1 2018-05-18 Richard Biener <rguenther@suse.de>
3 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
4 results when processing array refs with variable index.
6 2018-05-18 Toon Moene <toon@moene.org>
8 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
9 directly after that of -floop-interchange. Indicate that both
10 options are enabled by default when specifying -O3.
12 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
15 iterator. Delete separate integer-mode vec_set<mode> expander.
16 (aarch64_simd_vec_setv2di): Delete.
17 (vec_setv2di): Delete.
18 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
19 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
20 the "w, r" alternative.
22 2018-05-18 Martin Liska <mliska@suse.cz>
24 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
25 * tree-pass.h (make_pass_lower_switch_O0): New function.
26 * tree-switch-conversion.c (node_has_low_bound): Remove.
27 (node_has_high_bound): Likewise.
28 (node_is_bounded): Likewise.
29 (class pass_lower_switch): Make it a template type and create
31 (pass_lower_switch::execute): Add template argument.
32 (make_pass_lower_switch): New function.
33 (make_pass_lower_switch_O0): New function.
34 (do_jump_if_equal): Remove.
35 (emit_case_nodes): Simplify to just handle all 3 cases and leave
36 all the hard work to tree optimization passes.
38 2018-05-18 Martin Liska <mliska@suse.cz>
40 * dbgcnt.c (limit_low): Renamed from limit.
41 (limit_high): New variable.
42 (dbg_cnt_is_enabled): Check for upper limit.
43 (dbg_cnt): Adjust dumping.
44 (dbg_cnt_set_limit_by_index): Add new argument for high
46 (dbg_cnt_set_limit_by_name): Likewise.
47 (dbg_cnt_process_single_pair): Parse new format.
48 (dbg_cnt_process_opt): Use strtok.
49 (dbg_cnt_list_all_counters): Remove 'value' and add
51 * doc/invoke.texi: Document changes.
53 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
55 * doc/sourcebuild.texi (scalar_all_fma): Document.
56 * tree.def (FMA_EXPR): Delete.
57 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
58 * internal-fn.c (ternary_direct): New macro.
59 (expand_ternary_optab_fn): Likewise.
60 (direct_ternary_optab_supported_p): Likewise.
61 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
62 * builtins.c (fold_builtin_fma): Delete.
63 (fold_builtin_3): Don't call it.
64 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
65 * expr.c (expand_expr_real_2): Likewise.
66 * fold-const.c (operand_equal_p): Likewise.
67 (fold_ternary_loc): Likewise.
68 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
69 * gimple.c (DEFTREECODE): Likewise.
70 * gimplify.c (gimplify_expr): Likewise.
71 * optabs-tree.c (optab_for_tree_code): Likewise.
72 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
73 * tree-eh.c (operation_could_trap_p): Likewise.
74 (stmt_could_throw_1_p): Likewise.
75 * tree-inline.c (estimate_operator_cost): Likewise.
76 * tree-pretty-print.c (dump_generic_node): Likewise.
77 (op_code_prio): Likewise.
78 * tree-ssa-loop-im.c (stmt_cost): Likewise.
79 * tree-ssa-operands.c (get_expr_operands): Likewise.
80 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
81 * fold-const-call.h (fold_fma): Delete.
82 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
83 CFN_FNMA and CFN_FNMS.
85 * genmatch.c (combined_fn): New enum.
86 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
87 (commutative_op): New function.
88 (commutate): Use it. Handle more than 2 operands.
89 (dt_operand::gen_gimple_expr): Use commutative_op.
90 (parser::parse_expr): Allow :c to be used with non-binary
91 operators if the commutative operand is known.
92 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
93 CFN_FMS, CFN_FNMA and CFN_FNMS.
94 (backprop::process_assign_use): Remove FMA_EXPR handling.
95 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
96 (gen_hsa_fma): New function.
97 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
98 IFN_FNMA and IFN_FNMS.
99 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
100 * gimple-fold.h (follow_all_ssa_edges): Declare.
101 * gimple-fold.c (follow_all_ssa_edges): New function.
102 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
103 gimple_build interface and use follow_all_ssa_edges to fold the result.
104 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
105 instead of checking for optabs directly.
106 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
107 rather than FMA_EXPRs.
108 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
109 call to IFN_FMA instead of an FMA_EXPR.
111 2018-05-17 Jim Wilson <jimw@sifive.com>
113 * expr.c (do_tablejump): When converting index to Pmode, if we have a
114 sign extended promoted subreg, and the range does not have the sign bit
115 set, then do a sign extend.
117 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
118 test, check for sign extended subreg and/or constant operands, and
119 do a sign extend in that case.
121 2018-05-17 Steve Ellcey <sellcey@cavium.com>
123 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
124 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
126 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
127 Change logics_shift_reg to logics_shift_imm.
128 (thunderx2t99_fp_loadpair_basic): Delete.
129 (thunderx2t99_fp_storepair_basic): Delete.
130 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
131 (thunderx2t99_asimd_polynomial): Delete.
132 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
133 and neon_fp_mul_d_scalar_q.
134 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
135 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
136 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
137 (thunderx2t99_asimd_lut): Add missing tbl types.
138 (thunderx2t99_asimd_ext): Delete.
139 (thunderx2t99_asimd_load1_1_mult): Delete.
140 (thunderx2t99_asimd_load1_2_mult): Delete.
141 (thunderx2t99_asimd_load1_ldp): New.
142 (thunderx2t99_asimd_load1): New.
143 (thunderx2t99_asimd_load2): Add missing *load2* types.
144 (thunderx2t99_asimd_load3): New.
145 (thunderx2t99_asimd_load4): New.
146 (thunderx2t99_asimd_store1_1_mult): Delete.
147 (thunderx2t99_asimd_store1_2_mult): Delete.
148 (thunderx2t99_asimd_store2_mult): Delete.
149 (thunderx2t99_asimd_store2_onelane): Delete.
150 (thunderx2t99_asimd_store_stp): New.
151 (thunderx2t99_asimd_store1): New.
152 (thunderx2t99_asimd_store2): New.
153 (thunderx2t99_asimd_store3): New.
154 (thunderx2t99_asimd_store4): New.
156 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
158 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
159 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
161 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
162 Segher Boessenkool <segher@kernel.crashing.org>
165 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
168 2018-05-17 Richard Biener <rguenther@suse.de>
170 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
171 for pruning loop and prune defs feeding only already visited PHIs.
173 2018-05-17 Richard Biener <rguenther@suse.de>
175 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
177 2018-05-17 Bin Cheng <bin.cheng@arm.com>
178 Richard Biener <rguenther@suse.de>
180 PR tree-optimization/85793
181 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
182 for VMAT_ELEMENTWISE.
184 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
186 * internal-fn.h (lookup_internal_fn): Declare
187 * internal-fn.c (lookup_internal_fn): New function.
188 * gimple.c (gimple_build_call_from_tree): Handle calls to
190 * gimple-pretty-print.c (dump_gimple_call): Print "." before
191 internal function names.
192 * tree-pretty-print.c (dump_generic_node): Likewise.
193 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
195 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
197 * gimple-fold.h (gimple_build): Make the function forms take
198 combined_fn rather than built_in_function.
199 (gimple_simplify): Likewise.
200 * gimple-match-head.c (gimple_simplify): Likewise.
201 * gimple-fold.c (gimple_build): Likewise.
202 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
203 rather than gimple_build_call_internal.
204 (get_initial_defs_for_reduction): Likewise.
205 (vect_create_epilog_for_reduction): Likewise.
206 (vectorizable_live_operation): Likewise.
208 2018-05-17 Martin Liska <mliska@suse.cz>
210 * gimple-ssa-sprintf.c (format_directive): Do not use
211 space in between 'G_' and '('.
213 2018-05-17 Jakub Jelinek <jakub@redhat.com>
216 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
217 even if the mask is not all ones.
220 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
222 (ix86_gimple_fold_builtin): Likewise.
225 * config/i386/i386.c: Include tree-vector-builder.h.
226 (ix86_vector_shift_count): New function.
227 (ix86_fold_builtin): Fold shift builtins by scalar count.
228 (ix86_gimple_fold_builtin): Likewise.
230 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
231 _mm512_setzero): New intrinsics.
233 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
234 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
236 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
237 code generation for cases where splatting a value is not useful.
238 * simplify-rtx.c (simplify_ternary_operation): Simplify
239 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
240 mode to a vec_concat.
242 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
244 * config.gcc: Support "goldmont-plus".
245 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
247 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
248 PROCESSOR_GOLDMONT_PLUS.
249 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
250 (processor_target_table): Add "goldmont-plus".
251 (PTA_GOLDMONT_PLUS): Define.
252 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
253 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
254 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
255 (fold_builtin_cpu): Add "goldmont-plus".
256 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
257 (ix86_option_override_internal): Add "goldmont-plus".
258 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
259 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
260 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
261 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
263 2018-05-17 Richard Biener <rguenther@suse.de>
265 PR tree-optimization/85757
266 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
267 remove defs that only feed that PHI from further processing.
269 2018-05-16 Jim Wilson <jimw@sifive.com>
271 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
273 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
275 2018-05-16 Mark Wielaard <mark@klomp.org>
277 * dwarf2out.c (count_index_strings): New function.
278 (output_indirect_strings): Call count_index_strings and generate
279 header for dwarf_version >= 5.
281 2018-05-16 Mark Wielaard <mark@klomp.org>
283 * dwarf2out.c (dwarf_FORM): New function.
284 (set_indirect_string): Use dwarf_FORM.
285 (reset_indirect_string): Likewise.
286 (size_of_die): Likewise.
287 (value_format): Likewise.
288 (output_die): Likewise.
289 (add_skeleton_AT_string): Likewise.
290 (output_macinfo_op): Likewise.
291 (index_string): Likewise.
292 (output_index_string_offset): Likewise.
293 (output_index_string): Likewise.
294 (count_index_strings): Likewise.
296 2018-05-16 Carl Love <cel@us.ibm.com>
298 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
299 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
301 2018-05-16 Martin Jambor <mjambor@suse.cz>
303 * ipa-prop.c (ipa_free_all_edge_args): Remove.
304 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
306 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
308 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
309 (fnma<mode>4): Likewise.
310 (fms<mode>4): Likewise.
311 (fnms<mode>4): Likewise.
312 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
313 (aarch64_fnma<mode>4): Likewise.
314 (aarch64_fms<mode>4): Likewise.
315 (aarch64_fnms<mode>4): Likewise.
316 (aarch64_fnmadd<mode>4): Likewise.
318 2018-05-16 Jason Merrill <jason@redhat.com>
320 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
322 2018-05-16 Richard Biener <rguenther@suse.de>
324 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
325 (dump_stmt_cost): Declare.
326 (add_stmt_cost): Dump cost we add.
327 (add_stmt_costs): New function.
328 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
330 (vect_analyze_stmt): Adjust prototype.
331 (vectorizable_condition): Likewise.
332 (vectorizable_live_operation): Likewise.
333 (vectorizable_reduction): Likewise.
334 (vectorizable_induction): Likewise.
335 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
336 cost vector to pass to vectorizable_ and record afterwards.
337 (vect_model_reduction_cost): Take cost vector argument and adjust.
338 (vect_model_induction_cost): Likewise.
339 (vectorizable_reduction): Likewise.
340 (vectorizable_induction): Likewise.
341 (vectorizable_live_operation): Likewise.
342 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
343 SLP_TREE_NUMBER_OF_VEC_STMTS.
344 (vect_analyze_slp_cost_1): Remove.
345 (vect_analyze_slp_cost): Likewise.
346 (vect_slp_analyze_node_operations): Take visited args and
347 a target cost vector. Avoid processing already visited stmt sets.
348 (vect_slp_analyze_operations): Use a local cost vector to gather
349 costs and register those of non-discarded instances.
350 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
351 (vect_schedule_slp_instance): Remove copying of
352 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
354 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
355 adding cost. Record cost entry location.
356 (vect_prologue_cost_for_slp_op): Function to compute cost of
357 a constant or invariant generated for SLP vect in the prologue,
358 split out from vect_analyze_slp_cost_1.
359 (vect_model_simple_cost): Make static. Adjust for SLP costing.
360 (vect_model_promotion_demotion_cost): Likewise.
361 (vect_model_store_cost): Likewise, make static.
362 (vect_model_load_cost): Likewise.
363 (vectorizable_bswap): Add cost vector arg and adjust.
364 (vectorizable_call): Likewise.
365 (vectorizable_simd_clone_call): Likewise.
366 (vectorizable_conversion): Likewise.
367 (vectorizable_assignment): Likewise.
368 (vectorizable_shift): Likewise.
369 (vectorizable_operation): Likewise.
370 (vectorizable_store): Likewise.
371 (vectorizable_load): Likewise.
372 (vectorizable_condition): Likewise.
373 (vectorizable_comparison): Likewise.
374 (can_vectorize_live_stmts): Likewise.
375 (vect_analyze_stmt): Likewise.
376 (vect_transform_stmt): Adjust calls to vectorizable_*.
377 * tree-vectorizer.c: Include gimple-pretty-print.h.
378 (dump_stmt_cost): New function.
380 2018-05-16 Richard Biener <rguenther@suse.de>
382 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
383 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
384 * tree-ssa-dse.c: Include tree-ssa-loop.h.
385 (check_name): New callback.
386 (dse_classify_store): Track cycles via a visited bitmap of PHI
387 defs and simplify handling of in-loop and across loop dead stores
388 and properly fail for loop-variant refs. Handle byte-tracking with
389 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
392 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
394 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
395 (vect_get_mask_type_for_stmt): Likewise.
396 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
398 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
399 to determine the statement's vector type and the vector type that
400 should be used for calculating nunits. Deal with cases in which
401 the type has to be deferred.
402 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
403 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
404 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
405 (vect_determine_vf_for_stmt): New functions, split out from...
406 (vect_determine_vectorization_factor): ...here.
407 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
408 (vect_get_mask_type_for_stmt): New functions, split out from
409 vect_determine_vectorization_factor.
411 2018-05-16 Richard Biener <rguenther@suse.de>
413 * tree-cfg.c (verify_gimple_assign_ternary): Properly
414 verify the [VEC_]COND_EXPR embedded comparison.
416 2018-05-15 Martin Sebor <msebor@redhat.com>
418 PR tree-optimization/85753
419 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
420 RECORD_TYPE in addition to ARRAY_TYPE.
422 2018-05-15 Martin Sebor <msebor@redhat.com>
425 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
427 2018-05-15 Richard Biener <rguenther@suse.de>
429 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
430 add by_clobber_p one. Change algorithm to collect all defs
431 representing uses we need to walk and try reducing them to
432 a single one before failing.
433 (dse_dom_walker::dse_optimize_stmt): Adjust.
435 2018-05-13 Mark Wielaard <mark@klomp.org>
437 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
438 (size_of_loc_descr): Likewise.
439 (output_loc_operands): Likewise.
440 (output_loc_operands_raw): Likewise.
441 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
442 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
443 (hash_loc_operands): Likewise.
444 (compare_loc_operands): Likewise.
446 2018-05-14 Mark Wielaard <mark@klomp.org>
448 * dwarf2out.c (count_index_addrs): New function.
449 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
451 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
453 PR tree-optimization/83648
454 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
455 return value as malloc candidate.
457 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
460 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
461 as true in call to suggest_attribute.
463 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
465 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
468 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
470 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
472 (aarch64_crypto_aesd_fused): Likewise.
474 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
476 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
477 (movsi_aarch64): Likewise.
478 (load_pairsi): Likewise.
479 (load_pairdi): Likewise.
480 (store_pairsi): Likewise.
481 (store_pairdi): Likewise.
482 (load_pairsf): Likewise.
483 (load_pairdf): Likewise.
484 (store_pairsf): Likewise.
485 (store_pairdf): Likewise.
486 (zero_extend): Likewise.
487 (trunc): Swap alternatives.
488 (fcvt_target): Add '?' to prefer w over r.
490 2018-05-14 Jakub Jelinek <jakub@redhat.com>
493 * config/i386/i386.md: Disallow non-commutative arithmetics in
494 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
495 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
496 in the peephole2 before it.
498 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
500 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
501 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
502 (ix86_handle_option): Handle -mcldemote.
503 * config.gcc: New header.
504 * config/i386/cldemoteintrin.h: New file.
505 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
506 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
508 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
509 OPTION_MASK_ISA_CLDEMOTE.
510 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
511 (ix86_valid_target_attribute_inner_p): Ditto.
512 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
513 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
514 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
515 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
516 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
518 * config/i386/i386.opt: Add -mcldemote.
519 * config/i386/x86intrin.h: New header.
520 * doc/invoke.texi: Add -mcldemote.
522 2018-05-14 Richard Biener <rguenther@suse.de>
524 * doc/match-and-simplify.texi: Adjust :s documentation.
526 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
528 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
529 intended memcpy size.
530 (REORDER_45): Likewise.
532 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
535 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
536 * vec.c (qsort_chk): Use gcc_qsort.
537 * Makefile.in (OBJS-libcommon): Add sort.o.
538 (build/sort.o): New target. Use it...
539 (BUILD_RTL): ... here, and...
540 (build/gencfn-macros): ... here, and...
541 (build/genmatch): ... here.
543 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
544 Chung-Ju Wu <jasonwucj@gmail.com>
546 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
547 * config/nds32/nds32-graywolf.md: New file.
548 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
549 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
551 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
552 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
553 * config/nds32/nds32.md (pipeline_model): Add graywolf.
554 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
555 * config/nds32/pipelines.md: Include n15 settings.
557 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
558 Chung-Ju Wu <jasonwucj@gmail.com>
560 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
561 * config/nds32/nds32-n13.md: New file.
562 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
563 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
565 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
566 * config/nds32/nds32.md (pipeline_model): Add n13.
567 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
568 * config/nds32/pipelines.md: Include n13 settings.
570 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
571 Chung-Ju Wu <jasonwucj@gmail.com>
573 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
574 * config/nds32/nds32-n10.md: New file.
575 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
576 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
578 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
579 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
580 * config/nds32/nds32.md (pipeline_model): Add n10.
581 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
582 * config/nds32/pipelines.md: Include n10 settings.
584 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
585 Kito Cheng <kito.cheng@gmail.com>
586 Chung-Ju Wu <jasonwucj@gmail.com>
588 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
589 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
590 Add enum values for DSP extension instructions.
591 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
593 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
594 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
596 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
597 * config/nds32/nds32-dspext.md: New file for DSP implementation.
598 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
599 * config/nds32/nds32-intrinsic.md: Likewise.
600 * config/nds32/nds32_intrinsic.h: Likewise.
601 * config/nds32/nds32-md-auxiliary.c: Likewise.
602 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
603 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
604 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
605 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
606 * config/nds32/nds32-protos.h: New declarations for DSP extension.
607 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
608 TYPE_DMAC in switch statement.
609 * config/nds32/nds32.c: New checking and implementation for DSP
610 extension instructions.
611 * config/nds32/nds32.h: Likewise.
612 * config/nds32/nds32.md: Likewise.
613 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
614 * config/nds32/predicates.md: Implement new predicates for DSP
617 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
619 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
620 Reformat alternatives and attributes so it is easier to identify
621 which constraints/attributes go with which instruction.
622 (mov<mode>_hardfloat32, FMOVE64): Likewise.
623 (mov<mode>_softfloat32, FMOVE64): Likewise.
624 (mov<mode>_hardfloat64, FMOVE64): Likewise.
625 (mov<mode>_softfloat64, FMOVE64): Likewise.
627 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
629 * doc/extend.texi (PowerPC Built-in Functions): Rename this
631 (Basic PowerPC Built-in Functions): The new name of the
632 subsection previously known as "PowerPC Built-in Functions".
633 (Basic PowerPC Built-in Functions Available on all Configurations):
635 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
636 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
637 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
638 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
640 2018-05-11 Martin Jambor <mjambor@suse.cz>
643 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
646 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
649 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
651 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
653 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
654 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
655 (ix86_handle_option): Handle -mwaitpkg.
656 * config.gcc: New header.
657 * config/i386/cpuid.h (bit_WAITPKG): New bit.
658 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
659 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
661 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
662 OPTION_MASK_ISA_WAITPKG.
663 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
664 (ix86_option_override_internal): Add PTA_WAITPKG.
665 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
666 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
668 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
669 __builtin_ia32_umwait and __builtin_ia32_tpause.
670 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
671 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
672 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
673 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
674 UNSPECV_TPAUSE): New.
675 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
676 * config/i386/i386.opt: Add -mwaitpkg.
677 * config/i386/waitpkgintrin.h: New file.
678 * config/i386/x86intrin.h: New header.
679 * doc/invoke.texi: Add -mwaitpkg.
681 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
684 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
686 (cortex-m0): Use armv6s-m isa.
687 (cortex-m0plus): Likewise.
688 (cortex-m1): Likewise.
689 (cortex-m0.small-multiply): Likewise.
690 (cortex-m0plus.small-multiply): Likewise.
691 (cortex-m1.small-multiply): Likewise.
693 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
694 Jakub Jelinek <jakub@redhat.com>
696 PR tree-optimization/85692
697 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
698 source permute as well.
700 2018-05-11 Martin Liska <mliska@suse.cz>
703 * doc/extend.texi: Document LLVM style format for no_sanitize
706 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
708 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
709 mode_supports_vsx_dform_quad to mode_supports_dq_form.
710 (mode_supports_vsx_dform_quad): Likewise.
711 (mode_supports_vmx_dform): Move these functions to be next to the
712 other mode_supports functions.
713 (mode_supports_dq_form): Likewise.
714 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
715 mode_supports_dq_form.
716 (reg_offset_addressing_ok_p): Likewise.
717 (offsettable_ok_by_alignment): Likewise.
718 (rs6000_legitimate_offset_address_p): Likewise.
719 (legitimate_lo_sum_address_p): Likewise.
720 (rs6000_legitimize_address): Likewise.
721 (rs6000_legitimize_reload_address): Likewise.
722 (rs6000_secondary_reload_inner): Likewise.
723 (rs6000_preferred_reload_class): Likewise.
724 (rs6000_output_move_128bit): Likewise.
726 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
728 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
729 Generate SImode target register for null target.
730 <case IX86_BUILTIN_XGETBV>: Ditto.
731 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
732 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
734 2018-05-10 Carl Love <cel@us.ibm.com>
736 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
737 dcbtt and dcbtstt if operands[2] is 0.
739 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
742 * config/i386/sse.md (usadv64qi): New expander.
744 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
746 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
747 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
748 -maltivec=be support.
749 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
750 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
751 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
752 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
753 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
754 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
755 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
756 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
757 altivec_vsumsws): Adjust.
758 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
759 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
760 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
762 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
763 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
764 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
765 (altivec_lve<VI_char>x): Delete expand.
766 (*altivec_lve<VI_char>x_internal): Rename to...
767 (altivec_lve<VI_char>x): ... this.
768 (altivec_lvxl_<mode>): Delete expand.
769 (*altivec_lvxl_<mode>_internal): Rename to ...
770 (altivec_lvxl_<mode>): ... this.
771 (altivec_stvxl_<mode>): Delete expand.
772 (*altivec_stvxl_<mode>_internal): Rename to ...
773 (altivec_stvxl_<mode>): ... this.
774 (altivec_stve<VI_char>x): Delete expand.
775 (*altivec_stve<VI_char>x_internal): Rename to ...
776 (altivec_stve<VI_char>x): ... this.
777 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
778 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
779 reduc_plus_scal_<mode>): Adjust.
780 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
782 (rs6000_cpu_cpp_builtins): Adjust.
783 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
784 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
785 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
786 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
787 -maltivec=be support.
788 (rs6000_split_vec_extract_var): Adjust.
789 (rs6000_split_v4si_init): Adjust.
790 (swap_selector_for_mode): Delete.
791 (altivec_expand_lvx_be, altivec_expand_stvx_be,
792 altivec_expand_stvex_be): Delete.
793 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
794 -maltivec=be support.
795 (rs6000_gimple_fold_builtin): Ditto.
796 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
798 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
799 (TARGET_DIRECT_MOVE_64BIT): Adjust.
800 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
801 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
802 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
803 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
804 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
805 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
806 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
807 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
808 anonymous split): Adjust.
809 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
810 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
812 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
814 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
815 when --with-gxx-include-dir is also specified.
816 * configure: Regenerate.
818 2018-05-09 Jim Wilson <jimw@sifive.com>
821 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
822 * config/riscv/t-withmultilib: New.
823 * config/riscv/withmultilib.h: New.
824 * doc/install.texi: Document RISC-V --with-multilib-list support.
826 2018-05-09 Richard Biener <rguenther@suse.de>
828 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
830 (vect_bb_vectorization_profitable_p): Adjust. Compute
831 actual scalar cost using the cost vector and the add_stmt_cost
834 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
836 PR rtl-optimization/85645
837 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
838 in the REG_CFA_REGISTER note for LR, don't leave it empty.
840 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
842 PR rtl-optimization/85645
843 * shrink-wrap.c (spread_components): Return a boolean saying if
844 anything was changed.
845 (try_shrink_wrapping_separate): Iterate spread_components until
846 nothing changes anymore.
848 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
850 PR rtl-optimization/85645
851 * regrename.c (build_def_use): Also kill the chains that include the
852 destination of a REG_CFA_REGISTER note.
854 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
856 PR rtl-optimization/85645
857 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
858 insn that has a REG_CFA_REGISTER note.
860 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
862 * cfgexpand.c (expand_clobber): New function.
863 (expand_gimple_stmt_1): Use it.
864 * tree-vect-stmts.c (vect_clobber_variable): New function,
866 (vectorizable_simd_clone_call): ...here.
867 (vectorizable_store): Emit a clobber either side of an
868 IFN_STORE_LANES sequence.
869 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
871 2018-05-09 Tom de Vries <tom@codesourcery.com>
874 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
875 (define_insn "trap_if_false"): Add exit after trap.
877 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
879 PR rtl-optimization/85638
880 * bb-reorder.c: Include common/common-target.h.
881 (create_forwarder_block): New function extracted from...
882 (fix_up_crossing_landing_pad): ...here. Rename into...
883 (dw2_fix_up_crossing_landing_pad): ...this.
884 (sjlj_fix_up_crossing_landing_pad): New function.
885 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
886 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
887 from both partitions and exit the loop after one iteration.
889 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
892 * doc/extend.texi (PowerPC Built-in Functions): Rename this
894 (Basic PowerPC Built-in Functions): The new name of the
895 subsection previously known as "PowerPC Built-in Functions".
896 (Basic PowerPC Built-in Functions Available on all Configurations):
898 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
900 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
902 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
904 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
907 2018-05-08 Jim Wilson <jimw@sifive.com>
909 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
910 (LD_EMUL_SUFFIX): New.
913 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
915 * doc/extend.texi (PowerPC Built-in Functions): Rename this
917 (Basic PowerPC Built-in Functions): The new name of the
918 subsection previously known as "PowerPC Built-in Functions".
919 (Basic PowerPC Built-in Functions Available on all Configurations):
921 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
923 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
925 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
927 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
930 2018-05-08 Jakub Jelinek <jakub@redhat.com>
933 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
934 after cmpelim optimization.
936 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
938 * config.gcc: Support "goldmont".
939 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
940 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
942 * config/i386/i386.c (m_GOLDMONT): Define.
943 (processor_target_table): Add "goldmont".
944 (PTA_GOLDMONT): Define.
945 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
946 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
947 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
948 (fold_builtin_cpu): Add "goldmont".
949 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
950 (ix86_option_override_internal): Add "goldmont".
951 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
952 (processor_type): Add PROCESSOR_GOLDMONT.
953 * config/i386/i386.md: Add CPU "glm".
954 * config/i386/glm.md: New file.
955 * config/i386/x86-tune.def: Add m_GOLDMONT.
956 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
958 2018-05-08 Jakub Jelinek <jakub@redhat.com>
961 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
963 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
964 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
965 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
968 * config/i386/i386.c (ix86_fold_builtin): Handle
969 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
972 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
973 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
975 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
978 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
979 (check_arch): Likewise.
980 (check_fpu): Return the result rather than printing it.
981 (end arch): Fix operator precedence.
983 (END): Print the result from check_fpu.
985 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
986 Alan Hayward <alan.hayward@arm.com>
987 David Sherwood <david.sherwood@arm.com>
989 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
990 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
991 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
992 (*fcmuo<mode>_and): New patterns.
994 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
996 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
997 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
998 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
999 (cmp_op, sve_imm_con): New code attributes.
1000 (SVE_COND_INT_CMP, imm_con): Delete.
1001 (cmp_op): Remove above unspecs from int attribute.
1002 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
1004 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
1005 comparison-specific unspecs.
1006 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
1007 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
1008 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
1009 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
1010 (*vec_fcm<cmp_op><mode>): Rename to...
1011 (*fcm<cmp_op><mode>): ...this and adjust likewise.
1012 (*vec_fcmuo<mode>): Rename to...
1013 (*fcmuo<mode>): ...this and adjust likewise.
1014 (*pred_fcm<cmp_op><mode>): New pattern.
1015 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
1016 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
1018 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
1020 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
1021 (aarch64_emit_sve_predicated_cond): New function.
1022 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
1023 (aarch64_emit_unspec_cond_or): Replace with...
1024 (aarch64_emit_sve_or_conds): ...this new function. Use
1025 aarch64_emit_sve_ptrue_op for the individual comparisons and
1026 aarch64_emit_binop to OR them together.
1027 (aarch64_emit_inverted_unspec_cond): Replace with...
1028 (aarch64_emit_sve_inverted_cond): ...this new function. Use
1029 aarch64_emit_sve_ptrue_op for the comparison and
1030 aarch64_emit_unop to invert the result.
1031 (aarch64_expand_sve_vec_cmp_float): Update after the above
1032 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
1034 2018-05-07 Nathan Sidwell <nathan@acm.org>
1036 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
1037 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
1038 (Backwards Compatibility): Likewise.
1040 2018-05-07 Luis Machado <luis.machado@linaro.org>
1044 2018-05-07 Luis Machado <luis.machado@linaro.org>
1046 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1047 <prefetch_dynamic_strides>: New const bool field.
1048 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1049 prefetch_dynamic_strides.
1050 (exynosm1_prefetch_tune): Likewise.
1051 (thunderxt88_prefetch_tune): Likewise.
1052 (thunderx_prefetch_tune): Likewise.
1053 (thunderx2t99_prefetch_tune): Likewise.
1054 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1055 (aarch64_override_options_internal): Update to set
1056 PARAM_PREFETCH_DYNAMIC_STRIDES.
1057 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1058 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1059 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1060 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1061 prefetch-dynamic-strides setting.
1063 2018-05-07 Luis Machado <luis.machado@linaro.org>
1065 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1066 <minimum_stride>: New const int field.
1067 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1068 minimum_stride field.
1069 (exynosm1_prefetch_tune): Likewise.
1070 (thunderxt88_prefetch_tune): Likewise.
1071 (thunderx_prefetch_tune): Likewise.
1072 (thunderx2t99_prefetch_tune): Likewise.
1073 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1074 (aarch64_override_options_internal): Update to set
1075 PARAM_PREFETCH_MINIMUM_STRIDE.
1076 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1077 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1078 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1079 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1080 stride is constant and is below the minimum stride threshold.
1082 2018-05-07 Luis Machado <luis.machado@linaro.org>
1084 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
1087 2018-05-07 Luis Machado <luis.machado@linaro.org>
1089 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1090 <prefetch_dynamic_strides>: New const bool field.
1091 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1092 prefetch_dynamic_strides.
1093 (exynosm1_prefetch_tune): Likewise.
1094 (thunderxt88_prefetch_tune): Likewise.
1095 (thunderx_prefetch_tune): Likewise.
1096 (thunderx2t99_prefetch_tune): Likewise.
1097 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1098 (aarch64_override_options_internal): Update to set
1099 PARAM_PREFETCH_DYNAMIC_STRIDES.
1100 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1101 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1102 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1103 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1104 prefetch-dynamic-strides setting.
1106 2018-05-07 Luis Machado <luis.machado@linaro.org>
1108 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1109 <minimum_stride>: New const int field.
1110 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1111 minimum_stride field.
1112 (exynosm1_prefetch_tune): Likewise.
1113 (thunderxt88_prefetch_tune): Likewise.
1114 (thunderx_prefetch_tune): Likewise.
1115 (thunderx2t99_prefetch_tune): Likewise.
1116 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1117 (aarch64_override_options_internal): Update to set
1118 PARAM_PREFETCH_MINIMUM_STRIDE.
1119 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1120 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1121 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1122 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1123 stride is constant and is below the minimum stride threshold.
1125 2018-05-06 Jakub Jelinek <jakub@redhat.com>
1128 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
1129 the type is addressable. Don't force op into register if it has
1132 2018-05-05 Roland McGrath <mcgrathr@google.com>
1135 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
1136 any section for which we don't know a specific type it should have,
1137 regardless of name. Previously this was done only for the exact
1138 names ".init_array", ".fini_array", and ".preinit_array".
1139 (default_elf_asm_named_section): Add comment about
1140 relationship with default_section_type_flags and SECTION_NOTYPE.
1141 (get_section): Don't consider it a type conflict if one side has
1142 SECTION_NOTYPE and the other doesn't, as long as neither has the
1143 SECTION_BSS et al used in the default_section_type_flags logic.
1145 2018-05-05 Tom de Vries <tom@codesourcery.com>
1148 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
1149 (workaround_barsyncs): New function.
1150 (nvptx_reorg): Use workaround_barsyncs.
1151 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1152 (define_expand "nvptx_membar_cta"): New define_expand.
1153 (define_insn "*nvptx_membar_cta"): New insn.
1155 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
1157 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
1158 To improve optimization opportunities.
1159 * builtin-types.def: The new needed builtin types for the above.
1161 2018-05-04 Richard Biener <rguenther@suse.de>
1163 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
1164 * gimple-ssa-store-merging.c
1165 (imm_store_chain_info::output_merged_store): Remove redundant create,
1166 release split_store vector contents on failure.
1167 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
1168 scalar stmt vector on cache hit.
1170 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
1172 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
1174 * config.gcc (powerpc-xilinx-eabi*): Remove.
1175 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
1177 (fusion_addis_mem_combo_load): Ditto.
1178 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
1180 (rs6000_cpu_cpp_builtins): Ditto.
1181 * config/rs6000/rs6000-linux.c
1182 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
1183 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
1184 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
1186 (rs6000_setup_reg_addr_masks): Ditto.
1187 (rs6000_init_hard_regno_mode_ok): Ditto.
1188 (rs6000_option_override_internal): Ditto.
1189 (legitimate_lo_sum_address_p): Ditto.
1190 (rs6000_legitimize_address): Ditto.
1191 (rs6000_legitimize_reload_address): Ditto.
1192 (rs6000_legitimate_address_p): Ditto.
1193 (abi_v4_pass_in_fpr): Ditto.
1194 (setup_incoming_varargs): Ditto.
1195 (rs6000_gimplify_va_arg): Ditto.
1196 (rs6000_split_multireg_move): Ditto.
1197 (rs6000_savres_strategy): Ditto.
1198 (rs6000_emit_prologue_components): Ditto.
1199 (rs6000_emit_epilogue_components): Ditto.
1200 (rs6000_emit_prologue): Ditto.
1201 (rs6000_emit_epilogue): Ditto.
1202 (rs6000_elf_file_end): Ditto.
1203 (rs6000_function_value): Ditto.
1204 (rs6000_libcall_value): Ditto.
1205 * config/rs6000/rs6000.h: Ditto.
1206 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
1207 (TARGET_MINMAX): ... this. New.
1208 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
1209 * config/rs6000/rs6000.md: Remove Xilinx FP support.
1210 (*movsi_internal1_single): Delete.
1211 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
1212 mfpu=, mxilinx-fpu): Delete.
1213 * config/rs6000/singlefp.h: Delete.
1214 * config/rs6000/sysv4.h: Remove Xilinx FP support.
1215 * config/rs6000/t-rs6000: Ditto.
1216 * config/rs6000/t-xilinx: Delete.
1217 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
1218 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
1219 (VStype_simple): Delete.
1220 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
1221 * config/rs6000/xfpu.h: Delete.
1222 * config/rs6000/xfpu.md: Delete.
1223 * config/rs6000/xilinx.h: Delete.
1224 * config/rs6000/xilinx.opt: Delete.
1225 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
1226 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
1228 2018-05-04 Tom de Vries <tom@codesourcery.com>
1231 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
1234 2018-05-04 Richard Biener <rguenther@suse.de>
1237 * tree-complex.c (update_complex_assignment): We are always in SSA form.
1238 (expand_complex_div_wide): Likewise.
1239 (expand_complex_operations_1): Likewise.
1240 (expand_complex_libcall): Preserve EH info of the original stmt.
1241 (tree_lower_complex): Handle removed blocks.
1242 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
1243 on complex multiplication and division libcall builtins.
1245 2018-05-04 Richard Biener <rguenther@suse.de>
1248 * fold-const.c (negate_expr_p): Restrict negation of operand
1249 zero of a division to when we know that can happen without
1251 (fold_negate_expr_1): Likewise.
1253 2018-05-04 Jakub Jelinek <jakub@redhat.com>
1256 * real.h (real_nextafter): Declare.
1257 * real.c (real_nextafter): New function.
1258 * fold-const-call.c (fold_const_nextafter): New function.
1259 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
1260 CASE_CFN_NEXTTOWARD.
1261 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
1262 even when arg1_mode is different from arg0_mode.
1264 2018-05-03 Nathan Sidwell <nathan@acm.org>
1266 * doc/extend.texi (Deprecated Features): Remove
1268 (Backwards Compatibility): Likewise.
1269 * doc/invoke.texi (C++ Language Options): Likewise.
1270 (C++ Dialect Options): Likewise.
1272 2018-05-03 Jakub Jelinek <jakub@redhat.com>
1275 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1276 _mm512_mask_mullox_epi64): New intrinsics.
1278 2018-05-03 Tom de Vries <tom@codesourcery.com>
1281 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1282 dump files): Add offload-tree.
1284 2018-05-03 Richard Biener <rguenther@suse.de>
1286 PR tree-optimization/85615
1287 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1288 to loops not nested in BBs loop father to avoid creating multi-entry
1291 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1293 PR tree-optimization/70291
1294 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1295 arguments. Change return type to tree. Emit libcall as a new
1296 statement rather than replacing existing one when inplace_p is true.
1297 (expand_complex_multiplication_components): New function.
1298 (expand_complex_multiplication): Expand floating-point complex
1299 multiplication using the above.
1300 (expand_complex_division): Rename inner_type parameter to type.
1301 Update expand_complex_libcall call-site.
1302 (expand_complex_operations_1): Update expand_complex_multiplication
1303 and expand_complex_division call-sites.
1305 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1308 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1309 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1310 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1311 the highest significant bit of the shift count mask is clear. In
1312 check whether and[sq]i3 is needed verify that all significant bits
1313 of the shift count other than the highest are set.
1315 2018-05-02 Tom de Vries <tom@codesourcery.com>
1318 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1319 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1320 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1321 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1322 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1323 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1324 __builtin_goacc_parlevel_size.
1326 2018-05-02 Richard Biener <rguenther@suse.de>
1328 PR tree-optimization/85597
1329 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1330 do not use split vect_get_vec_defs call but call vect_get_slp_defs
1333 2018-05-02 Tom de Vries <tom@codesourcery.com>
1336 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1337 dump files): Add ltrans-tree.
1339 2018-05-02 Tom de Vries <tom@codesourcery.com>
1342 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1343 dump files): Add wpa-ipa.
1345 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
1347 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
1348 powerpc*-*-linux*paired* target.
1349 * config/rs6000/750cl.h: Delete.
1350 * config/rs6000/paired.h: Delete.
1351 * config/rs6000/paired.md: Delete.
1352 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1354 * config/rs6000/rs6000-builtin.def: Remove paired float support.
1355 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1356 comment. Remove paired float support.
