve-vector-bits=): New option.
9796 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
9797 SVE when these are disabled.
9798 (sve): New extension.
9799 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
9800 modes. Adjust their number of units based on aarch64_sve_vg.
9801 (MAX_BITSIZE_MODE_ANY_MODE): Define.
9802 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
9803 aarch64_addr_query_type.
9804 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
9805 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
9806 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
9807 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
9808 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
9809 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
9810 (aarch64_simd_imm_zero_p): Delete.
9811 (aarch64_check_zero_based_sve_index_immediate): Declare.
9812 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9813 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9814 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9815 (aarch64_sve_float_mul_immediate_p): Likewise.
9816 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9818 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
9819 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
9820 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
9821 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
9822 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
9823 (aarch64_regmode_natural_size): Likewise.
9824 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
9825 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
9827 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
9828 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
9829 for VG and the SVE predicate registers.
9830 (V_ALIASES): Add a "z"-prefixed alias.
9831 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
9832 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
9833 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
9834 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
9835 (REG_CLASS_NAMES): Add entries for them.
9836 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
9837 and the predicate registers.
9838 (aarch64_sve_vg): Declare.
9839 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
9840 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
9841 (REGMODE_NATURAL_SIZE): Define.
9842 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
9844 * config/aarch64/aarch64.c: Include cfgrtl.h.
9845 (simd_immediate_info): Add a constructor for series vectors,
9846 and an associated step field.
9847 (aarch64_sve_vg): New variable.
9848 (aarch64_dbx_register_number): Handle VG and the predicate registers.
9849 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
9850 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
9851 (VEC_ANY_DATA, VEC_STRUCT): New constants.
9852 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
9853 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
9854 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
9855 (aarch64_get_mask_mode): New functions.
9856 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
9857 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
9858 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
9859 predicate modes and predicate registers. Explicitly restrict
9860 GPRs to modes of 16 bytes or smaller. Only allow FP registers
9861 to store a vector mode if it is recognized by
9862 aarch64_classify_vector_mode.
9863 (aarch64_regmode_natural_size): New function.
9864 (aarch64_hard_regno_caller_save_mode): Return the original mode
9866 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
9867 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
9868 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
9869 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
9871 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
9872 does not overlap dest if the function is frame-related. Handle
9874 (aarch64_split_add_offset): New function.
9875 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
9876 them aarch64_add_offset.
9877 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
9878 and update call to aarch64_sub_sp.
9879 (aarch64_add_cfa_expression): New function.
9880 (aarch64_expand_prologue): Pass extra temporary registers to the
9881 functions above. Handle the case in which we need to emit new
9882 DW_CFA_expressions for registers that were originally saved
9883 relative to the stack pointer, but now have to be expressed
9884 relative to the frame pointer.
9885 (aarch64_output_mi_thunk): Pass extra temporary registers to the
9887 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
9888 IP0 and IP1 values for SVE frames.
9889 (aarch64_expand_vec_series): New function.
9890 (aarch64_expand_sve_widened_duplicate): Likewise.
9891 (aarch64_expand_sve_const_vector): Likewise.
9892 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
9893 Handle SVE constants. Use emit_move_insn to move a force_const_mem
9894 into the register, rather than emitting a SET directly.
9895 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
9896 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
9897 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
9898 (offset_9bit_signed_scaled_p): New functions.
9899 (aarch64_replicate_bitmask_imm): New function.
9900 (aarch64_bitmask_imm): Use it.
9901 (aarch64_cannot_force_const_mem): Reject expressions involving
9902 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
9903 (aarch64_classify_index): Handle SVE indices, by requiring
9904 a plain register index with a scale that matches the element size.
9905 (aarch64_classify_address): Handle SVE addresses. Assert that
9906 the mode of the address is VOIDmode or an integer mode.
9907 Update call to aarch64_classify_symbol.
9908 (aarch64_classify_symbolic_expression): Update call to
9909 aarch64_classify_symbol.
9910 (aarch64_const_vec_all_in_range_p): New function.
9911 (aarch64_print_vector_float_operand): Likewise.
9912 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
9913 "vN" for FP registers with SVE modes. Handle (const ...) vectors
9914 and the FP immediates 1.0 and 0.5.
9915 (aarch64_print_address_internal): Handle SVE addresses.
9916 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
9917 (aarch64_regno_regclass): Handle predicate registers.
9918 (aarch64_secondary_reload): Handle big-endian reloads of SVE
9920 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
9921 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
9922 (aarch64_convert_sve_vector_bits): New function.
9923 (aarch64_override_options): Use it to handle -msve-vector-bits=.
9924 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9926 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
9927 Handle SVE vector and predicate modes. Accept VL-based constants
9928 that need only one temporary register, and VL offsets that require
9929 no temporary registers.
9930 (aarch64_conditional_register_usage): Mark the predicate registers
9931 as fixed if SVE isn't available.
9932 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
9933 Return true for SVE vector and predicate modes.
9934 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
9935 rather than an unsigned int. Handle SVE modes.
9936 (aarch64_preferred_simd_mode): Update call accordingly. Handle
9938 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
9940 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9941 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9942 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9943 (aarch64_sve_float_mul_immediate_p): New functions.
9944 (aarch64_sve_valid_immediate): New function.
9945 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
9946 Explicitly reject structure modes. Check for INDEX constants.
9947 Handle PTRUE and PFALSE constants.
9948 (aarch64_check_zero_based_sve_index_immediate): New function.
9949 (aarch64_simd_imm_zero_p): Delete.
9950 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
9951 vector modes. Accept constants in the range of CNT[BHWD].
9952 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
9953 ask for an Advanced SIMD mode.
9954 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
9955 (aarch64_simd_vector_alignment): Handle SVE predicates.
9956 (aarch64_vectorize_preferred_vector_alignment): New function.
9957 (aarch64_simd_vector_alignment_reachable): Use it instead of
9959 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
9960 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
9962 (MAX_VECT_LEN): Delete.
9963 (expand_vec_perm_d): Add a vec_flags field.
9964 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
9965 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
9966 (aarch64_evpc_ext): Don't apply a big-endian lane correction
9968 (aarch64_evpc_rev): Rename to...
9969 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
9970 (aarch64_evpc_rev_global): New function.
9971 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
9972 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
9974 (aarch64_evpc_sve_tbl): New function.
9975 (aarch64_expand_vec_perm_const_1): Update after rename of
9976 aarch64_evpc_rev. Handle SVE permutes too, trying
9977 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
9978 than aarch64_evpc_tbl.
9979 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
9980 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
9981 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
9982 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
9983 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
9984 (aarch64_expand_sve_vcond): New functions.
9985 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
9986 of aarch64_vector_mode_p.
9987 (aarch64_dwarf_poly_indeterminate_value): New function.
9988 (aarch64_compute_pressure_classes): Likewise.
9989 (aarch64_can_change_mode_class): Likewise.
9990 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
9991 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
9992 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
9993 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
9994 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
9995 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
9996 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
9997 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
9999 (Dn, Dl, Dr): Accept const as well as const_vector.
10000 (Dz): Likewise. Compare against CONST0_RTX.
10001 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
10002 of "vector" where appropriate.
10003 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
10004 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
10005 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
10006 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
10007 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
10008 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
10009 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
10010 (v_int_equiv): Extend to SVE modes.
10011 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
10013 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
10014 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
10015 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
10016 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
10017 (SVE_COND_FP_CMP): New int iterators.
10018 (perm_hilo): Handle the new unpack unspecs.
10019 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
10021 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
10022 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
10023 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
10024 (aarch64_equality_operator, aarch64_constant_vector_operand)
10025 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
10026 (aarch64_sve_nonimmediate_operand): Likewise.
10027 (aarch64_sve_general_operand): Likewise.
10028 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
10029 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
10030 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
10031 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
10032 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
10033 (aarch64_sve_float_arith_immediate): Likewise.
10034 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
10035 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
10036 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
10037 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
10038 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
10039 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
10040 (aarch64_sve_float_arith_operand): Likewise.
10041 (aarch64_sve_float_arith_with_sub_operand): Likewise.
10042 (aarch64_sve_float_mul_operand): Likewise.
10043 (aarch64_sve_vec_perm_operand): Likewise.
10044 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
10045 (aarch64_mov_operand): Accept const_poly_int and const_vector.
10046 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
10047 as well as const_vector.
10048 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
10049 in file. Use CONST0_RTX and CONSTM1_RTX.
10050 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
10051 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
10052 Use aarch64_simd_imm_zero.
10053 * config/aarch64/aarch64-sve.md: New file.
10054 * config/aarch64/aarch64.md: Include it.
10055 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
10056 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
10057 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
10058 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
10059 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
10060 (sve): New attribute.
10061 (enabled): Disable instructions with the sve attribute unless
10063 (movqi, movhi): Pass CONST_POLY_INT operaneds through
10064 aarch64_expand_mov_immediate.
10065 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
10066 CNT[BHSD] immediates.
10067 (movti): Split CONST_POLY_INT moves into two halves.
10068 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
10069 Split additions that need a temporary here if the destination
10070 is the stack pointer.
10071 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
10072 (*add<mode>3_poly_1): New instruction.
10073 (set_clobber_cc): New expander.
10075 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10077 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
10078 parameter and use it instead of GET_MODE_SIZE (innermode). Use
10079 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
10080 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
10081 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
10082 Change innermode from fixed_mode_size to machine_mode.
10083 (simplify_subreg): Update call accordingly. Handle a constant-sized
10084 subreg of a variable-length CONST_VECTOR.
10086 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10087 Alan Hayward <alan.hayward@arm.com>
10088 David Sherwood <david.sherwood@arm.com>
10090 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
10091 (add_offset_to_base): New function, split out from...
10092 (create_mem_ref): ...here. When handling a scale other than 1,
10093 check first whether the address is valid without the offset.
10094 Add it into the base if so, leaving the index and scale as-is.
10096 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10099 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
10100 fold_for_warn before checking if arg2 is INTEGER_CST.
10102 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
10104 * config/rs6000/predicates.md (load_multiple_operation): Delete.
10105 (store_multiple_operation): Delete.
10106 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
10107 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
10108 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
10109 guarded by TARGET_STRING.
10110 (rs6000_output_load_multiple): Delete.
10111 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
10112 OPTION_MASK_STRING / TARGET_STRING handling.
10113 (print_operand) <'N', 'O'>: Add comment that these are unused now.
10114 (const rs6000_opt_masks) <"string">: Change mask to 0.
10115 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
10116 (MASK_STRING): Delete.
10117 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
10119 (load_multiple): Delete.
10126 (store_multiple): Delete.
10133 (movmemsi_8reg): Delete.
10134 (corresponding unnamed define_insn): Delete.
10135 (movmemsi_6reg): Delete.
10136 (corresponding unnamed define_insn): Delete.
10137 (movmemsi_4reg): Delete.
10138 (corresponding unnamed define_insn): Delete.
10139 (movmemsi_2reg): Delete.
10140 (corresponding unnamed define_insn): Delete.
10141 (movmemsi_1reg): Delete.
10142 (corresponding unnamed define_insn): Delete.
10143 * config/rs6000/rs6000.opt (mno-string): New.
10144 (mstring): Replace by deprecation warning stub.
10145 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
10147 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10149 * regrename.c (regrename_do_replace): If replacing the same
10150 reg multiple times, try to reuse last created gen_raw_REG.
10153 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
10154 main to workaround a bug in GDB.
10156 2018-01-12 Tom de Vries <tom@codesourcery.com>
10159 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
10161 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
10163 PR rtl-optimization/80481
10164 * ira-color.c (get_cap_member): New function.
10165 (allocnos_conflict_by_live_ranges_p): Use it.
10166 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
10167 (setup_slot_coalesced_allocno_live_ranges): Ditto.
10169 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
10172 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
10173 (*saddl_se_1): Ditto.
10174 (*ssubsi_1): Ditto.
10175 (*ssubl_se_1): Ditto.
10177 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10179 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
10180 rather than wi::to_widest for DR_INITs.
10181 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
10182 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
10183 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
10185 (vect_analyze_group_access_1): Note that here.
10187 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10189 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
10190 polynomial type sizes.
10192 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10194 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
10195 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
10196 (gimple_add_tmp_var): Likewise.
10198 2018-01-12 Martin Liska <mliska@suse.cz>
10200 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
10201 (gimple_alloc_sizes): Likewise.
10202 (dump_gimple_statistics): Use PRIu64 in printf format.
10203 * gimple.h: Change uint64_t to int.
10205 2018-01-12 Martin Liska <mliska@suse.cz>
10207 * tree-core.h: Use uint64_t instead of int.
10208 * tree.c (tree_node_counts): Likewise.
10209 (tree_node_sizes): Likewise.
10210 (dump_tree_statistics): Use PRIu64 in printf format.
10212 2018-01-12 Martin Liska <mliska@suse.cz>
10214 * Makefile.in: As qsort_chk is implemented in vec.c, add
10215 vec.o to linkage of gencfn-macros.
10216 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
10217 passing the info to record_node_allocation_statistics.
10218 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
10220 * ggc-common.c (struct ggc_usage): Add operator== and use
10221 it in operator< and compare function.
10222 * mem-stats.h (struct mem_usage): Likewise.
10223 * vec.c (struct vec_usage): Remove operator< and compare
10224 function. Can be simply inherited.