1357 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1358 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1360 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1361 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1363 * config/rs6000/rs6000.c: Remove paired float support.
1364 (paired_expand_vector_init, paired_expand_vector_move,
1365 paired_emit_vector_compare, paired_emit_vector_cond_expr,
1366 (paired_expand_lv_builtin, paired_expand_stv_builtin,
1367 paired_expand_builtin, paired_expand_predicate_builtin,
1368 paired_init_builtins): Delete.
1369 * config/rs6000/rs6000.h: Remove paired float support.
1370 * config/rs6000/rs6000.md: Remove paired float support.
1371 (move_from_CR_ov_bit): Delete.
1372 * config/rs6000/rs6000.opt (mpaired): Delete.
1373 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1374 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1376 2018-05-02 Richard Biener <rguenther@suse.de>
1379 * gimplify.c (gimplify_save_expr): When in SSA form allow
1380 SAVE_EXPRs to compute to SSA vars.
1382 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1385 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1386 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1387 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1388 clobber operands[2], instead use a new pseudo. Formatting fixes.
1390 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1392 PR tree-optimization/85586
1393 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1394 exit early for statements in the same group if the accesses are
1397 2018-05-02 Tom de Vries <tom@codesourcery.com>
1400 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1403 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
1405 PR tree-optimization/85143
1406 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1408 2018-05-01 Tom de Vries <tom@codesourcery.com>
1411 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1412 not found" error message.
1414 2018-05-01 Tom de Vries <tom@codesourcery.com>
1417 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1418 * vec.c (test_ordered_remove_if): New function.
1419 (vec_c_tests): Call test_ordered_remove_if.
1420 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1421 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1422 * tree-vect-patterns.c (vect_pattern_recog_1): Use
1423 VEC_ORDERED_REMOVE_IF.
1425 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1427 PR tree-optimization/82665
1428 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1429 pointer subtraction where arguments come from a memchr call.
1431 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1433 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1434 --push-state --as-needed and --pop-state instead of --as-needed and
1435 --no-as-needed if ld supports it.
1436 * configure: Regenerated.
1439 * doc/install.texi2html: Replace _002d with - and _002a with * in
1440 generated html files using sed.
1442 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1445 * gcc-rich-location.c (blank_line_before_p): New function.
1446 (use_new_line): New function.
1447 (gcc_rich_location::add_fixit_insert_formatted): New function.
1448 * gcc-rich-location.h
1449 (gcc_rich_location::add_fixit_insert_formatted): New function.
1451 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1453 * selftest.c (assert_streq): Rename "expected" and "actual" to
1454 "val1" and "val2". Extend NULL-handling to cover both inputs
1455 symmetrically, while still requiring both to be non-NULL for a pass.
1456 * selftest.h (assert_streq): Rename "expected" and "actual" to
1458 (ASSERT_EQ): Likewise.
1459 (ASSERT_EQ_AT): Likewise.
1460 (ASSERT_KNOWN_EQ): Likewise.
1461 (ASSERT_KNOWN_EQ_AT): Likewise.
1462 (ASSERT_NE): Likewise.
1463 (ASSERT_MAYBE_NE): Likewise.
1464 (ASSERT_MAYBE_NE_AT): Likewise.
1465 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1466 the assertion to pass.
1467 (ASSERT_STREQ_AT): Likewise.
1469 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1471 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1472 interaction with -pie.
1474 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1476 * selftest.h: Fix alphabetization of per-source-file selftest
1479 2018-04-30 Jason Merrill <jason@redhat.com>
1481 PR c++/61982 - dead stores to destroyed objects.
1482 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1485 2018-04-30 Jason Merrill <jason@redhat.com>
1487 * tree.c (build_clobber): New.
1488 * tree.h: Declare it.
1489 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1491 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1493 * diagnostic-show-locus.c (layout::layout): Update for
1494 location_get_source_line returning a char_span.
1495 (struct char_span): Move to input.h.
1496 (struct correction): Update for fields in char_span becoming
1498 (struct source_line): Update for location_get_source_line
1499 returning a char_span.
1500 (layout::print_line): Likewise.
1501 * edit-context.c (edited_file::print_content): Likewise.
1502 (edited_file::print_diff_hunk): Likewise.
1503 (edited_file::print_run_of_changed_lines): Likewise.
1504 (edited_file::get_num_lines): Likewise.
1505 (edited_line::edited_line): Likewise.
1506 * final.c (asm_show_source): Likewise.
1507 * input.c (location_get_source_line): Convert return type
1508 from const char * to char_span, losing the final "line_len"
1510 (dump_location_info): Update for the above.
1511 (get_substring_ranges_for_loc): Likewise. Use a char_span
1512 when handling the literal within the line.
1513 (test_reading_source_line): Update for location_get_source_line
1514 returning a char_span.
1515 * input.h (class char_span): Move here from
1516 diagnostic-show-locus.c, converting from a struct to a class.
1517 Make data members private.
1518 (char_span::operator bool): New.
1519 (char_span::length): New.
1520 (char_span::get_buffer): New.
1521 (char_span::operator[]): New.
1522 (char_span::subspan): Make const.
1523 (char_span::xstrdup): New.
1524 (location_get_source_line): Convert return type from const char *
1525 to char_span, losing the final "line_size" param.
1527 2018-04-30 Jan Hubicka <jh@suse.cz>
1529 * lto-wrapper.c (ltrans_priorities): New static var.
1530 (cmp_priority): New.
1531 (run_gcc): Read priorities and if doing parallel build order
1532 the Makefile by them.
1534 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1536 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
1538 2018-04-30 Richard Biener <rguenther@suse.de>
1540 * tree-cfg.c (verify_address): Remove base argument, add
1541 flag whether to check TREE_ADDRESSABLE and do that.
1542 (verify_expr): Remove.
1543 (verify_types_in_gimple_reference): Add pieces from verify_expr.
1544 (verify_gimple_assign_single): Likewise.
1545 (verify_gimple_switch): Likewise.
1546 (verify_expr_location_1): Dereference tp once. Add (disabled)
1547 piece from verify_expr.
1548 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1550 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1552 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1554 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1556 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1557 (small_data_pattern): Likewise.
1558 (arc_rewrite_small_data): Likewise.
1559 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1560 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1561 (get_symbol_alignment): New function.
1562 (legitimate_small_data_address_p): Likewise.
1563 (legitimate_scaled_address): Update, call
1564 legitimate_small_data_address_p.
1565 (output_sdata): New static variable.
1566 (arc_print_operand): Update how we handle small data operands.
1567 (arc_print_operand_address): Likewise.
1568 (arc_legitimate_address_p): Update, use
1569 legitimate_small_data_address_p.
1570 (arc_rewrite_small_data_p): Remove.
1571 (arc_rewrite_small_data_1): Likewise.
1572 (arc_rewrite_small_data): Likewise.
1573 (small_data_pattern): Likewise.
1574 (compact_sda_memory_operand): Update to use
1575 legitimate_small_data_address_p and get_symbol_alignment.
1576 (prepare_move_operands): Don't rewite sdata pattern.
1577 (prepare_extend_operands): Remove.
1578 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
1580 (zero_extendqisi2): Likewise.
1581 (zero_extendhisi2): Likewise.
1582 (extendqihi2): Likewise.
1583 (extendqisi2): Likewise.
1584 (extendhisi2): Likewise.
1588 * config/arc/constraints.md (Usd): Change it to memory constraint.
1590 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1592 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
1593 as source of std instructions.
1594 * config/arc/arc.md (movsi_insn): Update pattern predicate to
1595 allow 6-bit constants as source for store instructions.
1596 (movdi_insn): Update instruction pattern to allow 6-bit constants
1597 as source for store instructions.
1599 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1601 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
1603 2018-04-30 Nathan Sidwell <nathan@acm.org>
1604 Sandra Loosemore <sandra@codesourcery.com>
1606 * dumpfile.c (dump_open): Allow '-' for stdout.
1607 * doc/invoke.texi (Developer Options): Document dump filename
1608 determination early. Document stdin/stdout selection.
1610 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
1612 Microblaze Target: PIC data text relative
1614 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
1615 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
1617 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
1618 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
1619 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
1620 New addressing mode for data-text relative position indepenedent code.
1621 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
1622 'ADDRESS_SYMBOLIC_TXT_REL'.
1623 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
1624 (microblaze_legitimate_pic_operand): Exclude function calls from
1625 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
1626 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
1628 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1629 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1630 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
1631 for 'address + offset'.
1632 (microblaze_expand_prologue): Add new function prologue call for
1634 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
1635 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
1636 table in case of TARGET_PIC_DATA_TEXT_REL.
1637 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
1638 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
1639 Add new macros 'UNSPEC_TEXT',
1640 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
1641 + exclude function calls from 'UNSPEC_PLT' in case of data text
1643 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1644 new target hook for generating address diff vector tables in case of
1646 * doc/tm.texi : Regenerate.
1647 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
1648 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
1649 of addr diff vector generation.
1650 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1651 target hook definition.
1652 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
1653 Add default function for generate_pic_addr_diff_vec -> flag_pic.
1654 * doc/invoke.texi (Add new pic option): Add new microblaze pic
1655 option for data text relative.
1657 2018-04-30 Richard Biener <rguenther@suse.de>
1659 * tree-chrec.h (evolution_function_is_constant_p): Remove
1661 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
1663 2018-04-30 Richard Biener <rguenther@suse.de>
1666 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
1668 2018-04-30 Richard Biener <rguenther@suse.de>
1670 PR tree-optimization/28364
1671 PR tree-optimization/85275
1672 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
1673 copying first exit test.
1675 2018-04-28 Mark Wielaard <mark@klomp.org>
1677 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
1679 (dwarf_AT): Handle DW_AT_addr_base.
1680 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
1682 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
1685 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
1686 (*ashl<dwi>3_doubleword_mask_1): Ditto.
1687 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
1688 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
1690 2018-04-28 Richard Biener <rguenther@suse.de>
1692 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
1693 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
1694 to reflect use. Only add interesting stmts.
1696 2018-04-27 Martin Jambor <mjambor@suse.cz>
1699 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
1700 the jump function allows for passing through aggregate values.
1702 2018-04-27 David Malcolm <dmalcolm@redhat.com>
1704 * input.h (in_system_header_at): Convert from macro to inline
1706 (from_macro_expansion_at): Likewise.
1707 (from_macro_definition_at): Likewise.
1709 2018-04-27 Jeff Law <law@redhat.com>
1711 * config.gcc: Mark tile* targets as deprecated/obsolete.
1713 2018-04-27 Richard Biener <rguenther@suse.de>
1715 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
1718 2018-04-27 Richard Biener <rguenther@suse.de>
1720 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
1722 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
1724 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
1725 with Yd constraint. Set "preferred_for_speed" attribute from
1726 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
1728 (*movdi_internal): Ditto.
1729 (movti_interunit splitters): Remove
1730 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
1731 (movdi_interunit splitters): Ditto.
1732 * config/i386/constraints.md (Ye): Remove.
1733 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
1735 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1738 * config/aarch64/constraints.md (Usg): Limit to 31.
1741 2018-04-27 Jakub Jelinek <jakub@redhat.com>
1743 PR tree-optimization/85529
1744 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
1745 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
1746 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
1747 zero extension or masking of the MSB bit.
1748 (optimize_range_tests): Add FIRST_BB argument, pass it through
1749 to optimize_range_tests_var_bound.
1750 (maybe_optimize_range_tests, reassociate_bb): Adjust
1751 optimize_range_tests callers.
1753 2018-04-26 Richard Biener <rguenther@suse.de>
1754 Jakub Jelinek <jakub@redhat.com>
1756 * cgraph.h (symbol_table): Just declare debug method here.
1757 * symtab.c (symbol_table::debug): Define.
1759 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
1761 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
1763 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
1765 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
1766 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
1767 (*movdi_internal): Substitute Yi and Yj constraint with x
1768 and Ym and Yn constraint with y constraint. Update "isa"
1769 attribute and set "preferred_for_speed" attribute from
1770 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
1771 (*movsi_internal): Ditto.
1772 (*movdf_internal): Ditto.
1773 (*movsf_internal): Ditto.
1774 (*zero_extendsidi2): Ditto.
1775 * config/i386/sse.md (vec_set<mode>_0): Ditto.
1776 (sse2_loadld): Ditto.
1777 (*vec_extract<ssevecmodelower>_0): Ditto.
1778 (*vec_extractv4si_0_zext_sse4): Ditto.
1779 (vec_concatv2di): Ditto.
1780 (*vec_dup<mode>): Ditto.
1781 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
1782 * config/i386/constraints.md (Yi): Remove.
1787 2018-04-26 Nathan Sidwell <nathan@acm.org>
1789 * dumpfile.c (dump_open): New.
1790 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
1791 (dump_finish): Detect stdio/stderr by value not name.
1793 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
1795 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
1797 2018-04-26 Tom de Vries <tom@codesourcery.com>
1800 * config/nvptx/nvptx.c (verify_neutering_jumps)
1801 (verify_neutering_labels): New function
1802 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
1804 2018-04-26 Tom de Vries <tom@codesourcery.com>
1807 * config/nvptx/nvptx.c (needs_neutering_p): New function.
1808 (nvptx_single): Use needs_neutering_p to skip over insns that do not
1811 2018-04-26 Richard Biener <rguenther@suse.de>
1812 Tom de Vries <tom@codesourcery.com>
1815 * lto-streamer-out.c (output_function): Fixup loops if required to match
1816 discovery done in the reader.
1818 2018-04-26 Richard Biener <rguenther@suse.de>
1820 PR tree-optimization/85116
1821 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
1822 have a loop exit from the single latch predecessor. Remove
1823 case of header with just condition.
1824 (ch_base::copy_headers): Exclude infinite loops from any
1826 (pass_ch::execute): Record exits.
1828 2018-04-26 Richard Biener <rguenther@suse.de>
1830 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
1831 prologue cost vector and pass it to vect_get_load_cost.
1832 (vect_get_peeling_costs_all_drs): Likewise.
1833 (vect_peeling_hash_get_lowest_cost): Likewise.
1834 (vect_enhance_data_refs_alignment): Likewise.
1836 2018-04-26 Richard Biener <rguenther@suse.de>
1839 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
1840 checking of integer<->pointer conversions.
1841 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
1842 sign-/zero-extending pointer types.
1843 (expand_omp_for_static_chunk): Likewise.
1845 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
1846 Jean Lee <xiaoyur347@gmail.com>
1848 * config/mips/mips.c (mips_asan_shadow_offset): New function.
1849 (TARGET_ASAN_SHADOW_OFFSET): Define.
1850 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
1851 true for -fsanitize=address.
1853 2018-04-25 Mark Wielaard <mark@klomp.org>
1855 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
1858 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1860 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
1861 than "alu", remove explicit "memory" and "imm_disp" attributes.
1862 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
1865 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
1866 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
1867 gen_lowpart_no_emit.
1869 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
1872 * config/i386/i386.c (ix86_expand_builtin): Change memory
1873 operand to XI, extend p0 to Pmode.
1874 * config/i386/i386.md: Change unspec volatile and operand
1875 1 mode to XI, change operand 0 mode to P.
1877 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1879 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
1880 GET_MODE_MASK before any checking.
1881 (nds32_can_use_bset_p): Likewise.
1882 (nds32_can_use_btgl_p): Likewise.
1884 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1886 * config/nds32/nds32-doubleword.md: New define_split pattern for
1887 illegal register number.
1889 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1891 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
1893 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1895 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
1897 2018-04-25 Richard Biener <rguenther@suse.de>
1899 * lto-streamer.h (LTO_major_version): Bump to 8.
1901 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1903 * BASE-VER: Set to 9.0.0.
1905 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
1907 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
1908 in __abskf2 and __powikf2.
1910 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1913 * config/aarch64/constraints.md (Usg, Usj): New constraints.
1914 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
1915 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
1916 Use the above on operand 2. Reindent.
1917 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
1919 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1922 * common/config/i386/i386-common.c (ix86_handle_option): Don't
1924 * config/i386/i386.opt (mcet): Removed.
1925 * doc/install.texi: Remove -mcet documentation.
1926 * doc/invoke.texi: Likewise.
1928 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1931 * doc/install.texi: Remove -mcet from bootstrap-cet.
1933 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1936 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
1937 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
1941 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
1942 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
1943 containing a CONST_VECTOR.
1945 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
1947 * doc/install.texi: Update newlib dependency for nvptx.
1949 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1952 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
1953 instead of INTVAL when shifting x left.
1955 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
1957 PR tree-optimization/85478
1958 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
1959 vect_grouped_store_supported for single element vectors.
1961 2018-04-24 Richard Biener <rguenther@suse.de>
1964 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
1965 load cost increase to the case of non-constant step.
1967 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1970 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
1971 destination if any_malformed_asm.
1973 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
1976 * expr.c (store_field): In the bitfield case, if the value comes from
1977 a function call and is returned in registers by means of a PARALLEL,
1978 do not change the mode of the temporary unless BLKmode and VOIDmode.
1980 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
1982 PR rtl-optimization/85423
1983 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
1984 dependencies to debug insns when the previous insn is non-debug.
1986 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1988 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
1989 enums into a single definition.
1990 (fls): Fix predicates and printing.
1993 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1995 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
1996 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
1997 and short u6 immediate.
1998 (check_if_valid_sleep_operand): Remove.
1999 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
2001 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2003 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
2004 flag_always_save_lp condition.
2005 * config/nds32/nds32.opt (malways-save-lp): New option.
2007 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2009 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
2010 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
2011 * config/nds32/nds32.h
2012 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
2013 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
2015 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2017 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
2018 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
2020 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2021 Chung-Ju Wu <jasonwucj@gmail.com>
2023 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
2025 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
2026 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
2028 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2030 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
2032 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2034 * config/nds32/nds32-protos.h (nds32_data_alignment,
2035 nds32_local_alignment): Declare.
2036 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
2037 nds32_local_alignment): New functions.
2038 (TARGET_CONSTANT_ALIGNMENT): Define.
2039 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
2041 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2043 * config/nds32/nds32.c
2044 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
2045 (TARGET_MODES_TIEABLE_P): Likewise.
2047 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2049 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
2052 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
2053 Chung-Ju Wu <jasonwucj@gmail.com>
2055 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2056 for unaligned access.
2057 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2059 * config/nds32/nds32-intrinsic.md: Likewise.
2060 * config/nds32/nds32_intrinsic.h: Likewise.
2061 * config/nds32/nds32.h (nds32_builtins): Likewise.
2062 * config/nds32/nds32.opt (munaligned-access): New option.
2063 * config/nds32/nds32.c (nds32_asm_file_start): Display
2064 flag_unaligned_access status.
2066 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2068 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
2069 -mno-relax is present.
2070 * config/riscv/linux.h (LINK_SPEC): Ditto.
2072 2018-04-20 Martin Sebor <msebor@redhat.com>
2075 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
2077 (gimple_fold_builtin_stxcpy_chk): Same.
2078 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
2080 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
2083 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
2084 __powikf2 when long double is IEEE 128-bit.
2086 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2088 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
2089 step to make sure stack always aligned.
2091 2018-04-20 Carl Love <cel@us.ibm.com>
2094 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
2095 size check for arg0.
2097 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
2098 Tom de Vries <tom@codesourcery.com>
2101 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
2102 Emit insns for calls too.
2103 (nvptx_find_par): Always look for worker-level predecessor insn.
2104 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
2106 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
2107 (nvptx_process_pars): Propagate frames for calls.
2109 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
2112 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
2114 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
2115 (ix86_handle_option): Don't handle OPT_mibt.
2116 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
2118 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
2120 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
2121 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
2122 (ix86_target_macros): Define __CET__ with flag_cf_protection
2123 for -fcf-protection.
2124 * config/i386/i386.c (isa2_opts): Remove -mibt.
2125 * config/i386/i386.h (TARGET_IBT): Removed.
2126 (TARGET_IBT_P): Likewise.
2127 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
2128 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
2129 * config/i386/i386.opt (mcet): Update help message.
2132 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
2133 -mcet as an alias for -mshstk.
2135 2018-04-20 Richard Biener <rguenther@suse.de>
2138 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
2139 complexity by forcing a single use of the multiply operand.
2141 2018-04-20 Martin Jambor <mjambor@suse.cz>
2144 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
2145 recursion dependency to only apply to non-clones.
2147 2018-04-20 Martin Jambor <mjambor@suse.cz>
2150 * ipa-cp.c (create_specialized_node): Check that clones of
2151 self-recursive edges exist during IPA-CP.
2153 2018-04-19 Toon Moene <toon@moene.org>
2155 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
2158 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2160 PR tree-optimization/85467
2161 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
2162 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
2163 VECTOR_CST element to type.
2165 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2168 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
2169 * config/i386/i386.md (builtin_setjmp_setup): Removed.
2170 (builtin_longjmp): Likewise.
2171 (save_stack_nonlocal): New pattern.
2172 (restore_stack_nonlocal): Likewise.
2174 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2177 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2178 Replace ASM_OUTPUT_LABEL with fprintf.
2180 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2183 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2184 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
2185 * config/i386/i386-c.c (ix86_target_macros_internal): Also
2186 define __IBT__ and __SHSTK__ for -fcf-protection.
2187 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
2189 (ix86_trampoline_init): Likewise.
2190 (x86_output_mi_thunk): Likewise.
2191 (ix86_notrack_prefixed_insn_p): Likewise.
2192 (ix86_option_override_internal): Don't disallow -fcf-protection.
2193 * config/i386/i386.md (rdssp<mode>): Also enable for
2195 (incssp<mode>): Likewise.
2196 (nop_endbr): Likewise.
2197 * config/i386/i386.opt (mcet): Change help message to built-in
2201 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
2202 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
2203 enable CET built-in functions.
2205 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
2207 * common/config/i386/i386-common.c
2208 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
2209 OPTION_MASK_ISA_MOVDIRI_UNSET,
2210 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
2211 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
2212 * config.gcc (movdirintrin.h): New header.
2213 * config/i386/cpuid.h (bit_MOVDIRI,
2214 bit_MOVDIR64B): New bits.
2215 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
2217 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
2218 (VOID, PVOID, PCVOID)): New function types.
2219 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
2220 __builtin_ia32_directstoreu_u64,
2221 __builtin_ia32_movdir64b): New builtins.
2222 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
2223 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
2225 (ix86_valid_target_attribute_inner_p): Ditto.
2226 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
2227 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
2228 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
2229 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
2230 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
2231 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
2232 (movdiri<mode>, movdir64b_<mode>): New.
2233 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
2234 * config/i386/immintrin.h: Include movdirintrin.h.
2235 * config/i386/movdirintrin.h: New file.
2236 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
2238 2018-04-19 Richard Biener <rguenther@suse.de>
2241 * cfg.c (clear_bb_flags): When loop state says we have
2242 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
2244 2018-04-19 Richard Biener <rguenther@suse.de>
2246 PR tree-optimization/84737
2247 * tree-vect-data-refs.c (vect_copy_ref_info): New function
2248 copying restrict info.
2249 (vect_setup_realignment): Use it.
2250 * tree-vectorizer.h (vect_copy_ref_info): Declare.
2251 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
2252 the first DR to all generated stores.
2253 (vectorizable_load): Likewise for loads.
2255 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2257 PR tree-optimization/85446
2258 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
2259 the integral and pointer types to have the same precision.
2261 * doc/install.texi: Document --disable-cet being the default and
2264 2018-04-18 Martin Liska <mliska@suse.cz>
2266 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
2269 2018-04-18 Martin Liska <mliska@suse.cz>
2272 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2275 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2276 arguments if they are comparable.
2278 2018-04-18 Martin Liska <mliska@suse.cz>
2281 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2284 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2287 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
2290 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2291 ENDBR after calling __morestack.
2293 2018-04-18 David Malcolm <dmalcolm@redhat.com>
2296 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2297 by using gcc_base_ver to generate a gcc_driver_version, and use
2298 it when generating GCC_DRIVER_NAME.
2299 * configure: Regenerate.
2301 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2304 * config.gcc: Obsolete powerpc*-*-*spe*.
2306 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2309 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2310 (stabstr_D): Change type of unum from unsigned int to
2311 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
2314 2018-04-17 Jim Wilson <jimw@sifive.com>
2317 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2318 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2319 Set arg_pointer_offset after using pretend_args_size.
2321 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2323 PR rtl-optimization/85431
2324 * dse.c (record_store): Ignore zero width stores.
2327 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
2328 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2329 __builtin_stack_restore rather than after it.
2330 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2331 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2332 argument instead of virtual_dynamic_stack_rtx.
2334 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
2336 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2338 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2339 Add note to error message to explain internal mapping of overloaded
2340 built-in function name to non-overloaded built-in function name.
2341 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2344 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2347 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2348 where the inputs overlap with the output.
2350 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2352 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2353 (=v, v) alternative and explicit "memory" attribute.
2354 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
2355 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2357 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2358 "sselog1" type instead of "sselog".
2359 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2360 "sselog". Remove explicit "memory" attribute.
2361 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2362 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2364 (vec_extract_hi_v32hi): Merge all alternatives into one, use
2365 "sselog1" type instead of "sselog". Remove explicit "memory"
2367 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2368 use "sselog1" type instead of "sselog". Remove explicit "memory"
2370 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2371 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2373 (vec_extract_hi_v64qi): Merge all alternatives into one, use
2374 "sselog1" type instead of "sselog". Remove explicit "memory"
2376 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2377 use "sselog1" type instead of "sselog". Remove explicit "memory"
2381 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2384 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2387 2018-04-17 Martin Jambor <mjambor@suse.cz>
2390 * ipa-cp.c (create_specialized_node): Call
2391 expand_all_artificial_thunks if necessary.
2393 2018-04-17 Martin Liska <mliska@suse.cz>
2396 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2397 in message, remote space in between '_G' and '('.
2399 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2402 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2403 avx512f_vmcmp<mode>3<round_saeonly_name>,
2404 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2405 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2406 avx512f_rndscale<mode><round_saeonly_name>,
2407 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2408 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2409 Use %<iptr>2 instead of %2 for -masm=intel.
2410 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2411 avx512f_vcvttss2usi<round_saeonly_name>,
2412 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2414 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2415 avx512f_vcvttsd2usi<round_saeonly_name>,
2416 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2417 Use %q1 instead of %1 for -masm=intel.
2418 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2419 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2420 of %3 for -masm=intel.
2421 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2423 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2425 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2427 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2428 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2430 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2432 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2433 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2434 %g1 and one with %0 and %1.
2435 (avx512er_vmrcp28<mode><round_saeonly_name>,
2436 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2438 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2439 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2440 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2441 of %0 and %{%4%} for -masm=intel.
2442 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2443 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2444 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2445 order of %0 and %{%5%}%{z%} for -masm=intel.
2447 2018-04-17 Jan Hubicka <jh@suse.cz>
2450 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2452 2018-04-17 Martin Liska <mliska@suse.cz>
2455 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2456 for target_clone error message. Make default implementation
2457 clone to be a local declaration.
2458 (separate_attrs): Add new argument and check for an empty
2460 (expand_target_clones): Handle it.
2461 (ipa_target_clone): Make redirection just for target_clones
2464 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2465 Tom de Vries <tom@codesourcery.com>
2468 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2469 tiled basic blocks without omp continue statements.
2471 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2474 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2475 vec_extract expression as having side effects to make sure it gets
2478 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2481 * config/i386/i386.c (get_builtin_code_for_version): Check
2484 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2487 * gcc/config.gcc: Support "skylake".
2488 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2490 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2491 (processor_target_table): Add "skylake".
2492 (ix86_option_override_internal): Add "skylake".
2493 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2494 PROCESSOR_CANNONLAKE.
2495 (get_builtin_code_for_version): Fix priority for
2496 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2497 PROCESSOR_SKYLAKE-AVX512.
2498 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2499 (processor_type): Add PROCESSOR_SKYLAKE.
2501 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
2502 Jason Merrill <jason@redhat.com>
2505 * convert.c (convert_to_integer_1): Use direct recursion for
2506 enumeral types and types with a precision less than the number
2507 of bits in their mode.
2509 2018-04-16 Julia Koval <julia.koval@intel.com>
2512 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2513 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2515 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
2518 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2519 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2520 and -mno-direct-move.
2522 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
2525 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
2526 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
2527 Ensure negative shifts result in {0}.
2529 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
2531 PR rtl-optimization/79916
2532 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
2533 regs (if any) to define how to gnerate SD moves when LRA is in
2536 2018-04-13 Jakub Jelinek <jakub@redhat.com>
2538 PR rtl-optimization/85393
2539 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
2540 * except.c (expand_dw2_landing_pad_for_region): Make static.
2541 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2542 a label and unconditional jump to old_bb, rather than
2543 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2546 PR rtl-optimization/85376
2547 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2548 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2549 instead of a specific value.
2551 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2552 Bin Cheng <bin.cheng@arm.com>
2554 PR tree-optimization/82965
2555 PR tree-optimization/83991
2556 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2557 by_profile_only parameter.
2558 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2559 information if the loop was predicted to iterate too many times.
2560 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2562 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2565 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
2568 2018-04-13 Martin Liska <mliska@suse.cz>
2569 Jakub Jelinek <jakub@redhat.com>
2572 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
2573 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
2574 * builtins.c (expand_builtin_memory_copy_args): Use
2575 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
2576 handle dest_addr == pc_rtx.
2578 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
2581 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
2582 asked to not generate direct moves.
2583 (fix_trunc<mode>si2_stfiwx): Similar.
2584 (fix_trunc<mode>si2_internal): Similar.
2586 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2589 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
2590 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
2591 lookup if dest in some wider mode is known to be const0_rtx and
2592 if so, record permanent equivalence for it to be ZERO_EXTEND of
2593 the narrower mode destination.
2595 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2597 * lto-streamer-out.c (output_function): Revert 259346.
2598 * omp-expand.c (expand_oacc_for): Likewise.
2600 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
2602 PR rtl-optimization/85354
2603 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
2604 * sel-sched.c (sel_global_init): ... here.
2606 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
2609 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
2610 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2611 mode for PE-COFF targets.
2612 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
2613 (i386_pe_asm_lto_end): Likewise.
2614 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
2615 (TARGET_ASM_LTO_END): Likewise.
2616 * config/i386/winnt.c (saved_debug_info_level): New static variable.
2617 (i386_pe_asm_lto_start): New function.
2618 (i386_pe_asm_lto_end): Likewise.
2620 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2621 Richard Biener <rguenther@suse.de>
2624 * lto-streamer-out.c (output_function): Fix CFG loop state before
2626 * omp-expand.c (expand_oacc_for): Handle calls to internal
2627 functions like regular functions.
2629 2018-04-12 Richard Biener <rguenther@suse.de>
2632 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
2633 for the early LTO debug to properly generate references to it
2634 during DIE emission. Do not re-use that for the skeleton for
2636 (dwarf2out_early_finish): Likewise.
2638 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2641 * config/i386/sse.md
2642 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
2643 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
2644 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
2645 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
2646 and output is a reg, avoid creating invalid lowpart subreg, but
2647 instead split into a 512-bit move. Don't split if not AVX512VL,
2648 input is xmm16+ reg and output is a mem.
2649 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
2650 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
2651 xmm16+ reg and output is a mem.
2653 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2655 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
2656 also for flag_dwarf2_cfi_asm.
2658 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2660 PR rtl-optimization/85342
2661 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
2662 a bool scalar var inside of the loop instead. Don't try to update
2663 recog_data.operand after failed apply_change_group.
2665 2018-04-12 Tom de Vries <tom@codesourcery.com>
2668 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
2669 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
2670 array with flexible array member as array without given dimension.
2671 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
2672 argument for undefined param to true.
2674 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
2677 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2678 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
2679 from PowerPC section.
2680 * config/rs6000/sysv4.opt (mcall-): Improve help text.
2681 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
2682 help text that is too long.
2683 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
2684 help text that is too long.
2685 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
2686 help text that is too long.
2688 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
2690 * config/alpha/alpha.md (stack_probe_internal): Rename
2691 from "probe_stack". Update all callers.
2693 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2695 PR rtl-optimization/84566
2696 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
2697 sched_macro_fuse_insns.
2699 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2702 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
2703 (compute_block_dependences): ... from here.
2705 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2707 PR tree-optimization/85331
2708 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
2709 from int to HOST_WIDE_INT.
2711 2018-04-11 Martin Jambor <mjambor@suse.cz>
2714 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
2715 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
2716 not the same as the source val.
2717 (cgraph_edge_brings_value_p): New parameter.
2718 (gather_edges_for_value): Pass destination value to
2719 cgraph_edge_brings_value_p.
2720 (perhaps_add_new_callers): Likewise.
2721 (get_info_about_necessary_edges): Likewise and exclude values brought
2722 only by self-recursive edges.
2723 (create_specialized_node): Redirect only clones of self-calling edges.
2724 (+self_recursive_pass_through_p): New function.
2725 (find_more_scalar_values_for_callers_subset): Use it.
2726 (find_aggregate_values_for_callers_subset): Likewise.
2727 (known_aggs_to_agg_replacement_list): Removed.
2728 (decide_whether_version_node): Re-calculate known constants for all
2729 remaining context clones.
2731 2018-04-11 Richard Biener <rguenther@suse.de>
2734 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
2735 from early DWARF output.
2736 (dwarf2out_early_finish): Output line info unconditionally into
2737 early DWARF and add reference to it.
2739 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2742 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
2743 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
2744 other than V2DFmode using iptr mode attribute.
2745 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
2747 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2749 PR rtl-optimization/84659
2750 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
2752 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2755 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
2757 (output_loc_list): Pass address of a dummy size variable even in the
2758 locview handling loop.
2759 (index_location_lists): Add comment on why skip_loc_list_entry can't
2762 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
2765 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
2768 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
2771 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2772 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
2773 and -mstring-compare-inline-limit.
2775 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2778 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
2779 for stack clash protection in a register whenever we need it to be in
2782 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2784 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
2785 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
2787 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2790 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
2792 (mlong-double-): Ditto.
2793 * config/rs6000/sysv4.opt (msdata=): Ditto.
2794 (mtls-size=): Ditto.
2796 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
2798 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2799 erroneous entries for
2800 "vector int vec_ldl (int, long int *)", and
2801 "vector unsigned int vec_ldl (int, unsigned long int *)".
2802 Add comments and entries for
2803 "vector bool char vec_ldl (int, bool char *)",
2804 "vector bool short vec_ldl (int, bool short *)",
2805 "vector bool int vec_ldl (int, bool int *)",
2806 "vector bool long long vec_ldl (int, bool long long *)",
2807 "vector pixel vec_ldl (int, pixel *)",
2808 "vector long long vec_ldl (int, long long *)",
2809 "vector unsigned long long vec_ldl (int, unsigned long long *)".
2810 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
2811 type tree bool_long_long_type_node and correct definition of
2812 bool_V2DI_type_node to make reference to this new type tree.
2813 (rs6000_mangle_type): Replace erroneous reference to
2814 bool_long_type_node with bool_long_long_type_node.
2815 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
2816 comments to emphasize sign distinctions for char and int types and
2817 replace RS6000_BTI_bool_long constant with
2818 RS6000_BTI_bool_long_long constant. Also add comment to restrict
2819 use of RS6000_BTI_pixel.
2820 (bool_long_type_node): Remove this macro definition.
2821 (bool_long_long_type_node): New macro definition
2823 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2825 PR rtl-optimization/85300
2826 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
2827 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
2828 simplify_unary_operation fails.
2830 2018-04-10 Martin Liska <mliska@suse.cz>
2832 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
2833 cgraph_edge and ipa_ref.
2835 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2839 * config/i386/sse.md
2840 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
2841 computation of the VEC_MERGE selector from mask.
2842 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
2843 Fix decoding of the VEC_MERGE selector into mask.
2845 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2847 PR tree-optimization/85286
2848 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
2850 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2852 * final.c (final_1): Set insn_last_address as well as
2853 insn_current_address.
2855 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2858 * explow.c (emit_stack_probe): Call validize_mem on memory location
2859 before passing it to gen_probe_stack. Create address operand and
2860 legitimize it for the probe_stack_address case.
2862 2018-04-09 Jan Hubicka <jh@suse.cz>
2865 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
2866 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
2867 * tree.c (free_lang_data_in_type): Fix handling of binfos;
2869 (free_lang_data): Rebuild type inheritance graph.
2871 2018-04-09 Martin Sebor <msebor@redhat.com>
2873 * invoke.texi (-finline-small-functions): Mention other optimization
2875 (-findirect-inlining, -fpartial-inlining): Same.
2876 (-finline-functions-called-once): Same.
2877 (-freorder-blocks-and-partition): Same.
2879 2018-04-09 Jan Hubicka <jh@suse.cz>
2882 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
2883 jumps; choose last target that matches the criteria (i.e.
2884 no partition changes for non-crossing jumps).
2885 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
2886 support for redirecting crossing jumps to non-crossing.
2888 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
2890 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
2891 also for naked functions.
2893 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
2895 * config/arc/arc.md (add_shift): New pattern.
2896 (add_shift2): Likewise.
2897 (sub_shift): Likewise.
2898 (sub_shift_cmp0_noout): Likewise.
2899 (compare_si_ashiftsi): Likewise.
2900 (xbfu_cmp0_noout): New combine pattern.
2901 (xbfu_cmp0"): Likewise.
2902 (movsi_set_cc_insn): Place the predicable variant first.
2903 (commutative_binary_cmp0_noout): Remove clobber.
2904 (commutative_binary_cmp0): New pattern.
2905 (noncommutative_binary_cmp0): Likewise.
2906 (noncommutative_binary_cmp0_noout): Likewise.
2907 (noncommutative_binary_comparison_result_used): Removed.
2908 (rsub_cmp0): New pattern.
2909 (rsub_cmp0_noout): Likewise.
2910 (extzvsi): Changed, keep only meaningful variants.
2911 (SQH, SEZ): New iterators.
2912 (SQH_postfix): New mode attribute.
2913 (SEZ_prefix): New code attribute.
2914 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
2915 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
2916 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
2918 (noncommutative_operator): Check the availability of barrel
2921 2018-04-09 Richard Biener <rguenther@suse.de>
2923 PR tree-optimization/85284
2924 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
2925 Only use the niter constraining form of simple_iv when the exit
2928 2018-04-09 Tom de Vries <tom@codesourcery.com>
2931 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2932 (define_expand "*memory_barrier"): New define_expand.
2933 (define_insn "memory_barrier"): New insn.
2935 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2937 PR rtl-optimization/80463
2938 PR rtl-optimization/83972
2939 PR rtl-optimization/83480
2941 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
2942 correct producer for the insn.
2943 (tidy_control_flow): Fixup seqnos in case of debug insns.
2945 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2947 PR rtl-optimization/83913
2949 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
2950 different sched-times when merging exprs.
2952 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2954 PR rtl-optimization/83962
2956 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
2957 tidy_fallthru_edge and tidy_control_flow.
2959 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2961 PR rtl-optimization/83530
2963 * sel-sched.c (force_next_insn): New global variable.
2964 (remove_insn_for_debug): When force_next_insn is true, also leave only
2965 next insn in the ready list.
2966 (sel_sched_region): When the region wasn't scheduled, make another pass
2967 over it with force_next_insn set to 1.
2969 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2971 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
2973 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2974 for interrupt control.
2975 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2976 functions for interrupt control.
2977 * config/nds32/nds32-intrinsic.md: Likewise.
2978 * config/nds32/nds32_intrinsic.h: Likewise.
2979 * config/nds32/nds32.h (nds32_builtins): Likewise.
2981 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2983 * config/nds32/nds32.c (nds32_init_machine_status,
2984 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
2985 strict_aligned_p field.
2986 (nds32_expand_to_rtl_hook): New function.
2987 (TARGET_EXPAND_TO_RTL_HOOK): Define.
2988 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
2990 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2991 Chung-Ju Wu <jasonwucj@gmail.com>
2993 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
2994 * config/nds32/nds32-n7.md: New file.