10226 2018-01-12 Martin Jambor <mjambor@suse.cz>
10229 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
10230 * tree-ssa-math-opts.c: Include domwalk.h.
10231 (convert_mult_to_fma_1): New function.
10232 (fma_transformation_info): New type.
10233 (fma_deferring_state): Likewise.
10234 (cancel_fma_deferring): New function.
10235 (result_of_phi): Likewise.
10236 (last_fma_candidate_feeds_initial_phi): Likewise.
10237 (convert_mult_to_fma): Added deferring logic, split actual
10238 transformation to convert_mult_to_fma_1.
10239 (math_opts_dom_walker): New type.
10240 (math_opts_dom_walker::after_dom_children): New method, body moved
10241 here from pass_optimize_widening_mul::execute, added deferring logic
10243 (pass_optimize_widening_mul::execute): Moved most of code to
10244 math_opts_dom_walker::after_dom_children.
10245 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10246 * config/i386/i386.c (ix86_option_override_internal): Added
10247 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10249 2018-01-12 Richard Biener <rguenther@suse.de>
10252 * dwarf2out.c (gen_variable_die): Do not reset old_die for
10253 inline instance vars.
10255 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
10258 * config/rx/rx.c (rx_is_restricted_memory_address):
10259 Handle SUBREG case.
10261 2018-01-12 Richard Biener <rguenther@suse.de>
10263 PR tree-optimization/80846
10264 * target.def (split_reduction): New target hook.
10265 * targhooks.c (default_split_reduction): New function.
10266 * targhooks.h (default_split_reduction): Declare.
10267 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10268 target requests first reduce vectors by combining low and high
10270 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10271 (get_vectype_for_scalar_type_and_size): Export.
10272 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10273 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10274 * doc/tm.texi: Regenerate.
10275 * config/i386/i386.c (ix86_split_reduction): Implement
10276 TARGET_VECTORIZE_SPLIT_REDUCTION.
10278 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10281 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10282 in PIC mode except for TARGET_VXWORKS_RTP.
10283 * config/sparc/sparc.c: Include cfgrtl.h.
10284 (TARGET_INIT_PIC_REG): Define.
10285 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10286 (sparc_pic_register_p): New predicate.
10287 (sparc_legitimate_address_p): Use it.
10288 (sparc_legitimize_pic_address): Likewise.
10289 (sparc_delegitimize_address): Likewise.
10290 (sparc_mode_dependent_address_p): Likewise.
10291 (gen_load_pcrel_sym): Remove 4th parameter.
10292 (load_got_register): Adjust call to above. Remove obsolete stuff.
10293 (sparc_expand_prologue): Do not call load_got_register here.
10294 (sparc_flat_expand_prologue): Likewise.
10295 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10296 (sparc_use_pseudo_pic_reg): New function.
10297 (sparc_init_pic_reg): Likewise.
10298 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10299 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10301 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
10303 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10304 Add item for branch_cost.
10306 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10308 PR rtl-optimization/83565
10309 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10310 not extend the result to a larger mode for rotate operations.
10311 (num_sign_bit_copies1): Likewise.
10313 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10316 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10318 Use values-Xc.o for -pedantic.
10319 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10321 2018-01-12 Martin Liska <mliska@suse.cz>
10324 * ipa-devirt.c (final_warning_record::grow_type_warnings):
10326 (possible_polymorphic_call_targets): Use it.
10327 (ipa_devirt): Likewise.
10329 2018-01-12 Martin Liska <mliska@suse.cz>
10331 * profile-count.h (enum profile_quality): Use 0 as invalid
10332 enum value of profile_quality.
10334 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
10336 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10337 -mext-string options.
10339 2018-01-12 Richard Biener <rguenther@suse.de>
10341 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10342 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10343 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10345 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10347 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10349 * configure.ac (--with-long-double-format): Add support for the
10350 configuration option to change the default long double format on
10352 * config.gcc (powerpc*-linux*-*): Likewise.
10353 * configure: Regenerate.
10354 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10355 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10356 used without modification.
10358 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10360 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10361 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10362 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10363 MISC_BUILTIN_SPEC_BARRIER.
10364 (rs6000_init_builtins): Likewise.
10365 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10367 (speculation_barrier): New define_insn.
10368 * doc/extend.texi: Document __builtin_speculation_barrier.
10370 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10373 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10374 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10375 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10377 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10378 integral modes instead of "ss" and "sd".
10379 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10380 vectors with 32-bit and 64-bit elements.
10381 (vecdupssescalarmodesuffix): New mode attribute.
10382 (vec_dup<mode>): Use it.
10384 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
10387 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10388 frame if argument is passed on stack.
10390 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10393 * ree.c (combine_reaching_defs): Optimize also
10394 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10395 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10397 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10399 PR middle-end/83189
10400 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10402 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10404 PR middle-end/83718
10405 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10406 after they are computed.
10408 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10410 PR tree-optimization/83695
10411 * gimple-loop-linterchange.cc
10412 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10413 reset cached scev information after interchange.
10414 (pass_linterchange::execute): Remove call to scev_reset_htab.
10416 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10418 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10419 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10420 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10421 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10422 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10423 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10424 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10425 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10426 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10427 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10428 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10429 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10430 (V_lane_reg): Likewise.
10431 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10433 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10434 (vfmal_lane_low<mode>_intrinsic,
10435 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10436 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10437 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10438 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10439 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10440 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10442 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10444 * config/arm/arm-cpus.in (fp16fml): New feature.
10445 (ALL_SIMD): Add fp16fml.
10446 (armv8.2-a): Add fp16fml as an option.
10447 (armv8.3-a): Likewise.
10448 (armv8.4-a): Add fp16fml as part of fp16.
10449 * config/arm/arm.h (TARGET_FP16FML): Define.
10450 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10452 * config/arm/arm-modes.def (V2HF): Define.
10453 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10454 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10455 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10456 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10457 vfmsl_low, vfmsl_high): New set of builtins.
10458 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10459 (vfml_op): New code attribute.
10460 (VFMLHALVES): New int iterator.
10461 (VFML, VFMLSEL): New mode attributes.
10462 (V_reg): Define mapping for V2HF.
10463 (V_hi, V_lo): New mode attributes.
10464 (VF_constraint): Likewise.
10465 (vfml_half, vfml_half_selector): New int attributes.
10466 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10468 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10469 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10471 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10472 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10473 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10474 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
10476 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10477 Document new effective target and option set.
10479 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10481 * config/arm/arm-cpus.in (armv8_4): New feature.
10482 (ARMv8_4a): New fgroup.
10483 (armv8.4-a): New arch.
10484 * config/arm/arm-tables.opt: Regenerate.
10485 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10486 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
10487 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
10488 Add matching rules for -march=armv8.4-a and extensions.
10489 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
10491 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
10494 * config/rx/rx.md (BW): New mode attribute.
10495 (sync_lock_test_and_setsi): Add mode suffix to insn output.
10497 2018-01-11 Richard Biener <rguenther@suse.de>
10499 PR tree-optimization/83435
10500 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
10501 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
10502 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
10504 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10505 Alan Hayward <alan.hayward@arm.com>
10506 David Sherwood <david.sherwood@arm.com>
10508 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
10510 (aarch64_classify_address): Initialize it. Track polynomial offsets.
10511 (aarch64_print_address_internal): Use it to check for a zero offset.
10513 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10514 Alan Hayward <alan.hayward@arm.com>
10515 David Sherwood <david.sherwood@arm.com>
10517 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10518 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10519 Return a poly_int64 rather than a HOST_WIDE_INT.
10520 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10521 rather than a HOST_WIDE_INT.
10522 * config/aarch64/aarch64.h (aarch64_frame): Protect with
10523 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
10524 hard_fp_offset, frame_size, initial_adjust, callee_offset and
10525 final_offset from HOST_WIDE_INT to poly_int64.
10526 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
10527 to_constant when getting the number of units in an Advanced SIMD
10529 (aarch64_builtin_vectorized_function): Check for a constant number
10531 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
10533 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
10534 attribute instead of GET_MODE_NUNITS.
10535 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
10536 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
10537 GET_MODE_SIZE for fixed-size registers.
10538 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
10539 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
10540 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
10541 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
10542 (aarch64_print_operand, aarch64_print_address_internal)
10543 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
10544 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
10545 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
10546 Handle polynomial GET_MODE_SIZE.
10547 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
10548 wider than SImode without modification.
10549 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
10550 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
10551 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
10552 passing and returning SVE modes.
10553 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
10554 rather than GEN_INT.
10555 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
10556 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
10557 (aarch64_allocate_and_probe_stack_space): Likewise.
10558 (aarch64_layout_frame): Cope with polynomial offsets.
10559 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
10560 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
10561 polynomial offsets.
10562 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
10563 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
10564 poly_int64 rather than a HOST_WIDE_INT.
10565 (aarch64_get_separate_components, aarch64_process_components)
10566 (aarch64_expand_prologue, aarch64_expand_epilogue)
10567 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
10568 (aarch64_anchor_offset): New function, split out from...
10569 (aarch64_legitimize_address): ...here.
10570 (aarch64_builtin_vectorization_cost): Handle polynomial
10571 TYPE_VECTOR_SUBPARTS.
10572 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
10574 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
10575 number of elements from the PARALLEL rather than the mode.
10576 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
10577 rather than GET_MODE_BITSIZE.
10578 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
10579 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
10580 (aarch64_expand_vec_perm_const_1): Handle polynomial
10581 d->perm.length () and d->perm elements.
10582 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
10583 Apply to_constant to d->perm elements.
10584 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
10585 polynomial CONST_VECTOR_NUNITS.
10586 (aarch64_move_pointer): Take amount as a poly_int64 rather
10588 (aarch64_progress_pointer): Avoid temporary variable.
10589 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
10590 the mode attribute instead of GET_MODE.
10592 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10593 Alan Hayward <alan.hayward@arm.com>
10594 David Sherwood <david.sherwood@arm.com>
10596 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
10597 x exists before using it.
10598 (aarch64_add_constant_internal): Rename to...
10599 (aarch64_add_offset_1): ...this. Replace regnum with separate
10600 src and dest rtxes. Handle the case in which they're different,
10601 including when the offset is zero. Replace scratchreg with an rtx.
10602 Use 2 additions if there is no spare register into which we can
10603 move a 16-bit constant.
10604 (aarch64_add_constant): Delete.
10605 (aarch64_add_offset): Replace reg with separate src and dest
10606 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
10607 Use aarch64_add_offset_1.
10608 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
10609 an rtx rather than an int. Take the delta as a poly_int64
10610 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
10611 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
10612 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
10613 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
10614 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
10615 and aarch64_add_sp.
10616 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
10617 aarch64_add_constant.
10619 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10621 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
10622 Use scalar_float_mode.
10624 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10626 * config/aarch64/aarch64-simd.md
10627 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
10628 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
10629 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
10630 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
10631 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
10632 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
10633 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
10634 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
10635 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
10636 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
10638 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10641 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
10642 targ_options->x_arm_arch_string is non NULL.
10644 2018-01-11 Tamar Christina <tamar.christina@arm.com>
10646 * config/aarch64/aarch64.h
10647 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
10649 2018-01-11 Sudakshina Das <sudi.das@arm.com>
10652 * expmed.c (emit_store_flag_force): Swap if const op0
10653 and change VOIDmode to mode of op0.
10655 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10657 PR rtl-optimization/83761
10658 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
10659 than bytes to mode_for_size.
10661 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10663 PR middle-end/83189
10664 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
10665 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
10668 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10670 PR middle-end/83575
10671 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
10672 when in layout mode.
10673 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
10674 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
10677 2018-01-10 Michael Collison <michael.collison@arm.com>
10679 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
10680 * config/aarch64/aarch64-option-extension.def: Add
10681 AARCH64_OPT_EXTENSION of 'fp16fml'.
10682 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10683 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
10684 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
10685 * config/aarch64/constraints.md (Ui7): New constraint.
10686 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
10687 (VFMLA_SEL_W): Ditto.
10690 (VFMLA16_LOW): New int iterator.
10691 (VFMLA16_HIGH): Ditto.
10692 (UNSPEC_FMLAL): New unspec.
10693 (UNSPEC_FMLSL): Ditto.
10694 (UNSPEC_FMLAL2): Ditto.
10695 (UNSPEC_FMLSL2): Ditto.
10696 (f16mac): New code attribute.
10697 * config/aarch64/aarch64-simd-builtins.def
10698 (aarch64_fmlal_lowv2sf): Ditto.
10699 (aarch64_fmlsl_lowv2sf): Ditto.
10700 (aarch64_fmlalq_lowv4sf): Ditto.
10701 (aarch64_fmlslq_lowv4sf): Ditto.
10702 (aarch64_fmlal_highv2sf): Ditto.
10703 (aarch64_fmlsl_highv2sf): Ditto.
10704 (aarch64_fmlalq_highv4sf): Ditto.
10705 (aarch64_fmlslq_highv4sf): Ditto.
10706 (aarch64_fmlal_lane_lowv2sf): Ditto.
10707 (aarch64_fmlsl_lane_lowv2sf): Ditto.
10708 (aarch64_fmlal_laneq_lowv2sf): Ditto.
10709 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
10710 (aarch64_fmlalq_lane_lowv4sf): Ditto.