2995 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
2996 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
2998 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
2999 * config/nds32/nds32.md (pipeline_model): Add n7.
3000 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
3001 * config/nds32/pipelines.md: Include n7 settings.
3003 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3004 Chung-Ju Wu <jasonwucj@gmail.com>
3006 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
3007 * config/nds32/nds32-e8.md: New file.
3008 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
3009 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
3011 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
3012 * config/nds32/nds32.md (pipeline_model): Add e8.
3013 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
3014 * config/nds32/pipelines.md: Include e8 settings.
3016 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3017 Chung-Ju Wu <jasonwucj@gmail.com>
3019 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
3020 * config/nds32/nds32-n8.md: New file.
3021 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
3022 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
3024 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
3025 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
3026 * config/nds32/nds32.md (pipeline_model): Add n8.
3027 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
3028 * config/nds32/pipelines.md: Include n8 settings.
3030 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3031 Chung-Ju Wu <jasonwucj@gmail.com>
3033 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
3034 * config/nds32/nds32-n9-2r1w.md: New file.
3035 * config/nds32/nds32-n9-3r2w.md: New file.
3036 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
3037 nds32_register_ports): New or modify for cpu n9.
3038 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
3040 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
3041 * config/nds32/nds32-utils.c: New file.
3042 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
3043 TARGET_MUL_SLOW): Define.
3044 * config/nds32/nds32.md (pipeline_model): New attribute.
3045 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
3046 New options that support cpu n9.
3047 * config/nds32/pipelines.md: Include n9 settings.
3048 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
3050 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3052 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
3053 information if necessary.
3054 (output_cond_branch_compare_zero): Likewise.
3055 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
3056 (nds32_target_alignment): Refine for alignment.
3057 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
3058 (FUNCTION_BOUNDARY): Modify.
3059 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
3061 * config/nds32/nds32.opt (malways-align, malign-functions): New.
3063 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3065 * config/nds32/constants.md (unspec_volatile_element): Add values for
3066 TLB operation and data prefetch.
3067 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3068 functions for TLB operation and data prefetch.
3069 * config/nds32/nds32-intrinsic.md: Likewise.
3070 * config/nds32/nds32_intrinsic.h: Likewise.
3071 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
3072 (nds32_print_operand): Likewise.
3073 * config/nds32/nds32.h (nds32_builtins): Likewise.
3075 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
3076 Andrew Pinski <pinsika@gcc.gnu.org>
3079 * match.pd: Use constant_boolean_node of correct type instead of
3080 boolean_true_node or boolean_false_node for simplifying
3081 pointer comparisons to zero.
3083 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3085 PR tree-optimization/80021
3086 * tree.c (verify_type_variant): Make error call in verify_variant_match
3087 translatable and remove final full stop.
3089 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3091 * config/nds32/constants.md (unspec_volatile_element): Add
3092 UNSPEC_VOLATILE_EH_RETURN.
3093 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
3094 nds32_output_stack_pop): Support dwarf exception handling process.
3095 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
3096 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
3097 exception handling process.
3098 (nds32_compute_stack_frame): Likewise.
3099 (nds32_return_addr_rtx): Likewise.
3100 (nds32_initial_elimination_offset): Likewise.
3101 (nds32_expand_prologue): Likewise.
3102 (nds32_expand_epilogue): Likewise.
3103 (nds32_dynamic_chain_address): New function.
3104 * config/nds32/nds32.h (machine_function): Add fields for dwarf
3106 (DYNAMIC_CHAIN_ADDRESS): Define.
3107 (EH_RETURN_DATA_REGNO): Define.
3108 (EH_RETURN_STACKADJ_RTX): Define.
3109 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
3110 patterns for dwarf exception handling.
3112 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3114 * config/nds32/nds32.h: Clean up obsolete macros.
3116 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3118 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3119 Add enum values for particular instructions.
3120 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3121 particular intrinsic functions.
3122 * config/nds32/nds32-intrinsic.md: Likewise.
3123 * config/nds32/nds32_intrinsic.h: Likewise.
3124 * config/nds32/nds32.h (nds32_builtins): Likewise.
3125 * config/nds32/nds32.md (type): Add pbsad and pbsada.
3126 (btst, ave): New patterns for particular instructions.
3128 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3130 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3131 Add enum values for atomic load/store and memory sync.
3132 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
3134 * config/nds32/nds32-intrinsic.md: Likewise.
3135 * config/nds32/nds32_intrinsic.h: Likewise.
3136 * config/nds32/nds32.h (nds32_builtins): Likewise.
3138 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3140 PR tree-optimization/85257
3141 * fold-const.c (native_encode_vector): If not all elts could fit
3142 and off is -1, return 0 rather than offset.
3143 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
3144 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
3145 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
3146 adjust buffer in native_interpret_expr call.
3148 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3150 * config/nds32/constants.md (unspec_volatile_element): Add cache
3151 control enum values.
3152 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
3153 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
3154 * config/nds32/nds32.c (nds32_cctl_names): New.
3155 (nds32_print_operand): Handle cache control register names.
3156 * config/nds32/nds32.h (nds32_builtins): New enum values.
3157 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
3159 * config/nds32/nds32.md (type): Add mmu.
3160 * config/nds32/pipelines.md (simple_insn): Add mmu.
3162 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3164 * config/nds32/nds32.md (type): Remove call.
3165 * config/nds32/pipelines.md (simple_insn): Likewise.
3167 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3169 * config/nds32/constants.md (unspec_volatile_element): Add
3170 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
3171 UNSPEC_VOLATILE_FMFCFG.
3172 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
3173 description for fmfcfg and fmfcsr.
3174 (bdesc_1arg): Add fmtcsr.
3175 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
3176 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
3177 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
3178 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
3179 unspec_fmfcfg): New patterns.
3180 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
3181 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
3182 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
3183 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
3184 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
3185 __nds32__fmfcfg): Define.
3187 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3189 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
3190 intrinsic register names.
3191 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
3192 intrinsic register enum values and macros.
3194 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3196 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
3197 for load/store addressing form.
3198 (nds32_print_operand_address): Likewise.
3200 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
3203 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
3204 based on LABEL_REF. Remove useless assertion.
3205 (pic_address_needs_scratch): Fix formatting.
3206 (sparc_legitimize_pic_address): Minor tweaks.
3207 (sparc_delegitimize_address): Adjust assertion accordingly.
3208 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
3209 into symbolic_operand.
3210 (movsi_high_pic_label_ref): Likewise.
3211 (movsi_lo_sum_pic_label_ref): Likewise.
3212 (movdi_pic_label_ref): Likewise.
3213 (movdi_high_pic_label_ref): Likewise.
3214 (movdi_lo_sum_pic_label_ref): Likewise.
3216 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
3218 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
3219 custom LIB_SPEC setup.
3221 2018-04-06 Ruslan Bukin <br@bsdpad.com>
3222 Kito Cheng <kito.cheng@gmail.com>
3224 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
3225 * config/riscv/freebsd.h: New.
3227 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3229 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
3230 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
3233 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3234 Kito Cheng <kito.cheng@gmail.com>
3236 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
3237 nds32_output_call, nds32_symbol_binds_local_p): New functions.
3238 * config/nds32/nds32-protos.h (nds32_output_call,
3239 nds32_output_return): Declare.
3240 * config/nds32/nds32.md: Refine all the call and return patterns.
3242 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3245 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
3246 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
3248 PR rtl-optimization/84872
3249 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
3250 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
3253 2018-04-06 Tamar Christina <tamar.christina@arm.com>
3255 * expr.c (copy_blkmode_to_reg): Revert 254862.
3256 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
3258 2018-04-06 Richard Biener <rguenther@suse.de>
3261 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
3262 after seeing a component reference with an adjacent field. Treat
3263 refs to arrays at struct end of external decls similar to
3264 refs to unconstrained commons.
3266 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3269 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
3270 look through SAVE_EXPRs with non-side-effects argument. Adjust
3272 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3275 2018-04-06 Richard Biener <rguenther@suse.de>
3278 * alias.c (find_base_term): New wrapper around find_base_term
3279 unwinding CSELIB_VAL_PTR changes.
3280 (find_base_term): Do not restore CSELIB_VAL_PTR during the
3283 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3285 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3287 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3288 constant definitions.
3289 ("nop"): lr 0,0 -> nopr r0
3290 ("nop_lr0", "nop_lr1"): New insn definitions.
3292 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3294 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3295 NDS32_V3PUSH_AVAILABLE_P macro.
3297 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
3298 Chung-Ju Wu <jasonwucj@gmail.com>
3300 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3301 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3302 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3303 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3304 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3305 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3306 * config/nds32/constraints.md: New constraints and checking for hard
3307 float configuration.
3308 * config/nds32/iterators.md: New mode iterator and attribute for hard
3309 float configuration.
3310 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3312 * config/nds32/nds32-fpu.md: New file.
3313 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3314 deal with hard float code generation.
3315 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3317 (abi_type, float_reg_number): New enum type.
3318 * config/nds32/nds32-predicates.c: New predicates for hard float.
3319 * config/nds32/nds32-protos.h: Declare functions for hard float.
3320 * config/nds32/nds32.c: Implementation for hard float configuration.
3321 * config/nds32/nds32.h: Definitions for hard float configuration.
3322 * config/nds32/nds32.md: Include hard float machine description and
3323 modify patterns for hard float configuration.
3324 * config/nds32/nds32.opt: New options for hard float configuration.
3325 * config/nds32/predicates.md: New predicates for hard float
3328 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3330 * common/config/nds32/nds32-common.c
3331 (nds32_option_optimization_table): Enable -mreleax-hint by default.
3333 2018-04-05 Jakub Jelinek <jakub@redhat.com>
3336 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3337 CONSTRUCTOR_ELT (ctor, ...)->value.
3339 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
3342 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3344 2018-04-05 Tom de Vries <tom@codesourcery.com>
3347 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3350 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3351 Kito Cheng <kito.cheng@gmail.com>
3353 * config/nds32/constraints.md (U33): Fine-tune checking condition.
3354 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3355 * config/nds32/nds32.h (nds32_16bit_address_type): Add
3356 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3358 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3359 Kito Cheng <kito.cheng@gmail.com>
3361 * config/nds32/constraints.md (Ufe): New memory constraint.
3362 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3363 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3364 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3366 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3367 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3369 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3371 * config/nds32/nds32.md: Use optimize_size in the condition for
3372 alu-shift instructions.
3374 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3376 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3378 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3380 * config/nds32/nds32.md (negsi2): Refine pattern.
3382 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3383 Chung-Ju Wu <jasonwucj@gmail.com>
3385 * config/nds32/iterators.md (shift_rotate): New code iterator.
3386 (shift): New code attribute.
3387 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3388 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3389 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3390 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3391 bit-wise operations.
3392 (andsi3, *andsi3): Ditto.
3393 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3394 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3395 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3396 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3397 nds32_ior_operand, nds32_xor_operand): New predicates.
3399 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3401 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3402 (addsi3, subsi3): ... this.
3404 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3406 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3408 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3410 * config/nds32/nds32.md: Adjust indention.
3412 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3414 * config/nds32/nds32.md (feature): New attribute.
3416 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3418 * config/nds32/nds32.md (subtype): New attribute.
3420 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
3423 * config/arm/arm-builtins.c (arm_expand_builtin): Change
3424 expansion to perform a bitwise AND of the argument followed by a
3425 boolean negation of the result.
3427 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3429 PR rtl-optimization/84878
3430 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3431 the basic block. Assert the use reference is not artificial and that
3432 it has an associated insn.
3434 2018-04-04 Michael Matz <matz@suse.de>
3436 * builtins.c (compute_objsize): Pass correct operand
3437 to array_at_struct_end_p.
3439 2018-04-04 Richard Biener <rguenther@suse.de>
3442 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3443 from contexts for DINFO_LEVEL_TERSE and below.
3445 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3447 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3448 resiter_operand condition.
3449 * config/nds32/nds32.md (*move<mode>): Ditto.
3451 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3452 Monk Chiang <sh.chiang04@gmail.com>
3454 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3456 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3458 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3460 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3461 Kito Cheng <kito.cheng@gmail.com>
3463 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3464 nds32_cond_code_str, output_cond_branch,
3465 output_cond_branch_compare_zero, nds32_expand_cbranch,
3466 nds32_expand_cstore, nds32_expand_movcc,
3467 nds32_output_cbranchsi4_equality_zero,
3468 nds32_output_cbranchsi4_equality_reg,
3469 nds32_output_cbranchsi4_equality_reg_or_const_int,
3470 nds32_output_cbranchsi4_greater_less_zero: New functions.
3471 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3472 nds32_expand_cstore, nds32_expand_movcc,
3473 nds32_output_cbranchsi4_equality_zero,
3474 nds32_output_cbranchsi4_equality_reg,
3475 nds32_output_cbranchsi4_equality_reg_or_const_int,
3476 nds32_output_cbranchsi4_greater_less_zero): Declare.
3477 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3478 nds32_rimm11s_operand): New predicates.
3479 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3480 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3483 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3485 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3486 * config/nds32/nds32.md: Ditto.
3487 * config/nds32/pipelines.md: Ditto.
3489 2018-04-04 Richard Biener <rguenther@suse.de>
3491 PR tree-optimization/85168
3492 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3493 propagating abnormals.
3495 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3497 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3499 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3500 Kito Cheng <kito.cheng@gmail.com>
3502 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3503 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3504 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3505 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3506 * config/nds32/nds32.md (sibcall_internal): New.
3507 (sibcall_register): Remove.
3508 (sibcall_immediate): Remove.
3509 (sibcall_value_internal): New.
3510 (sibcall_value_register): Remove.
3511 (sibcall_value_immediate): Remove.
3512 * config/nds32/predicates.md (nds32_general_register_operand): New.
3513 (nds32_call_address_operand): New.
3515 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3517 PR rtl-optimization/85167
3518 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3519 bb_defs if *split_p, instead preinitialize it to NULL.
3521 PR tree-optimization/85156
3522 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3523 evaluating the argument multiple times.
3525 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
3527 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
3529 (_mm_cvtpd_ps): Likewise.
3530 (_mm_cvttpd_epi32): Likewise.
3531 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
3532 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
3533 vector, pixel, and bool following altivec.h include.
3535 2018-04-03 Martin Sebor <msebor@redhat.com>
3537 * doc/extend.texi (Common Function Attributes): Clarify.
3538 (const attribute): Likewise.
3539 (pure attribute): Likewise.
3541 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3544 * config/i386/i386.c (ix86_expand_vector_set): Use
3545 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
3547 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
3549 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3550 instructions when changing rounding bits to preserve precision bits
3551 in the x87 control word.
3553 2018-04-03 Martin Liska <mliska@suse.cz>
3555 PR tree-optimization/82491
3556 * rtl.h (strip_offset_and_add): Replace += suboffset with
3557 poly_uint64 () + suboffset.
3559 2018-03-29 Martin Liska <mliska@suse.cz>
3560 Martin Jambor <mjambor@suse.cz>
3563 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
3564 param_type is not an integral or pointer type.
3566 2018-04-03 Richard Biener <rguenther@suse.de>
3568 * sese.h (recompute_all_dominators): Remove.
3570 2018-04-02 Martin Sebor <msebor@redhat.com>
3572 * doc/invoke.texi (-Wrestrict): Fix typos.
3574 2018-04-02 Jim Wilson <jimw@sifive.com>
3576 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
3577 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
3578 (<optab>di3, <optab>si3_extend): Likewise.
3579 (<optab>si3_mask, <optab>si3_mask_1): New.
3580 (<optab>di3_mask, <optab>di3_mask_1): New.
3581 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
3582 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
3583 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
3585 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
3587 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
3590 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
3592 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
3593 (nds32_canonicalize_comparison): New function.
3595 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3596 Kito Cheng <kito.cheng@gmail.com>
3597 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3599 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
3600 * config/nds32/constants.md (unspec_volatile_element): Add
3601 UNSPEC_VOLATILE_RELAX_GROUP.
3602 * config/nds32/nds32-relax-opt.c: New file.
3603 * config/nds32/nds32-predicates.c
3604 (nds32_symbol_load_store_p): New function.
3605 * config/nds32/nds32-protos.h
3606 (nds32_symbol_load_store_p): Declare function.
3607 (make_pass_nds32_relax_opt): Declare new rtl pass function.
3608 * config/nds32/nds32.c
3609 (nds32_register_pass): New function to register pass.
3610 (nds32_register_passes): New function to register passes.
3611 * config/nds32/nds32.md (relax_group): New pattern.
3612 * config/nds32/nds32.opt (mrelax-hint): New option.
3613 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
3615 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
3617 * config/nds32/t-nds32: Modify files dependency.
3619 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3621 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
3622 (PROFILE_HOOK): Define its implementation.
3624 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3626 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
3627 type and 32-bit size.
3629 2018-04-01 Jakub Jelinek <jakub@redhat.com>
3632 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
3633 (V_128_256): New mode iterator.
3634 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
3635 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
3636 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
3638 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
3639 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
3641 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
3644 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
3645 NaN inputs correctly.
3647 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
3650 * config/rs6000/vsx.md (??r): New mode attribute.
3651 (*vsx_mov<mode>_64bit): Use it.
3652 (*vsx_mov<mode>_32bit): Likewise.
3654 2018-03-30 Martin Sebor <msebor@redhat.com>
3656 PR tree-optimization/84818
3657 * builtins.c (check_access): Use warning_n.
3659 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3662 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
3664 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
3667 2018-03-30 Julia Koval <julia.koval@intel.com>
3670 * x86-tune.def (movx, partial_reg_dependency): Enable for
3673 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
3676 * lra-constraints.c (process_alt_operands): Move setting
3677 this_alternative_matches below.
3679 2018-03-29 Martin Liska <mliska@suse.cz>
3682 * doc/invoke.texi: Document how LTO works with debug info.
3683 Describe auto-load support of binutils. Mention 'x86-64'
3684 as valid option value of -march option.
3686 2018-03-29 Jakub Jelinek <jakub@redhat.com>
3688 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
3691 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
3692 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
3693 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
3696 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3699 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
3700 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
3701 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
3702 for RS6000_BTM_POWERPC64.
3703 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
3704 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
3705 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
3710 2018-03-28 Carl Love <cel@us.ibm.com>
3713 2017-09-27 Carl Love <cel@us.ibm.com>
3715 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
3716 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
3717 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
3720 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3722 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
3723 instead of __vector bool.
3724 (_mm_max_pu8): Likewise.
3725 (_mm_min_pi16): Likewise.
3727 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3730 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
3731 (DIVWEUO): Likewise.
3733 (DIVDEUO): Likewise.
3734 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
3735 DIVWEUO and DIVDEUO.
3736 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
3737 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
3738 (div_extend): Likewise.
3739 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
3741 (__builtin_divweuo): Likewise.
3742 (__builtin_divdeo): Likewise.
3743 (__builtin_divdeuo): Likewise.
3745 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3748 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
3749 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
3751 PR tree-optimization/82004
3752 * gimple-match-head.c (optimize_pow_to_exp): New function.
3753 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
3754 Don't fold to exp if optimize_pow_to_exp is false.
3756 2018-03-28 Martin Liska <mliska@suse.cz>
3759 * calls.c (initialize_argument_information): Fix trailing space.
3760 * common.opt: Fix typo and provide better explanation for
3761 -fsanitize-coverage option.
3762 * config/i386/i386.opt: Fix typo.
3764 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3765 Martin Liska <mliska@suse.cz>
3768 * gimplify.c (asan_poison_variable): Don't do the check for
3769 gimplify_omp_ctxp here.
3770 (gimplify_decl_expr): Do it here.
3771 (gimplify_target_expr): Likewise.
3773 2018-03-28 Martin Liska <mliska@suse.cz>
3776 * config/i386/i386.c (ix86_function_arg_advance): Do not call
3777 chkp_type_bounds_count if MPX is not enabled.
3779 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
3781 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
3783 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
3786 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
3787 function to create the function decl for complex long double
3788 multiply and divide for -mabi=ieeelongdouble.
3789 (init_float128_ieee): Call it.
3791 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
3794 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
3795 -fcf-protection=branch -mibt.
3796 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
3798 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3801 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
3803 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
3806 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
3807 extern array declarations.
3809 2018-03-27 Richard Biener <rguenther@suse.de>
3812 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
3813 explicit single_use checks.
3815 2018-03-27 Richard Biener <rguenther@suse.de>
3817 PR tree-optimization/85082
3818 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
3821 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3823 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
3824 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
3825 Turn on fasynchronous-unwind-tables and funwind-tables.
3827 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
3830 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
3831 (*bmi_blsr_<mode>_ccz): Ditto.
3833 2018-03-26 Tom de Vries <tom@codesourcery.com>
3835 PR tree-optimization/85063
3836 * omp-general.c (offloading_function_p): New function. Factor out
3838 * omp-offload.c (pass_omp_target_link::gate): ... here.
3839 * omp-general.h (offloading_function_p): Declare.
3840 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
3841 with attribute omp declare target for offloading functions.
3843 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
3845 PR tree-optimization/84005
3846 * tree-data-ref.h (get_base_for_alignment): Declare.
3847 * tree-data-ref.c (get_base_for_alignment_1): New function.
3848 (get_base_for_alignment): Likewise.
3849 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
3850 get_base_for_alignment to find a suitable base object, instead
3851 of always using drb->base_address.
3853 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3856 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
3857 known size by default.
3859 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
3862 * lra-constraints.c (process_alt_operands): Don't match BLKmode
3863 and non BLKmode operands.
3865 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3868 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
3869 Clean up attributes.
3871 2018-03-23 Richard Biener <rguenther@suse.de>
3874 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
3875 we are going to emit early debug for LTO.
3877 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3880 * function.c (match_asm_constraints_1): Don't optimize if input
3881 doesn't satisfy general_operand predicate for output's mode.
3884 * alias.c (write_dependence_p): Don't require for x_canonicalized
3885 non-VOIDmode if x has VOIDmode.
3888 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
3889 just don't try to optimize it rather than assert it never happens.
3891 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3893 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
3894 macro expansions for definition of ST_INTERNAL_<mode> and
3895 LD_INTERNAL_<mode> builtins.
3896 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
3898 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
3900 (altivec_expand_st_builtin): Likewise.
3901 (altivec_expand_builtin): Remove calls to deleted functions.
3902 (rs6000_address_for_altivec): Delete this function.
3903 * config/rs6000/vector.md: Remove expands for
3904 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
3906 2018-03-22 Sudakshina Das <sudi.das@arm.com>
3909 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
3910 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
3911 re-computing once computed.
3912 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
3913 (arm_init_machine_status): Initialize
3914 machine->static_chain_stack_bytes.
3916 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3919 * doc/extend.texi: Add four new prototypes for vec_ld.
3920 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
3921 definitions for more logical presentation.
3922 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
3923 entries for V1TI variants of __builtin_altivec_ld builtin.
3924 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
3925 handling of V1TI variant of LVX icode pattern.
3926 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
3927 (rs6000_gimple_fold_builtin): Likewise.
3928 (altivec_init_builtins): Add code to define
3929 __builtin_altivec_lvx_v1ti function.
3931 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3934 * function.c (match_asm_constraints_1): Don't do the optimization
3935 if input isn't a REG, SUBREG, MEM or constant.
3937 2018-03-22 Tom de Vries <tom@codesourcery.com>
3939 PR tree-optimization/84956
3940 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
3941 bb_has_abnormal_pred.
3943 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3946 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
3947 DECL_INITIAL (decl) to decl at the end.
3948 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
3951 2018-03-21 Joseph Myers <joseph@codesourcery.com>
3953 * doc/extend.texi (__builtin_tgmath): Document when complex
3954 integer types are treated as _Complex _Float64.
3956 2018-03-21 Tom de Vries <tom@codesourcery.com>
3958 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
3960 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3962 PR tree-optimization/84960
3963 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
3964 if it is ENTRY block, move them into single succ of ENTRY in that case.
3966 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
3968 PR tree-optimization/84811
3969 * poly-int.h (poly_span_traits): Remove the T3 parameter and
3970 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
3971 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
3972 (known_subrange_p): Update accordingly. Cast each value involved
3973 in the size comparison, rather than casting the result of the
3976 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3978 PR tree-optimization/84982
3979 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
3980 by flipping the least significant bit rather than all bits from
3981 bitpos to bitpos + bitsize - 1.
3983 2018-03-21 Nathan Sidwell <nathan@acm.org>
3985 * doc/extend.texi (Deprecated Features): Remove mention of
3986 long-deleted deprecations.
3988 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3991 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
3992 * configure: Regenerate.
3994 2018-03-21 Tom de Vries <tom@codesourcery.com>
3996 PR tree-optimization/83126
3997 * tree-parloops.c (num_phis): New function.
3998 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
4000 2018-03-21 Nathan Sidwell <nathan@acm.org>
4002 * doc/extend.texi (Deprecated Features): Update deprecated flags,
4003 mention anon-struct/union members and trailing attributes.
4005 2018-03-21 Bin Cheng <bin.cheng@arm.com>
4007 PR tree-optimization/84969
4008 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
4009 builtin memset partitions if they set different rhs values.
4011 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4013 PR rtl-optimization/84989
4014 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
4015 VEC_DUPLICATE with scalar result mode.
4017 2018-03-21 Martin Liska <mliska@suse.cz>
4020 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
4021 not intended return statement.
4023 2018-03-21 Martin Liska <mliska@suse.cz>
4026 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
4027 (chkp_find_bound_slots_1): Limit number of iterations.
4029 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
4032 * Minor grammar fixes for x86 options.
4034 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4037 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
4038 holding REG_CFA_RESTORE notes, instead turn them into a USE.
4040 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
4043 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
4044 (altivec_lvx_<mode>_1op): Likewise.
4045 (altivec_stvx_<mode>_2op): Likewise.
4046 (altivec_stvx_<mode>_1op): Likewise.
4047 (altivec_lvx_<VM2:mode>): New define_expand.
4048 (altivec_stvx_<VM2:mode>): Likewise.
4049 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
4050 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4051 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
4052 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4053 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
4054 (rs6000_gen_lvx): Likewise.
4055 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
4056 (altivec_expand_stv_builtin): Likewise.
4057 (altivec_expand_builtin): Likewise.
4058 * config/rs6000/vector.md: Likewise.
4060 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4063 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
4066 2018-03-20 Richard Biener <rguenther@suse.de>
4069 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
4070 sign-conversions as zero, fall back to standard scalar_stmt
4073 2018-03-20 Martin Liska <mliska@suse.cz>
4076 * predict.c (rebuild_frequencies): Handle case when we have
4077 PROFILE_ABSENT, but flag_guess_branch_prob is false.
4079 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4082 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
4083 flag_section_anchors.
4084 * varasm.c (use_blocks_for_decl_p): Remove hack for
4085 dw2_force_const_mem.
4088 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
4090 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
4091 be created, use lowpart_subreg of operands[0] rather than operands[0]
4093 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
4094 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
4095 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
4096 and n constraint instead of aarch64_shift_imm_di and Usd.
4097 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
4098 (*aarch64_<optab>_reg_minus<mode>3): ... this.
4100 2018-03-20 Sudakshina Das <sudi.das@arm.com>
4103 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
4104 to favor GPR over NEON registers.
4105 (<shift>di3_neon): Likewise.
4107 2018-03-20 Tom de Vries <tom@codesourcery.com>
4110 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
4111 (nvptx_process_pars): Emit bar.sync asap and alap.
4113 2018-03-20 Tom de Vries <tom@codesourcery.com>
4116 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
4117 seen_label if seen_label is already set.
4119 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4122 * config/i386/i386.c (fold_builtin_cpu): For features above 31
4123 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
4124 Use 1U instead of 1. Formatting fixes.
4127 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
4128 instead of TREE_TYPE (s1) for the return value.
4130 2018-03-19 Jakub Jelinek <jakub@redhat.com>
4132 PR tree-optimization/84946
4133 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
4134 bitsize + bitsize in poly_uint64 rather than poly_int64.
4137 * dwarf2asm.c: Include fold-const.c.
4138 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
4139 of decl rather than decl itself.
4141 PR rtl-optimization/84643
4142 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
4144 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
4147 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
4148 calling assemble_variable.
4150 2018-03-19 Sudakshina Das <sudi.das@arm.com>
4153 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
4154 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
4156 2018-03-19 Jim Wilson <jimw@sifive.com>
4159 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
4160 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
4161 (riscv_first_stack_step): Likewise.
4162 (riscv_option_override): Use STACK_BOUNDARY instead of
4164 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
4166 (BIGGEST_ALIGNMENT): Set to 128.
4167 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
4168 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
4171 2018-03-19 Richard Biener <rguenther@suse.de>
4173 PR tree-optimization/84933
4174 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
4175 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
4177 2018-03-19 Richard Biener <rguenther@suse.de>
4179 PR tree-optimization/84859
4180 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
4181 (cond_if_else_store_replacement): Perform sinking operation on
4182 single-store BBs regardless of MAX_STORES_TO_SINK setting.
4183 Generalize what a BB with a single eligible store is.
4185 2018-03-19 Richard Biener <rguenther@suse.de>
4187 PR tree-optimization/84929
4188 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
4189 chrec_is_positive against non-chrec arg.
4191 2018-03-19 Tamar Christina <tamar.christina@arm.com>
4194 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
4196 2018-03-18 Martin Liska <mliska@suse.cz>
4198 PR rtl-optimization/84635
4199 * regrename.c (build_def_use): Use matches_mode only when
4202 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
4204 PR tree-optimization/84913
4205 * tree-vect-loop.c (vectorizable_reduction): Don't try to
4206 vectorize chains of COND_EXPRs.
4208 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4210 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
4212 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4214 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
4216 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4218 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
4220 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4221 Kito Cheng <kito.cheng@gmail.com>
4223 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
4224 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
4225 (nds32_adjust_reg_alloc_order): New function.
4226 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
4228 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
4230 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
4231 nds32_print_operand, nds32_print_operand_address): Use
4232 HOST_WIDE_INT_PRINT_DEC instead.
4234 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4236 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
4238 2018-03-17 Jakub Jelinek <jakub@redhat.com>
4241 * config/i386/i386.c (initial_ix86_tune_features,
4242 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
4244 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
4245 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
4246 rather than 1u << ix86_tune. Formatting fix.
4247 (ix86_option_override_internal): Change ix86_arch_mask from
4248 unsigned int to unsigned HOST_WIDE_INT, initialize to
4249 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
4250 (ix86_function_specific_restore): Likewise.
4252 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4255 * postreload.c (reload_combine_recognize_pattern): Perform
4256 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
4257 truncate_int_for_mode the result for the destination's mode.
4260 * hsa-gen.c (mem_type_for_type): Fix comment typo.
4261 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
4263 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4266 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
4269 * lra-assigns.c (lra_split_hard_reg_for): Don't use
4270 regno_allocno_class_array and sorted_pseudos.
4271 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
4272 insns where regno is used.
4274 2018-03-16 Martin Liska <mliska@suse.cz>
4277 * multiple_target.c (create_dispatcher_calls): Redirect
4278 reference in the symbol table.
4280 2018-03-16 Martin Liska <mliska@suse.cz>
4283 * multiple_target.c (create_dispatcher_calls): Redirect also
4286 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4290 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4291 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4292 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4294 2018-03-16 Julia Koval <julia.koval@intel.com>
4296 * doc/invoke.texi (Skylake Server): Add CLWB.
4297 Cannonlake): Remove CLWB.
4299 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4301 PR tree-optimization/84841
4302 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4304 (FLOAT_ONE_CONST_TYPE): Define.
4305 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4306 (sort_by_operand_rank): Put entries with higher constant_type last
4307 rather than first to match comments.
4309 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
4311 * config/nios2/nios2.md (movsi_internal): Fix thinko in
4314 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4317 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4318 check and use address of target always.
4320 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
4323 * config/i386/i386.c (indirect_thunk_needed): Update comments.
4324 (indirect_thunk_bnd_needed): Likewise.
4325 (indirect_thunks_used): Likewise.
4326 (indirect_thunks_bnd_used): Likewise.
4327 (indirect_return_needed): New.
4328 (indirect_return_bnd_needed): Likewise.
4329 (output_indirect_thunk_function): Add a bool argument for
4331 (output_indirect_thunk_function): Don't generate alias for
4332 function return thunk.
4333 (ix86_code_end): Call output_indirect_thunk_function to generate
4334 function return thunks.
4335 (ix86_output_function_return): Set indirect_return_bnd_needed
4336 and indirect_return_needed instead of indirect_thunk_bnd_needed
4337 and indirect_thunk_needed.
4339 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
4341 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4342 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4343 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4345 2018-03-15 David Malcolm <dmalcolm@redhat.com>
4346 Paul Hua <paul.hua.gm@gmail.com>
4349 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4351 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
4353 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4354 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4357 2018-03-15 Tamar Christina <tamar.christina@arm.com>
4360 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4361 instead of GET_MODE_SIZE when comparing Units.
4363 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4366 * varasm.c (hash_section): Return an unchangeble hash value
4367 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4368 Return !aarch64_can_use_per_function_literal_pools_p ().
4370 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4373 * optabs.c (emit_conditional_move): Pass address of cmode's copy
4374 rather than address of cmode as last argument to prepare_cmp_insn.
4376 2018-03-15 Julia Koval <julia.koval@intel.com>
4378 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4379 F_AVX512VNNI, F_AVX512BITALG): New.
4381 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4384 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
4385 insn for floating-point loads and stores.
4387 2018-03-14 Carl Love <cel@us.ibm.com>
4389 * config/rs6000/rs6000-c.c: Add macro definitions for
4390 ALTIVEC_BUILTIN_VEC_PERMXOR.
4391 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4392 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4393 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4394 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4396 * config/doc/extend.texi: Add prototypes for vec_permxor.
4398 2018-03-14 David Malcolm <dmalcolm@redhat.com>
4401 * diagnostic-show-locus.c (class layout_point): Convert m_line
4402 from int to linenum_type.
4403 (line_span::comparator): Use linenum "compare" function when
4404 comparing line numbers.
4405 (test_line_span): New function.
4406 (layout_range::contains_point): Convert param "row" from int to
4408 (layout_range::intersects_line_p): Likewise.
4409 (layout::will_show_line_p): Likewise.
4410 (layout::print_source_line): Likewise.
4411 (layout::should_print_annotation_line_p): Likewise.
4412 (layout::print_annotation_line): Likewise.
4413 (layout::print_leading_fixits): Likewise.
4414 (layout::annotation_line_showed_range_p): Likewise.
4415 (struct line_corrections): Likewise for field m_row.
4416 (line_corrections::line_corrections): Likewise for param "row".
4417 (layout::print_trailing_fixits): Likewise.
4418 (layout::get_state_at_point): Likewise.
4419 (layout::get_x_bound_for_row): Likewise.
4420 (layout::print_line): Likewise.
4421 (diagnostic_show_locus): Likewise for locals "last_line" and
4423 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4424 * input.c (selftest::test_linenum_comparisons): New function.
4425 (selftest::input_c_tests): Call it.
4426 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4427 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4428 * selftest.h (ASSERT_GT): New macro.
4429 (ASSERT_GT_AT): New macro.
4430 (ASSERT_LT): New macro.
4431 (ASSERT_LT_AT): New macro.
4433 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4435 PR rtl-optimization/84780
4436 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4438 2018-03-14 Martin Liska <mliska@suse.cz>
4440 * tree.c (record_node_allocation_statistics): Use
4441 get_stats_node_kind.
4442 (get_stats_node_kind): New function extracted from
4443 record_node_allocation_statistics.
4444 (free_node): Use get_stats_node_kind.
4446 2018-03-14 Richard Biener <rguenther@suse.de>
4448 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4449 that the value-set of ANTIC_IN doesn't grow.
4452 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4454 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4455 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4457 2018-03-14 Julia Koval <julia.koval@intel.com>
4459 * config.gcc (icelake-client, icelake-server): New.
4461 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4462 (initial_ix86_arch_features): Ditto.
4463 (PTA_SKYLAKE): Add SGX.
4464 (PTA_ICELAKE): Remove.
4465 (PTA_ICELAKE_CLIENT): New.
4466 (PTA_ICELAKE_SERVER): New.
4467 (ix86_option_override_internal): Split up icelake on icelake client and
4469 (get_builtin_code_for_version): Ditto.
4470 (fold_builtin_cpu): Ditto.
4471 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4472 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4473 * config/i386/i386.h (processor_type): Ditto.
4474 * doc/invoke.texi: Ditto.
4476 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4479 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4480 INTEGER_CST offset, add it together with bitpos / 8 and
4481 sign extend based on POINTER_SIZE.
4485 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4488 * config/i386/constraints.md (Yc): New register constraint.
4489 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4490 Use Yc constraint for alternative 2 of operand 0. Remove
4491 preferred_for_speed attribute.
4493 2018-03-14 Richard Biener <rguenther@suse.de>
4495 PR tree-optimization/84830
4496 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4497 with the old one to avoid oscillations.
4499 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
4502 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4504 (assign_by_spills): Return a flag of reload assignment failure.
4505 Do not process the reload assignment failures. Do not spill other
4506 reload pseudos if they has the same reg class. Update n if
4508 (lra_assign): Add a return arg. Set up from the result of
4509 assign_by_spills call.
4510 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4511 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4512 usage_insns if it is not NULL.
4513 (spill_hard_reg_in_range): New function.
4514 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4515 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4516 function prototypes.
4517 (lra_assign): Change prototype.
4518 * lra.c (lra): Add code to deal with fails by splitting hard reg
4521 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
4523 * config/riscv/riscv.opt (mrelax): New option.
4524 * config/riscv/riscv.c (riscv_file_start): Emit ".option
4525 "norelax" when riscv_mrelax is disabled.
4526 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
4528 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4531 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
4532 reassociation for int modes.
4534 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4536 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
4537 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
4539 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
4540 * config/aarch64/aarch64-sve.md
4541 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4542 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4543 (*extend<mode><Vwide>2): Rename to...
4544 (aarch64_sve_extend<mode><Vwide>2): ...this.
4545 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4546 renaming the old pattern to...
4547 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
4549 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4550 define_expand, renaming the old pattern to...
4551 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4552 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4553 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4554 account when deciding which SVE instruction the optab should use.
4555 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4557 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4559 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4560 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4561 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4562 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4563 (tlsdesc_small_<mode>): Turn a define_expand and use
4564 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
4565 (tlsdesc_small_advsimd_<mode>): ...this.
4566 (tlsdesc_small_sve_<mode>): New pattern.
4568 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4570 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
4571 (UNSPEC_UMUL_HIGHPART): New constants.
4572 (MUL_HIGHPART): New int iteraor.
4573 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
4574 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
4576 (*<su>mul<mode>3_highpart): New define_insn.
4578 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4581 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4584 2018-03-13 Martin Liska <mliska@suse.cz>
4587 * (sem_item_optimizer::sem_item_optimizer): Initialize new
4589 (sem_item_optimizer::~sem_item_optimizer): Release it.
4590 (sem_item_optimizer::merge_classes): Register variable aliases.
4591 (sem_item_optimizer::fixup_pt_set): New function.
4592 (sem_item_optimizer::fixup_points_to_sets): Likewise.
4593 * ipa-icf.h: Declare new variables and functions.
4595 2018-03-13 Jakub Jelinek <jakub@redhat.com>
4598 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
4599 integer_pow2p@2 and test integer_pow2p in condition.
4600 (A < 0 ? C : 0): Similarly for @1.
4603 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
4604 characters starting at p contain '\0' character, don't look beyond
4608 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
4609 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
4612 * reg-stack.c (change_stack): Change update_end var from int to
4613 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
4614 also call set_block_for_insn on the newly added insns and rescan.
4617 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
4618 on the last operand.
4621 * tree.c (stabilize_reference_1): Return save_expr (e) for
4622 STATEMENT_LIST even if it doesn't have side-effects.
4624 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
4626 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
4628 2018-03-12 Renlin Li <renlin.li@arm.com>
4630 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
4631 aarch64_output_scalar_simd_mov_immediate.
4633 2018-03-12 Martin Sebor <msebor@redhat.com>
4635 PR tree-optimization/83456
4636 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
4637 for perfectly overlapping calls to memcpy.
4638 (gimple_fold_builtin_memory_chk): Same.
4639 (gimple_fold_builtin_strcpy): Handle no-warning.
4640 (gimple_fold_builtin_stxcpy_chk): Same.
4641 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
4643 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4645 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
4646 parameter. Use it for SFmode.
4647 (rs6000_function_arg_advance_1): Adjust.
4648 (rs6000_function_arg): Adjust.
4649 (rs6000_gimplify_va_arg): Pass false for that new parameter.
4651 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4653 PR rtl-optimization/84169
4654 PR rtl-optimization/84780
4655 * combine.c (can_combine_p): Check for a 2-insn combination whether
4656 the destination register is used between the two insns, too.
4658 2018-03-12 Richard Biener <rguenther@suse.de>
4660 PR tree-optimization/84803
4661 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
4662 for refs DR analysis didn't process.
4664 2018-03-12 Richard Biener <rguenther@suse.de>
4666 PR tree-optimization/84777
4667 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
4668 force-vectorize loops ignore whether we are optimizing for size.
4670 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
4672 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
4673 (TARGET_MD_ASM_ADJUST): Define.
4675 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
4676 Kito Cheng <kito.cheng@gmail.com>
4677 Chung-Ju Wu <jasonwucj@gmail.com>
4679 * config/nds32/nds32.c (nds32_compute_stack_frame,
4680 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
4681 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
4682 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
4683 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
4684 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
4685 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
4686 * config/nds32/nds32.md (prologue, epilogue): Use macro
4687 NDS32_V3PUSH_AVAILABLE_P to do checking.
4689 2018-03-11 Jakub Jelinek <jakub@redhat.com>
4692 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
4693 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
4694 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
4695 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
4696 addition of most attributes on !orig_type_die or the attribute not
4697 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
4699 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4700 Chung-Ju Wu <jasonwucj@gmail.com>
4702 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
4704 * config/nds32/nds32.opt (mvh): New option.
4706 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4707 Chung-Ju Wu <jasonwucj@gmail.com>
4709 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
4711 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
4712 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
4715 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4716 Chung-Ju Wu <jasonwucj@gmail.com>
4718 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
4720 * config/nds32/nds32-multiple.md (strlensi): New pattern.
4721 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
4723 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
4724 Kito Cheng <kito.cheng@gmail.com>
4725 Chung-Ju Wu <jasonwucj@gmail.com>
4727 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
4728 UNSPEC_FFMISM and UNSPEC_FLMISM.
4729 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
4730 for ffb, ffmism and flmism.
4731 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
4732 (unspec_ffmism): Ditto.
4733 (unspec_flmism): Ditto.
4734 (nds32_expand_builtin_impl): Check if string extension is available.
4735 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
4736 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
4738 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
4741 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4744 * lra-assigns.c (assign_by_spills): Return a flag of reload
4745 assignment failure. Do not process the reload assignment
4746 failures. Do not spill other reload pseudos if they has the same
4748 (lra_assign): Add a return arg. Set up from the result of
4749 assign_by_spills call.
4750 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4751 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4752 usage_insns if it is not NULL.
4753 (spill_hard_reg_in_range): New function.
4754 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4755 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4756 function prototypes.
4757 (lra_assign): Change prototype.
4758 * lra.c (lra): Add code to deal with fails by splitting hard reg
4761 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
4764 * config/i386/i386.opt: Replace Enforcment with Enforcement.
4766 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
4769 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
4770 (dw_val_node): Add val_symbolic_view.
4771 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
4772 (symview_upper_bound): New.
4773 (new_line_info_table): Initialize symviews_since_reset.
4774 (dwarf2out_source_line): Count symviews_since_reset and set
4775 symview_upper_bound.
4776 (dw_val_equal_p): Handle symview.
4777 (add_AT_symview): New.
4778 (print_dw_val): Handle symview.
4779 (attr_checksum, attr_checksum_ordered): Likewise.
4780 (same_dw_val_p, size_of_die): Likewise.
4781 (value_format, output_die): Likewise.
4782 (add_high_low_attributes): Use add_AT_symview for entry_view.
4783 (dwarf2out_finish): Reset symview_upper_bound, clear
4786 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
4789 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
4790 Add strict argument and use it.
4791 (rs6000_split_multireg_move): Update for new strict argument.
4792 (mem_operand_gpr): Disallow all non-offsettable addresses.
4793 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
4795 2018-03-09 Jakub Jelinek <jakub@redhat.com>
4798 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
4799 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
4800 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
4803 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
4804 decl, use remap_type if we want to use the type.
4806 2018-03-09 Martin Sebor <msebor@redhat.com>
4808 PR tree-optimization/84526
4809 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4811 (builtin_access::generic_overlap): Be prepared to handle non-array
4814 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
4816 PR rtl-optimization/84682
4817 * lra-constraints.c (process_address_1): Check is_address flag
4818 for address constraints.
4819 (process_alt_operands): Likewise.
4820 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
4821 preprocess_constraints.
4822 * recog.h (preprocess_constraints): Add oploc parameter.
4824 * recog.c (preprocess_constraints): Test address_operand for
4825 CT_ADDRESS constraints.
4827 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4830 * lra-assigns.c (assign_by_spills): Return a flag of reload
4831 assignment failure. Do not process the reload assignment
4832 failures. Do not spill other reload pseudos if they has the same
4834 (lra_assign): Add a return arg. Set up from the result of
4835 assign_by_spills call.
4836 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4837 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4838 usage_insns if it is not NULL.
4839 (spill_hard_reg_in_range): New function.
4840 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4841 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4842 function prototypes.
4843 (lra_assign): Change prototype.
4844 * lra.c (lra): Add code to deal with fails by splitting hard reg
4847 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4850 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
4851 Accept complain bool parameter. Only emit errors if it is true.
4852 (arm_parse_cpu_option_name): Likewise.
4853 (arm_target_thumb_only): Adjust callers of the above.
4854 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
4855 prototype to take a default true bool parameter.
4856 (arm_parse_arch_option_name): Likewise.
4858 2018-03-09 David Malcolm <dmalcolm@redhat.com>
4859 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
4863 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
4864 * configure: Regenerate.
4865 * configure.ac ("linker --version-script option"): New.
4866 ("linker soname option"): New.
4868 2018-03-09 Richard Biener <rguenther@suse.de>
4870 PR tree-optimization/84775
4871 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
4872 immediate uses of predicate stmts and mark them modified.
4875 PR tree-optimization/84178
4876 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4878 (version_loop_for_if_conversion): Delay update_ssa call.
4879 (tree_if_conversion): Delay update_ssa until after predicate
4882 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
4885 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
4886 when the function accesses prior frames.
4888 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4891 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
4892 gen_llsym, otherwise call maybe_gen_llsym.
4895 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
4896 has ',' character inside of it.
4898 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4901 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
4902 as clobbering CC_REGNUM.
4904 2018-03-08 Richard Biener <rguenther@suse.de>
4907 * tree-scalar-evolution.c: Include tree-into-ssa.h.
4908 (follow_copies_to_constant): Do not follow SSA names registered
4911 2018-03-08 Richard Biener <rguenther@suse.de>
4913 PR tree-optimization/84178
4914 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4916 (version_loop_for_if_conversion): Delay update_ssa call.
4917 (tree_if_conversion): Delay update_ssa until after predicate
4920 2018-03-08 David Malcolm <dmalcolm@redhat.com>
4922 PR tree-optimization/84178
4923 * tree-if-conv.c (release_bb_predicate): Remove the
4924 the assertion that the stmts have NULL use_ops.
4925 Discard the statements, asserting that they haven't
4926 yet been added to a BB.
4928 2018-03-08 Richard Biener <rguenther@suse.de>
4930 PR tree-optimization/84746
4931 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
4932 (phi_translate): Pass in destination ANTIC_OUT set.
4933 (phi_translate_1): Likewise. For a simplified result lookup
4934 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
4935 (phi_translate_set): Adjust.
4936 (do_pre_regular_insertion): Likewise.
4937 (do_pre_partial_partial_insertion): Likewise.
4939 2018-03-08 Martin Liska <mliska@suse.cz>
4941 PR gcov-profile/84735
4942 * doc/gcov.texi: Document usage of profile files.
4943 * gcov-io.h: Document changes in the format.
4945 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
4949 * dwarf2out.c (struct dw_line_info_table): Update comments for
4951 (FORCE_RESET_NEXT_VIEW): New.
4952 (FORCE_RESETTING_VIEW_P): New.
4953 (RESETTING_VIEW_P): Check for -1 too.
4954 (ZERO_VIEW_P): Likewise.
4955 (new_line_info_table): Force-reset next view.
4956 (dwarf2out_begin_function): Likewise.
4957 (dwarf2out_source_line): Simplify zero_view_p initialization.
4958 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
4959 view directly. Omit view when omitting .loc at line 0.
4961 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4963 PR tree-optimization/84740
4964 * tree-switch-conversion.c (process_switch): Call build_constructors
4965 only if info.phi_count is non-zero.
4967 PR tree-optimization/84739
4968 * tree-tailcall.c (find_tail_calls): Check call arguments against
4969 DECL_ARGUMENTS (current_function_decl) rather than
4970 DECL_ARGUMENTS (func) when checking for tail recursion.
4972 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4974 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
4975 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
4976 Volker Reichelt's entry and add entries for people that perform
4977 GCC fuzzy testing and report numerous bugs.
4979 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
4982 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
4983 readonly data in sdata, if that is disabled.
4984 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
4985 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
4986 -mreadonly-in-sdata option.
4988 2018-03-07 Martin Sebor <msebor@redhat.com>
4990 PR tree-optimization/84468
4991 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
4992 basic block when looking for nul assignment.
4994 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
4997 * except.h (output_function_exception_table): Adjust prototype.
4998 * except.c (output_function_exception_table): Remove FNNAME parameter
4999 and add SECTION parameter. Ouput one part of the table at a time.
5000 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
5001 the first part of the exception table and emit unwind directives.
5002 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
5003 (i386_pe_seh_cold_init): Likewise.
5004 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
5005 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
5006 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
5007 (ix86_output_call_insn): Emit a nop in one more case for SEH.
5008 * config/i386/winnt.c: Include except.h.
5009 (struct seh_frame_state): Add reg_offset, after_prologue and
5010 in_cold_section fields.
5011 (i386_pe_seh_end_prologue): Set seh->after_prologue.
5012 (i386_pe_seh_cold_init): New function.
5013 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
5014 to seh->in_cold_section.
5015 (seh_emit_push): Record the offset of the push.
5016 (seh_emit_save): Record the offet of the save.
5017 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
5018 Test seh->after_prologue to disregard the epilogue.
5019 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
5020 (i386_pe_end_cold_function): New function.
5022 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5025 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
5026 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
5029 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
5031 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
5032 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
5035 * multiple_target.c: Include tree-inline.h and intl.h.
5036 (expand_target_clones): Diagnose and fail if node->definition and
5037 !tree_versionable_function_p (node->decl).
5039 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
5041 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
5043 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
5044 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
5045 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
5047 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5050 * combine.c (try_combine): Use reg_or_subregno instead of handling
5051 just paradoxical SUBREGs and REGs.
5053 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
5055 * config/arc/arc.c (arc_finalize_pic): Remove function.
5056 (arc_must_save_register): We use single base PIC register, remove
5057 checks to save/restore the PIC register.
5058 (arc_expand_prologue): Likewise.
5059 * config/arc/arc-protos.h (arc_set_default_type_attributes):
5061 (arc_verify_short): Likewise.
5062 (arc_attr_type): Likewise.
5063 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
5064 (walk_stores): Likewise.
5065 (arc_address_cost): Make it static.
5066 (arc_verify_short): Likewise.
5067 (branch_dest): Likewise.
5068 (arc_attr_type): Likewise.
5069 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
5070 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
5071 (arc_final_prescan_insn): Remove inserting the nops due to
5072 hardware hazards. It is done in reorg step.
5073 (insn_length_variant_t): Remove.
5074 (insn_length_parameters_t): Likewise.
5075 (arc_insn_length_parameters): Likewise.
5076 (arc_get_insn_variants): Likewise.
5077 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
5079 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5082 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
5085 PR tree-optimization/84687
5086 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
5088 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
5090 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5092 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
5093 Rename to ppc_speculation_barrier.
5094 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
5095 __builtin_ppc_speculation_barrier.
5097 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5100 * combine.c (combine_simplify_rtx): Don't try to simplify if
5101 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
5104 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
5106 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
5107 to 32 bytes when compiling for POWER9.
5109 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5112 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
5113 regparm >= 3 with no arg reg available also for calls with
5114 flag_force_indirect_call. Pass decl to ix86_function_regparm.
5117 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
5119 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
5121 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
5124 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
5126 2018-03-05 Richard Biener <rguenther@suse.de>
5128 PR tree-optimization/84486
5129 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
5130 When inserting a __builtin_assume_aligned call set the LHS
5131 SSA name alignment info accordingly.
5133 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
5135 PR tree-optimization/84114
5136 * config/aarch64/aarch64.c (aarch64_reassociation_width)
5137 Avoid reassociation of FLOAT_MODE addition.
5139 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
5141 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
5142 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
5143 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
5144 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
5145 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
5146 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
5147 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
5149 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
5150 __builtin_ia32_wbinvd): New builtins.
5151 (SPECIAL_ARGS2): New.
5152 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
5153 (SPECIAL_ARGS2): New.
5154 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
5155 (ix86_valid_target_attribute_inner_p): Ditto.
5156 (ix86_init_mmx_sse_builtins): Add special_args2.
5157 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
5158 TARGET_WBNOINVD_P): New.
5159 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
5160 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
5161 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
5162 * config/i386/immintrin.h (_wbinvd): New intrinsic.
5163 * config/i386/pconfigintrin.h: New file.
5164 * config/i386/wbnoinvdintrin.h: Ditto.
5165 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
5166 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
5168 2018-03-05 Richard Biener <rguenther@suse.de>
5170 PR tree-optimization/84670
5171 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5173 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5174 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5175 (compute_antic_aux): Only assert the number of values in ANTIC_IN
5176 doesn't grow if all successors (recursively) were visited at least
5179 2018-03-05 Richard Biener <rguenther@suse.de>
5181 PR tree-optimization/84650
5182 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
5183 if executed in the loop pipeline.
5185 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
5187 * doc/configfiles.texi (Configuration Files): Move info about
5188 conditionalizing $target-protos.h to...
5189 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
5190 differs from $target-protos.h.
5192 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
5193 Chung-Ju Wu <jasonwucj@gmail.com>
5195 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
5196 * config/nds32/nds32-multiple.md (setmemsi): Define.
5197 * config/nds32/nds32-memory-manipulation.c
5198 (nds32_gen_dup_4_byte_to_word_value): New.
5199 (emit_setmem_word_loop): New.
5200 (emit_setmem_byte_loop): New.
5201 (nds32_expand_setmem_loop): New.
5202 (nds32_expand_setmem_loop_v3m): New.
5203 (nds32_expand_setmem_unroll): New.
5204 (nds32_expand_setmem): New.
5206 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5207 Chung-Ju Wu <jasonwucj@gmail.com>
5209 * config/nds32/nds32-memory-manipulation.c
5210 (nds32_emit_load_store): New.
5211 (nds32_emit_post_inc_load_store): New.
5212 (nds32_emit_mem_move): New.
5213 (nds32_emit_mem_move_block): New.
5214 (nds32_expand_movmemsi_loop_unknown_size): New.
5215 (nds32_expand_movmemsi_loop_known_size): New.
5216 (nds32_expand_movmemsi_loop): New.
5217 (nds32_expand_movmemsi_unroll): New.
5218 (nds32_expand_movmemqi): Rename ...
5219 (nds32_expand_movmemsi): ... to this.
5220 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
5221 (movmemsi): ... to this.
5222 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
5223 (nds32_expand_movmemsi): ... to this.
5225 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5226 Monk Chiang <sh.chiang04@gmail.com>
5227 Chung-Ju Wu <jasonwucj@gmail.com>
5229 * config/nds32/nds32-protos.h
5230 (nds32_expand_load_multiple): New arguments.
5231 (nds32_expand_store_multiple): Ditto.
5232 (nds32_valid_multiple_load_store): Rename ...
5233 (nds32_valid_multiple_load_store_p): ... to this.
5234 * config/nds32/nds32-memory-manipulation.c
5235 (nds32_expand_load_multiple): Refine implementation.
5236 (nds32_expand_store_multiple): Ditto.
5237 * config/nds32/nds32-multiple.md
5238 (load_multiple): Update nds32_expand_load_multiple interface.
5239 (store_multiple): Update nds32_expand_store_multiple interface.
5240 * config/nds32/nds32-predicates.c
5241 (nds32_valid_multiple_load_store): Rename ...
5242 (nds32_valid_multiple_load_store_p): ... to this and refine
5244 * config/nds32/predicates.md
5245 (nds32_load_multiple_and_update_address_operation): New predicate.
5246 (nds32_store_multiple_and_update_address_operation): New predicate.
5248 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5249 Chung-Ju Wu <jasonwucj@gmail.com>
5251 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
5252 (combo): New attribute.
5253 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
5255 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
5257 * config/nds32/nds32.opt: Change -mcmodel= default value.
5259 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
5260 Monk Chiang <sh.chiang04@gmail.com>
5261 Chung-Ju Wu <jasonwucj@gmail.com>
5263 * config/nds32/constants.md (unspec_element): New enum.
5264 * config/nds32/constraints.md (Umw): New constraint.
5265 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
5266 * config/nds32/nds32-intrinsic.md: Likewise.
5267 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
5268 (nds32_valid_smw_lwm_base_p): New.
5269 (nds32_output_smw_single_word): New.
5270 (nds32_output_lmw_single_word): New.
5271 (nds32_expand_unaligned_load): New.
5272 (nds32_expand_unaligned_store): New.
5273 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5274 (nds32_output_smw_single_word): Declare.
5275 (nds32_output_lmw_single_word): Declare.
5276 (nds32_expand_unaligned_load): Declare.
5277 (nds32_expand_unaligned_store): Declare.
5278 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5279 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5280 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5281 NDS32_BUILTIN_UASTORE_DW.
5282 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5285 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
5286 Kito Cheng <kito.cheng@gmail.com>
5287 Chung-Ju Wu <jasonwucj@gmail.com>
5289 * config/nds32/nds32-intrinsic.c
5290 (nds32_expand_builtin_null_ftype_reg): Delete.
5291 (nds32_expand_builtin_reg_ftype_imm): Ditto.
5292 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5293 (nds32_read_argument): New.
5294 (nds32_legitimize_target): Ditto.
5295 (nds32_legitimize_argument): Ditto.
5296 (nds32_check_constant_argument): Ditto.
5297 (nds32_expand_unop_builtin): Ditto.
5298 (nds32_expand_unopimm_builtin): Ditto.
5299 (nds32_expand_binop_builtin): Ditto.
5300 (nds32_builtin_decl_impl): Ditto.
5301 (builtin_description): Ditto.
5302 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5303 (nds32_init_builtins_impl): Ditto.
5304 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5305 (nds32_builtin_decl): New.
5306 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5307 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5309 2018-03-02 Jeff Law <law@redhat.com>
5311 * reorg.c (stop_search_p): Handle DEBUG_INSN.
5312 (redundant_insn, fill_simple_delay_slots): Likewise.
5313 (fill_slots_from_thread): Likewise.
5314 * resource.c (mark_referenced_resources): Likewise.
5315 (mark_set_resources, find_dead_or_set_registers): Likewise.
5317 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5319 * substring-locations.h (format_warning_va): Formatting fix for
5321 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5323 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5324 * substring-locations.c: Include intl.h.
5325 (format_warning_va): Turned into small wrapper around
5326 format_warning_n_va, renamed to ...
5327 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5328 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5330 (format_warning_at_substring_n): New function.
5331 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5332 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
5333 format_warning_at_substring with just a shorter name instead of
5334 const function pointer.
5335 (fmtwarn_n): New function.
5336 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5337 appropriate, get rid of all the fmtstr temporaries, move conditionals
5338 with G_() wrapped string literals directly into fmtwarn arguments,
5339 cast dir.len to (int), formatting fixes.
5341 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
5343 * doc/invoke.texi: Remove "Cilk Plus" references.
5345 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5346 Richard Biener <rguenther@suse.de>
5349 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5350 for error or warning attributes if CALL_FROM_THUNK_P is set.
5353 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5356 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5357 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5360 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5361 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5363 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5364 instead of -1U in last predictors element's probability member.
5366 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5369 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5370 arguments if they are comparable.
5372 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5374 PR tree-optimization/84634
5375 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5376 masks and masked_loop_p with a single loop_masks, making sure it's
5377 null for bb vectorization.
5379 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5381 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5382 (vect_analyze_data_ref_access): Use loop->safe_len rather than
5383 loop->force_vectorize to check whether there is no alias.
5385 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5388 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5390 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5392 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5393 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5394 instead of a loop around prev_real_insn.
5395 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5399 * config/i386/i386.c (ix86_print_operand): Use conditional
5400 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5403 2018-03-02 Richard Biener <rguenther@suse.de>
5405 PR tree-optimization/84427
5406 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5407 (bitmap_set_subtract_values): Rewrite to handle multiple
5410 (prune_clobbered_mems): Likewise.
5411 (phi_translate): Take edge instead of pred/phiblock.
5412 (phi_translate_1): Likewise.
5413 (phi_translate_set): Likewise. Insert all translated
5414 exprs for a value into the set, keeping possibly multiple
5415 expressions per value.
5416 (compute_antic_aux): Adjust for phi_translate changes.
5417 When intersecting union the expressions and prune those
5418 not in the final value set, keeping possibly multiple
5419 expressions per value. Do not use value-insertion
5420 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5421 all expressions. Add verification that the value-sets
5422 only shrink during iteration.
5423 (compute_partial_antic_aux): Adjust for the phi_translate changes.
5424 (do_pre_regular_insertion): Likewise.
5425 (do_pre_partial_partial_insertion): Likewise.
5427 2018-03-02 Richard Biener <rguenther@suse.de>
5430 * config/darwin.c (saved_debug_info_level): New static global.
5431 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5432 (darwin_asm_lto_end): Restore debug info generation settings.
5434 2018-03-01 Martin Liska <mliska@suse.cz>
5437 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5440 2018-03-01 Richard Biener <rguenther@suse.de>
5443 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5444 type annotation with LTO when debug was disabled at compile-time.
5446 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5448 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5450 (mips_final_postscan_insn): Likewise.
5452 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5454 PR rtl-optimization/84528
5455 * alias.c (init_alias_target): Add commentary.
5456 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5457 a unique base value if the frame pointer is not eliminated
5458 to the stack pointer.
5460 2018-03-01 Tom de Vries <tom@codesourcery.com>
5462 PR rtl-optimization/83327
5463 * lra-int.h (hard_regs_spilled_into): Declare.
5464 * lra.c (hard_regs_spilled_into): Define.
5465 (init_reg_info): Init hard_regs_spilled_into.
5466 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5467 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5468 (process_bb_lives): Handle hard_regs_spilled_into.
5469 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5470 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5472 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5474 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5475 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5476 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5477 * config/rs6000/aix72.h: New file.
5479 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5481 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5482 instead of warning_at with conditional singular and plural messages
5486 * stor-layout.c (update_alignment_for_field): For
5487 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5488 && !DECL_PACKED (field), do the alignment update, just use
5489 only desired_align instead of MAX (type_align, desired_align)
5491 (place_field): Don't do known_align < desired_align handling
5492 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5493 is non-NULL, instead do it after rli->prev_field handling and
5494 only if not within a bitfield word. For DECL_PACKED (field)
5495 use type_align of BITS_PER_UNIT.
5497 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5499 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5500 superfluous parentheses and trailing spaces.
5502 2018-02-28 Richard Biener <rguenther@suse.de>
5504 PR tree-optimization/84584
5505 * graphite-scop-detection.c (scop_detection::add_scop): Discard
5506 SCoPs with fake exit edge.
5508 2018-02-28 Martin Liska <mliska@suse.cz>
5511 * timevar.c (timer::print): Fix format to properly print 100%
5514 2018-02-28 Richard Biener <rguenther@suse.de>
5517 * genmatch.c (capture_info::walk_match): Do not mark
5518 captured expressions without operands as expr_p given
5519 they act more like predicates and should be subject to
5520 "lost tail" side-effect preserving.
5522 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
5524 PR rtl-optimization/81611
5525 * auto-inc-dec.c (attempt_change): Move dead note from
5526 mem_insn if it's the next use of regno
5527 (find_address): Take address use of reg holding
5528 non-incremented value. Add parm to limit search to the named
5530 (merge_in_block): Attempt to use a mem insn that is the next
5531 use of the original regno.
5533 2018-02-27 Martin Sebor <msebor@redhat.com>
5536 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
5537 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
5539 2018-02-27 Martin Sebor <msebor@redhat.com>
5541 PR translation/84207
5542 * diagnostic-core.h (warning_n, error_n, inform_n): Change
5543 n argument to unsigned HOST_WIDE_INT.
5544 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5545 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
5546 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5547 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5549 2018-02-27 Richard Biener <rguenther@suse.de>
5551 PR tree-optimization/84512
5552 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5553 Do not use the estimate returned from record_stmt_cost for
5554 the scalar iteration cost but sum properly using add_stmt_cost.
5556 2018-02-27 Richard Biener <rguenther@suse.de>
5558 PR tree-optimization/84466
5559 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5560 Adjust last change to less strictly validate use operands.
5562 2018-02-27 Martin Liska <mliska@suse.cz>
5564 PR gcov-profile/84548
5565 * gcov.c (process_file): Allow partial overlap and consider it
5566 also as group functions.
5567 (output_lines): Properly calculate range of lines for a group.
5569 2018-02-27 Martin Liska <mliska@suse.cz>
5571 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
5572 'ggc' suffixes. Change first column width.
5573 (timer::print): Fix formatting of the column.
5575 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
5577 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
5578 preserve inline entry blocks for the sake of debug inline
5579 entry point markers alone.
5580 (remove_unused_locals): Suggest in comments a better place to
5581 force the preservation of inline entry blocks that are
5582 otherwise unused, but do not preserve them.
5584 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5586 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
5588 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5591 * config/i386/constraints.md (Bs): Replace
5592 ix86_indirect_branch_register with
5593 TARGET_INDIRECT_BRANCH_REGISTER.
5595 * config/i386/i386.md (indirect_jump): Likewise.
5596 (tablejump): Likewise.
5597 (*sibcall_memory): Likewise.
5598 (*sibcall_value_memory): Likewise.
5599 Peepholes of indirect call and jump via memory: Likewise.
5600 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
5601 (*sibcall_value_GOT_32): Likewise.
5602 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5603 (GOT_memory_operand): Likewise.
5604 (call_insn_operand): Likewise.
5605 (sibcall_insn_operand): Likewise.
5606 (GOT32_symbol_operand): Likewise.
5607 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
5609 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5611 PR rtl-optimization/83496
5612 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
5613 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
5614 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
5615 redundant insn, if any.
5616 (relax_delay_slots): Likewise.
5617 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
5619 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
5621 PR tree-optimization/83965
5622 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
5623 that grouped statements are part of a reduction chain. Return
5624 true if the statement is not marked as a reduction itself but
5626 (vect_recog_dot_prod_pattern): Don't check whether the statement
5627 is part of a group here.
5628 (vect_recog_sad_pattern): Likewise.
5629 (vect_recog_widen_sum_pattern): Likewise.
5631 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5634 * final.c (rest_of_clean_state): Also look for calls inside sequences.
5636 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5639 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
5641 (ix86_output_indirect_function_return): New prototype.
5642 (ix86_split_simple_return_pop_internal): Likewise.
5643 * config/i386/i386.c (indirect_return_via_cx): New.
5644 (indirect_return_via_cx_bnd): Likewise.
5645 (indirect_thunk_name): Handle return va CX_REG.
5646 (output_indirect_thunk_function): Create alias for
5647 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
5648 (ix86_output_indirect_jmp): Remove the bool argument.
5649 (ix86_output_indirect_function_return): New function.
5650 (ix86_split_simple_return_pop_internal): Likewise.
5651 * config/i386/i386.md (*indirect_jump): Don't pass false
5652 to ix86_output_indirect_jmp.
5653 (*tablejump_1): Likewise.
5654 (simple_return_pop_internal): Change it to define_insn_and_split.
5655 Call ix86_split_simple_return_pop_internal to split it for
5657 (simple_return_indirect_internal): Call
5658 ix86_output_indirect_function_return instead of
5659 ix86_output_indirect_jmp.
5661 2018-02-26 Jakub Jelinek <jakub@redhat.com>
5664 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
5665 memset and value initialization afterwards.
5667 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
5669 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
5671 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5674 * common/config/aarch64/aarch64-common.c
5675 (aarch_option_optimization_table[]): Switch
5676 off fomit-frame-pointer
5678 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5679 Chung-Ju Wu <jasonwucj@gmail.com>
5681 * config/nds32/nds32-multiple.md (load_multiple): Disallow
5683 (store_multiple): Ditto.
5685 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5687 * config.gcc: Add --with-cpu support for nds32 target.
5688 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
5689 * config/nds32/nds32.opt: Add -mcpu= option.
5691 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
5693 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
5694 isel=yes): Warn for these deprecated options.
5696 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
5698 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
5699 ISA_2_5_MASKS_EMBEDDED.
5701 2018-02-23 Jakub Jelinek <jakub@redhat.com>
5703 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
5704 p->max as pointers rather than using iterative_hash_expr.
5706 2018-02-23 Carl Love <cel@us.ibm.com>
5708 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
5709 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
5711 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
5712 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
5713 P8V_BUILTIN_VEC_VUNSIGNED2.
5715 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
5718 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
5719 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
5721 * lra-constraints.c (curr_insn_transform): Set up
5722 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
5724 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
5725 * lra-eliminations.c (spill_pseudos): Ditto.
5726 (process_insn_for_elimination): Ditto.
5727 * lra-lives.c (reg_early_clobber_p): Use the new macros.
5728 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
5729 LRA_NON_CLOBBERED_ALT.
5731 2018-02-22 Martin Sebor <msebor@redhat.com>
5733 PR tree-optimization/84480
5734 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
5735 to maybe_diag_stxncpy_trunc. Call it.
5736 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
5737 from gimple_fold_builtin_strcpy. Print inlining stack.
5738 (handle_builtin_stxncpy): Print inlining stack.
5739 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
5741 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
5744 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
5745 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
5746 and -fcheck-pointer-bounds are used together.
5747 (indirect_thunk_prefix): New enum.
5748 (indirect_thunk_need_prefix): New function.
5749 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
5750 "_nt" instead of "_bnd" for NOTRACK prefix.
5751 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
5752 (output_indirect_thunk_function): Likewise.
5754 (ix86_code_end): Update output_indirect_thunk_function calls.
5755 (ix86_output_indirect_branch_via_reg): Replace
5756 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
5757 (ix86_output_indirect_branch_via_push): Likewise.
5758 (ix86_output_function_return): Likewise.
5759 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
5760 incompatible with -fcf-protection=branch and
5761 -fcheck-pointer-bounds.
5763 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5766 * config/aarch64/aarch64.c (aarch64_print_address_internal):
5767 Change gcc_assert call to output_operand_lossage.
5769 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5771 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
5773 2018-02-22 DJ Delorie <dj@redhat.com>
5774 Sebastian Perta <sebastian.perta@renesas.com>
5775 Oleg Endo <olegendo@gcc.gnu.org>
5777 * config/rx/rx.c (rx_rtx_costs): New function.
5778 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
5780 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
5782 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
5784 2018-02-22 Martin Liska <mliska@suse.cz>
5787 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
5788 Add "native" as a possible value.
5790 2018-02-22 Martin Liska <mliska@suse.cz>
5793 * config/i386/i386.c (ix86_option_override_internal):
5794 Add "native" as a possible value for -march and -mtune.
5796 2018-02-22 Jakub Jelinek <jakub@redhat.com>
5799 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
5800 to all type variants.
5802 PR tree-optimization/84503
5803 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
5804 width as info->bitpos + info->bitsize - start.
5805 (merged_store_group::merge_overlapping): Simplify width computation.
5806 (check_no_overlap): New function.
5807 (imm_store_chain_info::try_coalesce_bswap): Compute expected
5808 start + width and last_order of the group, fail if check_no_overlap
5810 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
5811 to group if check_no_overlap fails.
5813 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5815 * config/rs6000/altivec.md: Delete contraint arguments to
5816 define_expand, define_split, and define_peephole2, and in
5817 define_insn_and_split if always unused.
5818 * config/rs6000/darwin.md: Ditto.
5819 * config/rs6000/dfp.md: Ditto.
5820 * config/rs6000/rs6000.md: Ditto.
5821 * config/rs6000/sync.md: Ditto.
5822 * config/rs6000/vector.md: Ditto.
5823 * config/rs6000/vsx.md: Ditto.
5825 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5827 * config/rs6000/altivec.md: Write output control strings as braced
5828 blocks instead of double-quoted strings.
5829 * config/rs6000/darwin.md: Ditto.
5830 * config/rs6000/rs6000.md: Ditto.
5831 * config/rs6000/vector.md: Ditto.
5832 * config/rs6000/vsx.md: Ditto.
5834 2018-02-21 Jason Merrill <jason@redhat.com>
5836 PR c++/84314 - ICE with templates and fastcall attribute.
5837 * attribs.c (build_type_attribute_qual_variant): Remove assert.
5839 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5841 * ipa-cp.c (determine_versionability): Fix comment typos.
5843 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5846 * ipa-cp.c (determine_versionability): Do not version functions caling
5849 2018-02-21 Martin Liska <mliska@suse.cz>
5852 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
5853 Add "native" as a possible value.
5854 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
5855 the macro when native cpu detection is available.
5857 2018-02-21 Martin Liska <mliska@suse.cz>
5860 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
5861 Add "native" as a possible value.
5862 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
5863 when native cpu detection is available.
5865 2018-02-21 Jakub Jelinek <jakub@redhat.com>
5866 Martin Sebor <msebor@redhat.com>
5868 PR tree-optimization/84478
5869 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
5871 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
5872 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
5873 support which is conservatively correct, for 2 only stay conservative
5874 for maxlen. Formatting and comment capitalization fixes. Add STRICT
5875 argument to the 2 argument get_range_strlen, adjust 6 arg
5876 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
5878 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
5879 (gimple_fold_builtin_strlen): Pass true as last argument to
5882 2018-02-20 Martin Sebor <msebor@redhat.com>
5885 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
5886 (builtin_memref::set_base_and_offset): Same. Handle inner references.
5887 (builtin_memref::builtin_memref): Factor out parts into
5888 set_base_and_offset and call it.
5890 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5893 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
5894 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
5895 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
5896 search at the associated MODE_INT.
5898 2018-02-20 Jeff Law <law@redhat.com>
5901 PR tree-optimization/81592
5903 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
5904 for range data rather than using global data.
5905 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
5906 range data rather than using global data.
5907 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
5908 pass it to children as needed.
5909 (struct directive::fmtresult): Similarly.
5910 (struct directive::set_width): Similarly.
5911 (struct directive::set_precision): Similarly.
5912 (format_integer, format_directive, parse_directive): Similarly.
5913 (format_none): Accept unnamed vr_values parameter.
5914 (format_percent, format_floating, format_character): Similarly.
5915 (format_string, format_plain): Similarly.
5916 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
5917 the EVRP range analyzer for range data rather than using global data.
5918 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
5919 gimple-ssa-evrp-analyze.h
5920 (class sprintf_dom_walker): Add after_dom_children member function.
5921 Add evrp_range_analyzer member.
5922 (sprintf_dom_walker::before_dom_children): Call into the EVRP
5923 range analyzer as needed.
5924 (sprintf_dom_walker::after_dom_children): New member function.
5925 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
5927 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5928 (evrp_range_analyzer::pop_to_marker): Likewise.
5930 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5932 PR tree-optimization/84419
5933 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
5934 with the required type if its current type is compatible but
5937 2018-02-20 Jakub Jelinek <jakub@redhat.com>
5940 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
5941 after vectorization.
5943 2018-02-20 Martin Liska <mliska@suse.cz>
5946 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
5947 possible values if we don't have a hint.
5949 2018-02-20 Martin Liska <mliska@suse.cz>
5953 * final.c (shorten_branches): Build align_tab array with one
5955 * opts.c (finish_options): Add alignment option limit check.
5956 (MAX_CODE_ALIGN): Likewise.
5957 (MAX_CODE_ALIGN_VALUE): Likewise.
5958 * doc/invoke.texi: Document maximum allowed option value for
5959 all -falign-* options.
5961 2018-02-19 Jakub Jelinek <jakub@redhat.com>
5964 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
5965 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
5966 * var-tracking.c (emit_note_insn_var_location): Remove all references
5967 to NOTE_INSN_CALL_ARG_LOCATION.
5968 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
5969 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
5970 Use copy_rtx_if_shared.
5971 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
5972 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
5973 (dwarf2out_var_location): Remove handling of
5974 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
5976 * final.c (final_scan_insn): Remove all references to
5977 NOTE_INSN_CALL_ARG_LOCATION.
5978 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
5979 before dumping final insns.
5980 * except.c (emit_note_eh_region_end): Remove all references to
5981 NOTE_INSN_CALL_ARG_LOCATION.
5982 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
5983 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
5984 * config/arc/arc.c (hwloop_optimize): Likewise.
5985 * config/arm/arm.c (create_fix_barrier): Likewise.
5986 * config/s390/s390.c (s390_chunkify_start): Likewise.
5987 * config/sh/sh.c (find_barrier): Likewise.
5988 * config/i386/i386.c (rest_of_insert_endbranch,
5989 ix86_seh_fixup_eh_fallthru): Likewise.
5990 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
5991 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
5992 * config/frv/frv.c (frv_function_prologue): Likewise.
5993 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
5995 (note_outside_basic_block_p): Remove all references to
5996 NOTE_INSN_CALL_ARG_LOCATION.
5997 * gengtype.c (adjust_field_rtx_def): Likewise.
5998 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
6000 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
6001 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
6004 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
6007 PR tree-optimization/84452
6008 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
6009 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
6012 2018-02-19 Martin Liska <mliska@suse.cz>
6015 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
6017 2018-02-19 Martin Liska <mliska@suse.cz>
6018 Richard Sandiford <richard.sandiford@linaro.org>
6020 PR tree-optimization/82491
6021 * gimple-fold.c (get_base_constructor): Make earlier bail out
6024 2018-02-19 Carl Love <cel@us.ibm.com>
6026 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
6027 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
6029 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
6030 P8V_BUILTIN_VEC_NEG.
6032 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
6034 * config/rl78/rl78.md (movdf): New define expand.
6036 2018-02-19 Martin Liska <mliska@suse.cz>
6039 * doc/invoke.texi: Fix typo.
6040 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
6042 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
6044 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
6045 handle rs6000_single_float and rs6000_double_float specially for
6048 2018-02-16 Jeff Law <law@redhat.com>
6050 * config/rx/rx.c (add_pop_cfi_notes): New function.;
6053 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6056 * ipa-inline.c (inline_small_functions): Fix a typo.
6058 2018-02-16 Nathan Sidwell <nathan@acm.org>
6060 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
6062 2018-02-16 Carl Love <cel@us.ibm.com>
6064 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
6065 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
6066 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
6067 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
6068 expansion to P8V_BUILTIN_VEC_FLOAT2.
6070 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
6072 PR rtl-optimization/70023
6073 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
6074 src_regno into account.