10711 (aarch64_fmlsl_lane_lowv4sf): Ditto.
10712 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
10713 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
10714 (aarch64_fmlal_lane_highv2sf): Ditto.
10715 (aarch64_fmlsl_lane_highv2sf): Ditto.
10716 (aarch64_fmlal_laneq_highv2sf): Ditto.
10717 (aarch64_fmlsl_laneq_highv2sf): Ditto.
10718 (aarch64_fmlalq_lane_highv4sf): Ditto.
10719 (aarch64_fmlsl_lane_highv4sf): Ditto.
10720 (aarch64_fmlalq_laneq_highv4sf): Ditto.
10721 (aarch64_fmlsl_laneq_highv4sf): Ditto.
10722 * config/aarch64/aarch64-simd.md:
10723 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
10724 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10725 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
10726 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10727 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
10728 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
10729 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
10730 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
10731 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
10732 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
10733 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
10734 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
10735 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
10736 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
10737 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
10738 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
10739 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
10740 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
10741 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
10742 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
10743 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
10744 (vfmlsl_low_u32): Ditto.
10745 (vfmlalq_low_u32): Ditto.
10746 (vfmlslq_low_u32): Ditto.
10747 (vfmlal_high_u32): Ditto.
10748 (vfmlsl_high_u32): Ditto.
10749 (vfmlalq_high_u32): Ditto.
10750 (vfmlslq_high_u32): Ditto.
10751 (vfmlal_lane_low_u32): Ditto.
10752 (vfmlsl_lane_low_u32): Ditto.
10753 (vfmlal_laneq_low_u32): Ditto.
10754 (vfmlsl_laneq_low_u32): Ditto.
10755 (vfmlalq_lane_low_u32): Ditto.
10756 (vfmlslq_lane_low_u32): Ditto.
10757 (vfmlalq_laneq_low_u32): Ditto.
10758 (vfmlslq_laneq_low_u32): Ditto.
10759 (vfmlal_lane_high_u32): Ditto.
10760 (vfmlsl_lane_high_u32): Ditto.
10761 (vfmlal_laneq_high_u32): Ditto.
10762 (vfmlsl_laneq_high_u32): Ditto.
10763 (vfmlalq_lane_high_u32): Ditto.
10764 (vfmlslq_lane_high_u32): Ditto.
10765 (vfmlalq_laneq_high_u32): Ditto.
10766 (vfmlslq_laneq_high_u32): Ditto.
10767 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
10768 (AARCH64_FL_FOR_ARCH8_4): New.
10769 (AARCH64_ISA_F16FML): New ISA flag.
10770 (TARGET_F16FML): New feature flag for fp16fml.
10771 (doc/invoke.texi): Document new fp16fml option.
10773 2018-01-10 Michael Collison <michael.collison@arm.com>
10775 * config/aarch64/aarch64-builtins.c:
10776 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
10777 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10778 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
10779 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
10780 (AARCH64_ISA_SHA3): New ISA flag.
10781 (TARGET_SHA3): New feature flag for sha3.
10782 * config/aarch64/iterators.md (sha512_op): New int attribute.
10783 (CRYPTO_SHA512): New int iterator.
10784 (UNSPEC_SHA512H): New unspec.
10785 (UNSPEC_SHA512H2): Ditto.
10786 (UNSPEC_SHA512SU0): Ditto.
10787 (UNSPEC_SHA512SU1): Ditto.
10788 * config/aarch64/aarch64-simd-builtins.def
10789 (aarch64_crypto_sha512hqv2di): New builtin.
10790 (aarch64_crypto_sha512h2qv2di): Ditto.
10791 (aarch64_crypto_sha512su0qv2di): Ditto.
10792 (aarch64_crypto_sha512su1qv2di): Ditto.
10793 (aarch64_eor3qv8hi): Ditto.
10794 (aarch64_rax1qv2di): Ditto.
10795 (aarch64_xarqv2di): Ditto.
10796 (aarch64_bcaxqv8hi): Ditto.
10797 * config/aarch64/aarch64-simd.md:
10798 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
10799 (aarch64_crypto_sha512su0qv2di): Ditto.
10800 (aarch64_crypto_sha512su1qv2di): Ditto.
10801 (aarch64_eor3qv8hi): Ditto.
10802 (aarch64_rax1qv2di): Ditto.
10803 (aarch64_xarqv2di): Ditto.
10804 (aarch64_bcaxqv8hi): Ditto.
10805 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
10806 (vsha512h2q_u64): Ditto.
10807 (vsha512su0q_u64): Ditto.
10808 (vsha512su1q_u64): Ditto.
10809 (veor3q_u16): Ditto.
10810 (vrax1q_u64): Ditto.
10811 (vxarq_u64): Ditto.
10812 (vbcaxq_u16): Ditto.
10813 * config/arm/types.md (crypto_sha512): New type attribute.
10814 (crypto_sha3): Ditto.
10815 (doc/invoke.texi): Document new sha3 option.
10817 2018-01-10 Michael Collison <michael.collison@arm.com>
10819 * config/aarch64/aarch64-builtins.c:
10820 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
10821 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10822 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
10823 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
10824 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
10825 (AARCH64_ISA_SM4): New ISA flag.
10826 (TARGET_SM4): New feature flag for sm4.
10827 * config/aarch64/aarch64-simd-builtins.def
10828 (aarch64_sm3ss1qv4si): Ditto.
10829 (aarch64_sm3tt1aq4si): Ditto.
10830 (aarch64_sm3tt1bq4si): Ditto.
10831 (aarch64_sm3tt2aq4si): Ditto.
10832 (aarch64_sm3tt2bq4si): Ditto.
10833 (aarch64_sm3partw1qv4si): Ditto.
10834 (aarch64_sm3partw2qv4si): Ditto.
10835 (aarch64_sm4eqv4si): Ditto.
10836 (aarch64_sm4ekeyqv4si): Ditto.
10837 * config/aarch64/aarch64-simd.md:
10838 (aarch64_sm3ss1qv4si): Ditto.
10839 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
10840 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
10841 (aarch64_sm4eqv4si): Ditto.
10842 (aarch64_sm4ekeyqv4si): Ditto.
10843 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
10844 (sm3part_op): Ditto.
10845 (CRYPTO_SM3TT): Ditto.
10846 (CRYPTO_SM3PART): Ditto.
10847 (UNSPEC_SM3SS1): New unspec.
10848 (UNSPEC_SM3TT1A): Ditto.
10849 (UNSPEC_SM3TT1B): Ditto.
10850 (UNSPEC_SM3TT2A): Ditto.
10851 (UNSPEC_SM3TT2B): Ditto.
10852 (UNSPEC_SM3PARTW1): Ditto.
10853 (UNSPEC_SM3PARTW2): Ditto.
10854 (UNSPEC_SM4E): Ditto.
10855 (UNSPEC_SM4EKEY): Ditto.
10856 * config/aarch64/constraints.md (Ui2): New constraint.
10857 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
10858 * config/arm/types.md (crypto_sm3): New type attribute.
10859 (crypto_sm4): Ditto.
10860 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
10861 (vsm3tt1aq_u32): Ditto.
10862 (vsm3tt1bq_u32): Ditto.
10863 (vsm3tt2aq_u32): Ditto.
10864 (vsm3tt2bq_u32): Ditto.
10865 (vsm3partw1q_u32): Ditto.
10866 (vsm3partw2q_u32): Ditto.
10867 (vsm4eq_u32): Ditto.
10868 (vsm4ekeyq_u32): Ditto.
10869 (doc/invoke.texi): Document new sm4 option.
10871 2018-01-10 Michael Collison <michael.collison@arm.com>
10873 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
10874 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
10875 (AARCH64_FL_FOR_ARCH8_4): New.
10876 (AARCH64_FL_V8_4): New flag.
10877 (doc/invoke.texi): Document new armv8.4-a option.
10879 2018-01-10 Michael Collison <michael.collison@arm.com>
10881 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10882 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
10883 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
10884 * config/aarch64/aarch64-option-extension.def: Add
10885 AARCH64_OPT_EXTENSION of 'sha2'.
10886 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
10887 (crypto): Disable sha2 and aes if crypto disabled.
10888 (crypto): Enable aes and sha2 if enabled.
10889 (simd): Disable sha2 and aes if simd disabled.
10890 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
10892 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
10893 (TARGET_SHA2): New feature flag for sha2.
10894 (TARGET_AES): New feature flag for aes.
10895 * config/aarch64/aarch64-simd.md:
10896 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
10897 conditional on TARGET_AES.
10898 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
10899 (aarch64_crypto_sha1hsi): Make pattern conditional
10901 (aarch64_crypto_sha1hv4si): Ditto.
10902 (aarch64_be_crypto_sha1hv4si): Ditto.
10903 (aarch64_crypto_sha1su1v4si): Ditto.
10904 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
10905 (aarch64_crypto_sha1su0v4si): Ditto.
10906 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
10907 (aarch64_crypto_sha256su0v4si): Ditto.
10908 (aarch64_crypto_sha256su1v4si): Ditto.
10909 (doc/invoke.texi): Document new aes and sha2 options.
10911 2018-01-10 Martin Sebor <msebor@redhat.com>
10913 PR tree-optimization/83781
10914 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
10917 2018-01-11 Martin Sebor <msebor@gmail.com>
10918 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10920 PR tree-optimization/83501
10921 PR tree-optimization/81703
10923 * tree-ssa-strlen.c (get_string_cst): Rename...
10924 (get_string_len): ...to this. Handle global constants.
10925 (handle_char_store): Adjust.
10927 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
10928 Jim Wilson <jimw@sifive.com>
10930 * config/riscv/riscv-protos.h (riscv_output_return): New.
10931 * config/riscv/riscv.c (struct machine_function): New naked_p field.
10932 (riscv_attribute_table, riscv_output_return),
10933 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
10934 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
10935 (riscv_compute_frame_info): Only compute frame->mask if not a naked
10937 (riscv_expand_prologue): Add early return for naked function.
10938 (riscv_expand_epilogue): Likewise.
10939 (riscv_function_ok_for_sibcall): Return false for naked function.
10940 (riscv_set_current_function): New.
10941 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
10942 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
10943 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
10944 * doc/extend.texi (RISC-V Function Attributes): New.
10946 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
10948 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
10949 check for 128-bit long double before checking TCmode.
10950 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
10951 128-bit long doubles before checking TFmode or TCmode.
10952 (FLOAT128_IBM_P): Likewise.
10954 2018-01-10 Martin Sebor <msebor@redhat.com>
10956 PR tree-optimization/83671
10957 * builtins.c (c_strlen): Unconditionally return zero for the empty
10959 Use -Warray-bounds for warnings.
10960 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
10961 for non-constant array indices with COMPONENT_REF, arrays of
10962 arrays, and pointers to arrays.
10963 (gimple_fold_builtin_strlen): Determine and set length range for
10964 non-constant character arrays.
10966 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
10968 PR middle-end/81897
10969 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
10972 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
10974 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
10976 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10979 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
10980 VECTOR_MEM_ALTIVEC_OR_VSX_P.
10981 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
10982 indexed_or_indirect_operand predicate.
10983 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
10984 (*vsx_le_perm_load_v8hi): Likewise.
10985 (*vsx_le_perm_load_v16qi): Likewise.
10986 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
10987 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
10988 (*vsx_le_perm_store_v8hi): Likewise.
10989 (*vsx_le_perm_store_v16qi): Likewise.
10990 (eight unnamed splitters): Likewise.
10992 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10994 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
10995 * config/rs6000/emmintrin.h: Likewise.
10996 * config/rs6000/mmintrin.h: Likewise.
10997 * config/rs6000/xmmintrin.h: Likewise.
10999 2018-01-10 David Malcolm <dmalcolm@redhat.com>
11002 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
11004 * tree.c (tree_nop_conversion): Return true for location wrapper
11006 (maybe_wrap_with_location): New function.
11007 (selftest::check_strip_nops): New function.
11008 (selftest::test_location_wrappers): New function.
11009 (selftest::tree_c_tests): Call it.
11010 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
11011 (maybe_wrap_with_location): New decl.
11012 (EXPR_LOCATION_WRAPPER_P): New macro.
11013 (location_wrapper_p): New inline function.
11014 (tree_strip_any_location_wrapper): New inline function.
11016 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
11019 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
11020 stack_realign_offset for the largest alignment of stack slot
11022 (ix86_find_max_used_stack_alignment): New function.
11023 (ix86_finalize_stack_frame_flags): Use it. Set
11024 max_used_stack_alignment if we don't realign stack.
11025 * config/i386/i386.h (machine_function): Add
11026 max_used_stack_alignment.
11028 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
11030 * config/arm/arm.opt (-mbranch-cost): New option.
11031 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
11034 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
11037 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
11038 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
11040 2018-01-10 Richard Biener <rguenther@suse.de>
11043 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
11044 early out so it also covers the case where we have a non-NULL
11047 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11049 PR tree-optimization/83753
11050 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
11051 for non-strided grouped accesses if the number of elements is 1.
11053 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11056 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
11057 * i386.h (TARGET_USE_GATHER): Define.
11058 * x86-tune.def (X86_TUNE_USE_GATHER): New.
11060 2018-01-10 Martin Liska <mliska@suse.cz>
11063 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
11064 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
11066 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
11067 CLEANUP_NO_PARTITIONING is not set.
11069 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11071 * doc/rtl.texi: Remove documentation of (const ...) wrappers
11072 for vectors, as a partial revert of r254296.