6076 2018-02-16 Carl Love <cel@us.ibm.com>
6078 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
6079 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
6080 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
6081 * config/rs6000/rs6000.c: Remove case statements for
6082 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
6083 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
6084 and P9V_BUILTIN_VEC_VINSERT4B.
6085 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
6086 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
6087 * config/rs6000/vsx.md:
6088 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
6091 2018-02-16 Carl Love <cel@us.ibm.com>
6093 * config/rs6000/altivec.h: Add builtin names vec_extract4b
6095 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
6097 * config/rs6000/rs6000-c.c: Add the definitions for
6098 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
6099 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
6100 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
6101 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
6102 definition for insert4b and define insn *insert3b_internal.
6103 * doc/extend.texi: Add documentation for vec_extract4b.
6105 2018-02-16 Nathan Sidwell <nathan@acm.org>
6107 * doc/extend.texi (Backwards Compatibility): Mention friend
6108 injection. Note for-scope is deprecated.
6109 * doc/invoke.texi (-ffriend-injection): Deprecate.
6111 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
6113 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
6114 that moved to I2, also allow destinations that are a paradoxical
6115 subreg (instead of a normal reg).
6117 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
6120 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
6123 2018-02-16 Richard Biener <rguenther@suse.de>
6125 PR tree-optimization/84037
6126 PR tree-optimization/84016
6128 * config/i386/i386.c (ix86_builtin_vectorization_cost):
6129 Adjust vec_construct for the fact we need additional higher latency
6130 128bit inserts for AVX256 and AVX512 vector builds.
6131 (ix86_add_stmt_cost): Scale vector construction cost for
6134 2018-02-16 Richard Biener <rguenther@suse.de>
6136 PR tree-optimization/84417
6137 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
6138 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
6139 (non_rewritable_lvalue_p): Likewise, use poly-ints.
6141 2018-02-16 Martin Liska <mliska@suse.cz>
6144 * internal-fn.def (ASAN_CHECK): Set proper flags.
6145 (ASAN_MARK): Likewise.
6147 2018-02-16 Julia Koval <julia.koval@intel.com>
6149 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
6150 from PTA_CANNONLAKE.
6152 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6155 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
6156 Use ++iter rather than iter++ for std::list iterators.
6157 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
6158 defer deleting them until all nodes in the forest are processed. Do
6159 free even leaf nodes. Change to_process into auto_vec.
6162 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
6163 * vec.h (vec_default_construct): Use memset instead of placement new
6164 if BROKEN_VALUE_INITIALIZATION is defined.
6165 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
6166 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
6169 PR rtl-optimization/83723
6170 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
6171 * lra.c (lra_substitute_pseudo): Likewise. If true, use
6172 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
6174 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
6176 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
6178 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
6180 PR rtl-optimization/81443
6181 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
6182 from inner REGs to paradoxical SUBREGs.
6184 2018-02-16 Richard Biener <rguenther@suse.de>
6186 PR tree-optimization/84399
6187 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6188 For operands we can analyze at their definition make sure we can
6189 analyze them at each use as well.
6191 2018-02-16 Richard Biener <rguenther@suse.de>
6193 PR tree-optimization/84190
6194 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
6195 volatile accesses if the decl isn't volatile.
6197 2018-02-15 Jason Merrill <jason@redhat.com>
6199 PR c++/84314 - ICE with templates and fastcall attribute.
6200 * attribs.c (build_type_attribute_qual_variant): Don't clobber
6201 TYPE_CANONICAL on an existing type.
6203 2018-02-15 Jakub Jelinek <jakub@redhat.com>
6205 PR tree-optimization/84383
6206 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
6207 dstoff nor call operand_equal_p if dstbase is NULL.
6209 PR tree-optimization/84334
6210 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
6211 also a CONSTANT_CLASS_P, punt.
6213 2018-02-14 Jim Wilson <jimw@sifive.com>
6215 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
6216 first SMALL_OPERAND check. New local min_second_step. Move assert
6217 to where locals are set. Add TARGET_RVC support.
6218 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
6220 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
6222 * doc/invoke.texi: Correct -Wformat-overflow code sample.
6224 2018-02-14 Martin Sebor <msebor@redhat.com>
6226 PR tree-optimization/83698
6227 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
6228 arrays constrain the offset range to their bounds.
6229 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
6230 (builtin_access::overlap): Avoid setting the size of overlap if it's
6232 (maybe_diag_overlap): Also consider arrays when deciding what values
6233 of offsets to include in diagnostics.
6235 2018-02-14 Martin Sebor <msebor@redhat.com>
6238 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
6239 that correspond to the kind of a declaration.
6241 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
6244 * config/pa/pa.md: Load address of PIC label using the linkage table
6245 if the label is nonlocal.
6247 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
6249 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
6250 warning message if user requests -maltivec=be.
6251 * doc/invoke.texi: Document deprecation of -maltivec=be.
6253 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
6256 * config/rs6000/rs6000-c.c: Update definitions for
6257 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
6258 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
6260 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6263 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
6264 add _get_ssp intrinsics. Remove argument from
6265 __builtin_ia32_rdssp[d|q].
6266 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
6267 * config/i386/i386-builtin.def: Remove argument from
6268 __builtin_ia32_rdssp[d|q].
6269 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
6270 ix86_expand_special_args_builtin for _rdssp[d|q].
6271 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
6272 Clear register before usage.
6273 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6274 Add documentation for new _get_ssp and _inc_ssp intrinsics.
6276 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
6278 PR tree-optimization/84357
6279 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6280 operand 1 of an ARRAY_REF too.
6282 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
6285 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6286 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6288 (set_of_reg): New struct.
6289 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6290 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6291 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6293 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6294 Split into bitclr, bitset, bitinvert patterns if appropriate.
6295 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6296 use rx_fuse_in_memory_bitop.
6297 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6298 to named insn, correct maximum insn length.
6300 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6303 * machmode.def: Define a complex mode for PARTIAL_INT.
6304 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6306 * doc/rtl.texi: Document CSPImode.
6307 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6309 (msp430_hard_regno_nregs_with_padding): Likewise.
6311 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
6314 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6316 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
6318 PR rtl-optimization/84169
6319 * combine.c (try_combine): New variable split_i2i3. Set it to true if
6320 we generated a parallel as new i3 and we split that to new i2 and i3
6321 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
6322 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
6323 those to i2, not i1. Partially rewrite this scan code.
6325 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6328 * stor-layout.c (place_field): For variable length fields, adjust
6329 offset_align afterwards not just based on the field's alignment,
6330 but also on the size.
6333 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6334 of exps and logs in the use_exp2 case.
6336 2018-02-13 Jeff Law <law@redhat.com>
6338 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6341 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6344 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
6348 * common.opt (gas-loc-support, gas-locview-support): New.
6349 (ginline-points, ginternal-reset-location-views): New.
6350 * doc/invoke.texi: Document them. Use @itemx where intended.
6351 (gvariable-location-views): Adjust.
6352 * target.def (reset_location_view): New.
6353 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6354 (TARGET_RESET_LOCATION_VIEW): New.
6355 * doc/tm.texi: Rebuilt.
6356 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6357 (dwarf2out_default_as_locview_support): New.
6358 (output_asm_line_debug_info): Use option variables.
6359 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6360 (output_loc_list): Likewise.
6361 (add_high_low_attributes): Check option variables.
6362 Don't output entry view attribute in strict mode.
6363 (gen_inlined_subroutine_die): Check option variables.
6364 (dwarf2out_inline_entry): Likewise.
6365 (init_sections_and_labels): Likewise.
6366 (dwarf2out_early_finish): Likewise.
6367 (maybe_reset_location_view): New, from...
6368 (dwarf2out_var_location): ... here. Call it.
6369 * debug.h (dwarf2out_default_as_loc_support): Declare.
6370 (dwarf2out_default_as_locview_support): Declare.
6371 * hooks.c (hook_int_rtx_insn_0): New.
6372 * hooks.h (hook_int_rtx_insn_0): Declare.
6373 * toplev.c (process_options): Take -gas-loc-support and
6374 -gas-locview-support from dwarf2out. Enable
6375 -gvariable-location-views by default only with locview
6376 assembler support. Enable -ginternal-reset-location-views by
6377 default only if the target defines the corresponding hook.
6378 Enable -ginline-points by default if location views are
6379 enabled; force it disabled if statement frontiers are
6381 * tree-inline.c (expand_call_inline): Check option variables.
6382 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6384 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
6386 PR tree-optimization/84321
6387 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6388 handling. Also check whether the anti-range contains any values
6389 that satisfy the mask; switch to a VR_RANGE if not.
6391 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
6394 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6396 2018-02-13 Martin Jambor <mjambor@suse.cz>
6399 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6400 of call statements, also set location of a load to a temporary.
6402 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
6404 * config/rl78/rl78.c (add_vector_labels): New function.
6405 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6406 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6407 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6408 which checks that no arguments are passed.
6409 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6410 * doc/extend.texi: Documentation for the new attribute.
6412 2018-02-13 Andreas Schwab <schwab@suse.de>
6414 * config/riscv/linux.h (CPP_SPEC): Define.
6416 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6419 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6420 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6421 OPTION_MASK_ISA_AES as first argument to def_builtin_const
6422 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6423 instead of OPTION_MASK_ISA_PCLMUL as first argument to
6424 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6425 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6426 temporarily for AES and PCLMUL builtins.
6428 PR tree-optimization/84339
6429 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6430 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6434 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6435 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6436 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6438 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6440 * omp-simd-clone.h: New file.
6441 * omp-simd-clone.c: Include omp-simd-clone.h.
6442 (expand_simd_clones): No longer static.
6443 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6444 cgraph.h and omp-simd-clone.h.
6445 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6446 (vect_recog_widen_shift_pattern): Formatting fix.
6447 (vect_pattern_recog_1): Don't check optab for calls.
6450 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6451 operands[2] into a REG before using gen_lowpart on it.
6453 2018-02-12 Jeff Law <law@redhat.com>
6456 * config/sh/sh.c (find_barrier): Consider a sibling call
6459 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6460 successfully back substituting a reg.
6462 2018-02-12 Richard Biener <rguenther@suse.de>
6464 PR tree-optimization/84037
6465 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6466 parameter, move visited init to caller.
6467 (vect_slp_analyze_operations): Separate cost from validity
6468 check, initialize visited once for all instances.
6469 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6471 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6473 (vect_model_promotion_demotion_cost): Likewise.
6474 (vectorizable_bswap): Guard cost modeling with !slp_node
6475 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6477 (vectorizable_call): Likewise.
6478 (vectorizable_conversion): Likewise.
6479 (vectorizable_assignment): Likewise.
6480 (vectorizable_shift): Likewise.
6481 (vectorizable_operation): Likewise.
6482 (vectorizable_store): Likewise.
6483 (vectorizable_load): Likewise.
6484 (vectorizable_condition): Likewise.
6485 (vectorizable_comparison): Likewise.
6487 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6490 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6491 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6492 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6494 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6497 * params.def (inline-min-speedup): Increase from 8 to 15.
6498 (max-inline-insns-auto): Decrease from 40 to 30.
6499 * ipa-split.c (consider_split): Add some buffer for function to
6500 be considered inlining candidate.
6501 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6504 2018-02-12 Richard Biener <rguenther@suse.de>
6506 PR tree-optimization/84037
6507 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6508 matched stmts if we cannot swap the non-matched ones.
6510 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6512 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6513 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6514 _mm_maskz_scalef_round_ss): New intrinsics.
6515 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6516 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6517 __builtin_ia32_scalefss_round): Remove.
6518 (__builtin_ia32_scalefsd_mask_round,
6519 __builtin_ia32_scalefss_mask_round): New intrinsics.
6520 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6521 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6522 ((match_operand:VF_128 2 "<round_nimm_predicate>"
6523 "<round_constraint>")): Changed to ...
6524 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
6525 "<round_scalar_constraint>")): ... this.
6526 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
6527 %0, %1, %2<round_op3>}"): Changed to ...
6528 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
6529 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
6530 %2<round_scalar_mask_op3>}"): ... this.
6531 * config/i386/subst.md (round_scalar_nimm_predicate): New.
6533 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6535 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
6536 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
6537 (_mm_maskz_sqrt_round_ss): New intrinsics.
6538 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
6539 (__builtin_ia32_sqrtsd_mask_round)
6540 (__builtin_ia32_sqrtss_mask_round): New builtins.
6541 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6542 (__builtin_ia32_sqrtss_round): Remove.
6543 (__builtin_ia32_sqrtsd_mask_round)
6544 (__builtin_ia32_sqrtss_mask_round): New builtins.
6545 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6546 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6547 ((match_operand:VF_128 1 "vector_operand"
6548 "xBm,<round_constraint>")): Changed to ...
6549 ((match_operand:VF_128 1 "vector_operand"
6550 "xBm,<round_scalar_constraint>")): ... this.
6551 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6552 %0, %2, %<iptr>1<round_op3>}): Changed to ...
6553 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6554 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6555 %<iptr>1<round_scalar_mask_op3>}): ... this.
6556 ((set_attr "prefix" "<round_prefix>")): Changed to ...
6557 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6559 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
6562 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6563 Cast vec_cmpeq result to correct type.
6564 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
6565 Cast vec_cmpgt result to correct type.
6567 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
6569 * final.c (final_scan_insn_1): Renamed from...
6570 (final_scan_insn): ... this. New wrapper, to recover
6571 seen from the outermost call in recursive ones.
6572 * config/sparc/sparc.c (output_return): Drop seen from call.
6573 (output_sibcall): Likewise.
6574 * config/visium/visium.c (output_branch): Likewise.
6576 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
6578 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
6581 2018-02-10 Alan Modra <amodra@gmail.com>
6584 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
6585 Specify LR as an input.
6587 2018-02-10 Jakub Jelinek <jakub@redhat.com>
6590 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
6591 remove_member_access_dummy_vars): New functions.
6592 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
6593 lower_omp_1, execute_lower_omp): Use them.
6595 PR rtl-optimization/84308
6596 * shrink-wrap.c (spread_components): Release todo vector.
6598 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
6600 PR rtl-optimization/57193
6601 * ira-color.c (struct allocno_color_data): Add member
6602 conflict_allocno_hard_prefs.
6603 (update_conflict_allocno_hard_prefs): New.
6604 (bucket_allocno_compare_func): Add a preference based on
6605 conflict_allocno_hard_prefs.
6606 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
6607 (color_allocnos): Remove a dead code. Initiate
6608 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
6610 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6613 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
6614 constraint from =wa to wa. Avoid a subreg on the output operand,
6615 instead use a pseudo and subreg it in a move.
6616 (p9_xxbrd_<mode>): Changed to ...
6617 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
6618 (p9_xxbrd_v2df): New expander.
6619 (p9_xxbrw_<mode>): Changed to ...
6620 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
6621 (p9_xxbrw_v4sf): New expander.
6623 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6625 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
6627 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
6630 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
6631 multiply in 32-bit mode.
6632 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
6633 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
6636 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6638 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
6639 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
6640 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
6641 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
6643 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
6646 * dwarf2out.c (is_trivial_indirect_ref): New function.
6647 (dwarf2out_late_global_decl): Do not generate a location
6648 attribute for variables that have a non-trivial DECL_VALUE_EXPR
6649 and that are not defined in the current unit.
6651 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6653 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
6654 instead of a libcall for UNORDERED.
6656 2018-02-09 Tamar Christina <tamar.christina@arm.com>
6659 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
6660 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
6662 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6665 * config/s390/s390.c (s390_set_current_function): Invoke
6666 s390_indirect_branch_settings also if fndecl didn't change.
6668 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6670 * config/rs6000/rs6000.md (blockage): Set length to zero.
6672 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6674 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
6676 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6679 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
6680 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
6684 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
6685 PARALLEL incoming that failed vt_get_decl_and_offset check.
6688 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
6689 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
6691 (get_variable_section): For decls in named .bss* sections pass true as
6692 second argument to bss_initializer_p.
6694 2018-02-09 Marek Polacek <polacek@redhat.com>
6695 Jakub Jelinek <jakub@redhat.com>
6698 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
6699 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
6700 Sync some changes from cxx_fold_indirect_ref.
6702 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6704 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
6706 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
6707 (BLOCK_INLINE_ENTRY_LABEL): New.
6708 (dwarf2out_var_location): Disregard inline entry markers.
6709 (inline_entry_data): New struct.
6710 (inline_entry_data_hasher): New hashtable type.
6711 (inline_entry_data_hasher::hash): New.
6712 (inline_entry_data_hasher::equal): New.
6713 (inline_entry_data_table): New variable.
6714 (add_high_low_attributes): Add DW_AT_entry_pc and
6715 DW_AT_GNU_entry_view attributes if a pending entry is found
6716 in inline_entry_data_table. Add old entry_pc attribute only
6717 if debug nonbinding markers are disabled.
6718 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
6719 markers are enabled.
6720 (block_within_block_p, dwarf2out_inline_entry): New.
6721 (dwarf2out_finish): Check that no entries remained in
6722 inline_entry_data_table.
6723 * final.c (reemit_insn_block_notes): Handle inline entry notes.
6724 (final_scan_insn, notice_source_line): Likewise.
6725 (rest_of_clean_state): Skip inline entry markers.
6726 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
6728 * gimple.c (gimple_build_debug_inline_entry): New.
6729 * gimple.h (enum gimple_debug_subcode): Add
6730 GIMPLE_DEBUG_INLINE_ENTRY.
6731 (gimple_build_debug_inline_entry): Declare.
6732 (gimple_debug_inline_entry_p): New.
6733 (gimple_debug_nonbind_marker_p): Adjust.
6734 * insn-notes.def (INLINE_ENTRY): New.
6735 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
6736 inline entry marker notes.
6737 (print_insn): Likewise.
6738 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
6739 (INSN_DEBUG_MARKER_KIND): Likewise.
6740 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
6741 * tree-inline.c (expand_call_inline): Build and insert
6742 debug_inline_entry stmt.
6743 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
6744 inline entry blocks early, if nonbind markers are enabled.
6745 (dump_scope_block): Dump fragment info.
6746 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
6747 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
6748 (gimple_build_debug_inline_entry): New.
6749 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
6750 Enable/disable inline entry points too.
6751 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
6752 (DEBUG_INSN): Describe inline entry markers.
6754 * common.opt (gvariable-location-views): New.
6755 (gvariable-location-views=incompat5): New.
6756 * config.in: Rebuilt.
6757 * configure: Rebuilt.
6758 * configure.ac: Test assembler for view support.
6759 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
6760 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
6761 * dwarf2out.c (var_loc_view): New typedef.
6762 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
6763 (dwarf2out_locviews_in_attribute): New.
6764 (dwarf2out_locviews_in_loclist): New.
6765 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
6766 (enum dw_line_info_opcode): Add LI_adv_address.
6767 (struct dw_line_info_table): Add view.
6768 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
6769 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
6770 (zero_view_p): New variable.
6771 (ZERO_VIEW_P): New macro.
6772 (output_asm_line_debug_info): New.
6773 (struct var_loc_node): Add view.
6774 (add_AT_view_list, AT_loc_list): New.
6775 (add_var_loc_to_decl): Add view param. Test it against last.
6776 (new_loc_list): Add view params. Record them.
6777 (AT_loc_list_ptr): Handle loc and view lists.
6778 (view_list_to_loc_list_val_node): New.
6779 (print_dw_val): Handle dw_val_class_view_list.
6780 (size_of_die): Likewise.
6781 (value_format): Likewise.
6782 (loc_list_has_views): New.
6783 (gen_llsym): Set vl_symbol too.
6784 (maybe_gen_llsym, skip_loc_list_entry): New.
6785 (dwarf2out_maybe_output_loclist_view_pair): New.
6786 (output_loc_list): Output view list or entries too.
6787 (output_view_list_offset): New.
6788 (output_die): Handle dw_val_class_view_list.
6789 (output_dwarf_version): New.
6790 (output_compilation_unit_header): Use it.
6791 (output_skeleton_debug_sections): Likewise.
6792 (output_rnglists, output_line_info): Likewise.
6793 (output_pubnames, output_aranges): Update version comments.
6794 (output_one_line_info_table): Output view numbers in asm comments.
6795 (dw_loc_list): Determine current endview, pass it to new_loc_list.
6796 Call maybe_gen_llsym.
6797 (loc_list_from_tree_1): Adjust.
6798 (add_AT_location_description): Create view list attribute if
6799 needed, check it's absent otherwise.
6800 (convert_cfa_to_fb_loc_list): Adjust.
6801 (maybe_emit_file): Call output_asm_line_debug_info for test.
6802 (dwarf2out_var_location): Reset views as needed. Precompute
6803 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
6804 attribute. Set view.
6805 (new_line_info_table): Reset next view.
6806 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
6807 (dwarf2out_source_line): Likewise. Output view resets and labels to
6808 the assembler, or select appropriate line info opcodes.
6809 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
6810 (optimize_string_length): Catch it. Adjust.
6811 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
6812 dw_val_class_view_list, and remove it if no longer needed.
6813 (hash_loc_list): Hash view numbers.
6814 (loc_list_hasher::equal): Compare them.
6815 (optimize_location_lists): Check whether a view list symbol is
6816 needed, and whether the locview attribute is present, and
6817 whether they match. Remove the locview attribute if no longer
6819 (index_location_lists): Call skip_loc_list_entry for test.
6820 (dwarf2out_finish): Call output_asm_line_debug_info for test.
6821 Use output_dwarf_version.
6822 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
6823 (struct dw_val_node): Add val_view_list.
6824 * final.c (SEEN_NEXT_VIEW): New.
6825 (set_next_view_needed): New.
6826 (clear_next_view_needed): New.
6827 (maybe_output_next_view): New.
6828 (final_start_function): Rename to...
6829 (final_start_function_1): ... this. Take pointer to FIRST,
6830 add SEEN parameter. Emit param bindings in the initial view.
6831 (final_start_function): Reintroduce SEEN-less interface.
6832 (final): Rename to...
6833 (final_1): ... this. Take SEEN parameter. Output final pending
6834 next view at the end.
6835 (final): Reintroduce seen-less interface.
6836 (final_scan_insn): Output pending next view before switching
6837 sections or ending a block. Mark the next view as needed when
6838 outputting variable locations. Notify debug backend of section
6839 changes, and of location view changes.
6840 (rest_of_handle_final): Adjust.
6841 * toplev.c (process_options): Autodetect value for debug variable
6842 location views option. Warn on incompat5 without -gdwarf-5.
6843 * doc/invoke.texi (gvariable-location-views): New.
6844 (gvariable-location-views=incompat5): New.
6845 (gno-variable-location-views): New.
6847 2018-02-08 David Malcolm <dmalcolm@redhat.com>
6849 PR tree-optimization/84136
6850 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
6851 that the result of find_edge is non-NULL.
6853 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6856 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
6857 storing integer register in SImode. Fix cost of 256 and 512
6858 byte aligned SSE register store.
6860 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6862 * config/i386/i386.c (ix86_multiplication_cost): Fix
6863 multiplication cost for TARGET_AVX512DQ.
6865 2018-02-08 Marek Polacek <polacek@redhat.com>
6867 PR tree-optimization/84238
6868 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
6871 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6873 PR tree-optimization/84265
6874 * tree-vect-stmts.c (vectorizable_store): Don't treat
6875 VMAT_CONTIGUOUS accesses as grouped.
6876 (vectorizable_load): Likewise.
6878 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6880 PR tree-optimization/81635
6881 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
6882 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
6883 (test_round_for_mask): New functions.
6884 (wide_int_cc_tests): Call test_round_for_mask.
6885 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
6886 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
6887 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
6888 range returned by get_range_info.
6890 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6893 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
6894 * symtab.c: Include builtins.h
6895 (symtab_node::output_to_lto_symbol_table_p): Move here
6896 from lto-streamer-out.c:output_symbol_p.
6897 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
6898 (output_symbol_p): Move all logic to symtab.c
6899 (produce_symtab): Update.
6901 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6903 * config/s390/s390-opts.h (enum indirect_branch): Define.
6904 * config/s390/s390-protos.h (s390_return_addr_from_memory)
6905 (s390_indirect_branch_via_thunk)
6906 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
6907 (enum s390_indirect_branch_type): Define.
6908 * config/s390/s390.c (struct s390_frame_layout, struct
6909 machine_function): Remove.
6910 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
6911 (indirect_branch_table_label_no, indirect_branch_table_name):
6913 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
6914 (enum s390_indirect_branch_option): Define.
6915 (s390_return_addr_from_memory): New function.
6916 (s390_handle_string_attribute): New function.
6917 (s390_attribute_table): Add new attribute handler.
6918 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
6919 (s390_indirect_branch_via_thunk): New function.
6920 (s390_indirect_branch_via_inline_thunk): New function.
6921 (s390_function_ok_for_sibcall): When jumping via thunk disallow
6922 sibling call optimization for non z10 compiles.
6923 (s390_emit_call): Force indirect branch target to be a single
6924 register. Add r1 clobber for non-z10 compiles.
6925 (s390_emit_epilogue): Emit return jump via return_use expander.
6926 (s390_reorg): Handle JUMP_INSNs as execute targets.
6927 (s390_option_override_internal): Perform validity checks for the
6928 new command line options.
6929 (s390_indirect_branch_attrvalue): New function.
6930 (s390_indirect_branch_settings): New function.
6931 (s390_set_current_function): Invoke s390_indirect_branch_settings.
6932 (s390_output_indirect_thunk_function): New function.
6933 (s390_code_end): Implement target hook.
6934 (s390_case_values_threshold): Implement target hook.
6935 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
6937 * config/s390/s390.h (struct s390_frame_layout)
6938 (struct machine_function): Move here from s390.c.
6939 (TARGET_INDIRECT_BRANCH_NOBP_RET)
6940 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
6941 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
6942 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
6943 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
6944 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
6945 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
6946 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
6947 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
6948 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
6949 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
6950 (mnemonic attribute): Add values which aren't recognized
6952 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
6953 pattern for branch conversion. Fix mnemonic attribute.
6954 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
6955 indirect branch via thunk if requested.
6956 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
6957 ("*indirect_jump"): Disable for branch conversion using out of
6959 ("indirect_jump_via_thunk<mode>_z10")
6960 ("indirect_jump_via_thunk<mode>")
6961 ("indirect_jump_via_inlinethunk<mode>_z10")
6962 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
6963 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
6964 ("casesi_jump_via_inlinethunk<mode>_z10")
6965 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
6966 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
6967 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
6968 ("*indirect2_jump"): Disable for branch conversion.
6969 ("casesi_jump"): Turn into expander and expand patterns for branch
6971 ("return_use"): New expander.
6972 ("*return"): Emit return via thunk and rename it to ...
6973 ("*return<mode>"): ... this one.
6974 * config/s390/s390.opt: Add new options and and enum for the
6977 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6979 * lra-constraints.c (match_reload): Unconditionally use
6980 gen_lowpart_SUBREG, rather than selecting between that
6981 and equivalent gen_rtx_SUBREG code.
6983 2018-02-08 Richard Biener <rguenther@suse.de>
6985 PR tree-optimization/84233
6986 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
6987 changed flag instead of boguously re-using phi_inserted.
6989 2018-02-08 Martin Jambor <mjambor@suse.cz>
6991 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
6992 static local variables.
6994 2018-02-08 Richard Biener <rguenther@suse.de>
6996 PR tree-optimization/84278
6997 * tree-vect-stmts.c (vectorizable_store): When looking for
6998 smaller vector types to perform grouped strided loads/stores
6999 make sure the mode is supported by the target.
7000 (vectorizable_load): Likewise.
7002 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7004 * config/aarch64/aarch64.c (aarch64_components_for_bb):
7005 Increase LDP/STP opportunities by adding adjacent callee-saves.
7007 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7009 PR rtl-optimization/84068
7010 PR rtl-optimization/83459
7011 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
7013 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
7015 PR tree-optimization/84224
7016 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
7017 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
7020 2018-02-07 Iain Sandoe <iain@codesourcery.com>
7023 * config/rs6000/altivec.md (*restore_world): Remove LR use.
7024 * config/rs6000/predicates.md (restore_world_operation): Adjust op
7025 count, remove one USE.
7027 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
7029 * doc/install.texi (Configuration): Document the
7030 --with-long-double-format={ibm,ieee} PowerPC configuration
7034 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
7035 Convert from define_expand to be define_insn_and_split. Rework
7036 float/double/_Float128 conversions to QI/HI/SImode to work with
7037 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
7038 conversions to QI/HImode types did a store and then a load to
7039 truncate the value. For conversions to VSX registers, don't split
7040 the insn, instead emit the code directly. Use the code iterator
7041 any_fix to combine signed and unsigned conversions.
7042 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
7043 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
7044 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7045 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
7046 (fix_<mode>di2_hw): Likewise.
7047 (fixuns_<mode>di2_hw): Likewise.
7048 (fix_<mode>si2_hw): Likewise.
7049 (fixuns_<mode>si2_hw): Likewise.
7050 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
7051 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7052 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
7053 fix<uns>_trunc<SFDF:mode>si2_p8.
7054 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
7056 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
7057 (fix<uns>_<mode>_mem): Likewise.
7058 (fctiw<u>z_<mode>_mem): Likewise.
7059 (fix<uns>_<mode>_mem): Likewise.
7060 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
7061 the register allocator from doing a direct move to the GPRs to do
7062 a store, and instead use the ISA 3.0 store byte/half-word from
7063 vector register instruction. For IEEE 128-bit floating point,
7064 also optimize stores of 32-bit ints.
7065 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
7067 2018-02-07 Alan Hayward <alan.hayward@arm.com>
7069 * genextract.c (push_pathstr_operand): New function to support
7071 (walk_rtx): Call push_pathstr_operand.
7072 (print_path): Support [a-zA-Z].
7074 2018-02-07 Richard Biener <rguenther@suse.de>
7076 PR tree-optimization/84037
7077 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
7078 (cse_and_gimplify_to_preheader): Declare.
7079 (vect_get_place_in_interleaving_chain): Likewise.
7080 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7082 (_loop_vec_info::~_loop_vec_info): Delete it.
7083 (cse_and_gimplify_to_preheader): New function.
7084 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
7085 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
7086 (vectorizable_load): Likewise. For grouped stores always base
7087 the IV on the first element.
7088 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
7089 condition before gimplifying.
7091 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7093 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
7094 *DIV_EXPR and *MOD_EXPR.
7096 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
7099 * config/i386/i386.c (ix86_option_override_internal): Mask out
7100 the CF_SET bit when checking -fcf-protection.
7102 2018-02-07 Tom de Vries <tom@codesourcery.com>
7105 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
7108 2018-02-07 Richard Biener <rguenther@suse.de>
7110 PR tree-optimization/84204
7111 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
7114 PR tree-optimization/84205
7115 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
7116 special-case isl_ast_op_zdiv_r.
7118 PR tree-optimization/84223
7119 * graphite-scop-detection.c (gather_bbs::before_dom_children):
7120 Only add conditions from within the region.
7121 (gather_bbs::after_dom_children): Adjust.
7123 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
7126 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
7127 * config/avr/avr.md: Only post-reload split REG-REG moves if
7128 either register is GENERAL_REG_P.
7130 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7132 PR tree-optimization/84235
7133 * tree-ssa-scopedtables.c
7134 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
7135 if the subtraction is performed in floating point type where NaNs are
7136 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
7137 build 1. Formatting fix.
7139 2018-02-06 Jakub Jelinek <jakub@redhat.com>
7142 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
7143 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
7144 and skip it regardless of bb boundaries. Use CALL_P macro,
7145 don't test INSN_P (insn) together with CALL_P or JUMP_P check
7146 unnecessarily, formatting fix.
7148 2018-02-06 Michael Collison <michael.collison@arm.com>
7150 * config/arm/thumb2.md:
7151 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
7152 (*thumb_mov_notscc): Ditto.
7154 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
7157 * config/rs6000/rs6000.md (su code attribute): Use "u" for
7158 unsigned_fix, not "s".
7160 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7162 * configure.ac (gcc_fn_eh_frame_ro): New function.
7163 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
7164 correct .eh_frame permissions.
7165 * configure: Regenerate.
7167 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
7169 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
7172 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7174 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7175 Display warning message for -mno-speculate-indirect-jumps.
7177 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
7179 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
7181 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
7183 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
7185 PR tree-optimization/84225
7186 * tree-eh.c (find_trapping_overflow): Only call
7187 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
7189 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7192 * config/i386/i386.c: Reimplement the check of possible options
7193 -mibt/-mshstk conbination. Change error messages.
7194 * doc/invoke.texi: Fix a typo: remove extra '='.
7196 2018-02-06 Marek Polacek <polacek@redhat.com>
7198 PR tree-optimization/84228
7199 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
7201 2018-02-06 Tamar Christina <tamar.christina@arm.com>
7204 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
7205 emitted arch directives.
7206 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
7207 __ARM_FEATURE_COPROC before changing architectures.
7209 2018-02-06 Richard Biener <rguenther@suse.de>
7211 * config/i386/i386.c (print_reg): Fix typo.
7212 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
7214 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
7216 * configure: Regenerate.
7218 2018-02-05 Martin Sebor <msebor@redhat.com>
7220 PR tree-optimization/83369
7221 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
7224 2018-02-05 Martin Liska <mliska@suse.cz>
7226 * doc/invoke.texi: Cherry-pick upstream r323995.
7228 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
7230 * ira.c (ira_init_register_move_cost): Adjust comment.
7232 2018-02-05 Martin Liska <mliska@suse.cz>
7234 PR gcov-profile/84137
7235 * doc/gcov.texi: Fix typo in documentation.
7237 2018-02-05 Martin Liska <mliska@suse.cz>
7239 PR gcov-profile/83879
7240 * doc/gcov.texi: Document necessity of --dynamic-list-data when
7241 using dlopen functionality.
7243 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
7245 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
7246 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
7247 _mm_maskz_range_ss, _mm_mask_range_round_ss,
7248 _mm_maskz_range_round_ss): New intrinsics.
7249 (__builtin_ia32_rangesd128_round)
7250 (__builtin_ia32_rangess128_round): Remove.
7251 (__builtin_ia32_rangesd128_mask_round,
7252 __builtin_ia32_rangess128_mask_round): New builtins.
7253 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
7254 __builtin_ia32_rangess128_round): Remove.
7255 (__builtin_ia32_rangesd128_mask_round,
7256 __builtin_ia32_rangess128_mask_round): New builtins.
7257 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
7258 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
7259 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
7260 "<round_saeonly_constraint>")): Changed to ...
7261 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
7262 "<round_saeonly_scalar_constraint>")): ... this.
7263 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
7264 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
7265 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
7266 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
7267 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
7269 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
7271 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
7273 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7274 Remove all values except native, 8540 and 8548.
7276 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
7278 * config/i386/i386.c (ix86_output_function_return): Pass
7279 INVALID_REGNUM, instead of -1, as invalid register number to
7280 indirect_thunk_name and output_indirect_thunk.
7282 2018-02-02 Julia Koval <julia.koval@intel.com>
7284 * config.gcc: Add -march=icelake.
7285 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7286 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7287 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7288 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7289 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7290 (processor_target_table): Add icelake.
7291 (ix86_option_override_internal): Handle new PTAs.
7292 (get_builtin_code_for_version): Handle icelake.
7293 (M_INTEL_COREI7_ICELAKE): New.
7294 (fold_builtin_cpu): Handle icelake.
7295 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7296 * doc/invoke.texi: Add -march=icelake.
7298 2018-02-02 Julia Koval <julia.koval@intel.com>
7300 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7301 to wide_int_bitmask.
7302 * wide-int-bitmask.h: New.
7304 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7307 * config/i386/i386.md: Replace Pmode with word_mode in
7308 builtin_setjmp_setup and builtin_longjmp to support x32.
7310 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
7314 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7316 (rs6000_supported_cpu_names): New static variable.
7317 (linux_cpu_translation_table): Likewise.
7318 (elf_platform) <cpu>: Define new static variable and use it.
7319 Translate kernel AT_PLATFORM name to canonical name if needed.
7320 Error if platform name is unknown.
7322 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
7325 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7327 2018-02-01 Jeff Law <law@redhat.com>
7330 * config/i386/i386.c (release_scratch_register_on_entry): Add new
7331 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
7332 the scratch if RELEASE_VIA_POP is false.
7333 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7334 If we have to save a temporary register, decrement SIZE appropriately.
7335 Pass new arguments to release_scratch_register_on_entry.
7336 (ix86_adjust_stack_and_probe): Likewise.
7337 (ix86_emit_probe_stack_range): Pass new arguments to
7338 release_scratch_register_on_entry.
7340 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
7342 PR rtl-optimization/84157
7343 * combine.c (change_zero_ext): Use REG_P predicate in
7344 front of HARD_REGISTER_P predicate.
7346 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
7348 * config/avr/avr.c (avr_option_override): Move disabling of
7349 -fdelete-null-pointer-checks to...
7350 * common/config/avr/avr-common.c (avr_option_optimization_table):
7353 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7355 PR tree-optimization/81635
7356 * tree-data-ref.c (split_constant_offset_1): For types that
7357 wrap on overflow, try to use range info to prove that wrapping
7360 2018-02-01 Renlin Li <renlin.li@arm.com>
7363 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7365 (aarch64_register_move_cost): Likewise.
7366 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7368 (REG_CLASS_NAMES): Likewise.
7369 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7370 TAILCALL_ADDR_REGS. Remove IP registers.
7371 * config/aarch64/aarch64.md (Ucs): Update register constraint.
7373 2018-02-01 Richard Biener <rguenther@suse.de>
7375 * domwalk.h (dom_walker::dom_walker): Add additional constructor
7376 for specifying RPO order and allow NULL for that.
7377 * domwalk.c (dom_walker::dom_walker): Likewise.
7378 (dom_walker::walk): Handle NULL RPO order.
7379 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7381 (rewrite_update_dom_walker): Likewise.
7382 (mark_def_dom_walker): Likewise.
7384 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7386 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7387 (aarch64_maybe_expand_sve_subreg_move): Declare.
7388 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7389 * config/aarch64/predicates.md (aarch64_any_register_operand): New
7391 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7392 that are semantically a reverse operation.
7393 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7394 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7395 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7397 (aarch64_can_change_mode_class): For big-endian, forbid changes
7398 between two SVE modes if they have different element sizes.
7400 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7402 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7403 the TImode handling for big-endian targets.
7405 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7407 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7408 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
7410 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7411 Remove BSWAP handing for big-endian targets and use the form of
7412 LD1RQ appropariate for the mode.
7414 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7416 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7417 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7420 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7423 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7424 check for operands that need to go through aarch64_sve_reload_be.
7426 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7428 PR tree-optimization/81661
7429 PR tree-optimization/84117
7430 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7431 * tree-eh.c: Include gimplify.h.
7432 (find_trapping_overflow, replace_trapping_overflow,
7433 rewrite_to_non_trapping_overflow): New functions.
7434 * tree-vect-loop.c: Include tree-eh.h.
7435 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7436 * tree-data-ref.c: Include tree-eh.h.
7437 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7439 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7441 PR rtl-optimization/84123
7442 * combine.c (change_zero_ext): Check if hard register satisfies
7443 can_change_dest_mode before calling gen_lowpart_SUBREG.
7445 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7448 * ira.c (ira_init_register_move_cost): Remove assert.
7450 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7452 PR rtl-optimization/84071
7453 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7454 * doc/tm.texi: Regenerate.
7456 2018-01-31 Richard Biener <rguenther@suse.de>
7458 PR tree-optimization/84132
7459 * tree-data-ref.c (analyze_miv_subscript): Properly
7460 check whether evolution_function_is_affine_multivariate_p
7461 before calling gcd_of_steps_may_divide_p.
7463 2018-01-31 Julia Koval <julia.koval@intel.com>
7466 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7467 * config/i386/i386.md (rdpid_rex64) New.
7468 (rdpid): Make 32bit only.
7470 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7473 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7474 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7476 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7479 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7481 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7483 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7485 PR rtl-optimization/84071
7486 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7487 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7489 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7491 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7492 (arc_attribute_table): Add 'aux' attribute.
7493 (arc_in_small_data_p): Consider aux like variables.
7494 (arc_is_aux_reg_p): New function.
7495 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7496 (arc_get_aux_arg): New function.
7497 (prepare_move_operands): Handle aux-register access.
7498 (arc_handle_aux_attribute): New function.
7499 * doc/extend.texi (ARC Variable attributes): Add subsection.
7501 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7503 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7504 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7505 (arc_attribute_table): Add 'uncached' attribute.
7506 (arc_print_operand): Print '.di' flag for uncached memory
7508 (arc_in_small_data_p): Do not consider for small data the uncached
7510 (arc_is_uncached_mem_p): New function.
7511 * config/arc/predicates.md (compact_store_memory_operand): Check
7512 for uncached memory accesses.
7513 (nonvol_nonimm_operand): Likewise.
7514 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7516 2018-01-31 Jakub Jelinek <jakub@redhat.com>
7519 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7520 falign-loops=): Add Optimization flag.
7522 2018-01-30 Jeff Law <law@redhat.com>
7525 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
7526 INT_REGISTERS_SAVED. Check it prior to calling
7527 get_scratch_register_on_entry.
7528 (ix86_adjust_stack_and_probe): Similarly.
7529 (ix86_emit_probe_stack_range): Similarly.
7530 (ix86_expand_prologue): Corresponding changes.
7532 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7535 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
7536 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
7538 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
7541 * lra-constraints.c (curr_insn_transform): Process AND in the
7544 2018-01-30 Jakub Jelinek <jakub@redhat.com>
7546 PR rtl-optimization/83986
7547 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7548 dependence against last_pending_memory_flush in addition to
7551 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
7553 PR tree-optimization/81611
7554 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7557 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7560 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7563 2018-01-30 Richard Biener <rguenther@suse.de>
7564 Jakub Jelinek <jakub@redhat.com>
7566 PR tree-optimization/84111
7567 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
7568 inner loops added during recursion, as they don't have up-to-date
7571 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7574 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
7575 (can_inline_edge_by_limits_p): ... here.
7576 (can_early_inline_edge_p, check_callers,
7577 update_caller_keys, update_callee_keys, recursive_inlining,
7578 add_new_edges_to_heap, speculation_useful_p,
7579 inline_small_functions,
7580 inline_small_functions, flatten_function,
7581 inline_to_all_callers_1): Update.
7583 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7585 * profile-count.c (profile_count::combine_with_ipa_count): Handle
7588 2018-01-30 Richard Biener <rguenther@suse.de>
7590 PR tree-optimization/83008
7591 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
7592 invariant and constant vector uses in stmts when they need
7595 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7598 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
7599 * configure: Regenerate.
7601 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7603 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
7605 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
7606 Use gen_rtx_REG rather than gen_lowpart.
7608 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7610 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
7611 rather than 0 when creating partial subregs.
7613 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7615 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
7618 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
7621 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
7622 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
7623 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
7624 flags. This restores the settings used before the 2017-07-24.
7625 Turning off pre increment/decrement/modify allows IVOPTS to
7626 optimize DF/SF loops where the index is an int.
7628 2018-01-29 Richard Biener <rguenther@suse.de>
7629 Kelvin Nilsen <kelvin@gcc.gnu.org>
7632 * tree-vect-stmts.c (vectorizable_call): Don't call
7633 targetm.vectorize_builtin_md_vectorized_function if callee is
7636 2018-01-22 Carl Love <cel@us.ibm.com>
7638 * doc/extend.tex: Fix typo in second arg in
7639 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
7641 2018-01-29 Richard Biener <rguenther@suse.de>
7643 PR tree-optimization/84086
7644 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
7645 (flush_ssaname_freelist): When SSA names were released reset
7646 the SCEV hash table.
7648 2018-01-29 Richard Biener <rguenther@suse.de>
7650 PR tree-optimization/84057
7651 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
7652 removed paths when removing edges.
7654 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
7656 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
7657 -mfunction-return=@var{choice}.
7659 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
7662 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
7664 (layout::print_source_line): Likewise.
7665 (test_get_line_width_without_trailing_whitespace): Add test cases.
7667 2018-01-27 Jakub Jelinek <jakub@redhat.com>
7670 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
7673 2018-01-26 Jim Wilson <jimw@sifive.com>
7675 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
7677 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
7680 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7682 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
7683 and CMP + SUB-immediate -> SUBS.
7685 2018-01-26 Martin Sebor <msebor@redhat.com>
7687 PR tree-optimization/83896
7688 * tree-ssa-strlen.c (get_string_len): Rename...
7689 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
7690 Avoid assuming length is constant.
7691 (handle_char_store): Use HOST_WIDE_INT for string length.
7693 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
7696 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
7697 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
7699 2018-01-26 Richard Biener <rguenther@suse.de>
7701 PR rtl-optimization/84003
7702 * dse.c (record_store): Only record redundant stores when
7703 the earlier store aliases at least all accesses the later one does.
7705 2018-01-26 Jakub Jelinek <jakub@redhat.com>
7707 PR rtl-optimization/83985
7708 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
7709 REG_CFA_RESTORE insns.
7710 (delete_unmarked_insns): Don't ignore separate shrink wrapping
7711 REG_CFA_RESTORE insns here.
7714 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
7715 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
7717 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7719 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
7720 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
7721 (arc_init): Likewise.
7722 (arc_override_options): Likewise.
7723 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
7725 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
7727 * config/arc/arc.h (TARGET_DBNZ): Define.
7728 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
7729 properly set the tune attribute.
7730 (dbnz): Use TARGET_DBNZ guard.
7731 * config/arc/arc.opt (mtune): Add core3 option.
7733 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7735 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
7736 recognize new pic like addresses.
7737 (arc_delegitimize_address): Clean up.
7739 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7741 * config/arc/arc-arches.def: Option mrf16 valid for all
7743 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
7744 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
7745 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
7746 * config/arc/arc-tables.opt: Regenerate.
7747 * config/arc/arc.c (arc_conditional_register_usage): Handle
7748 reduced register file case.
7749 (arc_file_start): Set must have build attributes.
7750 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
7752 * config/arc/arc.opt (mrf16): Add new option.
7753 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
7754 * config/arc/genmultilib.awk: Handle new mrf16 option.
7755 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
7756 * config/arc/t-multilib: Regenerate.
7757 * doc/invoke.texi (ARC Options): Document mrf16 option.
7759 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7761 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
7762 * config/arc/arc.c (arc_handle_secure_attribute): New function.
7763 (arc_attribute_table): Add 'secure_call' attribute.
7764 (arc_print_operand): Print secure call operand.
7765 (arc_function_ok_for_sibcall): Don't optimize tail calls when
7767 (arc_is_secure_call_p): New function. * config/arc/arc.md
7768 (call_i): Add support for sjli instruction.
7769 (call_value_i): Likewise.
7770 * config/arc/constraints.md (Csc): New constraint.
7772 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7773 John Eric Martin <John.Martin@emmicro-us.com>
7775 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
7776 * config/arc/arc.c (_arc_jli_section): New struct.
7777 (arc_jli_section): New type.
7778 (rc_jli_sections): New static variable.
7779 (arc_handle_jli_attribute): New function.
7780 (arc_attribute_table): Add jli_always and jli_fixed attribute.
7781 (arc_file_end): New function.
7782 (TARGET_ASM_FILE_END): Define.
7783 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
7784 (arc_add_jli_section): New function.
7785 (jli_call_scan): Likewise.
7786 (arc_reorg): Call jli_call_scan.
7787 (arc_output_addsi): Remove 'S' from printing asm operand.
7788 (arc_is_jli_call_p): New function.
7789 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
7791 (movhi_insn): Likewise.
7792 (movsi_insn): Likewise.
7793 (movsi_set_cc_insn): Likewise.
7794 (loadqi_update): Likewise.
7795 (load_zeroextendqisi_update): Likewise.
7796 (load_signextendqisi_update): Likewise.
7797 (loadhi_update): Likewise.
7798 (load_zeroextendhisi_update): Likewise.
7799 (load_signextendhisi_update): Likewise.
7800 (loadsi_update): Likewise.
7801 (loadsf_update): Likewise.
7802 (movsicc_insn): Likewise.
7803 (bset_insn): Likewise.
7804 (bxor_insn): Likewise.
7805 (bclr_insn): Likewise.
7806 (bmsk_insn): Likewise.
7807 (bicsi3_insn): Likewise.
7808 (cmpsi_cc_c_insn): Likewise.
7809 (movsi_ne): Likewise.
7810 (movsi_cond_exec): Likewise.
7811 (clrsbsi2): Likewise.
7822 (call_i): Remove 'S' asm letter, add jli instruction.
7823 (call_value_i): Likewise.
7824 * config/arc/arc.op (mjli-always): New option.
7825 * config/arc/constraints.md (Cji): New constraint.
7826 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
7828 (subsf3_fpx): Likewise.
7829 (mulsf3_fpx): Likewise.
7830 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
7832 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
7834 * doc/invoke.texi (ARC): Document mjli-always option.
7836 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
7838 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
7839 avoid addition with 0 and use incw and decw where possible.
7841 2018-01-26 Richard Biener <rguenther@suse.de>
7843 PR tree-optimization/81082
7844 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
7845 association if it requires casting to unsigned.
7846 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
7847 from fold_plusminus_mult_expr to catch important cases late when
7848 range info is available.
7850 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7852 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
7853 * configure.ac (hidden_linkonce): New test.
7854 * configure: Regenerate.
7855 * config.in: Regenerate.
7857 2018-01-26 Julia Koval <julia.koval@intel.com>
7859 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
7860 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
7861 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
7862 _mm_mask_bitshuffle_epi64_mask): Fix type.
7863 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
7864 USI_FTYPE_V4DI_V4DI_USI): Remove.
7865 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
7866 __builtin_ia32_vpshufbitqmb256_mask,
7867 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
7868 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
7869 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
7871 2018-01-26 Alan Modra <amodra@gmail.com>
7874 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
7875 UNSPEC_VBPERMQ. Sort other unspecs.
7877 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
7879 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
7881 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
7884 * predict.c (drop_profile): Do not push/pop cfun; update also
7886 (handle_missing_profiles): Fix logic looking for zero profiles.
7888 2018-01-25 Jakub Jelinek <jakub@redhat.com>
7891 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
7892 on functions with #pragma omp declare simd or functions with simd
7894 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
7895 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
7896 Remove trailing \n from warning_at calls.
7898 2018-01-25 Tom de Vries <tom@codesourcery.com>
7901 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7902 for neutered workers.
7904 2018-01-24 Joseph Myers <joseph@codesourcery.com>
7907 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
7908 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
7910 2018-01-24 Jeff Law <law@redhat.com>
7913 * i386.c (get_probe_interval): Move to earlier point.
7914 (ix86_compute_frame_layout): If -fstack-clash-protection and
7915 the frame is larger than the probe interval, then use pushes
7916 to save registers rather than reg->mem moves.
7917 (ix86_expand_prologue): Remove conditional for int_registers_saved
7920 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
7923 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
7924 min/max for never referenced object.
7926 2018-01-24 Jakub Jelinek <jakub@redhat.com>
7929 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
7931 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
7932 attributes from DECL_ATTRIBUTES (decl) without affecting
7933 DECL_ATTRIBUTES (current_function_decl).
7934 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
7935 functions with non-NULL DECL_ABSTRACT_ORIGIN.
7937 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
7939 PR tree-optimization/83979
7940 * fold-const.c (fold_comparison): Use constant_boolean_node
7941 instead of boolean_{true,false}_node.
7943 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
7945 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
7948 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7950 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7951 Simplify the clause that sets the length attribute.
7952 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7953 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
7954 clause that sets the length attribute.
7955 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7957 2018-01-24 Tom de Vries <tom@codesourcery.com>
7960 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
7961 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
7962 Add strict parameter.
7963 (prevent_branch_around_nothing): Insert dummy insn between branch to
7964 label and label with no ptx insn inbetween.
7965 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
7967 2018-01-24 Tom de Vries <tom@codesourcery.com>
7970 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7971 for neutered threads in warp.
7972 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
7974 2018-01-24 Richard Biener <rguenther@suse.de>
7976 PR tree-optimization/83176
7977 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
7980 2018-01-24 Richard Biener <rguenther@suse.de>
7982 PR tree-optimization/82819
7983 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
7984 code generating pluses that are no-ops in the target precision.
7986 2018-01-24 Richard Biener <rguenther@suse.de>
7989 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
7991 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7993 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
7994 to merge probabilities.
7995 * predict.c (probably_never_executed): Also mark as cold functions
7996 with global 0 profile and guessed local profile.
7997 * profile-count.c (profile_probability::combine_with_count): New
7999 * profile-count.h (profile_probability::operator*,
8000 profile_probability::operator*=, profile_probability::operator/,
8001 profile_probability::operator/=): Reduce precision to adjusted
8002 and set value to guessed on contradictory divisions.
8003 (profile_probability::combine_with_freq): Remove.
8004 (profile_probability::combine_wiht_count): Declare.
8005 (profile_count::force_nonzero):: Set to adjusted.
8006 (profile_count::probability_in):: Set quality to adjusted.
8007 * tree-ssa-tail-merge.c (replace_block_by): Use
8010 2018-01-23 Andrew Waterman <andrew@sifive.com>
8011 Jim Wilson <jimw@sifive.com>
8013 * config/riscv/riscv.c (riscv_stack_boundary): New.
8014 (riscv_option_override): Set riscv_stack_boundary. Handle
8015 riscv_preferred_stack_boundary_arg.
8016 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
8017 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
8018 (STACK_BOUNDARY): Set to riscv_stack_boundary.
8019 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
8020 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
8021 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
8023 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
8026 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
8027 of struct ix86_frame.
8028 (ix86_expand_epilogue): Likewise. Add a local variable for
8029 the reg_save_offset field in struct ix86_frame.
8031 2018-01-23 Bin Cheng <bin.cheng@arm.com>
8033 PR tree-optimization/82604
8034 * tree-loop-distribution.c (enum partition_kind): New enum item
8035 PKIND_PARTIAL_MEMSET.
8036 (partition_builtin_p): Support above new enum item.
8037 (generate_code_for_partition): Ditto.
8038 (compute_access_range): Differentiate cases that equality can be
8039 proven at all loops, the innermost loops or no loops.
8040 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
8041 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
8042 (finalize_partitions, distribute_loop): Don't fuse partition of
8043 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
8044 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
8047 2018-01-23 Martin Liska <mliska@suse.cz>
8049 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
8050 order to ignore the predictor.
8051 (PRED_POLYMORPHIC_CALL): Likewise.
8052 (PRED_RECURSIVE_CALL): Likewise.
8054 2018-01-23 Martin Liska <mliska@suse.cz>
8056 * tree-profile.c (tree_profiling): Print function header to
8057 aware reader which function we are working on.
8058 * value-prof.c (gimple_find_values_to_profile): Do not print
8059 not interesting value histograms.
8061 2018-01-23 Martin Liska <mliska@suse.cz>
8063 * profile-count.h (enum profile_quality): Add
8064 profile_uninitialized as the first value. Do not number values
8065 as they are zero based.
8066 (profile_count::verify): Update sanity check.
8067 (profile_probability::verify): Likewise.
8069 2018-01-23 Nathan Sidwell <nathan@acm.org>
8071 * doc/invoke.texi (ffor-scope): Deprecate.
8073 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8075 PR tree-optimization/83510
8076 * domwalk.c (set_all_edges_as_executable): New function.
8077 (dom_walker::dom_walker): Convert bool param
8078 "skip_unreachable_blocks" to enum reachability. Move setup of
8079 edge flags to set_all_edges_as_executable and only do it when
8080 reachability is REACHABLE_BLOCKS.
8081 * domwalk.h (enum dom_walker::reachability): New enum.
8082 (dom_walker::dom_walker): Convert bool param
8083 "skip_unreachable_blocks" to enum reachability.
8084 (set_all_edges_as_executable): New decl.
8085 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
8086 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
8088 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
8089 but converting true to REACHABLE_BLOCKS.
8090 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
8092 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
8093 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
8094 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
8096 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
8097 if check_all_array_refs will be called.
8099 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8101 * tree.c (selftest::test_location_wrappers): Add more test
8104 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8106 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
8107 (selftest::test_bit_in_range): Likewise.
8109 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8112 * doc/sourcebuild.texi (vect_float): Say that the selector
8113 only describes the situation when -funsafe-math-optimizations is on.
8114 (vect_float_strict): Document.
8116 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8118 PR tree-optimization/83965
8119 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
8120 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
8121 instead of checking only for a reduction.
8122 (vect_recog_widen_sum_pattern): Likewise.
8124 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8126 * predict.c (probably_never_executed): Only use precise profile info.
8127 (compute_function_frequency): Skip after inlining hack since we now
8128 have quality checking.
8130 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8132 * profile-count.h (profile_probability::very_unlikely,
8133 profile_probability::unlikely, profile_probability::even): Set
8134 precision to guessed.
8136 2018-01-23 Richard Biener <rguenther@suse.de>
8138 PR tree-optimization/83963
8139 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
8140 Properly terminate dominator walk when crossing the exit edge not
8141 when visiting its source block.
8143 2018-01-23 Jakub Jelinek <jakub@redhat.com>
8146 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
8147 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
8149 2018-01-22 Jakub Jelinek <jakub@redhat.com>
8151 PR tree-optimization/83957
8152 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
8153 semicolon after for body surrounded by braces.
8155 PR tree-optimization/83081
8156 * profile-count.h (profile_probability::split): New method.
8157 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
8158 Use profile_probability::split.
8159 (do_compare_rtx_and_jump): Fix adjustment of probabilities
8160 when splitting a single conditional jump into 2.
8162 2018-01-22 David Malcolm <dmalcolm@redhat.com>
8164 PR tree-optimization/69452
8165 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
8168 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8170 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
8171 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
8172 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
8174 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8176 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
8177 * config/rl78/rl78.md (movdi): New define_expand.
8178 * config/rl78/rl78.c (rl78_split_movdi): New function.
8180 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
8183 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
8185 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
8186 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
8187 128-bit to produce an UNSPEC move to get the double word with the
8188 signbit and then a shift directly to do signbit.
8189 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
8190 implementation with a new version that just does either a direct
8191 move or a regular move. Move memory interface to separate insns.
8192 Move insns so they are next to the expander.
8193 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
8194 with signbit move. Split big and little endian case.
8195 (signbit<mode>2_dm_mem_le): Likewise.
8196 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
8197 (signbit<mode>2_dm2): Likewise.
8199 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8201 * config/rl78/rl78.md (anddi3): New define_expand.
8203 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8205 * config/rl78/rl78.md (umindi3): New define_expand.
8207 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8209 * config/rl78/rl78.md (smindi3): New define_expand.
8211 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8213 * config/rl78/rl78.md (smaxdi3): New define_expand.
8215 2018-01-22 Carl Love <cel@us.ibm.com>
8217 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
8218 LVX_V1TI): Add macro expansion.
8219 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
8220 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
8221 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
8222 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
8223 Change check to determine if the instruction is a byte reversing
8224 entry. Fix typo in comment.
8225 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
8226 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
8227 Add def_builtin calls for new builtins.
8228 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
8229 Add define_insn expansion.
8231 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8233 * config/rl78/rl78.md (umaxdi3): New define_expand.
8235 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8237 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
8238 for non-QImode registers.
8240 2018-01-22 Richard Biener <rguenther@suse.de>
8242 PR tree-optimization/83963
8243 * graphite-scop-detection.c (scop_detection::get_sese): Delay
8244 including the loop exit block.
8245 (scop_detection::merge_sese): Likewise.
8246 (scop_detection::add_scop): Do it here instead.
8248 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8250 * doc/sourcebuild.texi (arm_softfloat): Document.
8252 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
8255 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
8256 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
8257 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
8259 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8260 David Edelsohn <dje.gcc@gmail.com>
8263 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8264 Change "crset eq" to "crset 2".
8265 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8266 (*call_indirect_aix<mode>_nospec): Likewise.
8267 (*call_value_indirect_aix<mode>_nospec): Likewise.
8268 (*call_indirect_elfv2<mode>_nospec): Likewise.
8269 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
8270 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
8271 change assembly output from . to $.
8272 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8273 (indirect_jump<mode>_nospec): Change assembly output from . to $.
8274 (*tablejump<mode>_internal1_nospec): Likewise.
8276 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
8279 * config/sh/sh_optimize_sett_clrt.cc:
8280 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8282 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
8284 PR tree-optimization/83940
8285 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8286 offset_dt to vect_constant_def rather than vect_unknown_def_type.
8287 (vect_check_load_store_mask): Add a mask_dt_out parameter and
8288 use it to pass back the definition type.
8289 (vect_check_store_rhs): Likewise rhs_dt_out.
8290 (vect_build_gather_load_calls): Add a mask_dt argument and use
8291 it instead of a call to vect_is_simple_use.
8292 (vectorizable_store): Update calls to vect_check_load_store_mask
8293 and vect_check_store_rhs. Use the dt returned by the latter instead
8294 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
8295 instead of calls to vect_is_simple_use. Pass the scalar rather
8296 than the vector operand to vect_is_simple_use when handling
8297 second and subsequent copies of an rhs value.
8298 (vectorizable_load): Update calls to vect_check_load_store_mask
8299 and vect_build_gather_load_calls. Use the cached mask_dt and
8300 gs_info.offset_dt instead of calls to vect_is_simple_use.
8302 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8305 * tree-emutls.c: Include gimplify.h.
8306 (lower_emutls_2): New function.
8307 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8308 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8309 it before further processing.
8312 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8313 UINTVAL (trueop1) instead of INTVAL (op1).
8315 2018-01-19 Jakub Jelinek <jakub@redhat.com>
8319 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8320 INCOMING_FRAME_SP_OFFSET if not defined.
8321 (scan_trace): Add ENTRY argument. If true and
8322 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8323 emit a note to adjust the CFA offset.
8324 (create_cfi_notes): Adjust scan_trace callers.
8325 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8326 INCOMING_FRAME_SP_OFFSET in the CIE.
8327 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8328 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8330 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8331 * doc/tm.texi: Regenerated.
8333 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8335 PR rtl-optimization/83147
8336 * lra-constraints.c (remove_inheritance_pseudos): Use
8337 lra_substitute_pseudo_within_insn.
8339 2018-01-19 Tom de Vries <tom@codesourcery.com>
8340 Cesar Philippidis <cesar@codesourcery.com>
8343 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8345 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
8348 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8349 spaces for function labels.
8351 2018-01-19 Martin Liska <mliska@suse.cz>
8353 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8354 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8355 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8356 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8357 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8358 (PRED_CONST_RETURN): Change from 69 to 65.
8359 (PRED_NULL_RETURN): Change from 91 to 71.
8360 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8361 (PRED_LOOP_GUARD): Change from 66 to 73.
8363 2018-01-19 Martin Liska <mliska@suse.cz>
8365 * predict.c (predict_insn_def): Add new assert.
8366 (struct branch_predictor): Change type to signed integer.
8367 (test_prediction_value_range): Amend test to cover
8369 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8370 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8371 (PRED_LOOP_ITERATIONS_MAX): Likewise.
8372 (PRED_LOOP_IV_COMPARE): Likewise.
8373 * predict.h (PROB_UNINITIALIZED): Define new constant.
8375 2018-01-19 Martin Liska <mliska@suse.cz>
8377 * predict.c (dump_prediction): Add new format for
8378 analyze_brprob.py script which is enabled with -details
8380 * profile-count.h (precise_p): New function.
8382 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8384 PR tree-optimization/83922
8385 * tree-vect-loop.c (vect_verify_full_masking): Return false if
8386 there are no statements that need masking.
8387 (vect_active_double_reduction_p): New function.
8388 (vect_analyze_loop_operations): Use it when handling phis that
8389 are not in the loop header.
8391 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8393 PR tree-optimization/83914
8394 * tree-vect-loop.c (vectorizable_induction): Don't convert
8395 init_expr or apply the peeling adjustment for inductions
8396 that are nested within the vectorized loop.
8398 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8400 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8403 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8407 * function.h (gimplify_parameters): Add gimple_seq * argument.
8408 * function.c: Include gimple.h and options.h.
8409 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8410 for the added local temporaries if needed.
8411 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8412 if there are any parameter cleanups, wrap whole body into a
8413 try/finally with the cleanups.
8415 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8418 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8419 Use GET_MODE_CLASS for scalar floating point.
8421 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8425 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8426 Fix call of call_cgraph_insertion_hooks.
8428 2018-01-18 Martin Sebor <msebor@redhat.com>
8430 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8432 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8435 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8438 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8441 * common.opt: (-ffile-prefix-map): New option.
8442 * opts.c (common_handle_option): Defer it.
8443 * opts-global.c (handle_common_deferred_options): Handle it.
8444 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8445 * file-prefix-map.h: New file.
8446 (remap_debug_filename, add_debug_prefix_map): ...here.
8447 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8448 * final.c (debug_prefix_map, add_debug_prefix_map
8449 remap_debug_filename): Move to...
8450 * file-prefix-map.c: New file.
8451 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8452 generalize, get rid of alloca(), use strrchr() instead of strchr().
8453 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8454 Implement in terms of add_prefix_map().
8455 (remap_macro_filename, remap_debug_filename): Implement in term of
8457 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8458 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8459 * dbxout.c: Include file-prefix-map.h.
8460 * varasm.c: Likewise.
8461 * vmsdbgout.c: Likewise.
8462 * xcoffout.c: Likewise.
8463 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8464 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8465 * doc/invoke.texi (-ffile-prefix-map): Document.
8466 (-fdebug-prefix-map): Update description.
8468 2018-01-18 Martin Liska <mliska@suse.cz>
8470 * config/i386/i386.c (indirect_thunk_name): Document that also
8472 (output_indirect_thunk): Document why both instructions
8473 (pause and lfence) are generated.
8475 2018-01-18 Richard Biener <rguenther@suse.de>
8477 PR tree-optimization/83887
8478 * graphite-scop-detection.c
8479 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8480 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8481 (scop_detection::merge_sese): Re-implement with a flood-fill
8482 algorithm that properly finds a SESE region if it exists.
8484 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8487 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8488 pointer_diff optimizations use view_convert instead of convert.
8490 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8492 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8493 Generate different code for -mno-speculate-indirect-jumps.
8494 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8495 (*call_indirect_aix<mode>): Disable for
8496 -mno-speculate-indirect-jumps.
8497 (*call_indirect_aix<mode>_nospec): New define_insn.
8498 (*call_value_indirect_aix<mode>): Disable for
8499 -mno-speculate-indirect-jumps.
8500 (*call_value_indirect_aix<mode>_nospec): New define_insn.
8501 (*sibcall_nonlocal_sysv<mode>): Generate different code for
8502 -mno-speculate-indirect-jumps.
8503 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8505 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8507 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8508 long double type, set the flags for noting the default long double
8509 type, even if we don't pass or return a long double type.
8511 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
8514 * ipa-inline.c (flatten_function): Do not overwrite final inlining
8517 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
8519 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8520 support for merge[hl].
8521 (fold_mergehl_helper): New helper function.
8522 (tree-vector-builder.h): New #include for tree_vector_builder usage.
8523 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
8524 (altivec_vmrglw_direct): Add xxmrglw insn.
8526 2018-01-17 Andrew Waterman <andrew@sifive.com>
8528 * config/riscv/riscv.c (riscv_conditional_register_usage): If
8529 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
8531 2018-01-17 David Malcolm <dmalcolm@redhat.com>
8534 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
8535 call the lto_location_cache before reading the
8536 DECL_SOURCE_LOCATION of the types.
8538 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
8539 Richard Sandiford <richard.sandiford@linaro.org>
8541 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8542 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8543 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8544 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8545 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8547 * config/aarch64/constraints.md (aarch64_movti_operand):
8549 * config/aarch64/predicates.md (Uti): Add new constraint.
8551 2018-01-17 Carl Love <cel@us.ibm.com>
8553 * config/rs6000/vsx.md (define_expand xl_len_r,
8554 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8555 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8557 (define_expand, define_insn): Move the shift left from the
8558 define_insn to the define_expand for lxvl and stxvl instructions.
8559 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8560 and XL_LEN_R definitions to PURE.
8562 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
8564 * config/i386/i386.c (indirect_thunk_name): Declare regno
8565 as unsigned int. Compare regno with INVALID_REGNUM.
8566 (output_indirect_thunk): Ditto.
8567 (output_indirect_thunk_function): Ditto.
8568 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
8569 in the call to output_indirect_thunk_function.
8571 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
8574 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
8575 rather than the size of inner_type to determine the stack slot size
8576 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
8578 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
8581 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
8584 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
8586 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
8587 endian Linux systems to optionally enable multilibs for selecting
8588 the long double type if the user configured an explicit type.
8589 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
8590 have no long double multilibs if not defined.
8591 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
8592 warn if the user used -mabi={ieee,ibm}longdouble and we built
8593 multilibs for long double.
8594 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
8595 appropriate multilib option.
8596 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
8598 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
8599 for building long double multilibs.
8600 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
8602 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
8604 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
8607 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
8609 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
8612 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
8615 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
8618 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8620 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
8621 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
8623 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8625 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
8626 different rtl trees depending on TARGET_64BIT.
8627 (rs6000_gen_lvx): Likewise.
8629 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8631 * config/visium/visium.md (nop): Tweak comment.
8632 (hazard_nop): Likewise.
8634 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8636 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
8637 -mspeculate-indirect-jumps.
8638 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
8639 for -mno-speculate-indirect-jumps.
8640 (*call_indirect_elfv2<mode>_nospec): New define_insn.
8641 (*call_value_indirect_elfv2<mode>): Disable for
8642 -mno-speculate-indirect-jumps.
8643 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
8644 (indirect_jump): Emit different RTL for
8645 -mno-speculate-indirect-jumps.
8646 (*indirect_jump<mode>): Disable for
8647 -mno-speculate-indirect-jumps.
8648 (*indirect_jump<mode>_nospec): New define_insn.
8649 (tablejump): Emit different RTL for
8650 -mno-speculate-indirect-jumps.
8651 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
8652 (tablejumpsi_nospec): New define_expand.
8653 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
8654 (tablejumpdi_nospec): New define_expand.
8655 (*tablejump<mode>_internal1): Disable for
8656 -mno-speculate-indirect-jumps.
8657 (*tablejump<mode>_internal1_nospec): New define_insn.
8658 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
8661 2018-01-16 Artyom Skrobov tyomitch@gmail.com
8663 * caller-save.c (insert_save): Drop unnecessary parameter. All
8666 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8667 Richard Biener <rguenth@suse.de>
8670 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
8671 return early, inline manually is_gimple_sizepos. Make sure if we
8672 call gimplify_expr we don't end up with a gimple constant.
8673 * tree.c (variably_modified_type_p): Don't return true for
8674 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
8675 * gimplify.h (is_gimple_sizepos): Remove.
8677 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8679 PR tree-optimization/83857
8680 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
8681 vectorizable_live_operation for pure SLP statements.
8682 (vectorizable_live_operation): Handle PHIs.
8684 2018-01-16 Richard Biener <rguenther@suse.de>
8686 PR tree-optimization/83867
8687 * tree-vect-stmts.c (vect_transform_stmt): Precompute
8688 nested_in_vect_loop_p since the scalar stmt may get invalidated.
8690 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8693 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
8694 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
8695 If off is not INTEGER_CST, issue a may not be aligned warning
8696 rather than isn't aligned. Use isn%'t rather than isn't.
8697 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
8699 <case MULT_EXPR>: Improve the case when bottom and one of the
8700 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
8701 operand, in that case check if the other operand is multiple of
8702 bottom divided by the INTEGER_CST operand.
8704 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8707 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
8708 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
8709 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
8710 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
8711 * config/pa/pa.c (pa_function_arg_advance): Likewise.
8712 (pa_function_arg, pa_arg_partial_bytes): Likewise.
8713 (pa_function_arg_size): New function.
8715 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8717 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
8718 in a separate statement.
8720 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8722 PR tree-optimization/83847
8723 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
8724 group gathers and scatters.
8726 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8728 PR rtl-optimization/86620
8729 * params.def (max-sched-ready-insns): Bump minimum value to 1.
8731 PR rtl-optimization/83213
8732 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
8733 to last if both are JUMP_INSNs.
8735 PR tree-optimization/83843
8736 * gimple-ssa-store-merging.c
8737 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
8738 store_immediate_info for bswap/nop orig_stores.
8740 2018-01-15 Andrew Waterman <andrew@sifive.com>
8742 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
8744 <UDIV>: Increase cost if !TARGET_DIV.
8746 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
8748 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
8749 (define_attr "cr_logical_3op"): New.
8750 (cceq_ior_compare): Adjust.
8751 (cceq_ior_compare_complement): Adjust.
8752 (*cceq_rev_compare): Adjust.
8753 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
8754 (is_cracked_insn): Adjust.
8755 (insn_must_be_first_in_group): Adjust.
8756 * config/rs6000/40x.md: Adjust.
8757 * config/rs6000/440.md: Adjust.
8758 * config/rs6000/476.md: Adjust.
8759 * config/rs6000/601.md: Adjust.
8760 * config/rs6000/603.md: Adjust.
8761 * config/rs6000/6xx.md: Adjust.
8762 * config/rs6000/7450.md: Adjust.
8763 * config/rs6000/7xx.md: Adjust.
8764 * config/rs6000/8540.md: Adjust.
8765 * config/rs6000/cell.md: Adjust.
8766 * config/rs6000/e300c2c3.md: Adjust.
8767 * config/rs6000/e500mc.md: Adjust.
8768 * config/rs6000/e500mc64.md: Adjust.
8769 * config/rs6000/e5500.md: Adjust.
8770 * config/rs6000/e6500.md: Adjust.
8771 * config/rs6000/mpc.md: Adjust.
8772 * config/rs6000/power4.md: Adjust.
8773 * config/rs6000/power5.md: Adjust.
8774 * config/rs6000/power6.md: Adjust.
8775 * config/rs6000/power7.md: Adjust.
8776 * config/rs6000/power8.md: Adjust.
8777 * config/rs6000/power9.md: Adjust.
8778 * config/rs6000/rs64.md: Adjust.
8779 * config/rs6000/titan.md: Adjust.
8781 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8783 * config/i386/predicates.md (indirect_branch_operand): Rewrite
8784 ix86_indirect_branch_register logic.
8786 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8788 * config/i386/constraints.md (Bs): Update
8789 ix86_indirect_branch_register check. Don't check
8790 ix86_indirect_branch_register with GOT_memory_operand.
8792 * config/i386/predicates.md (GOT_memory_operand): Don't check
8793 ix86_indirect_branch_register here.
8794 (GOT32_symbol_operand): Likewise.
8796 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8798 * config/i386/predicates.md (constant_call_address_operand):
8799 Rewrite ix86_indirect_branch_register logic.
8800 (sibcall_insn_operand): Likewise.
8802 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8804 * config/i386/constraints.md (Bs): Replace
8805 ix86_indirect_branch_thunk_register with
8806 ix86_indirect_branch_register.
8808 * config/i386/i386.md (indirect_jump): Likewise.
8809 (tablejump): Likewise.
8810 (*sibcall_memory): Likewise.
8811 (*sibcall_value_memory): Likewise.
8812 Peepholes of indirect call and jump via memory: Likewise.
8813 * config/i386/i386.opt: Likewise.
8814 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8815 (GOT_memory_operand): Likewise.
8816 (call_insn_operand): Likewise.
8817 (sibcall_insn_operand): Likewise.
8818 (GOT32_symbol_operand): Likewise.
8820 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8823 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
8824 type rather than type addr's type points to.
8825 (expand_omp_atomic_mutex): Likewise.
8826 (expand_omp_atomic): Likewise.
8828 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8831 * config/i386/i386.c (output_indirect_thunk_function): Use
8832 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
8833 for __x86_return_thunk.
8835 2018-01-15 Richard Biener <rguenther@suse.de>
8838 * expmed.c (extract_bit_field_1): Fix typo.
8840 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8843 * config/arm/iterators.md (VF): New mode iterator.
8844 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
8845 Remove integer-related logic from pattern.
8846 (neon_vabd<mode>_3): Likewise.
8848 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8851 * common.opt (fstrict-overflow): No longer an alias.
8852 (fwrapv-pointer): New option.
8853 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
8854 also for pointer types based on flag_wrapv_pointer.
8855 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
8856 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
8857 opts->x_flag_wrapv got set.
8858 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
8859 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
8860 POINTER_TYPE_OVERFLOW_UNDEFINED.
8861 * match.pd: Likewise in address comparison pattern.
8862 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
8864 2018-01-15 Richard Biener <rguenther@suse.de>
8867 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
8868 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
8869 Reset type names to their identifier if their TYPE_DECL doesn't
8870 have linkage (and thus is used for ODR and devirt).
8871 (save_debug_info_for_decl): Remove.
8872 (save_debug_info_for_type): Likewise.
8873 (add_tree_to_fld_list): Adjust.
8874 * tree-pretty-print.c (dump_generic_node): Make dumping of
8875 type names more robust.
8877 2018-01-15 Richard Biener <rguenther@suse.de>
8879 * BASE-VER: Bump to 8.0.1.
8881 2018-01-14 Martin Sebor <msebor@redhat.com>
8884 * builtins.c (check_access): Avoid warning when the no-warning bit
8887 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
8889 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
8890 * ira-color (allocno_hard_regs_compare): Likewise.
8892 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
8895 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
8896 Use .pushsection/.popsection.
8898 2018-01-14 Martin Sebor <msebor@redhat.com>
8901 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
8903 2018-01-14 Jakub Jelinek <jakub@redhat.com>
8905 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
8906 entry from extra_headers.
8907 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
8908 extra_headers, make the list bitwise identical to the i?86-*-* one.
8910 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8912 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
8913 -mcmodel=large with -mindirect-branch=thunk,
8914 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
8915 -mfunction-return=thunk-extern.
8916 * doc/invoke.texi: Document -mcmodel=large is incompatible with
8917 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
8918 -mfunction-return=thunk and -mfunction-return=thunk-extern.
8920 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8922 * config/i386/i386.c (print_reg): Print the name of the full
8923 integer register without '%'.
8924 (ix86_print_operand): Handle 'V'.
8925 * doc/extend.texi: Document 'V' modifier.
8927 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8929 * config/i386/constraints.md (Bs): Disallow memory operand for
8930 -mindirect-branch-register.
8932 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8933 (GOT_memory_operand): Likewise.
8934 (call_insn_operand): Likewise.
8935 (sibcall_insn_operand): Likewise.
8936 (GOT32_symbol_operand): Likewise.
8937 * config/i386/i386.md (indirect_jump): Call convert_memory_address
8938 for -mindirect-branch-register.
8939 (tablejump): Likewise.
8940 (*sibcall_memory): Likewise.
8941 (*sibcall_value_memory): Likewise.
8942 Disallow peepholes of indirect call and jump via memory for
8943 -mindirect-branch-register.
8944 (*call_pop): Replace m with Bw.
8945 (*call_value_pop): Likewise.
8946 (*sibcall_pop_memory): Replace m with Bs.
8947 * config/i386/i386.opt (mindirect-branch-register): New option.
8948 * doc/invoke.texi: Document -mindirect-branch-register option.
8950 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8952 * config/i386/i386-protos.h (ix86_output_function_return): New.
8953 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
8954 set function_return_type.
8955 (indirect_thunk_name): Add ret_p to indicate thunk for function
8957 (output_indirect_thunk_function): Pass false to
8958 indirect_thunk_name.
8959 (ix86_output_indirect_branch_via_reg): Likewise.
8960 (ix86_output_indirect_branch_via_push): Likewise.
8961 (output_indirect_thunk_function): Create alias for function
8962 return thunk if regno < 0.
8963 (ix86_output_function_return): New function.