11073 * rtl.h (const_vec_p): Delete.
11074 (const_vec_duplicate_p): Don't test for vector CONSTs.
11075 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
11076 * expmed.c (make_tree): Likewise.
11079 * common.md (E, F): Use CONSTANT_P instead of checking for
11081 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
11082 checking for CONST_VECTOR.
11084 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11086 PR middle-end/83575
11087 * predict.c (force_edge_cold): Handle in more sane way edges
11088 with no prediction.
11090 2018-01-09 Carl Love <cel@us.ibm.com>
11092 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
11094 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
11095 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
11096 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
11097 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
11098 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
11099 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
11100 * config/rs6000/rs6000-protos.h: Add extern defition for
11101 rs6000_generate_float2_double_code.
11102 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
11104 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
11105 (float2_v2df): Add define_expand.
11107 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
11110 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
11111 op_mode in the force_to_mode call.
11113 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11115 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
11116 instead of checking each element individually.
11117 (aarch64_evpc_uzp): Likewise.
11118 (aarch64_evpc_zip): Likewise.
11119 (aarch64_evpc_ext): Likewise.
11120 (aarch64_evpc_rev): Likewise.
11121 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
11122 instead of checking each element individually. Return true without
11124 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
11125 whether all selected elements come from the same input, instead of
11126 checking each element individually. Remove calls to gen_rtx_REG,
11127 start_sequence and end_sequence and instead assert that no rtl is
11130 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11132 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
11133 order of HIGH and CONST checks.
11135 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11137 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
11138 if the destination isn't an SSA_NAME.
11140 2018-01-09 Richard Biener <rguenther@suse.de>
11142 PR tree-optimization/83668
11143 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
11145 (canonicalize_loop_form): ... here, renamed from ...
11146 (canonicalize_loop_closed_ssa_form): ... this and amended to
11147 swap successor edges for loop exit blocks to make us use
11148 the RPO order we need for initial schedule generation.
11150 2018-01-09 Joseph Myers <joseph@codesourcery.com>
11152 PR tree-optimization/64811
11153 * match.pd: When optimizing comparisons with Inf, avoid
11154 introducing or losing exceptions from comparisons with NaN.
11156 2018-01-09 Martin Liska <mliska@suse.cz>
11159 * asan.c (shadow_mem_size): Add gcc_assert.
11161 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
11163 Don't save registers in main().
11166 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
11167 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
11168 * config/avr/avr.c (avr_set_current_function): Don't error if
11169 naked, OS_task or OS_main are specified at the same time.
11170 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
11172 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
11174 * common/config/avr/avr-common.c (avr_option_optimization_table):
11175 Switch on -mmain-is-OS_task for optimizing compilations.
11177 2018-01-09 Richard Biener <rguenther@suse.de>
11179 PR tree-optimization/83572
11180 * graphite.c: Include cfganal.h.
11181 (graphite_transform_loops): Connect infinite loops to exit
11182 and remove fake edges at the end.
11184 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11186 * ipa-inline.c (edge_badness): Revert accidental checkin.
11188 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11191 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
11192 symbols; not inline clones.
11194 2018-01-09 Jakub Jelinek <jakub@redhat.com>
11197 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
11198 hard registers. Formatting fixes.
11200 PR preprocessor/83722
11201 * gcc.c (try_generate_repro): Pass
11202 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
11203 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
11206 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
11207 Kito Cheng <kito.cheng@gmail.com>
11209 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
11210 (riscv_leaf_function_p): Delete.
11211 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
11213 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11215 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
11217 (do_ifelse): New function.
11218 (do_isel): New function.
11219 (do_sub3): New function.
11220 (do_add3): New function.
11221 (do_load_mask_compare): New function.
11222 (do_overlap_load_compare): New function.
11223 (expand_compare_loop): New function.
11224 (expand_block_compare): Call expand_compare_loop() when appropriate.
11225 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
11226 option description.
11227 (-mblock-compare-inline-loop-limit): New option.
11229 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11232 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
11233 Reverse order of second and third operands in first alternative.
11234 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
11235 of first and second elements in UNSPEC_VPERMR vector.
11236 (altivec_expand_vec_perm_le): Likewise.
11238 2018-01-08 Jeff Law <law@redhat.com>
11240 PR rtl-optimizatin/81308
11241 * tree-switch-conversion.c (cfg_altered): New file scoped static.
11242 (process_switch): If group_case_labels makes a change, then set
11244 (pass_convert_switch::execute): If a switch is converted, then
11245 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
11247 PR rtl-optimization/81308
11248 * recog.c (split_all_insns): Conditionally cleanup the CFG after
11251 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
11253 PR target/83663 - Revert r255946
11254 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11255 generation for cases where splatting a value is not useful.
11256 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11257 across a vec_duplicate and a paradoxical subreg forming a vector
11258 mode to a vec_concat.
11260 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11262 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11263 -march=armv8.3-a variants.
11264 * config/arm/t-multilib: Likewise.
11265 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11267 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11269 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11271 (cceq_ior_compare_complement): Give it a name so I can use it, and
11272 change boolean_or_operator predicate to boolean_operator so it can
11273 be used to generate a crand.
11274 (eqne): New code iterator.
11275 (bd/bd_neg): New code_attrs.
11276 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11277 a single define_insn.
11278 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11279 decrement (bdnzt/bdnzf/bdzt/bdzf).
11280 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11281 with the new names of the branch decrement patterns, and added the
11282 names of the branch decrement conditional patterns.
11284 2018-01-08 Richard Biener <rguenther@suse.de>
11286 PR tree-optimization/83563
11287 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11290 2018-01-08 Richard Biener <rguenther@suse.de>
11292 PR middle-end/83713
11293 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11295 2018-01-08 Richard Biener <rguenther@suse.de>
11297 PR tree-optimization/83685
11298 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11299 references to abnormals.
11301 2018-01-08 Richard Biener <rguenther@suse.de>
11304 * dwarf2out.c (output_indirect_strings): Handle empty
11305 skeleton_debug_str_hash.
11306 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11308 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11310 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11311 (emit_store_direct): Likewise.
11312 (arc_trampoline_adjust_address): Likewise.
11313 (arc_asm_trampoline_template): New function.
11314 (arc_initialize_trampoline): Use asm_trampoline_template.
11315 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11316 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11317 * config/arc/arc.md (flush_icache): Delete pattern.
11319 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11321 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11322 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11325 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11328 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11329 by not USED_FOR_TARGET.
11330 (make_pass_resolve_sw_modes): Likewise.
11332 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11334 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11337 2018-01-08 Richard Biener <rguenther@suse.de>
11339 PR middle-end/83580
11340 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11342 2018-01-08 Richard Biener <rguenther@suse.de>
11344 PR middle-end/83517
11345 * match.pd ((t * 2) / 2) -> t): Add missing :c.
11347 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
11349 PR middle-end/81897
11350 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11351 basic blocks with a small number of successors.
11352 (convert_control_dep_chain_into_preds): Improve handling of
11354 (dump_predicates): Split apart into...
11355 (dump_pred_chain): ...here...
11356 (dump_pred_info): ...and here.
11357 (can_one_predicate_be_invalidated_p): Add debugging printfs.
11358 (can_chain_union_be_invalidated_p): Improve check for invalidation
11360 (uninit_uses_cannot_happen): Avoid unnecessary if
11361 convert_control_dep_chain_into_preds yielded nothing.
11363 2018-01-06 Martin Sebor <msebor@redhat.com>
11365 PR tree-optimization/83640
11366 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11367 subtracting negative offset from size.
11368 (builtin_access::overlap): Adjust offset bounds of the access to fall
11369 within the size of the object if possible.
11371 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
11373 PR rtl-optimization/83699
11374 * expmed.c (extract_bit_field_1): Restrict the vector usage of
11375 extract_bit_field_as_subreg to cases in which the extracted
11376 value is also a vector.
11378 * lra-constraints.c (process_alt_operands): Test for the equivalence
11379 substitutions when detecting a possible reload cycle.
11381 2018-01-06 Jakub Jelinek <jakub@redhat.com>
11384 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11385 by default if flag_selective_schedling{,2}. Formatting fixes.
11387 PR rtl-optimization/83682
11388 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11389 if it has non-VECTOR_MODE element mode.
11390 (vec_duplicate_p): Likewise.
11392 PR middle-end/83694
11393 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11394 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11396 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11399 * config/i386/i386-builtin.def
11400 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11401 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11402 Require also OPTION_MASK_ISA_AVX512F in addition to
11403 OPTION_MASK_ISA_GFNI.
11404 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11405 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11406 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11407 to OPTION_MASK_ISA_GFNI.
11408 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11409 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11410 OPTION_MASK_ISA_AVX512BW.
11411 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11412 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11413 addition to OPTION_MASK_ISA_GFNI.
11414 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11415 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11416 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11417 to OPTION_MASK_ISA_GFNI.
11418 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11419 a requirement for all ISAs rather than any of them with a few
11421 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11423 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11424 bitmasks to be enabled with 3 exceptions, instead of requiring any
11425 enabled ISA with lots of exceptions.
11426 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11427 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11428 Change avx512bw in isa attribute to avx512f.
11429 * config/i386/sgxintrin.h: Add license boilerplate.
11430 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11431 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11432 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11433 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11435 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11436 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11437 temporarily sse2 rather than sse if not enabled already.
11440 * config/i386/sse.md (VI248_VLBW): Rename to ...
11441 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11442 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11443 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11444 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11445 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11446 mode iterator instead of VI248_VLBW.
11448 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11450 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11451 (record_modified): Skip clobbers; add debug output.
11452 (param_change_prob): Use sreal frequencies.
11454 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11456 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11457 punt for user-aligned variables.
11459 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11461 * tree-chrec.c (chrec_contains_symbols): Return true for
11464 2018-01-05 Sudakshina Das <sudi.das@arm.com>
11467 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11468 of (x|y) == x for BICS pattern.
11470 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11472 PR tree-optimization/83605
11473 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11474 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11477 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
11479 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11480 * config/epiphany/rtems.h: New file.
11482 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11483 Uros Bizjak <ubizjak@gmail.com>
11486 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
11487 QIreg_operand instead of register_operand predicate.
11488 * config/i386/i386.c (ix86_rop_should_change_byte_p,
11489 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
11490 comments instead of -fmitigate[-_]rop.
11492 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11495 * cgraphunit.c (symbol_table::compile): Switch to text_section
11496 before calling assembly_start debug hook.
11497 * run-rtl-passes.c (run_rtl_passes): Likewise.
11500 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11502 * tree-vrp.c (extract_range_from_binary_expr_1): Check
11503 range_int_cst_p rather than !symbolic_range_p before calling
11504 extract_range_from_multiplicative_op_1.
11506 2018-01-04 Jeff Law <law@redhat.com>
11508 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11509 redundant test in assertion.
11511 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11513 * doc/rtl.texi: Document machine_mode wrapper classes.
11515 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11517 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11518 using tree_to_uhwi.
11520 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11522 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11523 the VEC_PERM_EXPR fold to fail.
11525 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11528 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
11529 to switched_sections.
11531 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11534 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
11535 test for d.testing.
11537 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
11540 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
11541 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
11543 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11546 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
11547 is BLKmode and bitpos not zero or mode change is needed.
11549 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11552 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
11555 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
11558 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
11559 instead of MULT rtx. Update all corresponding splitters.
11560 (*saddl_se): Ditto.
11561 (*ssub<modesuffix>): Ditto.
11562 (*ssubl_se): Ditto.
11563 (*cmp_sadd_di): Update split patterns.
11564 (*cmp_sadd_si): Ditto.
11565 (*cmp_sadd_sidi): Ditto.
11566 (*cmp_ssub_di): Ditto.
11567 (*cmp_ssub_si): Ditto.
11568 (*cmp_ssub_sidi): Ditto.
11569 * config/alpha/predicates.md (const23_operand): New predicate.
11570 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
11571 Look for ASHIFT, not MULT inner operand.
11572 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
11574 2018-01-04 Martin Liska <mliska@suse.cz>
11576 PR gcov-profile/83669
11577 * gcov.c (output_intermediate_file): Add version to intermediate
11579 * doc/gcov.texi: Document new field 'version' in intermediate
11580 file format. Fix location of '-k' option of gcov command.
11582 2018-01-04 Martin Liska <mliska@suse.cz>
11585 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
11587 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11589 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
11591 2018-01-03 Martin Sebor <msebor@redhat.com>
11593 PR tree-optimization/83655
11594 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
11595 checking calls with invalid arguments.
11597 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11599 * tree-vect-stmts.c (vect_get_store_rhs): New function.
11600 (vectorizable_mask_load_store): Delete.
11601 (vectorizable_call): Return false for masked loads and stores.
11602 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
11603 instead of gimple_assign_rhs1.
11604 (vectorizable_load): Handle IFN_MASK_LOAD.
11605 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
11607 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11609 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
11611 (vectorizable_mask_load_store): ...here.
11612 (vectorizable_load): ...and here.
11614 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11616 * tree-vect-stmts.c (vect_build_all_ones_mask)
11617 (vect_build_zero_merge_argument): New functions, split out from...
11618 (vectorizable_load): ...here.