8964 (ix86_handle_fndecl_attribute): Handle function_return.
8965 (ix86_attribute_table): Add function_return.
8966 * config/i386/i386.h (machine_function): Add
8967 function_return_type.
8968 * config/i386/i386.md (simple_return_internal): Use
8969 ix86_output_function_return.
8970 (simple_return_internal_long): Likewise.
8971 * config/i386/i386.opt (mfunction-return=): New option.
8972 (indirect_branch): Mention -mfunction-return=.
8973 * doc/extend.texi: Document function_return function attribute.
8974 * doc/invoke.texi: Document -mfunction-return= option.
8976 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8978 * config/i386/i386-opts.h (indirect_branch): New.
8979 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
8980 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
8981 with local indirect jump when converting indirect call and jump.
8982 (ix86_set_indirect_branch_type): New.
8983 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
8984 (indirectlabelno): New.
8985 (indirect_thunk_needed): Likewise.
8986 (indirect_thunk_bnd_needed): Likewise.
8987 (indirect_thunks_used): Likewise.
8988 (indirect_thunks_bnd_used): Likewise.
8989 (INDIRECT_LABEL): Likewise.
8990 (indirect_thunk_name): Likewise.
8991 (output_indirect_thunk): Likewise.
8992 (output_indirect_thunk_function): Likewise.
8993 (ix86_output_indirect_branch_via_reg): Likewise.
8994 (ix86_output_indirect_branch_via_push): Likewise.
8995 (ix86_output_indirect_branch): Likewise.
8996 (ix86_output_indirect_jmp): Likewise.
8997 (ix86_code_end): Call output_indirect_thunk_function if needed.
8998 (ix86_output_call_insn): Call ix86_output_indirect_branch if
9000 (ix86_handle_fndecl_attribute): Handle indirect_branch.
9001 (ix86_attribute_table): Add indirect_branch.
9002 * config/i386/i386.h (machine_function): Add indirect_branch_type
9003 and has_local_indirect_jump.
9004 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
9006 (tablejump): Likewise.
9007 (*indirect_jump): Use ix86_output_indirect_jmp.
9008 (*tablejump_1): Likewise.
9009 (simple_return_indirect_internal): Likewise.
9010 * config/i386/i386.opt (mindirect-branch=): New option.
9011 (indirect_branch): New.
9014 (thunk-inline): Likewise.
9015 (thunk-extern): Likewise.
9016 * doc/extend.texi: Document indirect_branch function attribute.
9017 * doc/invoke.texi: Document -mindirect-branch= option.
9019 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
9022 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
9024 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
9026 * ipa-inline.c (want_inline_small_function_p): Return false if
9027 inlining has already failed with CIF_FINAL_ERROR.
9028 (update_caller_keys): Call want_inline_small_function_p before
9030 (update_callee_keys): Likewise.
9032 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9034 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
9036 (rs6000_quadword_masked_address_p): Likewise.
9037 (quad_aligned_load_p): Likewise.
9038 (quad_aligned_store_p): Likewise.
9039 (const_load_sequence_p): Add comment to describe the outer-most loop.
9040 (mimic_memory_attributes_and_flags): New function.
9041 (rs6000_gen_stvx): Likewise.
9042 (replace_swapped_aligned_store): Likewise.
9043 (rs6000_gen_lvx): Likewise.
9044 (replace_swapped_aligned_load): Likewise.
9045 (replace_swapped_load_constant): Capitalize argument name in
9046 comment describing this function.
9047 (rs6000_analyze_swaps): Add a third pass to search for vector loads
9048 and stores that access quad-word aligned addresses and replace
9049 with stvx or lvx instructions when appropriate.
9050 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
9051 New function prototype.
9052 (rs6000_quadword_masked_address_p): Likewise.
9053 (rs6000_gen_lvx): Likewise.
9054 (rs6000_gen_stvx): Likewise.
9055 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
9056 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
9057 when memory address is aligned.
9058 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
9059 this split to select lvx instruction when memory address is aligned.
9060 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
9061 instruction when memory address is aligned.
9062 (*vsx_le_perm_load_v16qi): Likewise.
9063 (four unnamed splitters): Modify to select the stvx instruction
9064 when memory is aligned.
9066 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
9068 * predict.c (determine_unlikely_bbs): Handle correctly BBs
9069 which appears in the queue multiple times.
9071 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9072 Alan Hayward <alan.hayward@arm.com>
9073 David Sherwood <david.sherwood@arm.com>
9075 * tree-vectorizer.h (vec_lower_bound): New structure.
9076 (_loop_vec_info): Add check_nonzero and lower_bounds.
9077 (LOOP_VINFO_CHECK_NONZERO): New macro.
9078 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
9079 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
9080 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
9081 fields. Make seg_len the distance travelled, not including the
9083 (dr_direction_indicator): Declare.
9084 (dr_zero_step_indicator): Likewise.
9085 (dr_known_forward_stride_p): Likewise.
9086 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
9088 (runtime_alias_check_p): Allow runtime alias checks with
9090 (operator ==): Compare access_size and align.
9091 (prune_runtime_alias_test_list): Rework for new distinction between
9092 the access_size and seg_len.
9093 (create_intersect_range_checks_index): Likewise. Cope with polynomial
9095 (get_segment_min_max): New function.
9096 (create_intersect_range_checks): Use it.
9097 (dr_step_indicator): New function.
9098 (dr_direction_indicator): Likewise.
9099 (dr_zero_step_indicator): Likewise.
9100 (dr_known_forward_stride_p): Likewise.
9101 * tree-loop-distribution.c (data_ref_segment_size): Return
9102 DR_STEP * (niters - 1).
9103 (compute_alias_check_pairs): Update call to the dr_with_seg_len
9105 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
9106 (vect_preserves_scalar_order_p): New function, split out from...
9107 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
9108 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
9109 (vect_vfa_access_size): New function.
9110 (vect_vfa_align): Likewise.
9111 (vect_compile_time_alias): Take access_size_a and access_b arguments.
9112 (dump_lower_bound): New function.
9113 (vect_check_lower_bound): Likewise.
9114 (vect_small_gap_p): Likewise.
9115 (vectorizable_with_step_bound_p): Likewise.
9116 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
9117 depencies if the vectorization factor is 1. Convert the checks
9118 for nonzero steps into checks on the bounds of DR_STEP. Try using
9119 a bunds check for variable steps if the minimum required step is
9120 relatively small. Update calls to the dr_with_seg_len
9121 constructor and to vect_compile_time_alias.
9122 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
9124 (vect_loop_versioning): Call it.
9125 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
9127 (vect_estimate_min_profitable_iters): Account for any bounds checks.
9129 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9130 Alan Hayward <alan.hayward@arm.com>
9131 David Sherwood <david.sherwood@arm.com>
9133 * doc/sourcebuild.texi (vect_scatter_store): Document.
9134 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
9136 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
9138 * genopinit.c (main): Add supports_vec_scatter_store and
9139 supports_vec_scatter_store_cached to target_optabs.
9140 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
9141 IFN_MASK_SCATTER_STORE.
9142 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
9144 * internal-fn.h (internal_store_fn_p): Declare.
9145 (internal_fn_stored_value_index): Likewise.
9146 * internal-fn.c (scatter_store_direct): New macro.
9147 (expand_scatter_store_optab_fn): New function.
9148 (direct_scatter_store_optab_supported_p): New macro.
9149 (internal_store_fn_p): New function.
9150 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
9151 IFN_MASK_SCATTER_STORE.
9152 (internal_fn_mask_index): Likewise.
9153 (internal_fn_stored_value_index): New function.
9154 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
9156 * optabs-query.h (supports_vec_scatter_store_p): Declare.
9157 * optabs-query.c (supports_vec_scatter_store_p): New function.
9158 * tree-vectorizer.h (vect_get_store_rhs): Declare.
9159 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
9160 true for scatter stores.
9161 (vect_gather_scatter_fn_p): Handle scatter stores too.
9162 (vect_check_gather_scatter): Consider using scatter stores if
9163 supports_vec_scatter_store_p.
9164 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
9166 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9167 internal_fn_stored_value_index.
9168 (check_load_store_masking): Handle scatter stores too.
9169 (vect_get_store_rhs): Make public.
9170 (vectorizable_call): Use internal_store_fn_p.
9171 (vectorizable_store): Handle scatter store internal functions.
9172 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
9173 when deciding whether the end of the group has been reached.
9174 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
9175 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
9176 (mask_scatter_store<mode>): New insns.
9178 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9179 Alan Hayward <alan.hayward@arm.com>
9180 David Sherwood <david.sherwood@arm.com>
9182 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
9183 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
9184 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
9186 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
9187 Use vect_truncate_gather_scatter_offset if we can't treat the
9188 operation as a normal gather load or scatter store.
9189 (get_group_load_store_type): Take the gather_scatter_info
9190 as argument. Try using a gather load or scatter store for
9191 single-element groups.
9192 (get_load_store_type): Update calls to get_group_load_store_type
9193 and vect_use_strided_gather_scatters_p.
9195 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9196 Alan Hayward <alan.hayward@arm.com>
9197 David Sherwood <david.sherwood@arm.com>
9199 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
9200 optional tree argument.
9201 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
9203 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
9204 but continue to use the current value as a fallback.
9205 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
9206 to compare the updates.
9207 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
9208 (get_load_store_type): Use it when handling a strided access.
9209 (vect_get_strided_load_store_ops): New function.
9210 (vect_get_data_ptr_increment): Likewise.
9211 (vectorizable_load): Handle strided gather loads. Always pass
9212 a step to vect_create_data_ref_ptr and bump_vector_ptr.
9214 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9215 Alan Hayward <alan.hayward@arm.com>
9216 David Sherwood <david.sherwood@arm.com>
9218 * doc/md.texi (gather_load@var{m}): Document.
9219 (mask_gather_load@var{m}): Likewise.
9220 * genopinit.c (main): Add supports_vec_gather_load and
9221 supports_vec_gather_load_cached to target_optabs.
9222 * optabs-tree.c (init_tree_optimization_optabs): Use
9223 ggc_cleared_alloc to allocate target_optabs.
9224 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
9225 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
9227 * internal-fn.h (internal_load_fn_p): Declare.
9228 (internal_gather_scatter_fn_p): Likewise.
9229 (internal_fn_mask_index): Likewise.
9230 (internal_gather_scatter_fn_supported_p): Likewise.
9231 * internal-fn.c (gather_load_direct): New macro.
9232 (expand_gather_load_optab_fn): New function.
9233 (direct_gather_load_optab_supported_p): New macro.
9234 (direct_internal_fn_optab): New function.
9235 (internal_load_fn_p): Likewise.
9236 (internal_gather_scatter_fn_p): Likewise.
9237 (internal_fn_mask_index): Likewise.
9238 (internal_gather_scatter_fn_supported_p): Likewise.
9239 * optabs-query.c (supports_at_least_one_mode_p): New function.
9240 (supports_vec_gather_load_p): Likewise.
9241 * optabs-query.h (supports_vec_gather_load_p): Declare.
9242 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
9243 and memory_type field.
9244 (NUM_PATTERNS): Bump to 15.
9245 * tree-vect-data-refs.c: Include internal-fn.h.
9246 (vect_gather_scatter_fn_p): New function.
9247 (vect_describe_gather_scatter_call): Likewise.
9248 (vect_check_gather_scatter): Try using internal functions for
9249 gather loads. Recognize existing calls to a gather load function.
9250 (vect_analyze_data_refs): Consider using gather loads if
9251 supports_vec_gather_load_p.
9252 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
9253 (vect_get_gather_scatter_offset_type): Likewise.
9254 (vect_convert_mask_for_vectype): Likewise.
9255 (vect_add_conversion_to_patterm): Likewise.
9256 (vect_try_gather_scatter_pattern): Likewise.
9257 (vect_recog_gather_scatter_pattern): New pattern recognizer.
9258 (vect_vect_recog_func_ptrs): Add it.
9259 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9260 internal_fn_mask_index and internal_gather_scatter_fn_p.
9261 (check_load_store_masking): Take the gather_scatter_info as an
9262 argument and handle gather loads.
9263 (vect_get_gather_scatter_ops): New function.
9264 (vectorizable_call): Check internal_load_fn_p.
9265 (vectorizable_load): Likewise. Handle gather load internal
9267 (vectorizable_store): Update call to check_load_store_masking.
9268 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
9269 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
9270 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
9271 (aarch64_gather_scale_operand_d): New predicates.
9272 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
9273 (mask_gather_load<mode>): New insns.
9275 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9276 Alan Hayward <alan.hayward@arm.com>
9277 David Sherwood <david.sherwood@arm.com>
9279 * optabs.def (fold_left_plus_optab): New optab.
9280 * doc/md.texi (fold_left_plus_@var{m}): Document.
9281 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9282 * internal-fn.c (fold_left_direct): Define.
9283 (expand_fold_left_optab_fn): Likewise.
9284 (direct_fold_left_optab_supported_p): Likewise.
9285 * fold-const-call.c (fold_const_fold_left): New function.
9286 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9287 * tree-parloops.c (valid_reduction_p): New function.
9288 (gather_scalar_reductions): Use it.
9289 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9290 (vect_finish_replace_stmt): Declare.
9291 * tree-vect-loop.c (fold_left_reduction_fn): New function.
9292 (needs_fold_left_reduction_p): New function, split out from...
9293 (vect_is_simple_reduction): ...here. Accept reductions that
9294 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9295 (vect_force_simple_reduction): Also store the reduction type in
9296 the assignment's STMT_VINFO_REDUC_TYPE.
9297 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9298 (merge_with_identity): New function.
9299 (vect_expand_fold_left): Likewise.
9300 (vectorize_fold_left_reduction): Likewise.
9301 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9302 scalar phi in place for it. Check for target support and reject
9303 cases that would reassociate the operation. Defer the transform
9304 phase to vectorize_fold_left_reduction.
9305 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9306 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9307 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9309 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9311 * tree-if-conv.c (predicate_mem_writes): Remove redundant
9312 call to ifc_temp_var.
9314 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9315 Alan Hayward <alan.hayward@arm.com>
9316 David Sherwood <david.sherwood@arm.com>
9318 * target.def (legitimize_address_displacement): Take the original
9319 offset as a poly_int.
9320 * targhooks.h (default_legitimize_address_displacement): Update
9322 * targhooks.c (default_legitimize_address_displacement): Likewise.
9323 * doc/tm.texi: Regenerate.
9324 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9325 as an argument, moving assert of ad->disp == ad->disp_term to...
9326 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
9327 Try calling targetm.legitimize_address_displacement before expanding
9328 the address rather than afterwards, and adjust for the new interface.
9329 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9330 Match the new hook interface. Handle SVE addresses.
9331 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9334 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9336 * Makefile.in (OBJS): Add early-remat.o.
9337 * target.def (select_early_remat_modes): New hook.
9338 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9339 * doc/tm.texi: Regenerate.
9340 * targhooks.h (default_select_early_remat_modes): Declare.
9341 * targhooks.c (default_select_early_remat_modes): New function.
9342 * timevar.def (TV_EARLY_REMAT): New timevar.
9343 * passes.def (pass_early_remat): New pass.
9344 * tree-pass.h (make_pass_early_remat): Declare.
9345 * early-remat.c: New file.
9346 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9348 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
9350 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9351 Alan Hayward <alan.hayward@arm.com>
9352 David Sherwood <david.sherwood@arm.com>
9354 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9355 vfm1 with a bound_epilog parameter.
9356 (vect_do_peeling): Update calls accordingly, and move the prologue
9357 call earlier in the function. Treat the base bound_epilog as 0 for
9358 fully-masked loops and retain vf - 1 for other loops. Add 1 to
9359 this base when peeling for gaps.
9360 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9361 with fully-masked loops.
9362 (vect_estimate_min_profitable_iters): Handle the single peeled
9363 iteration in that case.
9365 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9366 Alan Hayward <alan.hayward@arm.com>
9367 David Sherwood <david.sherwood@arm.com>
9369 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9370 single-element interleaving even if the size is not a power of 2.
9371 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9372 accesses for single-element interleaving if the group size is
9375 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9376 Alan Hayward <alan.hayward@arm.com>
9377 David Sherwood <david.sherwood@arm.com>
9379 * doc/md.texi (fold_extract_last_@var{m}): Document.
9380 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9381 * optabs.def (fold_extract_last_optab): New optab.
9382 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9383 * internal-fn.c (fold_extract_direct): New macro.
9384 (expand_fold_extract_optab_fn): Likewise.
9385 (direct_fold_extract_optab_supported_p): Likewise.
9386 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9387 * tree-vect-loop.c (vect_model_reduction_cost): Handle
9388 EXTRACT_LAST_REDUCTION.
9389 (get_initial_def_for_reduction): Do not create an initial vector
9390 for EXTRACT_LAST_REDUCTION reductions.
9391 (vectorizable_reduction): Leave the scalar phi in place for
9392 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
9393 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
9394 epilogue code for EXTRACT_LAST_REDUCTION and defer the
9395 transform phase to vectorizable_condition.
9396 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9398 (vect_finish_stmt_generation): ...here.
9399 (vect_finish_replace_stmt): New function.
9400 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9401 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9403 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9405 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9406 Alan Hayward <alan.hayward@arm.com>
9407 David Sherwood <david.sherwood@arm.com>
9409 * doc/md.texi (extract_last_@var{m}): Document.
9410 * optabs.def (extract_last_optab): New optab.
9411 * internal-fn.def (EXTRACT_LAST): New internal function.
9412 * internal-fn.c (cond_unary_direct): New macro.
9413 (expand_cond_unary_optab_fn): Likewise.
9414 (direct_cond_unary_optab_supported_p): Likewise.
9415 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9416 loops using EXTRACT_LAST.
9417 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9418 (extract_last_<mode>): ...this optab.
9419 (vec_extract<mode><Vel>): Update accordingly.
9421 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9422 Alan Hayward <alan.hayward@arm.com>
9423 David Sherwood <david.sherwood@arm.com>
9425 * target.def (empty_mask_is_expensive): New hook.
9426 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9427 * doc/tm.texi: Regenerate.
9428 * targhooks.h (default_empty_mask_is_expensive): Declare.
9429 * targhooks.c (default_empty_mask_is_expensive): New function.
9430 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9431 if the target says that empty masks are expensive.
9432 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9434 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9436 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9437 Alan Hayward <alan.hayward@arm.com>
9438 David Sherwood <david.sherwood@arm.com>
9440 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9441 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9442 (vect_use_loop_mask_for_alignment_p): New function.
9443 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9444 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9445 niters_skip argument. Make sure that the first niters_skip elements
9446 of the first iteration are inactive.
9447 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9448 Update call to vect_set_loop_masks_directly.
9449 (get_misalign_in_elems): New function, split out from...
9450 (vect_gen_prolog_loop_niters): ...here.
9451 (vect_update_init_of_dr): Take a code argument that specifies whether
9452 the adjustment should be added or subtracted.
9453 (vect_update_init_of_drs): Likewise.
9454 (vect_prepare_for_masked_peels): New function.
9455 (vect_do_peeling): Skip prologue peeling if we're using a mask
9456 instead. Update call to vect_update_inits_of_drs.
9457 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9459 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9460 alignment. Do not include the number of peeled iterations in
9461 the minimum threshold in that case.
9462 (vectorizable_induction): Adjust the start value down by
9463 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9464 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9465 Take the number of skipped iterations into account when calculating
9467 * tree-vect-stmts.c (vect_gen_while_not): New function.
9469 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9470 Alan Hayward <alan.hayward@arm.com>
9471 David Sherwood <david.sherwood@arm.com>
9473 * doc/sourcebuild.texi (vect_fully_masked): Document.
9474 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9476 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9478 (vect_analyze_loop_2): ...here. Don't check the vectorization
9479 factor against the number of loop iterations if the loop is
9482 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9483 Alan Hayward <alan.hayward@arm.com>
9484 David Sherwood <david.sherwood@arm.com>
9486 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9487 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9488 (dump_groups): Update accordingly.
9489 (iv_use::mem_type): New member variable.
9490 (address_p): New function.
9491 (record_use): Add a mem_type argument and initialize the new
9493 (record_group_use): Add a mem_type argument. Use address_p.
9494 Remove obsolete null checks of base_object. Update call to record_use.
9495 (find_interesting_uses_op): Update call to record_group_use.
9496 (find_interesting_uses_cond): Likewise.
9497 (find_interesting_uses_address): Likewise.
9498 (get_mem_type_for_internal_fn): New function.
9499 (find_address_like_use): Likewise.
9500 (find_interesting_uses_stmt): Try find_address_like_use before
9501 calling find_interesting_uses_op.
9502 (addr_offset_valid_p): Use the iv mem_type field as the type
9503 of the addressed memory.
9504 (add_autoinc_candidates): Likewise.
9505 (get_address_cost): Likewise.
9506 (split_small_address_groups_p): Use address_p.
9507 (split_address_groups): Likewise.
9508 (add_iv_candidate_for_use): Likewise.
9509 (autoinc_possible_for_pair): Likewise.
9510 (rewrite_groups): Likewise.
9511 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9512 (determine_group_iv_cost): Update after split of USE_ADDRESS.
9513 (get_alias_ptr_type_for_ptr_address): New function.
9514 (rewrite_use_address): Rewrite address uses in calls that were
9515 identified by find_address_like_use.
9517 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9518 Alan Hayward <alan.hayward@arm.com>
9519 David Sherwood <david.sherwood@arm.com>
9521 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9523 * gimple-expr.h (is_gimple_addressable: Likewise.
9524 * gimple-expr.c (is_gimple_address): Likewise.
9525 * internal-fn.c (expand_call_mem_ref): New function.
9526 (expand_mask_load_optab_fn): Use it.
9527 (expand_mask_store_optab_fn): Likewise.
9529 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9530 Alan Hayward <alan.hayward@arm.com>
9531 David Sherwood <david.sherwood@arm.com>
9533 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
9534 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
9535 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
9536 (cond_umax@var{mode}): Document.
9537 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
9538 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
9539 (cond_umin_optab, cond_umax_optab): New optabs.
9540 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
9541 (COND_IOR, COND_XOR): New internal functions.
9542 * internal-fn.h (get_conditional_internal_fn): Declare.
9543 * internal-fn.c (cond_binary_direct): New macro.
9544 (expand_cond_binary_optab_fn): Likewise.
9545 (direct_cond_binary_optab_supported_p): Likewise.
9546 (get_conditional_internal_fn): New function.
9547 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9548 Cope with reduction statements that are vectorized as calls rather
9550 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9551 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9552 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9553 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9554 (UNSPEC_COND_EOR): New unspecs.
9555 (optab): Add mappings for them.
9556 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9557 (sve_int_op, sve_fp_op): New int attributes.
9559 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9560 Alan Hayward <alan.hayward@arm.com>
9561 David Sherwood <david.sherwood@arm.com>
9563 * optabs.def (while_ult_optab): New optab.
9564 * doc/md.texi (while_ult@var{m}@var{n}): Document.
9565 * internal-fn.def (WHILE_ULT): New internal function.
9566 * internal-fn.h (direct_internal_fn_supported_p): New override
9567 that takes two types as argument.
9568 * internal-fn.c (while_direct): New macro.
9569 (expand_while_optab_fn): New function.
9570 (convert_optab_supported_p): Likewise.
9571 (direct_while_optab_supported_p): New macro.
9572 * wide-int.h (wi::udiv_ceil): New function.
9573 * tree-vectorizer.h (rgroup_masks): New structure.
9574 (vec_loop_masks): New typedef.
9575 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
9577 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
9578 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
9579 (vect_max_vf): New function.
9580 (slpeel_make_loop_iterate_ntimes): Delete.
9581 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
9582 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
9583 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
9584 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
9585 internal-fn.h, stor-layout.h and optabs-query.h.
9586 (vect_set_loop_mask): New function.
9587 (add_preheader_seq): Likewise.
9588 (add_header_seq): Likewise.
9589 (interleave_supported_p): Likewise.
9590 (vect_maybe_permute_loop_masks): Likewise.
9591 (vect_set_loop_masks_directly): Likewise.
9592 (vect_set_loop_condition_masked): Likewise.
9593 (vect_set_loop_condition_unmasked): New function, split out from
9594 slpeel_make_loop_iterate_ntimes.
9595 (slpeel_make_loop_iterate_ntimes): Rename to..
9596 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
9597 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
9598 (vect_do_peeling): Update call accordingly.
9599 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
9601 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9602 mask_compare_type, can_fully_mask_p and fully_masked_p.
9603 (release_vec_loop_masks): New function.
9604 (_loop_vec_info): Use it to free the loop masks.
9605 (can_produce_all_loop_masks_p): New function.
9606 (vect_get_max_nscalars_per_iter): Likewise.
9607 (vect_verify_full_masking): Likewise.
9608 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
9609 retries, and free the mask rgroups before retrying. Check loop-wide
9610 reasons for disallowing fully-masked loops. Make the final decision
9611 about whether use a fully-masked loop or not.
9612 (vect_estimate_min_profitable_iters): Do not assume that peeling
9613 for the number of iterations will be needed for fully-masked loops.
9614 (vectorizable_reduction): Disable fully-masked loops.
9615 (vectorizable_live_operation): Likewise.
9616 (vect_halve_mask_nunits): New function.
9617 (vect_double_mask_nunits): Likewise.
9618 (vect_record_loop_mask): Likewise.
9619 (vect_get_loop_mask): Likewise.
9620 (vect_transform_loop): Handle the case in which the final loop
9621 iteration might handle a partial vector. Call vect_set_loop_condition
9622 instead of slpeel_make_loop_iterate_ntimes.
9623 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
9624 (check_load_store_masking): New function.
9625 (prepare_load_store_mask): Likewise.
9626 (vectorizable_store): Handle fully-masked loops.
9627 (vectorizable_load): Likewise.
9628 (supportable_widening_operation): Use vect_halve_mask_nunits for
9630 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
9631 (vect_gen_while): New function.
9632 * config/aarch64/aarch64.md (umax<mode>3): New expander.
9633 (aarch64_uqdec<mode>): New insn.
9635 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9636 Alan Hayward <alan.hayward@arm.com>
9637 David Sherwood <david.sherwood@arm.com>
9639 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
9640 (reduc_xor_scal_optab): New optabs.
9641 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
9642 (reduc_xor_scal_@var{m}): Document.
9643 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
9644 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
9646 * fold-const-call.c (fold_const_call): Handle them.
9647 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
9648 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
9649 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
9650 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
9651 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
9652 (UNSPEC_XORV): New unspecs.
9653 (optab): Add entries for them.
9654 (BITWISEV): New int iterator.
9655 (bit_reduc_op): New int attributes.
9657 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9658 Alan Hayward <alan.hayward@arm.com>
9659 David Sherwood <david.sherwood@arm.com>
9661 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
9662 * internal-fn.def (VEC_SHL_INSERT): New internal function.
9663 * optabs.def (vec_shl_insert_optab): New optab.
9664 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
9665 (duplicate_and_interleave): Likewise.
9666 * tree-vect-loop.c: Include internal-fn.h.
9667 (neutral_op_for_slp_reduction): New function, split out from
9668 get_initial_defs_for_reduction.
9669 (get_initial_def_for_reduction): Handle option 2 for variable-length
9670 vectors by loading the neutral value into a vector and then shifting
9671 the initial value into element 0.
9672 (get_initial_defs_for_reduction): Replace the code argument with
9673 the neutral value calculated by neutral_op_for_slp_reduction.
9674 Use gimple_build_vector for constant-length vectors.
9675 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
9676 but the first group_size elements have a neutral value.
9677 Use duplicate_and_interleave otherwise.
9678 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
9679 Update call to get_initial_defs_for_reduction. Handle SLP
9680 reductions for variable-length vectors by creating one vector
9681 result for each scalar result, with the elements associated
9682 with other scalar results stubbed out with the neutral value.
9683 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
9684 Require IFN_VEC_SHL_INSERT for double reductions on
9685 variable-length vectors, or SLP reductions that have
9686 a neutral value. Require can_duplicate_and_interleave_p
9687 support for variable-length unchained SLP reductions if there
9688 is no neutral value, such as for MIN/MAX reductions. Also require
9689 the number of vector elements to be a multiple of the number of
9690 SLP statements when doing variable-length unchained SLP reductions.
9691 Update call to vect_create_epilog_for_reduction.
9692 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
9693 and remove initial values.
9694 (duplicate_and_interleave): Make public.
9695 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
9696 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
9698 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9699 Alan Hayward <alan.hayward@arm.com>
9700 David Sherwood <david.sherwood@arm.com>
9702 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
9703 (can_duplicate_and_interleave_p): New function.
9704 (vect_get_and_check_slp_defs): Take the vector of statements
9705 rather than just the current one. Remove excess parentheses.
9706 Restriction rejectinon of vect_constant_def and vect_external_def
9707 for variable-length vectors to boolean types, or types for which
9708 can_duplicate_and_interleave_p is false.
9709 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
9710 (duplicate_and_interleave): New function.
9711 (vect_get_constant_vectors): Use gimple_build_vector for
9712 constant-length vectors and suitable variable-length constant
9713 vectors. Use duplicate_and_interleave for other variable-length
9714 vectors. Don't defer the update when inserting new statements.
9716 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9717 Alan Hayward <alan.hayward@arm.com>
9718 David Sherwood <david.sherwood@arm.com>
9720 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
9721 min_profitable_iters doesn't go negative.
9723 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9724 Alan Hayward <alan.hayward@arm.com>
9725 David Sherwood <david.sherwood@arm.com>
9727 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
9728 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
9729 * optabs.def (vec_mask_load_lanes_optab): New optab.
9730 (vec_mask_store_lanes_optab): Likewise.
9731 * internal-fn.def (MASK_LOAD_LANES): New internal function.
9732 (MASK_STORE_LANES): Likewise.
9733 * internal-fn.c (mask_load_lanes_direct): New macro.
9734 (mask_store_lanes_direct): Likewise.
9735 (expand_mask_load_optab_fn): Handle masked operations.
9736 (expand_mask_load_lanes_optab_fn): New macro.
9737 (expand_mask_store_optab_fn): Handle masked operations.
9738 (expand_mask_store_lanes_optab_fn): New macro.
9739 (direct_mask_load_lanes_optab_supported_p): Likewise.
9740 (direct_mask_store_lanes_optab_supported_p): Likewise.
9741 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
9743 (vect_load_lanes_supported): Likewise.
9744 * tree-vect-data-refs.c (strip_conversion): New function.
9745 (can_group_stmts_p): Likewise.
9746 (vect_analyze_data_ref_accesses): Use it instead of checking
9747 for a pair of assignments.
9748 (vect_store_lanes_supported): Take a masked_p parameter.
9749 (vect_load_lanes_supported): Likewise.
9750 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
9751 vect_store_lanes_supported and vect_load_lanes_supported.
9752 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9753 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
9754 parameter. Don't allow gaps for masked accesses.
9755 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
9756 and vect_load_lanes_supported.
9757 (get_load_store_type): Take a masked_p parameter and update
9758 call to get_group_load_store_type.
9759 (vectorizable_store): Update call to get_load_store_type.
9760 Handle IFN_MASK_STORE_LANES.
9761 (vectorizable_load): Update call to get_load_store_type.
9762 Handle IFN_MASK_LOAD_LANES.
9764 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9765 Alan Hayward <alan.hayward@arm.com>
9766 David Sherwood <david.sherwood@arm.com>
9768 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
9770 * config/aarch64/aarch64-protos.h
9771 (aarch64_sve_struct_memory_operand_p): Declare.
9772 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
9773 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
9774 (VPRED, vpred): Handle SVE structure modes.
9775 * config/aarch64/constraints.md (Utx): New constraint.
9776 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
9777 (aarch64_sve_struct_nonimmediate_operand): New predicates.
9778 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
9779 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
9780 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
9781 structure modes. Split into pieces after RA.
9782 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
9783 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
9785 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
9786 SVE structure modes.
9787 (aarch64_classify_address): Likewise.
9788 (sizetochar): Move earlier in file.
9789 (aarch64_print_operand): Handle SVE register lists.
9790 (aarch64_array_mode): New function.
9791 (aarch64_sve_struct_memory_operand_p): Likewise.
9792 (TARGET_ARRAY_MODE): Redefine.
9794 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9795 Alan Hayward <alan.hayward@arm.com>
9796 David Sherwood <david.sherwood@arm.com>
9798 * target.def (array_mode): New target hook.
9799 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
9800 * doc/tm.texi: Regenerate.
9801 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
9802 * hooks.c (hook_optmode_mode_uhwi_none): New function.
9803 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
9805 * stor-layout.c (mode_for_array): Likewise. Support polynomial
9808 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9809 Alan Hayward <alan.hayward@arm.com>
9810 David Sherwood <david.sherwood@arm.com>
9812 * fold-const.c (fold_binary_loc): Check the argument types
9813 rather than the result type when testing for a vector operation.
9815 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9817 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
9818 * doc/tm.texi: Regenerate.
9820 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9821 Alan Hayward <alan.hayward@arm.com>
9822 David Sherwood <david.sherwood@arm.com>
9824 * doc/invoke.texi (-msve-vector-bits=): Document new option.
9825 (sve): Document new AArch64 extension.
9826 * doc/md.texi (w): Extend the description of the AArch64
9827 constraint to include SVE vectors.
9828 (Upl, Upa): Document new AArch64 predicate constraints.
9829 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
9831 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
9832 (msve-vector-bits=): New option.
9833 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
9834 SVE when these are disabled.
9835 (sve): New extension.
9836 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
9837 modes. Adjust their number of units based on aarch64_sve_vg.
9838 (MAX_BITSIZE_MODE_ANY_MODE): Define.
9839 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
9840 aarch64_addr_query_type.
9841 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
9842 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
9843 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
9844 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
9845 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
9846 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
9847 (aarch64_simd_imm_zero_p): Delete.
9848 (aarch64_check_zero_based_sve_index_immediate): Declare.
9849 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9850 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9851 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9852 (aarch64_sve_float_mul_immediate_p): Likewise.
9853 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9855 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
9856 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
9857 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
9858 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
9859 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
9860 (aarch64_regmode_natural_size): Likewise.
9861 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
9862 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
9864 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
9865 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
9866 for VG and the SVE predicate registers.
9867 (V_ALIASES): Add a "z"-prefixed alias.
9868 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
9869 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
9870 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
9871 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
9872 (REG_CLASS_NAMES): Add entries for them.
9873 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
9874 and the predicate registers.
9875 (aarch64_sve_vg): Declare.
9876 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
9877 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
9878 (REGMODE_NATURAL_SIZE): Define.
9879 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
9881 * config/aarch64/aarch64.c: Include cfgrtl.h.
9882 (simd_immediate_info): Add a constructor for series vectors,
9883 and an associated step field.
9884 (aarch64_sve_vg): New variable.
9885 (aarch64_dbx_register_number): Handle VG and the predicate registers.
9886 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
9887 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
9888 (VEC_ANY_DATA, VEC_STRUCT): New constants.
9889 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
9890 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
9891 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
9892 (aarch64_get_mask_mode): New functions.
9893 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
9894 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
9895 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
9896 predicate modes and predicate registers. Explicitly restrict
9897 GPRs to modes of 16 bytes or smaller. Only allow FP registers
9898 to store a vector mode if it is recognized by
9899 aarch64_classify_vector_mode.
9900 (aarch64_regmode_natural_size): New function.
9901 (aarch64_hard_regno_caller_save_mode): Return the original mode
9903 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
9904 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
9905 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
9906 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
9908 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
9909 does not overlap dest if the function is frame-related. Handle
9911 (aarch64_split_add_offset): New function.
9912 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
9913 them aarch64_add_offset.
9914 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
9915 and update call to aarch64_sub_sp.
9916 (aarch64_add_cfa_expression): New function.
9917 (aarch64_expand_prologue): Pass extra temporary registers to the
9918 functions above. Handle the case in which we need to emit new
9919 DW_CFA_expressions for registers that were originally saved
9920 relative to the stack pointer, but now have to be expressed
9921 relative to the frame pointer.
9922 (aarch64_output_mi_thunk): Pass extra temporary registers to the
9924 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
9925 IP0 and IP1 values for SVE frames.
9926 (aarch64_expand_vec_series): New function.
9927 (aarch64_expand_sve_widened_duplicate): Likewise.
9928 (aarch64_expand_sve_const_vector): Likewise.
9929 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
9930 Handle SVE constants. Use emit_move_insn to move a force_const_mem
9931 into the register, rather than emitting a SET directly.
9932 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
9933 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
9934 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
9935 (offset_9bit_signed_scaled_p): New functions.
9936 (aarch64_replicate_bitmask_imm): New function.
9937 (aarch64_bitmask_imm): Use it.
9938 (aarch64_cannot_force_const_mem): Reject expressions involving
9939 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
9940 (aarch64_classify_index): Handle SVE indices, by requiring
9941 a plain register index with a scale that matches the element size.
9942 (aarch64_classify_address): Handle SVE addresses. Assert that
9943 the mode of the address is VOIDmode or an integer mode.
9944 Update call to aarch64_classify_symbol.
9945 (aarch64_classify_symbolic_expression): Update call to
9946 aarch64_classify_symbol.
9947 (aarch64_const_vec_all_in_range_p): New function.
9948 (aarch64_print_vector_float_operand): Likewise.
9949 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
9950 "vN" for FP registers with SVE modes. Handle (const ...) vectors
9951 and the FP immediates 1.0 and 0.5.
9952 (aarch64_print_address_internal): Handle SVE addresses.
9953 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
9954 (aarch64_regno_regclass): Handle predicate registers.
9955 (aarch64_secondary_reload): Handle big-endian reloads of SVE
9957 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
9958 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
9959 (aarch64_convert_sve_vector_bits): New function.
9960 (aarch64_override_options): Use it to handle -msve-vector-bits=.
9961 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9963 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
9964 Handle SVE vector and predicate modes. Accept VL-based constants
9965 that need only one temporary register, and VL offsets that require
9966 no temporary registers.
9967 (aarch64_conditional_register_usage): Mark the predicate registers
9968 as fixed if SVE isn't available.
9969 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
9970 Return true for SVE vector and predicate modes.
9971 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
9972 rather than an unsigned int. Handle SVE modes.
9973 (aarch64_preferred_simd_mode): Update call accordingly. Handle
9975 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
9977 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9978 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9979 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9980 (aarch64_sve_float_mul_immediate_p): New functions.
9981 (aarch64_sve_valid_immediate): New function.
9982 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
9983 Explicitly reject structure modes. Check for INDEX constants.
9984 Handle PTRUE and PFALSE constants.
9985 (aarch64_check_zero_based_sve_index_immediate): New function.
9986 (aarch64_simd_imm_zero_p): Delete.
9987 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
9988 vector modes. Accept constants in the range of CNT[BHWD].
9989 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
9990 ask for an Advanced SIMD mode.
9991 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
9992 (aarch64_simd_vector_alignment): Handle SVE predicates.
9993 (aarch64_vectorize_preferred_vector_alignment): New function.
9994 (aarch64_simd_vector_alignment_reachable): Use it instead of
9996 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
9997 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
9999 (MAX_VECT_LEN): Delete.
10000 (expand_vec_perm_d): Add a vec_flags field.
10001 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
10002 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
10003 (aarch64_evpc_ext): Don't apply a big-endian lane correction
10005 (aarch64_evpc_rev): Rename to...
10006 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
10007 (aarch64_evpc_rev_global): New function.
10008 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
10009 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
10011 (aarch64_evpc_sve_tbl): New function.
10012 (aarch64_expand_vec_perm_const_1): Update after rename of
10013 aarch64_evpc_rev. Handle SVE permutes too, trying
10014 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
10015 than aarch64_evpc_tbl.
10016 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
10017 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
10018 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
10019 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
10020 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
10021 (aarch64_expand_sve_vcond): New functions.
10022 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
10023 of aarch64_vector_mode_p.
10024 (aarch64_dwarf_poly_indeterminate_value): New function.