11620 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11622 * tree-vect-stmts.c (vect_check_store_rhs): New function,
11624 (vectorizable_mask_load_store): ...here.
11625 (vectorizable_store): ...and here.
11627 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11629 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
11631 (vectorizable_mask_load_store): ...here.
11633 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11635 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
11636 (vect_model_store_cost): Take a vec_load_store_type instead of a
11638 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
11639 (vect_model_store_cost): Take a vec_load_store_type instead of a
11641 (vectorizable_mask_load_store): Update accordingly.
11642 (vectorizable_store): Likewise.
11643 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
11645 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11647 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
11648 IFN_MASK_LOAD calls here rather than...
11649 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
11651 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11652 Alan Hayward <alan.hayward@arm.com>
11653 David Sherwood <david.sherwood@arm.com>
11655 * expmed.c (extract_bit_field_1): For vector extracts,
11656 fall back to extract_bit_field_as_subreg if vec_extract
11659 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11660 Alan Hayward <alan.hayward@arm.com>
11661 David Sherwood <david.sherwood@arm.com>
11663 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
11664 they are variable or constant sized.
11665 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
11666 slots for constant-sized data.
11668 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11669 Alan Hayward <alan.hayward@arm.com>
11670 David Sherwood <david.sherwood@arm.com>
11672 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
11673 handling COND_EXPRs with boolean comparisons, try to find a better
11674 basis for the mask type than the boolean itself.
11676 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11678 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
11679 is calculated and how it can be overridden.
11680 * genmodes.c (max_bitsize_mode_any_mode): New variable.
11681 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
11683 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
11686 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11687 Alan Hayward <alan.hayward@arm.com>
11688 David Sherwood <david.sherwood@arm.com>
11690 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
11691 Remove the mode argument.
11692 (aarch64_simd_valid_immediate): Remove the mode and inverse
11694 * config/aarch64/iterators.md (bitsize): New iterator.
11695 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
11696 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
11697 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
11698 aarch64_simd_valid_immediate.
11699 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
11700 (aarch64_reg_or_bic_imm): Likewise.
11701 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
11702 with an insn_type enum and msl with a modifier_type enum.
11703 Replace element_width with a scalar_mode. Change the shift
11704 to unsigned int. Add constructors for scalar_float_mode and
11705 scalar_int_mode elements.
11706 (aarch64_vect_float_const_representable_p): Delete.
11707 (aarch64_can_const_movi_rtx_p)
11708 (aarch64_simd_scalar_immediate_valid_for_move)
11709 (aarch64_simd_make_constant): Update call to
11710 aarch64_simd_valid_immediate.
11711 (aarch64_advsimd_valid_immediate_hs): New function.
11712 (aarch64_advsimd_valid_immediate): Likewise.
11713 (aarch64_simd_valid_immediate): Remove mode and inverse
11714 arguments. Rewrite to use the above. Use const_vec_duplicate_p
11715 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
11716 and aarch64_float_const_representable_p on the result.
11717 (aarch64_output_simd_mov_immediate): Remove mode argument.
11718 Update call to aarch64_simd_valid_immediate and use of
11719 simd_immediate_info.
11720 (aarch64_output_scalar_simd_mov_immediate): Update call
11723 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11724 Alan Hayward <alan.hayward@arm.com>
11725 David Sherwood <david.sherwood@arm.com>
11727 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
11728 (mode_nunits): Likewise CONST_MODE_NUNITS.
11729 * machmode.def (ADJUST_NUNITS): Document.
11730 * genmodes.c (mode_data::need_nunits_adj): New field.
11731 (blank_mode): Update accordingly.
11732 (adj_nunits): New variable.
11733 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
11735 (emit_mode_size_inline): Set need_bytesize_adj for all modes
11736 listed in adj_nunits.
11737 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
11738 listed in adj_nunits. Don't emit case statements for such modes.
11739 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
11740 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
11741 nothing if adj_nunits is nonnull.
11742 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
11743 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
11744 (emit_mode_fbit): Update use of print_maybe_const_decl.
11745 (emit_move_size): Likewise. Treat the array as non-const
11747 (emit_mode_adjustments): Handle adj_nunits.
11749 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11751 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
11752 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
11753 (VECTOR_MODES): Use it.
11754 (make_vector_modes): Take the prefix as an argument.
11756 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11757 Alan Hayward <alan.hayward@arm.com>
11758 David Sherwood <david.sherwood@arm.com>
11760 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
11761 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
11762 for MODE_VECTOR_BOOL.
11763 * machmode.def (VECTOR_BOOL_MODE): Document.
11764 * genmodes.c (VECTOR_BOOL_MODE): New macro.
11765 (make_vector_bool_mode): New function.
11766 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
11768 * lto-streamer-in.c (lto_input_mode_table): Likewise.
11769 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
11771 * stor-layout.c (int_mode_for_mode): Likewise.
11772 * tree.c (build_vector_type_for_mode): Likewise.
11773 * varasm.c (output_constant_pool_2): Likewise.
11774 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
11775 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
11776 for MODE_VECTOR_BOOL.
11777 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
11778 of mode class checks.
11779 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
11780 instead of a list of mode class checks.
11781 (expand_vector_scalar_condition): Likewise.
11782 (type_for_widest_vector_mode): Handle BImode as an inner mode.
11784 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11785 Alan Hayward <alan.hayward@arm.com>
11786 David Sherwood <david.sherwood@arm.com>
11788 * machmode.h (mode_size): Change from unsigned short to
11790 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
11791 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11792 or if measurement_type is not polynomial.
11793 (fixed_size_mode::includes_p): Check for constant-sized modes.
11794 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
11795 return a poly_uint16 rather than an unsigned short.
11796 (emit_mode_size): Change the type of mode_size from unsigned short
11797 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
11798 (emit_mode_adjustments): Cope with polynomial vector sizes.
11799 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11801 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11803 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
11804 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
11805 * caller-save.c (setup_save_areas): Likewise.
11806 (replace_reg_with_saved_mem): Likewise.
11807 * calls.c (emit_library_call_value_1): Likewise.
11808 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
11809 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
11810 (gen_lowpart_for_combine): Likewise.
11811 * convert.c (convert_to_integer_1): Likewise.
11812 * cse.c (equiv_constant, cse_insn): Likewise.
11813 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
11814 (cselib_subst_to_values): Likewise.
11815 * dce.c (word_dce_process_block): Likewise.
11816 * df-problems.c (df_word_lr_mark_ref): Likewise.
11817 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
11818 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
11819 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
11820 (rtl_for_decl_location): Likewise.
11821 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
11822 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
11823 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
11824 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
11825 (expand_expr_real_1): Likewise.
11826 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
11827 (pad_below): Likewise.
11828 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11829 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
11830 * ira.c (get_subreg_tracking_sizes): Likewise.
11831 * ira-build.c (ira_create_allocno_objects): Likewise.
11832 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
11833 (ira_sort_regnos_for_alter_reg): Likewise.
11834 * ira-costs.c (record_operand_costs): Likewise.
11835 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
11836 (resolve_simple_move): Likewise.
11837 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
11838 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
11839 (lra_constraints): Likewise.
11840 (CONST_POOL_OK_P): Reject variable-sized modes.
11841 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
11842 (add_pseudo_to_slot, lra_spill): Likewise.
11843 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11844 * optabs-query.c (get_best_extraction_insn): Likewise.
11845 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11846 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
11847 (expand_mult_highpart, valid_multiword_target_p): Likewise.
11848 * recog.c (offsettable_address_addr_space_p): Likewise.
11849 * regcprop.c (maybe_mode_change): Likewise.
11850 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
11851 * regrename.c (build_def_use): Likewise.
11852 * regstat.c (dump_reg_info): Likewise.
11853 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
11854 (find_reloads, find_reloads_subreg_address): Likewise.
11855 * reload1.c (eliminate_regs_1): Likewise.
11856 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
11857 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
11858 (simplify_binary_operation_1, simplify_subreg): Likewise.
11859 * targhooks.c (default_function_arg_padding): Likewise.
11860 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
11861 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
11862 (verify_gimple_assign_ternary): Likewise.
11863 * tree-inline.c (estimate_move_cost): Likewise.
11864 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11865 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
11866 (get_address_cost_ainc): Likewise.
11867 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
11868 (vect_supportable_dr_alignment): Likewise.
11869 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11870 (vectorizable_reduction): Likewise.
11871 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
11872 (vectorizable_operation, vectorizable_load): Likewise.
11873 * tree.c (build_same_sized_truth_vector_type): Likewise.
11874 * valtrack.c (cleanup_auto_inc_dec): Likewise.
11875 * var-tracking.c (emit_note_insn_var_location): Likewise.
11876 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
11877 (ADDR_VEC_ALIGN): Likewise.
11879 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11880 Alan Hayward <alan.hayward@arm.com>
11881 David Sherwood <david.sherwood@arm.com>
11883 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
11885 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11886 or if measurement_type is polynomial.
11887 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
11888 * combine.c (make_extraction): Likewise.
11889 * dse.c (find_shift_sequence): Likewise.
11890 * dwarf2out.c (mem_loc_descriptor): Likewise.
11891 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
11892 (extract_bit_field, extract_low_bits): Likewise.
11893 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
11894 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
11895 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
11896 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
11897 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11898 * reload.c (find_reloads): Likewise.
11899 * reload1.c (alter_reg): Likewise.
11900 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
11901 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
11902 * tree-if-conv.c (predicate_mem_writes): Likewise.
11903 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
11904 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
11905 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
11906 * valtrack.c (dead_debug_insert_temp): Likewise.
11907 * varasm.c (mergeable_constant_section): Likewise.
11908 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
11910 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11911 Alan Hayward <alan.hayward@arm.com>
11912 David Sherwood <david.sherwood@arm.com>
11914 * expr.c (expand_assignment): Cope with polynomial mode sizes
11915 when assigning to a CONCAT.
11917 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11918 Alan Hayward <alan.hayward@arm.com>
11919 David Sherwood <david.sherwood@arm.com>
11921 * machmode.h (mode_precision): Change from unsigned short to
11923 (mode_to_precision): Return a poly_uint16 rather than an unsigned
11925 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
11926 or if measurement_type is not polynomial.
11927 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
11928 in which the mode is already known to be a scalar_int_mode.
11929 * genmodes.c (emit_mode_precision): Change the type of mode_precision
11930 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
11932 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11933 for GET_MODE_PRECISION.
11934 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11935 for GET_MODE_PRECISION.
11936 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
11938 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
11939 (expand_field_assignment, make_extraction): Likewise.
11940 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
11941 (get_last_value): Likewise.
11942 * convert.c (convert_to_integer_1): Likewise.
11943 * cse.c (cse_insn): Likewise.
11944 * expr.c (expand_expr_real_1): Likewise.
11945 * lra-constraints.c (simplify_operand_subreg): Likewise.
11946 * optabs-query.c (can_atomic_load_p): Likewise.
11947 * optabs.c (expand_atomic_load): Likewise.
11948 (expand_atomic_store): Likewise.
11949 * ree.c (combine_reaching_defs): Likewise.
11950 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
11951 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
11952 * tree.h (type_has_mode_precision_p): Likewise.
11953 * ubsan.c (instrument_si_overflow): Likewise.
11955 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11956 Alan Hayward <alan.hayward@arm.com>
11957 David Sherwood <david.sherwood@arm.com>
11959 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
11960 polynomial numbers of units.
11961 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
11962 (valid_vector_subparts_p): New function.
11963 (build_vector_type): Remove temporary shim and take the number
11964 of units as a poly_uint64 rather than an int.
11965 (build_opaque_vector_type): Take the number of units as a
11966 poly_uint64 rather than an int.
11967 * tree.c (build_vector_from_ctor): Handle polynomial
11968 TYPE_VECTOR_SUBPARTS.
11969 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
11970 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
11971 (build_vector_from_val): If the number of units is variable,
11972 use build_vec_duplicate_cst for constant operands and
11973 VEC_DUPLICATE_EXPR otherwise.
11974 (make_vector_type): Remove temporary is_constant ().
11975 (build_vector_type, build_opaque_vector_type): Take the number of
11976 units as a poly_uint64 rather than an int.
11977 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
11979 * cfgexpand.c (expand_debug_expr): Likewise.
11980 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
11981 (store_constructor, expand_expr_real_1): Likewise.
11982 (const_scalar_mask_from_tree): Likewise.
11983 * fold-const-call.c (fold_const_reduction): Likewise.
11984 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
11985 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
11986 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
11987 (fold_relational_const): Likewise.
11988 (native_interpret_vector): Likewise. Change the size from an
11989 int to an unsigned int.
11990 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
11991 TYPE_VECTOR_SUBPARTS.
11992 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
11993 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
11994 duplicating a non-constant operand into a variable-length vector.
11995 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
11996 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
11997 * ipa-icf.c (sem_variable::equals): Likewise.
11998 * match.pd: Likewise.
11999 * omp-simd-clone.c (simd_clone_subparts): Likewise.
12000 * print-tree.c (print_node): Likewise.
12001 * stor-layout.c (layout_type): Likewise.
12002 * targhooks.c (default_builtin_vectorization_cost): Likewise.
12003 * tree-cfg.c (verify_gimple_comparison): Likewise.
12004 (verify_gimple_assign_binary): Likewise.
12005 (verify_gimple_assign_ternary): Likewise.
12006 (verify_gimple_assign_single): Likewise.
12007 * tree-pretty-print.c (dump_generic_node): Likewise.
12008 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12009 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
12010 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
12011 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
12012 (vect_shift_permute_load_chain): Likewise.
12013 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
12014 (expand_vector_condition, optimize_vector_constructor): Likewise.
12015 (lower_vec_perm, get_compute_type): Likewise.
12016 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12017 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
12018 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
12019 (vect_recog_mask_conversion_pattern): Likewise.
12020 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
12021 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
12022 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12023 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
12024 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
12025 (vectorizable_shift, vectorizable_operation, vectorizable_store)
12026 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
12027 (supportable_widening_operation): Likewise.
12028 (supportable_narrowing_operation): Likewise.
12029 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
12031 * varasm.c (output_constant): Likewise.
12033 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12034 Alan Hayward <alan.hayward@arm.com>
12035 David Sherwood <david.sherwood@arm.com>
12037 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
12038 so that both the length == 3 and length != 3 cases set up their
12039 own permute vectors. Add comments explaining why we know the
12040 number of elements is constant.
12041 (vect_permute_load_chain): Likewise.
12043 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12044 Alan Hayward <alan.hayward@arm.com>
12045 David Sherwood <david.sherwood@arm.com>
12047 * machmode.h (mode_nunits): Change from unsigned char to
12049 (ONLY_FIXED_SIZE_MODES): New macro.
12050 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
12051 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
12052 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
12054 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
12055 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
12056 or if measurement_type is not polynomial.
12057 * genmodes.c (ZERO_COEFFS): New macro.
12058 (emit_mode_nunits_inline): Make mode_nunits_inline return a
12060 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
12061 Use ZERO_COEFFS when emitting initializers.
12062 * data-streamer.h (bp_pack_poly_value): New function.
12063 (bp_unpack_poly_value): Likewise.
12064 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12065 for GET_MODE_NUNITS.
12066 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12067 for GET_MODE_NUNITS.
12068 * tree.c (make_vector_type): Remove temporary shim and make
12069 the real function take the number of units as a poly_uint64
12070 rather than an int.
12071 (build_vector_type_for_mode): Handle polynomial nunits.
12072 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
12073 * emit-rtl.c (const_vec_series_p_1): Likewise.
12074 (gen_rtx_CONST_VECTOR): Likewise.
12075 * fold-const.c (test_vec_duplicate_folding): Likewise.
12076 * genrecog.c (validate_pattern): Likewise.
12077 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
12078 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12079 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
12080 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
12081 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
12082 * rtlanal.c (subreg_get_info): Likewise.
12083 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12084 (vect_grouped_load_supported): Likewise.
12085 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
12086 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
12087 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12088 (simplify_const_unary_operation, simplify_binary_operation_1)
12089 (simplify_const_binary_operation, simplify_ternary_operation)
12090 (test_vector_ops_duplicate, test_vector_ops): Likewise.
12091 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
12092 instead of CONST_VECTOR_NUNITS.
12093 * varasm.c (output_constant_pool_2): Likewise.
12094 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
12095 explicit-encoded elements in the XVEC for variable-length vectors.
12097 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12099 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
12101 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12102 Alan Hayward <alan.hayward@arm.com>
12103 David Sherwood <david.sherwood@arm.com>
12105 * coretypes.h (fixed_size_mode): Declare.
12106 (fixed_size_mode_pod): New typedef.
12107 * builtins.h (target_builtins::x_apply_args_mode)
12108 (target_builtins::x_apply_result_mode): Change type to
12109 fixed_size_mode_pod.
12110 * builtins.c (apply_args_size, apply_result_size, result_vector)
12111 (expand_builtin_apply_args_1, expand_builtin_apply)
12112 (expand_builtin_return): Update accordingly.
12114 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12116 * cse.c (hash_rtx_cb): Hash only the encoded elements.
12117 * cselib.c (cselib_hash_rtx): Likewise.
12118 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
12119 CONST_VECTOR encoding.
12121 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12122 Jeff Law <law@redhat.com>
12125 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
12126 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
12127 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
12128 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
12131 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
12132 explicitly probe *sp in a noreturn function if there were any callee
12133 register saves or frame pointer is needed.
12135 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12138 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
12139 BLKmode for ternary, binary or unary expressions.
12142 * var-tracking.c (delete_vta_debug_insn): New inline function.
12143 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
12144 insns from get_insns () to NULL instead of each bb separately.
12145 Use delete_vta_debug_insn. No longer static.
12146 (vt_debug_insns_local, variable_tracking_main_1): Adjust
12147 delete_vta_debug_insns callers.
12148 * rtl.h (delete_vta_debug_insns): Declare.
12149 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
12150 instead of variable_tracking_main.
12152 2018-01-03 Martin Sebor <msebor@redhat.com>
12154 PR tree-optimization/83603
12155 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
12156 arguments past the endof the argument list in functions declared
12157 without a prototype.
12158 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
12159 Avoid checking when arguments are null.
12161 2018-01-03 Martin Sebor <msebor@redhat.com>
12164 * doc/extend.texi (attribute const): Fix a typo.
12165 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
12166 issuing -Wsuggest-attribute for void functions.
12168 2018-01-03 Martin Sebor <msebor@redhat.com>
12170 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
12171 offset_int::from instead of wide_int::to_shwi.
12172 (maybe_diag_overlap): Remove assertion.
12173 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
12174 * gimple-ssa-sprintf.c (format_directive): Same.
12175 (parse_directive): Same.
12176 (sprintf_dom_walker::compute_format_length): Same.
12177 (try_substitute_return_value): Same.
12179 2018-01-03 Jeff Law <law@redhat.com>
12181 PR middle-end/83654
12182 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
12183 non-constant residual for zero at runtime and avoid probing in
12184 that case. Reorganize code for trailing problem to mirror handling
12187 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12189 PR tree-optimization/83501
12190 * tree-ssa-strlen.c (get_string_cst): New.
12191 (handle_char_store): Call get_string_cst.
12193 2018-01-03 Martin Liska <mliska@suse.cz>
12195 PR tree-optimization/83593
12196 * tree-ssa-strlen.c: Include tree-cfg.h.
12197 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
12198 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
12199 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
12201 (strlen_dom_walker::before_dom_children): Call
12202 gimple_purge_dead_eh_edges. Dump tranformation with details
12204 (strlen_dom_walker::before_dom_children): Update call by adding
12205 new argument cleanup_eh.
12206 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
12208 2018-01-03 Martin Liska <mliska@suse.cz>
12211 * cif-code.def (VARIADIC_THUNK): New enum value.
12212 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
12215 2018-01-03 Jan Beulich <jbeulich@suse.com>
12217 * sse.md (mov<mode>_internal): Tighten condition for when to use
12218 vmovdqu<ssescalarsize> for TI and OI modes.
12220 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12222 Update copyright years.
12224 2018-01-03 Martin Liska <mliska@suse.cz>
12227 * ipa-visibility.c (function_and_variable_visibility): Skip
12228 functions with noipa attribure.
12230 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12232 * gcc.c (process_command): Update copyright notice dates.
12233 * gcov-dump.c (print_version): Ditto.
12234 * gcov.c (print_version): Ditto.
12235 * gcov-tool.c (print_version): Ditto.
12236 * gengtype.c (create_file): Ditto.
12237 * doc/cpp.texi: Bump @copying's copyright year.
12238 * doc/cppinternals.texi: Ditto.
12239 * doc/gcc.texi: Ditto.
12240 * doc/gccint.texi: Ditto.
12241 * doc/gcov.texi: Ditto.
12242 * doc/install.texi: Ditto.
12243 * doc/invoke.texi: Ditto.
12245 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12247 * vector-builder.h (vector_builder::m_full_nelts): Change from
12248 unsigned int to poly_uint64.
12249 (vector_builder::full_nelts): Update prototype accordingly.
12250 (vector_builder::new_vector): Likewise.
12251 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12252 (vector_builder::operator ==): Likewise.
12253 (vector_builder::finalize): Likewise.
12254 * int-vector-builder.h (int_vector_builder::int_vector_builder):
12255 Take the number of elements as a poly_uint64 rather than an
12257 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12258 from unsigned int to poly_uint64.
12259 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12260 (vec_perm_indices::new_vector): Likewise.
12261 (vec_perm_indices::length): Likewise.
12262 (vec_perm_indices::nelts_per_input): Likewise.
12263 (vec_perm_indices::input_nelts): Likewise.
12264 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12265 number of elements per input as a poly_uint64 rather than an
12266 unsigned int. Use the original encoding for variable-length
12267 vectors, rather than clamping each individual element.
12268 For the second and subsequent elements in each pattern,
12269 clamp the step and base before clamping their sum.
12270 (vec_perm_indices::series_p): Handle polynomial element counts.
12271 (vec_perm_indices::all_in_range_p): Likewise.
12272 (vec_perm_indices_to_tree): Likewise.
12273 (vec_perm_indices_to_rtx): Likewise.
12274 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12275 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12276 (tree_vector_builder::new_binary_operation): Handle polynomial
12277 element counts. Return false if we need to know the number
12278 of elements at compile time.
12279 * fold-const.c (fold_vec_perm): Punt if the number of elements
12280 isn't known at compile time.
12282 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12284 * vec-perm-indices.h (vec_perm_builder): Change element type
12285 from HOST_WIDE_INT to poly_int64.
12286 (vec_perm_indices::element_type): Update accordingly.
12287 (vec_perm_indices::clamp): Handle polynomial element_types.
12288 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12289 (vec_perm_indices::all_in_range_p): Likewise.
12290 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12292 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12293 polynomial vec_perm_indices element types.
12294 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12295 * fold-const.c (fold_vec_perm): Likewise.
12296 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12297 * tree-vect-generic.c (lower_vec_perm): Likewise.
12298 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12299 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12300 element type to HOST_WIDE_INT.
12302 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12303 Alan Hayward <alan.hayward@arm.com>
12304 David Sherwood <david.sherwood@arm.com>
12306 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12307 rather than an int. Use plus_constant.
12308 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12309 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12311 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12312 Alan Hayward <alan.hayward@arm.com>
12313 David Sherwood <david.sherwood@arm.com>
12315 * calls.c (emit_call_1, expand_call): Change struct_value_size from
12316 a HOST_WIDE_INT to a poly_int64.
12318 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12319 Alan Hayward <alan.hayward@arm.com>
12320 David Sherwood <david.sherwood@arm.com>
12322 * calls.c (load_register_parameters): Cope with polynomial
12323 mode sizes. Require a constant size for BLKmode parameters
12324 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
12325 forces a parameter to be padded at the lsb end in order to
12326 fill a complete number of words, require the parameter size
12327 to be ordered wrt UNITS_PER_WORD.
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 * reload1.c (spill_stack_slot_width): Change element type
12334 from unsigned int to poly_uint64_pod.
12335 (alter_reg): Treat mode sizes as polynomial.
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 * reload.c (complex_word_subreg_p): New function.
12342 (reload_inner_reg_of_subreg, push_reload): Use it.
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 * lra-constraints.c (process_alt_operands): Reject matched
12349 operands whose sizes aren't ordered.
12350 (match_reload): Refer to this check here.
12352 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12353 Alan Hayward <alan.hayward@arm.com>
12354 David Sherwood <david.sherwood@arm.com>
12356 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12357 that the mode size is in the set {1, 2, 4, 8, 16}.
12359 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12360 Alan Hayward <alan.hayward@arm.com>
12361 David Sherwood <david.sherwood@arm.com>
12363 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12364 Use plus_constant instead of gen_rtx_PLUS.
12366 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12367 Alan Hayward <alan.hayward@arm.com>
12368 David Sherwood <david.sherwood@arm.com>
12370 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12371 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12372 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12373 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12374 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12375 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12376 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12377 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12378 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12379 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12381 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12382 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12383 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12384 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12385 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12386 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12387 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12388 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12389 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12390 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12392 * expr.c (emit_move_resolve_push): Treat the input and result
12393 of PUSH_ROUNDING as a poly_int64.
12394 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12395 (emit_push_insn): Likewise.
12396 * lra-eliminations.c (mark_not_eliminable): Likewise.
12397 * recog.c (push_operand): Likewise.
12398 * reload1.c (elimination_effects): Likewise.
12399 * rtlanal.c (nonzero_bits1): Likewise.
12400 * calls.c (store_one_arg): Likewise. Require the padding to be
12401 known at compile time.
12403 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12404 Alan Hayward <alan.hayward@arm.com>
12405 David Sherwood <david.sherwood@arm.com>
12407 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12408 Use plus_constant instead of gen_rtx_PLUS.
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 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12415 rather than an int.
12417 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12418 Alan Hayward <alan.hayward@arm.com>
12419 David Sherwood <david.sherwood@arm.com>
12421 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12422 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12423 via stack temporaries. Treat the mode size as polynomial too.
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 * expr.c (expand_expr_real_2): When handling conversions involving
12430 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12431 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12432 as a poly_uint64 too.
12434 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12435 Alan Hayward <alan.hayward@arm.com>
12436 David Sherwood <david.sherwood@arm.com>
12438 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
12440 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12441 Alan Hayward <alan.hayward@arm.com>
12442 David Sherwood <david.sherwood@arm.com>
12444 * combine.c (can_change_dest_mode): Handle polynomial
12445 REGMODE_NATURAL_SIZE.
12446 * expmed.c (store_bit_field_1): Likewise.
12447 * expr.c (store_constructor): Likewise.
12448 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12449 and polynomial REGMODE_NATURAL_SIZE.
12450 (gen_lowpart_common): Likewise.
12451 * reginfo.c (record_subregs_of_mode): Likewise.
12452 * rtlanal.c (read_modify_subreg_p): Likewise.
12454 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12455 Alan Hayward <alan.hayward@arm.com>
12456 David Sherwood <david.sherwood@arm.com>
12458 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12459 numbers of elements.
12461 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12462 Alan Hayward <alan.hayward@arm.com>
12463 David Sherwood <david.sherwood@arm.com>
12465 * match.pd: Cope with polynomial numbers of vector elements.
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 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12472 in a POINTER_PLUS_EXPR.
12474 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12475 Alan Hayward <alan.hayward@arm.com>
12476 David Sherwood <david.sherwood@arm.com>
12478 * omp-simd-clone.c (simd_clone_subparts): New function.
12479 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12480 (ipa_simd_modify_function_body): Likewise.
12482 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12483 Alan Hayward <alan.hayward@arm.com>
12484 David Sherwood <david.sherwood@arm.com>
12486 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
12487 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
12488 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
12489 (expand_vector_condition, vector_element): Likewise.
12490 (subparts_gt): New function.
12491 (get_compute_type): Use subparts_gt.
12492 (count_type_subparts): Delete.
12493 (expand_vector_operations_1): Use subparts_gt instead of
12494 count_type_subparts.
12496 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12497 Alan Hayward <alan.hayward@arm.com>
12498 David Sherwood <david.sherwood@arm.com>
12500 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
12501 (vect_compile_time_alias): ...this new function. Do the calculation
12502 on poly_ints rather than trees.
12503 (vect_prune_runtime_alias_test_list): Update call accordingly.
12505 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12506 Alan Hayward <alan.hayward@arm.com>
12507 David Sherwood <david.sherwood@arm.com>
12509 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
12511 (vect_schedule_slp_instance): Likewise.
12513 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12514 Alan Hayward <alan.hayward@arm.com>
12515 David Sherwood <david.sherwood@arm.com>
12517 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12518 constant and extern definitions for variable-length vectors.
12519 (vect_get_constant_vectors): Note that the number of units
12520 is known to be constant.
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_conversion): Treat the number
12527 of units as polynomial. Choose between WIDE and NARROW based
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 (simd_clone_subparts): New function.
12535 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
12537 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12538 Alan Hayward <alan.hayward@arm.com>
12539 David Sherwood <david.sherwood@arm.com>
12541 * tree-vect-stmts.c (vectorizable_call): Treat the number of
12542 vectors as polynomial. Use build_index_vector for
12543 IFN_GOMP_SIMD_LANE.
12545 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12546 Alan Hayward <alan.hayward@arm.com>
12547 David Sherwood <david.sherwood@arm.com>
12549 * tree-vect-stmts.c (get_load_store_type): Treat the number of
12550 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
12551 for variable-length vectors.
12552 (vectorizable_mask_load_store): Treat the number of units as
12553 polynomial, asserting that it is constant if the condition has
12554 already been enforced.
12555 (vectorizable_store, vectorizable_load): Likewise.
12557 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12558 Alan Hayward <alan.hayward@arm.com>
12559 David Sherwood <david.sherwood@arm.com>
12561 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
12562 of units as polynomial. Punt if we can't tell at compile time
12563 which vector contains the final result.
12565 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12566 Alan Hayward <alan.hayward@arm.com>
12567 David Sherwood <david.sherwood@arm.com>
12569 * tree-vect-loop.c (vectorizable_induction): Treat the number
12570 of units as polynomial. Punt on SLP inductions. Use an integer
12571 VEC_SERIES_EXPR for variable-length integer reductions. Use a
12572 cast of such a series for variable-length floating-point
12575 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12576 Alan Hayward <alan.hayward@arm.com>
12577 David Sherwood <david.sherwood@arm.com>
12579 * tree.h (build_index_vector): Declare.
12580 * tree.c (build_index_vector): New function.
12581 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
12582 of units as polynomial, forcibly converting it to a constant if
12583 vectorizable_reduction has already enforced the condition.
12584 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
12585 to create a {1,2,3,...} vector.
12586 (vectorizable_reduction): Treat the number of units as polynomial.
12587 Choose vectype_in based on the largest scalar element size rather
12588 than the smallest number of units. Enforce the restrictions
12591 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12592 Alan Hayward <alan.hayward@arm.com>
12593 David Sherwood <david.sherwood@arm.com>
12595 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
12596 number of units as polynomial.
12598 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12599 Alan Hayward <alan.hayward@arm.com>
12600 David Sherwood <david.sherwood@arm.com>
12602 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
12603 * target.def (autovectorize_vector_sizes): Return the vector sizes
12604 by pointer, using vector_sizes rather than a bitmask.
12605 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
12606 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
12607 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
12609 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
12610 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
12611 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
12612 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
12613 * omp-general.c (omp_max_vf): Likewise.
12614 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12615 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
12616 * tree-vect-loop.c (vect_analyze_loop): Likewise.
12617 * tree-vect-slp.c (vect_slp_bb): Likewise.
12618 * doc/tm.texi: Regenerate.
12619 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
12621 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
12622 the vector size as a poly_uint64 rather than an unsigned int.
12623 (current_vector_size): Change from an unsigned int to a poly_uint64.
12624 (get_vectype_for_scalar_type): Update accordingly.
12625 * tree.h (build_truth_vector_type): Take the size and number of
12626 units as a poly_uint64 rather than an unsigned int.
12627 (build_vector_type): Add a temporary overload that takes
12628 the number of units as a poly_uint64 rather than an unsigned int.
12629 * tree.c (make_vector_type): Likewise.
12630 (build_truth_vector_type): Take the number of units as a poly_uint64
12631 rather than an unsigned int.
12633 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12634 Alan Hayward <alan.hayward@arm.com>
12635 David Sherwood <david.sherwood@arm.com>
12637 * target.def (get_mask_mode): Take the number of units and length
12638 as poly_uint64s rather than unsigned ints.
12639 * targhooks.h (default_get_mask_mode): Update accordingly.
12640 * targhooks.c (default_get_mask_mode): Likewise.
12641 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
12642 * doc/tm.texi: Regenerate.
12644 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12645 Alan Hayward <alan.hayward@arm.com>
12646 David Sherwood <david.sherwood@arm.com>
12648 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
12649 * omp-general.c (omp_max_vf): Likewise.
12650 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
12651 (expand_omp_simd): Handle polynomial safelen.
12652 * omp-low.c (omplow_simd_context): Add a default constructor.
12653 (omplow_simd_context::max_vf): Change from int to poly_uint64.
12654 (lower_rec_simd_input_clauses): Update accordingly.
12655 (lower_rec_input_clauses): Likewise.
12657 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12658 Alan Hayward <alan.hayward@arm.com>
12659 David Sherwood <david.sherwood@arm.com>
12661 * tree-vectorizer.h (vect_nunits_for_cost): New function.
12662 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
12663 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
12664 (vect_analyze_slp_cost): Likewise.
12665 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
12666 (vect_model_load_cost): Likewise.
12668 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12669 Alan Hayward <alan.hayward@arm.com>
12670 David Sherwood <david.sherwood@arm.com>
12672 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
12673 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
12674 from an unsigned int * to a poly_uint64_pod *.
12675 (calculate_unrolling_factor): New function.
12676 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
12678 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12679 Alan Hayward <alan.hayward@arm.com>
12680 David Sherwood <david.sherwood@arm.com>
12682 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
12683 from an unsigned int to a poly_uint64.
12684 (_loop_vec_info::slp_unrolling_factor): Likewise.
12685 (_loop_vec_info::vectorization_factor): Change from an int
12687 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
12688 (vect_get_num_vectors): New function.
12689 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
12690 (vect_get_num_copies): Use vect_get_num_vectors.
12691 (vect_analyze_data_ref_dependences): Change max_vf from an int *
12692 to an unsigned int *.
12693 (vect_analyze_data_refs): Change min_vf from an int * to a
12695 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12696 than an unsigned HOST_WIDE_INT.
12697 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
12698 (vect_analyze_data_ref_dependence): Change max_vf from an int *
12699 to an unsigned int *.
12700 (vect_analyze_data_ref_dependences): Likewise.
12701 (vect_compute_data_ref_alignment): Handle polynomial vf.
12702 (vect_enhance_data_refs_alignment): Likewise.
12703 (vect_prune_runtime_alias_test_list): Likewise.
12704 (vect_shift_permute_load_chain): Likewise.
12705 (vect_supportable_dr_alignment): Likewise.
12706 (dependence_distance_ge_vf): Take the vectorization factor as a
12707 poly_uint64 rather than an unsigned HOST_WIDE_INT.
12708 (vect_analyze_data_refs): Change min_vf from an int * to a
12710 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
12711 vfm1 as a poly_uint64 rather than an int. Make the same change
12712 for the returned bound_scalar.
12713 (vect_gen_vector_loop_niters): Handle polynomial vf.
12714 (vect_do_peeling): Likewise. Update call to
12715 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
12716 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
12718 * tree-vect-loop.c (vect_determine_vectorization_factor)
12719 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
12720 (vect_get_known_peeling_cost): Likewise.
12721 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
12722 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
12723 (vect_transform_loop): Likewise. Use the lowest possible VF when
12724 updating the upper bounds of the loop.
12725 (vect_min_worthwhile_factor): Make static. Return an unsigned int
12726 rather than an int.
12727 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
12728 polynomial unroll factors.
12729 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
12730 (vect_make_slp_decision): Likewise.
12731 (vect_supported_load_permutation_p): Likewise, and polynomial
12733 (vect_analyze_slp_cost): Handle polynomial vf.
12734 (vect_slp_analyze_node_operations): Likewise.
12735 (vect_slp_analyze_bb_1): Likewise.
12736 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12737 than an unsigned HOST_WIDE_INT.
12738 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
12739 (vectorizable_load): Handle polynomial vf.
12740 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
12742 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
12744 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12745 Alan Hayward <alan.hayward@arm.com>
12746 David Sherwood <david.sherwood@arm.com>
12748 * match.pd: Handle bit operations involving three constants
12749 and try to fold one pair.
12751 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12753 * tree-vect-loop-manip.c: Include gimple-fold.h.
12754 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
12755 niters_maybe_zero parameters. Handle other cases besides a step of 1.
12756 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
12757 Add a path that uses a step of VF instead of 1, but disable it
12759 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
12760 and niters_no_overflow parameters. Update calls to
12761 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
12762 Create a new SSA name if the latter choses to use a ste other
12763 than zero, and return it via niters_vector_mult_vf_var.
12764 * tree-vect-loop.c (vect_transform_loop): Update calls to
12765 vect_do_peeling, vect_gen_vector_loop_niters and
12766 slpeel_make_loop_iterate_ntimes.
12767 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
12768 (vect_gen_vector_loop_niters): Update declarations after above changes.
12770 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
12772 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
12773 128-bit round to integer instructions.
12774 (ceil<mode>2): Likewise.
12775 (btrunc<mode>2): Likewise.
12776 (round<mode>2): Likewise.
12778 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12780 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
12781 unaligned VSX load/store on P8/P9.
12782 (expand_block_clear): Allow the use of unaligned VSX
12783 load/store on P8/P9.
12785 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12787 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
12789 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
12790 swap associated with both a load and a store.
12792 2018-01-02 Andrew Waterman <andrew@sifive.com>
12794 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
12795 * config/riscv/riscv.md (clear_cache): Use it.
12797 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
12799 * web.c: Remove out-of-date comment.
12801 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12803 * expr.c (fixup_args_size_notes): Check that any existing
12804 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
12805 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
12806 (emit_single_push_insn): ...here.
12808 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12810 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
12811 (const_vector_encoded_nelts): New function.
12812 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
12813 (const_vector_int_elt, const_vector_elt): Declare.
12814 * emit-rtl.c (const_vector_int_elt_1): New function.
12815 (const_vector_elt): Likewise.
12816 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
12817 of CONST_VECTOR_ELT.
12819 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12821 * expr.c: Include rtx-vector-builder.h.
12822 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
12823 directly on the tree encoding.
12824 (const_vector_from_tree): Likewise.
12825 * optabs.c: Include rtx-vector-builder.h.
12826 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
12827 sequence of "u" values.
12828 * vec-perm-indices.c: Include rtx-vector-builder.h.
12829 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
12830 directly on the vec_perm_indices encoding.
12832 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12834 * doc/rtl.texi (const_vector): Describe new encoding scheme.
12835 * Makefile.in (OBJS): Add rtx-vector-builder.o.
12836 * rtx-vector-builder.h: New file.
12837 * rtx-vector-builder.c: Likewise.
12838 * rtl.h (rtx_def::u2): Add a const_vector field.
12839 (CONST_VECTOR_NPATTERNS): New macro.
12840 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
12841 (CONST_VECTOR_DUPLICATE_P): Likewise.
12842 (CONST_VECTOR_STEPPED_P): Likewise.
12843 (CONST_VECTOR_ENCODED_ELT): Likewise.
12844 (const_vec_duplicate_p): Check for a duplicated vector encoding.
12845 (unwrap_const_vec_duplicate): Likewise.
12846 (const_vec_series_p): Check for a non-duplicated vector encoding.
12847 Say that the function only returns true for integer vectors.
12848 * emit-rtl.c: Include rtx-vector-builder.h.
12849 (gen_const_vec_duplicate_1): Delete.
12850 (gen_const_vector): Call gen_const_vec_duplicate instead of
12851 gen_const_vec_duplicate_1.
12852 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
12853 (gen_const_vec_duplicate): Use rtx_vector_builder.
12854 (gen_const_vec_series): Likewise.
12855 (gen_rtx_CONST_VECTOR): Likewise.
12856 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
12857 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12858 Build a new vector rather than modifying a CONST_VECTOR in-place.
12859 (handle_special_swappables): Update call accordingly.
12860 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
12861 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12862 Build a new vector rather than modifying a CONST_VECTOR in-place.
12863 (handle_special_swappables): Update call accordingly.
12865 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12867 * simplify-rtx.c (simplify_const_binary_operation): Use
12868 CONST_VECTOR_ELT instead of XVECEXP.
12870 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12872 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
12873 the selector elements to be different from the data elements
12874 if the selector is a VECTOR_CST.
12875 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
12876 ssizetype for the selector.
12878 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12880 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
12881 before testing each element individually.
12882 * tree-vect-generic.c (lower_vec_perm): Likewise.
12884 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12886 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
12887 * selftest-run-tests.c (selftest::run_tests): Call it.
12888 * vector-builder.h (vector_builder::operator ==): New function.
12889 (vector_builder::operator !=): Likewise.
12890 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
12891 (vec_perm_indices::all_from_input_p): New function.
12892 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12893 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
12894 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
12895 instead of reading the VECTOR_CST directly. Detect whether both
12896 vector inputs are the same before constructing the vec_perm_indices,
12897 and update the number of inputs argument accordingly. Use the
12898 utility functions added above. Only construct sel2 if we need to.
12900 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12902 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
12903 the broadcast of the low byte.
12904 (expand_mult_highpart): Use an explicit encoding for the permutes.
12905 * optabs-query.c (can_mult_highpart_p): Likewise.
12906 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
12907 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12908 (vectorizable_bswap): Likewise.
12909 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
12910 explicit encoding for the power-of-2 permutes.
12911 (vect_permute_store_chain): Likewise.
12912 (vect_grouped_load_supported): Likewise.
12913 (vect_permute_load_chain): Likewise.
12915 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12917 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
12918 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
12919 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
12920 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12921 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
12922 (vect_gen_perm_mask_any): Likewise.
12924 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12926 * int-vector-builder.h: New file.
12927 * vec-perm-indices.h: Include int-vector-builder.h.
12928 (vec_perm_indices): Redefine as an int_vector_builder.
12929 (auto_vec_perm_indices): Delete.
12930 (vec_perm_builder): Redefine as a stand-alone class.
12931 (vec_perm_indices::vec_perm_indices): New function.
12932 (vec_perm_indices::clamp): Likewise.
12933 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
12934 (vec_perm_indices::new_vector): New function.
12935 (vec_perm_indices::new_expanded_vector): Update for new
12936 vec_perm_indices class.
12937 (vec_perm_indices::rotate_inputs): New function.
12938 (vec_perm_indices::all_in_range_p): Operate directly on the
12939 encoded form, without computing elided elements.
12940 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
12941 encoding. Update for new vec_perm_indices class.
12942 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
12943 the given vec_perm_builder.
12944 (expand_vec_perm_var): Update vec_perm_builder constructor.
12945 (expand_mult_highpart): Use vec_perm_builder instead of
12946 auto_vec_perm_indices.
12947 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
12948 vec_perm_indices instead of auto_vec_perm_indices. Use a single
12949 or double series encoding as appropriate.
12950 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
12951 vec_perm_indices instead of auto_vec_perm_indices.
12952 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12953 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12954 (vect_permute_store_chain): Likewise.
12955 (vect_grouped_load_supported): Likewise.
12956 (vect_permute_load_chain): Likewise.
12957 (vect_shift_permute_load_chain): Likewise.
12958 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12959 (vect_transform_slp_perm_load): Likewise.
12960 (vect_schedule_slp_instance): Likewise.
12961 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12962 (vectorizable_mask_load_store): Likewise.
12963 (vectorizable_bswap): Likewise.
12964 (vectorizable_store): Likewise.
12965 (vectorizable_load): Likewise.
12966 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
12967 vec_perm_indices instead of auto_vec_perm_indices. Use
12968 tree_to_vec_perm_builder to read the vector from a tree.
12969 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
12970 vec_perm_builder instead of a vec_perm_indices.
12971 (have_whole_vector_shift): Use vec_perm_builder and
12972 vec_perm_indices instead of auto_vec_perm_indices. Leave the
12973 truncation to calc_vec_perm_mask_for_shift.
12974 (vect_create_epilog_for_reduction): Likewise.
12975 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
12976 from auto_vec_perm_indices to vec_perm_indices.
12977 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12978 instead of changing individual elements.
12979 (aarch64_vectorize_vec_perm_const): Use new_vector to install
12980 the vector in d.perm.
12981 * config/arm/arm.c (expand_vec_perm_d::perm): Change
12982 from auto_vec_perm_indices to vec_perm_indices.
12983 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12984 instead of changing individual elements.
12985 (arm_vectorize_vec_perm_const): Use new_vector to install
12986 the vector in d.perm.
12987 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
12988 Update vec_perm_builder constructor.
12989 (rs6000_expand_interleave): Likewise.
12990 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
12991 (rs6000_expand_interleave): Likewise.
12993 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12995 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
12996 to qimode could truncate the indices.
12997 * optabs.c (expand_vec_perm_var): Likewise.
12999 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13001 * Makefile.in (OBJS): Add vec-perm-indices.o.
13002 * vec-perm-indices.h: New file.
13003 * vec-perm-indices.c: Likewise.
13004 * target.h (vec_perm_indices): Replace with a forward class
13006 (auto_vec_perm_indices): Move to vec-perm-indices.h.
13007 * optabs.h: Include vec-perm-indices.h.
13008 (expand_vec_perm): Delete.
13009 (selector_fits_mode_p, expand_vec_perm_var): Declare.
13010 (expand_vec_perm_const): Declare.
13011 * target.def (vec_perm_const_ok): Replace with...
13012 (vec_perm_const): ...this new hook.
13013 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
13014 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
13015 * doc/tm.texi: Regenerate.
13016 * optabs.def (vec_perm_const): Delete.
13017 * doc/md.texi (vec_perm_const): Likewise.
13018 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
13019 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
13020 expand_vec_perm for constant permutation vectors. Assert that
13021 the mode of variable permutation vectors is the integer equivalent
13022 of the mode that is being permuted.
13023 * optabs-query.h (selector_fits_mode_p): Declare.
13024 * optabs-query.c: Include vec-perm-indices.h.
13025 (selector_fits_mode_p): New function.
13026 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
13027 is defined, instead of checking whether the vec_perm_const_optab
13028 exists. Use targetm.vectorize.vec_perm_const instead of
13029 targetm.vectorize.vec_perm_const_ok. Check whether the indices
13030 fit in the vector mode before using a variable permute.
13031 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
13032 vec_perm_indices instead of an rtx.
13033 (expand_vec_perm): Replace with...
13034 (expand_vec_perm_const): ...this new function. Take the selector
13035 as a vec_perm_indices rather than an rtx. Also take the mode of
13036 the selector. Update call to shift_amt_for_vec_perm_mask.
13037 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
13038 Use vec_perm_indices::new_expanded_vector to expand the original
13039 selector into bytes. Check whether the indices fit in the vector
13040 mode before using a variable permute.
13041 (expand_vec_perm_var): Make global.
13042 (expand_mult_highpart): Use expand_vec_perm_const.
13043 * fold-const.c: Includes vec-perm-indices.h.
13044 * tree-ssa-forwprop.c: Likewise.
13045 * tree-vect-data-refs.c: Likewise.
13046 * tree-vect-generic.c: Likewise.
13047 * tree-vect-loop.c: Likewise.
13048 * tree-vect-slp.c: Likewise.
13049 * tree-vect-stmts.c: Likewise.
13050 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
13052 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
13053 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
13054 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
13055 (aarch64_vectorize_vec_perm_const): ...this new function.
13056 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13057 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13058 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
13059 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
13060 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13061 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13062 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
13064 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
13065 check for NEON modes.
13066 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
13067 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
13068 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
13069 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
13071 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
13072 the old VEC_PERM_CONST conditions.
13073 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
13074 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
13075 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
13076 (ia64_vectorize_vec_perm_const_ok): Merge into...
13077 (ia64_vectorize_vec_perm_const): ...this new function.
13078 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
13079 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
13080 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
13081 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
13082 * config/mips/mips.c (mips_expand_vec_perm_const)
13083 (mips_vectorize_vec_perm_const_ok): Merge into...
13084 (mips_vectorize_vec_perm_const): ...this new function.
13085 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
13086 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
13087 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
13088 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
13089 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
13090 (rs6000_expand_vec_perm_const): Delete.
13091 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
13093 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13094 (altivec_expand_vec_perm_const_le): Take each operand individually.
13095 Operate on constant selectors rather than rtxes.
13096 (altivec_expand_vec_perm_const): Likewise. Update call to
13097 altivec_expand_vec_perm_const_le.
13098 (rs6000_expand_vec_perm_const): Delete.
13099 (rs6000_vectorize_vec_perm_const_ok): Delete.
13100 (rs6000_vectorize_vec_perm_const): New function.
13101 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13102 an element count and rtx array.
13103 (rs6000_expand_extract_even): Update call accordingly.
13104 (rs6000_expand_interleave): Likewise.
13105 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
13106 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
13107 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
13108 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
13109 (rs6000_expand_vec_perm_const): Delete.
13110 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13111 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13112 (altivec_expand_vec_perm_const_le): Take each operand individually.
13113 Operate on constant selectors rather than rtxes.
13114 (altivec_expand_vec_perm_const): Likewise. Update call to
13115 altivec_expand_vec_perm_const_le.
13116 (rs6000_expand_vec_perm_const): Delete.
13117 (rs6000_vectorize_vec_perm_const_ok): Delete.
13118 (rs6000_vectorize_vec_perm_const): New function. Remove stray
13119 reference to the SPE evmerge intructions.
13120 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13121 an element count and rtx array.
13122 (rs6000_expand_extract_even): Update call accordingly.
13123 (rs6000_expand_interleave): Likewise.
13124 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
13125 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
13127 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13129 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13131 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
13132 vector mode and that that mode matches the mode of the data
13134 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
13135 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
13136 directly using expand_vec_perm_1 when forcing selectors into
13138 (expand_vec_perm_var): New function, split out from expand_vec_perm.
13140 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13142 * optabs-query.h (can_vec_perm_p): Delete.
13143 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
13144 * optabs-query.c (can_vec_perm_p): Split into...
13145 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
13146 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
13147 particular selector is valid.
13148 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13149 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13150 (vect_grouped_load_supported): Likewise.
13151 (vect_shift_permute_load_chain): Likewise.
13152 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13153 (vect_transform_slp_perm_load): Likewise.
13154 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13155 (vectorizable_bswap): Likewise.
13156 (vect_gen_perm_mask_checked): Likewise.
13157 * fold-const.c (fold_ternary_loc): Likewise. Don't take
13158 implementations of variable permutation vectors into account
13159 when deciding which selector to use.
13160 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
13161 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
13162 with a false third argument.
13163 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
13164 to test whether the constant selector is valid and can_vec_perm_var_p
13165 to test whether a variable selector is valid.
13167 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13169 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
13170 * optabs-query.c (can_vec_perm_p): Likewise.
13171 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
13172 instead of vec_perm_indices.
13173 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
13174 (vect_gen_perm_mask_checked): Likewise,
13175 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
13176 (vect_gen_perm_mask_checked): Likewise,
13178 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13180 * optabs-query.h (qimode_for_vec_perm): Declare.
13181 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
13182 (qimode_for_vec_perm): ...this new function.
13183 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
13185 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13187 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
13188 does not have a conditional at the top.
13190 2018-01-02 Richard Biener <rguenther@suse.de>
13192 * ipa-inline.c (big_speedup_p): Fix expression.
13194 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13197 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
13200 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13204 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
13205 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
13206 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
13207 cond_taken_branch_cost 3->4.
13209 2018-01-01 Jakub Jelinek <jakub@redhat.com>
13211 PR tree-optimization/83581
13212 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
13213 TODO_cleanup_cfg if any changes have been made.
13215 PR middle-end/83608
13216 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
13217 convert_modes if target mode has the right side, but different mode
13220 PR middle-end/83609
13221 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
13222 last argument when extracting from CONCAT. If either from_real or
13223 from_imag is NULL, use expansion through memory. If result is not
13224 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
13225 the parts directly to inner mode, if even that fails, use expansion
13228 PR middle-end/83623
13229 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
13230 check for bswap in mode rather than HImode and use that in expand_unop
13233 Copyright (C) 2018 Free Software Foundation, Inc.
13235 Copying and distribution of this file, with or without modification,
13236 are permitted in any medium without royalty provided the copyright
13237 notice and this notice are preserved.