10025 (aarch64_compute_pressure_classes): Likewise.
10026 (aarch64_can_change_mode_class): Likewise.
10027 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
10028 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
10029 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
10030 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
10031 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
10032 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
10033 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
10034 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
10036 (Dn, Dl, Dr): Accept const as well as const_vector.
10037 (Dz): Likewise. Compare against CONST0_RTX.
10038 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
10039 of "vector" where appropriate.
10040 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
10041 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
10042 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
10043 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
10044 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
10045 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
10046 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
10047 (v_int_equiv): Extend to SVE modes.
10048 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
10050 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
10051 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
10052 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
10053 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
10054 (SVE_COND_FP_CMP): New int iterators.
10055 (perm_hilo): Handle the new unpack unspecs.
10056 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
10058 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
10059 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
10060 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
10061 (aarch64_equality_operator, aarch64_constant_vector_operand)
10062 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
10063 (aarch64_sve_nonimmediate_operand): Likewise.
10064 (aarch64_sve_general_operand): Likewise.
10065 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
10066 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
10067 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
10068 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
10069 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
10070 (aarch64_sve_float_arith_immediate): Likewise.
10071 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
10072 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
10073 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
10074 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
10075 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
10076 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
10077 (aarch64_sve_float_arith_operand): Likewise.
10078 (aarch64_sve_float_arith_with_sub_operand): Likewise.
10079 (aarch64_sve_float_mul_operand): Likewise.
10080 (aarch64_sve_vec_perm_operand): Likewise.
10081 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
10082 (aarch64_mov_operand): Accept const_poly_int and const_vector.
10083 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
10084 as well as const_vector.
10085 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
10086 in file. Use CONST0_RTX and CONSTM1_RTX.
10087 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
10088 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
10089 Use aarch64_simd_imm_zero.
10090 * config/aarch64/aarch64-sve.md: New file.
10091 * config/aarch64/aarch64.md: Include it.
10092 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
10093 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
10094 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
10095 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
10096 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
10097 (sve): New attribute.
10098 (enabled): Disable instructions with the sve attribute unless
10100 (movqi, movhi): Pass CONST_POLY_INT operaneds through
10101 aarch64_expand_mov_immediate.
10102 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
10103 CNT[BHSD] immediates.
10104 (movti): Split CONST_POLY_INT moves into two halves.
10105 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
10106 Split additions that need a temporary here if the destination
10107 is the stack pointer.
10108 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
10109 (*add<mode>3_poly_1): New instruction.
10110 (set_clobber_cc): New expander.
10112 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10114 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
10115 parameter and use it instead of GET_MODE_SIZE (innermode). Use
10116 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
10117 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
10118 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
10119 Change innermode from fixed_mode_size to machine_mode.
10120 (simplify_subreg): Update call accordingly. Handle a constant-sized
10121 subreg of a variable-length CONST_VECTOR.
10123 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10124 Alan Hayward <alan.hayward@arm.com>
10125 David Sherwood <david.sherwood@arm.com>
10127 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
10128 (add_offset_to_base): New function, split out from...
10129 (create_mem_ref): ...here. When handling a scale other than 1,
10130 check first whether the address is valid without the offset.
10131 Add it into the base if so, leaving the index and scale as-is.
10133 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10136 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
10137 fold_for_warn before checking if arg2 is INTEGER_CST.
10139 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
10141 * config/rs6000/predicates.md (load_multiple_operation): Delete.
10142 (store_multiple_operation): Delete.
10143 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
10144 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
10145 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
10146 guarded by TARGET_STRING.
10147 (rs6000_output_load_multiple): Delete.
10148 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
10149 OPTION_MASK_STRING / TARGET_STRING handling.
10150 (print_operand) <'N', 'O'>: Add comment that these are unused now.
10151 (const rs6000_opt_masks) <"string">: Change mask to 0.
10152 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
10153 (MASK_STRING): Delete.
10154 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
10156 (load_multiple): Delete.
10163 (store_multiple): Delete.
10170 (movmemsi_8reg): Delete.
10171 (corresponding unnamed define_insn): Delete.
10172 (movmemsi_6reg): Delete.
10173 (corresponding unnamed define_insn): Delete.
10174 (movmemsi_4reg): Delete.
10175 (corresponding unnamed define_insn): Delete.
10176 (movmemsi_2reg): Delete.
10177 (corresponding unnamed define_insn): Delete.
10178 (movmemsi_1reg): Delete.
10179 (corresponding unnamed define_insn): Delete.
10180 * config/rs6000/rs6000.opt (mno-string): New.
10181 (mstring): Replace by deprecation warning stub.
10182 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
10184 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10186 * regrename.c (regrename_do_replace): If replacing the same
10187 reg multiple times, try to reuse last created gen_raw_REG.
10190 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
10191 main to workaround a bug in GDB.
10193 2018-01-12 Tom de Vries <tom@codesourcery.com>
10196 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
10198 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
10200 PR rtl-optimization/80481
10201 * ira-color.c (get_cap_member): New function.
10202 (allocnos_conflict_by_live_ranges_p): Use it.
10203 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
10204 (setup_slot_coalesced_allocno_live_ranges): Ditto.
10206 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
10209 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
10210 (*saddl_se_1): Ditto.
10211 (*ssubsi_1): Ditto.
10212 (*ssubl_se_1): Ditto.
10214 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10216 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
10217 rather than wi::to_widest for DR_INITs.
10218 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
10219 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
10220 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
10222 (vect_analyze_group_access_1): Note that here.
10224 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10226 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
10227 polynomial type sizes.
10229 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10231 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
10232 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
10233 (gimple_add_tmp_var): Likewise.
10235 2018-01-12 Martin Liska <mliska@suse.cz>
10237 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
10238 (gimple_alloc_sizes): Likewise.
10239 (dump_gimple_statistics): Use PRIu64 in printf format.
10240 * gimple.h: Change uint64_t to int.
10242 2018-01-12 Martin Liska <mliska@suse.cz>
10244 * tree-core.h: Use uint64_t instead of int.
10245 * tree.c (tree_node_counts): Likewise.
10246 (tree_node_sizes): Likewise.
10247 (dump_tree_statistics): Use PRIu64 in printf format.
10249 2018-01-12 Martin Liska <mliska@suse.cz>
10251 * Makefile.in: As qsort_chk is implemented in vec.c, add
10252 vec.o to linkage of gencfn-macros.
10253 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
10254 passing the info to record_node_allocation_statistics.
10255 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
10257 * ggc-common.c (struct ggc_usage): Add operator== and use
10258 it in operator< and compare function.
10259 * mem-stats.h (struct mem_usage): Likewise.
10260 * vec.c (struct vec_usage): Remove operator< and compare
10261 function. Can be simply inherited.
10263 2018-01-12 Martin Jambor <mjambor@suse.cz>
10266 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
10267 * tree-ssa-math-opts.c: Include domwalk.h.
10268 (convert_mult_to_fma_1): New function.
10269 (fma_transformation_info): New type.
10270 (fma_deferring_state): Likewise.
10271 (cancel_fma_deferring): New function.
10272 (result_of_phi): Likewise.
10273 (last_fma_candidate_feeds_initial_phi): Likewise.
10274 (convert_mult_to_fma): Added deferring logic, split actual
10275 transformation to convert_mult_to_fma_1.
10276 (math_opts_dom_walker): New type.
10277 (math_opts_dom_walker::after_dom_children): New method, body moved
10278 here from pass_optimize_widening_mul::execute, added deferring logic
10280 (pass_optimize_widening_mul::execute): Moved most of code to
10281 math_opts_dom_walker::after_dom_children.
10282 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10283 * config/i386/i386.c (ix86_option_override_internal): Added
10284 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10286 2018-01-12 Richard Biener <rguenther@suse.de>
10289 * dwarf2out.c (gen_variable_die): Do not reset old_die for
10290 inline instance vars.
10292 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
10295 * config/rx/rx.c (rx_is_restricted_memory_address):
10296 Handle SUBREG case.
10298 2018-01-12 Richard Biener <rguenther@suse.de>
10300 PR tree-optimization/80846
10301 * target.def (split_reduction): New target hook.
10302 * targhooks.c (default_split_reduction): New function.
10303 * targhooks.h (default_split_reduction): Declare.
10304 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10305 target requests first reduce vectors by combining low and high
10307 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10308 (get_vectype_for_scalar_type_and_size): Export.
10309 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10310 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10311 * doc/tm.texi: Regenerate.
10312 * config/i386/i386.c (ix86_split_reduction): Implement
10313 TARGET_VECTORIZE_SPLIT_REDUCTION.
10315 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10318 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10319 in PIC mode except for TARGET_VXWORKS_RTP.
10320 * config/sparc/sparc.c: Include cfgrtl.h.
10321 (TARGET_INIT_PIC_REG): Define.
10322 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10323 (sparc_pic_register_p): New predicate.
10324 (sparc_legitimate_address_p): Use it.
10325 (sparc_legitimize_pic_address): Likewise.
10326 (sparc_delegitimize_address): Likewise.
10327 (sparc_mode_dependent_address_p): Likewise.
10328 (gen_load_pcrel_sym): Remove 4th parameter.
10329 (load_got_register): Adjust call to above. Remove obsolete stuff.
10330 (sparc_expand_prologue): Do not call load_got_register here.
10331 (sparc_flat_expand_prologue): Likewise.
10332 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10333 (sparc_use_pseudo_pic_reg): New function.
10334 (sparc_init_pic_reg): Likewise.
10335 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10336 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10338 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
10340 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10341 Add item for branch_cost.
10343 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10345 PR rtl-optimization/83565
10346 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10347 not extend the result to a larger mode for rotate operations.
10348 (num_sign_bit_copies1): Likewise.
10350 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10353 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10355 Use values-Xc.o for -pedantic.
10356 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10358 2018-01-12 Martin Liska <mliska@suse.cz>
10361 * ipa-devirt.c (final_warning_record::grow_type_warnings):
10363 (possible_polymorphic_call_targets): Use it.
10364 (ipa_devirt): Likewise.
10366 2018-01-12 Martin Liska <mliska@suse.cz>
10368 * profile-count.h (enum profile_quality): Use 0 as invalid
10369 enum value of profile_quality.
10371 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
10373 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10374 -mext-string options.
10376 2018-01-12 Richard Biener <rguenther@suse.de>
10378 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10379 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10380 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10382 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10384 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10386 * configure.ac (--with-long-double-format): Add support for the
10387 configuration option to change the default long double format on
10389 * config.gcc (powerpc*-linux*-*): Likewise.
10390 * configure: Regenerate.
10391 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10392 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10393 used without modification.
10395 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10397 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10398 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10399 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10400 MISC_BUILTIN_SPEC_BARRIER.
10401 (rs6000_init_builtins): Likewise.
10402 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10404 (speculation_barrier): New define_insn.
10405 * doc/extend.texi: Document __builtin_speculation_barrier.
10407 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10410 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10411 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10412 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10414 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10415 integral modes instead of "ss" and "sd".
10416 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10417 vectors with 32-bit and 64-bit elements.
10418 (vecdupssescalarmodesuffix): New mode attribute.
10419 (vec_dup<mode>): Use it.
10421 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
10424 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10425 frame if argument is passed on stack.
10427 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10430 * ree.c (combine_reaching_defs): Optimize also
10431 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10432 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10434 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10436 PR middle-end/83189
10437 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10439 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10441 PR middle-end/83718
10442 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10443 after they are computed.
10445 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10447 PR tree-optimization/83695
10448 * gimple-loop-linterchange.cc
10449 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10450 reset cached scev information after interchange.
10451 (pass_linterchange::execute): Remove call to scev_reset_htab.
10453 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10455 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10456 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10457 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10458 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10459 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10460 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10461 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10462 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10463 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10464 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10465 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10466 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10467 (V_lane_reg): Likewise.
10468 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10470 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10471 (vfmal_lane_low<mode>_intrinsic,
10472 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10473 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10474 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10475 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10476 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10477 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10479 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10481 * config/arm/arm-cpus.in (fp16fml): New feature.
10482 (ALL_SIMD): Add fp16fml.
10483 (armv8.2-a): Add fp16fml as an option.
10484 (armv8.3-a): Likewise.
10485 (armv8.4-a): Add fp16fml as part of fp16.
10486 * config/arm/arm.h (TARGET_FP16FML): Define.
10487 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10489 * config/arm/arm-modes.def (V2HF): Define.
10490 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10491 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10492 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10493 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10494 vfmsl_low, vfmsl_high): New set of builtins.
10495 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10496 (vfml_op): New code attribute.
10497 (VFMLHALVES): New int iterator.
10498 (VFML, VFMLSEL): New mode attributes.
10499 (V_reg): Define mapping for V2HF.
10500 (V_hi, V_lo): New mode attributes.
10501 (VF_constraint): Likewise.
10502 (vfml_half, vfml_half_selector): New int attributes.
10503 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10505 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10506 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10508 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10509 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10510 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10511 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
10513 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10514 Document new effective target and option set.
10516 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10518 * config/arm/arm-cpus.in (armv8_4): New feature.
10519 (ARMv8_4a): New fgroup.
10520 (armv8.4-a): New arch.
10521 * config/arm/arm-tables.opt: Regenerate.
10522 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10523 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
10524 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
10525 Add matching rules for -march=armv8.4-a and extensions.
10526 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
10528 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
10531 * config/rx/rx.md (BW): New mode attribute.
10532 (sync_lock_test_and_setsi): Add mode suffix to insn output.
10534 2018-01-11 Richard Biener <rguenther@suse.de>
10536 PR tree-optimization/83435
10537 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
10538 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
10539 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
10541 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10542 Alan Hayward <alan.hayward@arm.com>
10543 David Sherwood <david.sherwood@arm.com>
10545 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
10547 (aarch64_classify_address): Initialize it. Track polynomial offsets.
10548 (aarch64_print_address_internal): Use it to check for a zero offset.
10550 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10551 Alan Hayward <alan.hayward@arm.com>
10552 David Sherwood <david.sherwood@arm.com>
10554 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10555 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10556 Return a poly_int64 rather than a HOST_WIDE_INT.
10557 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10558 rather than a HOST_WIDE_INT.
10559 * config/aarch64/aarch64.h (aarch64_frame): Protect with
10560 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
10561 hard_fp_offset, frame_size, initial_adjust, callee_offset and
10562 final_offset from HOST_WIDE_INT to poly_int64.
10563 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
10564 to_constant when getting the number of units in an Advanced SIMD
10566 (aarch64_builtin_vectorized_function): Check for a constant number
10568 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
10570 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
10571 attribute instead of GET_MODE_NUNITS.
10572 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
10573 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
10574 GET_MODE_SIZE for fixed-size registers.
10575 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
10576 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
10577 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
10578 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
10579 (aarch64_print_operand, aarch64_print_address_internal)
10580 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
10581 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
10582 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
10583 Handle polynomial GET_MODE_SIZE.
10584 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
10585 wider than SImode without modification.
10586 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
10587 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
10588 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
10589 passing and returning SVE modes.
10590 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
10591 rather than GEN_INT.
10592 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
10593 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
10594 (aarch64_allocate_and_probe_stack_space): Likewise.
10595 (aarch64_layout_frame): Cope with polynomial offsets.
10596 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
10597 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
10598 polynomial offsets.
10599 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
10600 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
10601 poly_int64 rather than a HOST_WIDE_INT.
10602 (aarch64_get_separate_components, aarch64_process_components)
10603 (aarch64_expand_prologue, aarch64_expand_epilogue)
10604 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
10605 (aarch64_anchor_offset): New function, split out from...
10606 (aarch64_legitimize_address): ...here.
10607 (aarch64_builtin_vectorization_cost): Handle polynomial
10608 TYPE_VECTOR_SUBPARTS.
10609 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
10611 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
10612 number of elements from the PARALLEL rather than the mode.
10613 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
10614 rather than GET_MODE_BITSIZE.
10615 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
10616 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
10617 (aarch64_expand_vec_perm_const_1): Handle polynomial
10618 d->perm.length () and d->perm elements.
10619 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
10620 Apply to_constant to d->perm elements.
10621 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
10622 polynomial CONST_VECTOR_NUNITS.
10623 (aarch64_move_pointer): Take amount as a poly_int64 rather
10625 (aarch64_progress_pointer): Avoid temporary variable.
10626 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
10627 the mode attribute instead of GET_MODE.
10629 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10630 Alan Hayward <alan.hayward@arm.com>
10631 David Sherwood <david.sherwood@arm.com>
10633 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
10634 x exists before using it.
10635 (aarch64_add_constant_internal): Rename to...
10636 (aarch64_add_offset_1): ...this. Replace regnum with separate
10637 src and dest rtxes. Handle the case in which they're different,
10638 including when the offset is zero. Replace scratchreg with an rtx.
10639 Use 2 additions if there is no spare register into which we can
10640 move a 16-bit constant.
10641 (aarch64_add_constant): Delete.
10642 (aarch64_add_offset): Replace reg with separate src and dest
10643 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
10644 Use aarch64_add_offset_1.
10645 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
10646 an rtx rather than an int. Take the delta as a poly_int64
10647 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
10648 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
10649 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
10650 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
10651 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
10652 and aarch64_add_sp.
10653 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
10654 aarch64_add_constant.
10656 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10658 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
10659 Use scalar_float_mode.
10661 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10663 * config/aarch64/aarch64-simd.md
10664 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
10665 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
10666 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
10667 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
10668 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
10669 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
10670 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
10671 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
10672 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
10673 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
10675 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10678 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
10679 targ_options->x_arm_arch_string is non NULL.
10681 2018-01-11 Tamar Christina <tamar.christina@arm.com>
10683 * config/aarch64/aarch64.h
10684 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
10686 2018-01-11 Sudakshina Das <sudi.das@arm.com>
10689 * expmed.c (emit_store_flag_force): Swap if const op0
10690 and change VOIDmode to mode of op0.
10692 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10694 PR rtl-optimization/83761
10695 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
10696 than bytes to mode_for_size.
10698 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10700 PR middle-end/83189
10701 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
10702 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
10705 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10707 PR middle-end/83575
10708 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
10709 when in layout mode.
10710 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
10711 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
10714 2018-01-10 Michael Collison <michael.collison@arm.com>
10716 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
10717 * config/aarch64/aarch64-option-extension.def: Add
10718 AARCH64_OPT_EXTENSION of 'fp16fml'.
10719 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10720 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
10721 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
10722 * config/aarch64/constraints.md (Ui7): New constraint.
10723 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
10724 (VFMLA_SEL_W): Ditto.
10727 (VFMLA16_LOW): New int iterator.
10728 (VFMLA16_HIGH): Ditto.
10729 (UNSPEC_FMLAL): New unspec.
10730 (UNSPEC_FMLSL): Ditto.
10731 (UNSPEC_FMLAL2): Ditto.
10732 (UNSPEC_FMLSL2): Ditto.
10733 (f16mac): New code attribute.
10734 * config/aarch64/aarch64-simd-builtins.def
10735 (aarch64_fmlal_lowv2sf): Ditto.
10736 (aarch64_fmlsl_lowv2sf): Ditto.
10737 (aarch64_fmlalq_lowv4sf): Ditto.
10738 (aarch64_fmlslq_lowv4sf): Ditto.
10739 (aarch64_fmlal_highv2sf): Ditto.
10740 (aarch64_fmlsl_highv2sf): Ditto.
10741 (aarch64_fmlalq_highv4sf): Ditto.
10742 (aarch64_fmlslq_highv4sf): Ditto.
10743 (aarch64_fmlal_lane_lowv2sf): Ditto.
10744 (aarch64_fmlsl_lane_lowv2sf): Ditto.
10745 (aarch64_fmlal_laneq_lowv2sf): Ditto.
10746 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
10747 (aarch64_fmlalq_lane_lowv4sf): Ditto.
10748 (aarch64_fmlsl_lane_lowv4sf): Ditto.
10749 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
10750 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
10751 (aarch64_fmlal_lane_highv2sf): Ditto.
10752 (aarch64_fmlsl_lane_highv2sf): Ditto.
10753 (aarch64_fmlal_laneq_highv2sf): Ditto.
10754 (aarch64_fmlsl_laneq_highv2sf): Ditto.
10755 (aarch64_fmlalq_lane_highv4sf): Ditto.
10756 (aarch64_fmlsl_lane_highv4sf): Ditto.
10757 (aarch64_fmlalq_laneq_highv4sf): Ditto.
10758 (aarch64_fmlsl_laneq_highv4sf): Ditto.
10759 * config/aarch64/aarch64-simd.md:
10760 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
10761 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10762 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
10763 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10764 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
10765 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
10766 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
10767 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
10768 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
10769 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
10770 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
10771 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
10772 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
10773 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
10774 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
10775 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
10776 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
10777 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
10778 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
10779 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
10780 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
10781 (vfmlsl_low_u32): Ditto.
10782 (vfmlalq_low_u32): Ditto.
10783 (vfmlslq_low_u32): Ditto.
10784 (vfmlal_high_u32): Ditto.
10785 (vfmlsl_high_u32): Ditto.
10786 (vfmlalq_high_u32): Ditto.
10787 (vfmlslq_high_u32): Ditto.
10788 (vfmlal_lane_low_u32): Ditto.
10789 (vfmlsl_lane_low_u32): Ditto.
10790 (vfmlal_laneq_low_u32): Ditto.
10791 (vfmlsl_laneq_low_u32): Ditto.
10792 (vfmlalq_lane_low_u32): Ditto.
10793 (vfmlslq_lane_low_u32): Ditto.
10794 (vfmlalq_laneq_low_u32): Ditto.
10795 (vfmlslq_laneq_low_u32): Ditto.
10796 (vfmlal_lane_high_u32): Ditto.
10797 (vfmlsl_lane_high_u32): Ditto.
10798 (vfmlal_laneq_high_u32): Ditto.
10799 (vfmlsl_laneq_high_u32): Ditto.
10800 (vfmlalq_lane_high_u32): Ditto.
10801 (vfmlslq_lane_high_u32): Ditto.
10802 (vfmlalq_laneq_high_u32): Ditto.
10803 (vfmlslq_laneq_high_u32): Ditto.
10804 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
10805 (AARCH64_FL_FOR_ARCH8_4): New.
10806 (AARCH64_ISA_F16FML): New ISA flag.
10807 (TARGET_F16FML): New feature flag for fp16fml.
10808 (doc/invoke.texi): Document new fp16fml option.
10810 2018-01-10 Michael Collison <michael.collison@arm.com>
10812 * config/aarch64/aarch64-builtins.c:
10813 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
10814 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10815 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
10816 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
10817 (AARCH64_ISA_SHA3): New ISA flag.
10818 (TARGET_SHA3): New feature flag for sha3.
10819 * config/aarch64/iterators.md (sha512_op): New int attribute.
10820 (CRYPTO_SHA512): New int iterator.
10821 (UNSPEC_SHA512H): New unspec.
10822 (UNSPEC_SHA512H2): Ditto.
10823 (UNSPEC_SHA512SU0): Ditto.
10824 (UNSPEC_SHA512SU1): Ditto.
10825 * config/aarch64/aarch64-simd-builtins.def
10826 (aarch64_crypto_sha512hqv2di): New builtin.
10827 (aarch64_crypto_sha512h2qv2di): Ditto.
10828 (aarch64_crypto_sha512su0qv2di): Ditto.
10829 (aarch64_crypto_sha512su1qv2di): Ditto.
10830 (aarch64_eor3qv8hi): Ditto.
10831 (aarch64_rax1qv2di): Ditto.
10832 (aarch64_xarqv2di): Ditto.
10833 (aarch64_bcaxqv8hi): Ditto.
10834 * config/aarch64/aarch64-simd.md:
10835 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
10836 (aarch64_crypto_sha512su0qv2di): Ditto.
10837 (aarch64_crypto_sha512su1qv2di): Ditto.
10838 (aarch64_eor3qv8hi): Ditto.
10839 (aarch64_rax1qv2di): Ditto.
10840 (aarch64_xarqv2di): Ditto.
10841 (aarch64_bcaxqv8hi): Ditto.
10842 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
10843 (vsha512h2q_u64): Ditto.
10844 (vsha512su0q_u64): Ditto.
10845 (vsha512su1q_u64): Ditto.
10846 (veor3q_u16): Ditto.
10847 (vrax1q_u64): Ditto.
10848 (vxarq_u64): Ditto.
10849 (vbcaxq_u16): Ditto.
10850 * config/arm/types.md (crypto_sha512): New type attribute.
10851 (crypto_sha3): Ditto.
10852 (doc/invoke.texi): Document new sha3 option.
10854 2018-01-10 Michael Collison <michael.collison@arm.com>
10856 * config/aarch64/aarch64-builtins.c:
10857 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
10858 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10859 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
10860 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
10861 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
10862 (AARCH64_ISA_SM4): New ISA flag.
10863 (TARGET_SM4): New feature flag for sm4.
10864 * config/aarch64/aarch64-simd-builtins.def
10865 (aarch64_sm3ss1qv4si): Ditto.
10866 (aarch64_sm3tt1aq4si): Ditto.
10867 (aarch64_sm3tt1bq4si): Ditto.
10868 (aarch64_sm3tt2aq4si): Ditto.
10869 (aarch64_sm3tt2bq4si): Ditto.
10870 (aarch64_sm3partw1qv4si): Ditto.
10871 (aarch64_sm3partw2qv4si): Ditto.
10872 (aarch64_sm4eqv4si): Ditto.
10873 (aarch64_sm4ekeyqv4si): Ditto.
10874 * config/aarch64/aarch64-simd.md:
10875 (aarch64_sm3ss1qv4si): Ditto.
10876 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
10877 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
10878 (aarch64_sm4eqv4si): Ditto.
10879 (aarch64_sm4ekeyqv4si): Ditto.
10880 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
10881 (sm3part_op): Ditto.
10882 (CRYPTO_SM3TT): Ditto.
10883 (CRYPTO_SM3PART): Ditto.
10884 (UNSPEC_SM3SS1): New unspec.
10885 (UNSPEC_SM3TT1A): Ditto.
10886 (UNSPEC_SM3TT1B): Ditto.
10887 (UNSPEC_SM3TT2A): Ditto.
10888 (UNSPEC_SM3TT2B): Ditto.
10889 (UNSPEC_SM3PARTW1): Ditto.
10890 (UNSPEC_SM3PARTW2): Ditto.
10891 (UNSPEC_SM4E): Ditto.
10892 (UNSPEC_SM4EKEY): Ditto.
10893 * config/aarch64/constraints.md (Ui2): New constraint.
10894 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
10895 * config/arm/types.md (crypto_sm3): New type attribute.
10896 (crypto_sm4): Ditto.
10897 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
10898 (vsm3tt1aq_u32): Ditto.
10899 (vsm3tt1bq_u32): Ditto.
10900 (vsm3tt2aq_u32): Ditto.
10901 (vsm3tt2bq_u32): Ditto.
10902 (vsm3partw1q_u32): Ditto.
10903 (vsm3partw2q_u32): Ditto.
10904 (vsm4eq_u32): Ditto.
10905 (vsm4ekeyq_u32): Ditto.
10906 (doc/invoke.texi): Document new sm4 option.
10908 2018-01-10 Michael Collison <michael.collison@arm.com>
10910 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
10911 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
10912 (AARCH64_FL_FOR_ARCH8_4): New.
10913 (AARCH64_FL_V8_4): New flag.
10914 (doc/invoke.texi): Document new armv8.4-a option.
10916 2018-01-10 Michael Collison <michael.collison@arm.com>
10918 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10919 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
10920 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
10921 * config/aarch64/aarch64-option-extension.def: Add
10922 AARCH64_OPT_EXTENSION of 'sha2'.
10923 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
10924 (crypto): Disable sha2 and aes if crypto disabled.
10925 (crypto): Enable aes and sha2 if enabled.
10926 (simd): Disable sha2 and aes if simd disabled.
10927 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
10929 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
10930 (TARGET_SHA2): New feature flag for sha2.
10931 (TARGET_AES): New feature flag for aes.
10932 * config/aarch64/aarch64-simd.md:
10933 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
10934 conditional on TARGET_AES.
10935 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
10936 (aarch64_crypto_sha1hsi): Make pattern conditional
10938 (aarch64_crypto_sha1hv4si): Ditto.
10939 (aarch64_be_crypto_sha1hv4si): Ditto.
10940 (aarch64_crypto_sha1su1v4si): Ditto.
10941 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
10942 (aarch64_crypto_sha1su0v4si): Ditto.
10943 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
10944 (aarch64_crypto_sha256su0v4si): Ditto.
10945 (aarch64_crypto_sha256su1v4si): Ditto.
10946 (doc/invoke.texi): Document new aes and sha2 options.
10948 2018-01-10 Martin Sebor <msebor@redhat.com>
10950 PR tree-optimization/83781
10951 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
10954 2018-01-11 Martin Sebor <msebor@gmail.com>
10955 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10957 PR tree-optimization/83501
10958 PR tree-optimization/81703
10960 * tree-ssa-strlen.c (get_string_cst): Rename...
10961 (get_string_len): ...to this. Handle global constants.
10962 (handle_char_store): Adjust.
10964 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
10965 Jim Wilson <jimw@sifive.com>
10967 * config/riscv/riscv-protos.h (riscv_output_return): New.
10968 * config/riscv/riscv.c (struct machine_function): New naked_p field.
10969 (riscv_attribute_table, riscv_output_return),
10970 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
10971 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
10972 (riscv_compute_frame_info): Only compute frame->mask if not a naked
10974 (riscv_expand_prologue): Add early return for naked function.
10975 (riscv_expand_epilogue): Likewise.
10976 (riscv_function_ok_for_sibcall): Return false for naked function.
10977 (riscv_set_current_function): New.
10978 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
10979 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
10980 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
10981 * doc/extend.texi (RISC-V Function Attributes): New.
10983 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
10985 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
10986 check for 128-bit long double before checking TCmode.
10987 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
10988 128-bit long doubles before checking TFmode or TCmode.
10989 (FLOAT128_IBM_P): Likewise.
10991 2018-01-10 Martin Sebor <msebor@redhat.com>
10993 PR tree-optimization/83671
10994 * builtins.c (c_strlen): Unconditionally return zero for the empty
10996 Use -Warray-bounds for warnings.
10997 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
10998 for non-constant array indices with COMPONENT_REF, arrays of
10999 arrays, and pointers to arrays.
11000 (gimple_fold_builtin_strlen): Determine and set length range for
11001 non-constant character arrays.
11003 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
11005 PR middle-end/81897
11006 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
11009 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
11011 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
11013 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11016 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
11017 VECTOR_MEM_ALTIVEC_OR_VSX_P.
11018 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
11019 indexed_or_indirect_operand predicate.
11020 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
11021 (*vsx_le_perm_load_v8hi): Likewise.
11022 (*vsx_le_perm_load_v16qi): Likewise.
11023 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
11024 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
11025 (*vsx_le_perm_store_v8hi): Likewise.
11026 (*vsx_le_perm_store_v16qi): Likewise.
11027 (eight unnamed splitters): Likewise.
11029 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11031 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
11032 * config/rs6000/emmintrin.h: Likewise.
11033 * config/rs6000/mmintrin.h: Likewise.
11034 * config/rs6000/xmmintrin.h: Likewise.
11036 2018-01-10 David Malcolm <dmalcolm@redhat.com>
11039 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
11041 * tree.c (tree_nop_conversion): Return true for location wrapper
11043 (maybe_wrap_with_location): New function.
11044 (selftest::check_strip_nops): New function.
11045 (selftest::test_location_wrappers): New function.
11046 (selftest::tree_c_tests): Call it.
11047 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
11048 (maybe_wrap_with_location): New decl.
11049 (EXPR_LOCATION_WRAPPER_P): New macro.
11050 (location_wrapper_p): New inline function.
11051 (tree_strip_any_location_wrapper): New inline function.
11053 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
11056 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
11057 stack_realign_offset for the largest alignment of stack slot
11059 (ix86_find_max_used_stack_alignment): New function.
11060 (ix86_finalize_stack_frame_flags): Use it. Set
11061 max_used_stack_alignment if we don't realign stack.
11062 * config/i386/i386.h (machine_function): Add
11063 max_used_stack_alignment.
11065 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
11067 * config/arm/arm.opt (-mbranch-cost): New option.
11068 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
11071 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
11074 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
11075 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
11077 2018-01-10 Richard Biener <rguenther@suse.de>
11080 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
11081 early out so it also covers the case where we have a non-NULL
11084 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11086 PR tree-optimization/83753
11087 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
11088 for non-strided grouped accesses if the number of elements is 1.
11090 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11093 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
11094 * i386.h (TARGET_USE_GATHER): Define.
11095 * x86-tune.def (X86_TUNE_USE_GATHER): New.
11097 2018-01-10 Martin Liska <mliska@suse.cz>
11100 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
11101 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
11103 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
11104 CLEANUP_NO_PARTITIONING is not set.
11106 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11108 * doc/rtl.texi: Remove documentation of (const ...) wrappers
11109 for vectors, as a partial revert of r254296.
11110 * rtl.h (const_vec_p): Delete.
11111 (const_vec_duplicate_p): Don't test for vector CONSTs.
11112 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
11113 * expmed.c (make_tree): Likewise.
11116 * common.md (E, F): Use CONSTANT_P instead of checking for
11118 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
11119 checking for CONST_VECTOR.
11121 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11123 PR middle-end/83575
11124 * predict.c (force_edge_cold): Handle in more sane way edges
11125 with no prediction.
11127 2018-01-09 Carl Love <cel@us.ibm.com>
11129 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
11131 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
11132 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
11133 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
11134 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
11135 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
11136 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
11137 * config/rs6000/rs6000-protos.h: Add extern defition for
11138 rs6000_generate_float2_double_code.
11139 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
11141 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
11142 (float2_v2df): Add define_expand.
11144 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
11147 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
11148 op_mode in the force_to_mode call.
11150 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11152 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
11153 instead of checking each element individually.
11154 (aarch64_evpc_uzp): Likewise.
11155 (aarch64_evpc_zip): Likewise.
11156 (aarch64_evpc_ext): Likewise.
11157 (aarch64_evpc_rev): Likewise.
11158 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
11159 instead of checking each element individually. Return true without
11161 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
11162 whether all selected elements come from the same input, instead of
11163 checking each element individually. Remove calls to gen_rtx_REG,
11164 start_sequence and end_sequence and instead assert that no rtl is
11167 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11169 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
11170 order of HIGH and CONST checks.
11172 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11174 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
11175 if the destination isn't an SSA_NAME.
11177 2018-01-09 Richard Biener <rguenther@suse.de>
11179 PR tree-optimization/83668
11180 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
11182 (canonicalize_loop_form): ... here, renamed from ...
11183 (canonicalize_loop_closed_ssa_form): ... this and amended to
11184 swap successor edges for loop exit blocks to make us use
11185 the RPO order we need for initial schedule generation.
11187 2018-01-09 Joseph Myers <joseph@codesourcery.com>
11189 PR tree-optimization/64811
11190 * match.pd: When optimizing comparisons with Inf, avoid
11191 introducing or losing exceptions from comparisons with NaN.
11193 2018-01-09 Martin Liska <mliska@suse.cz>
11196 * asan.c (shadow_mem_size): Add gcc_assert.
11198 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
11200 Don't save registers in main().
11203 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
11204 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
11205 * config/avr/avr.c (avr_set_current_function): Don't error if
11206 naked, OS_task or OS_main are specified at the same time.
11207 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
11209 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
11211 * common/config/avr/avr-common.c (avr_option_optimization_table):
11212 Switch on -mmain-is-OS_task for optimizing compilations.
11214 2018-01-09 Richard Biener <rguenther@suse.de>
11216 PR tree-optimization/83572
11217 * graphite.c: Include cfganal.h.
11218 (graphite_transform_loops): Connect infinite loops to exit
11219 and remove fake edges at the end.
11221 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11223 * ipa-inline.c (edge_badness): Revert accidental checkin.
11225 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11228 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
11229 symbols; not inline clones.
11231 2018-01-09 Jakub Jelinek <jakub@redhat.com>
11234 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
11235 hard registers. Formatting fixes.
11237 PR preprocessor/83722
11238 * gcc.c (try_generate_repro): Pass
11239 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
11240 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
11243 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
11244 Kito Cheng <kito.cheng@gmail.com>
11246 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
11247 (riscv_leaf_function_p): Delete.
11248 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
11250 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11252 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
11254 (do_ifelse): New function.
11255 (do_isel): New function.
11256 (do_sub3): New function.
11257 (do_add3): New function.
11258 (do_load_mask_compare): New function.
11259 (do_overlap_load_compare): New function.
11260 (expand_compare_loop): New function.
11261 (expand_block_compare): Call expand_compare_loop() when appropriate.
11262 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
11263 option description.
11264 (-mblock-compare-inline-loop-limit): New option.
11266 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11269 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
11270 Reverse order of second and third operands in first alternative.
11271 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
11272 of first and second elements in UNSPEC_VPERMR vector.
11273 (altivec_expand_vec_perm_le): Likewise.
11275 2018-01-08 Jeff Law <law@redhat.com>
11277 PR rtl-optimizatin/81308
11278 * tree-switch-conversion.c (cfg_altered): New file scoped static.
11279 (process_switch): If group_case_labels makes a change, then set
11281 (pass_convert_switch::execute): If a switch is converted, then
11282 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
11284 PR rtl-optimization/81308
11285 * recog.c (split_all_insns): Conditionally cleanup the CFG after
11288 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
11290 PR target/83663 - Revert r255946
11291 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11292 generation for cases where splatting a value is not useful.
11293 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11294 across a vec_duplicate and a paradoxical subreg forming a vector
11295 mode to a vec_concat.
11297 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11299 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11300 -march=armv8.3-a variants.
11301 * config/arm/t-multilib: Likewise.
11302 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11304 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11306 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11308 (cceq_ior_compare_complement): Give it a name so I can use it, and
11309 change boolean_or_operator predicate to boolean_operator so it can
11310 be used to generate a crand.
11311 (eqne): New code iterator.
11312 (bd/bd_neg): New code_attrs.
11313 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11314 a single define_insn.
11315 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11316 decrement (bdnzt/bdnzf/bdzt/bdzf).
11317 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11318 with the new names of the branch decrement patterns, and added the
11319 names of the branch decrement conditional patterns.
11321 2018-01-08 Richard Biener <rguenther@suse.de>
11323 PR tree-optimization/83563
11324 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11327 2018-01-08 Richard Biener <rguenther@suse.de>
11329 PR middle-end/83713
11330 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11332 2018-01-08 Richard Biener <rguenther@suse.de>
11334 PR tree-optimization/83685
11335 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11336 references to abnormals.
11338 2018-01-08 Richard Biener <rguenther@suse.de>
11341 * dwarf2out.c (output_indirect_strings): Handle empty
11342 skeleton_debug_str_hash.
11343 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11345 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11347 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11348 (emit_store_direct): Likewise.
11349 (arc_trampoline_adjust_address): Likewise.
11350 (arc_asm_trampoline_template): New function.
11351 (arc_initialize_trampoline): Use asm_trampoline_template.
11352 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11353 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11354 * config/arc/arc.md (flush_icache): Delete pattern.
11356 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11358 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11359 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11362 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11365 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11366 by not USED_FOR_TARGET.
11367 (make_pass_resolve_sw_modes): Likewise.
11369 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11371 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11374 2018-01-08 Richard Biener <rguenther@suse.de>
11376 PR middle-end/83580
11377 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11379 2018-01-08 Richard Biener <rguenther@suse.de>
11381 PR middle-end/83517
11382 * match.pd ((t * 2) / 2) -> t): Add missing :c.
11384 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
11386 PR middle-end/81897
11387 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11388 basic blocks with a small number of successors.
11389 (convert_control_dep_chain_into_preds): Improve handling of
11391 (dump_predicates): Split apart into...
11392 (dump_pred_chain): ...here...
11393 (dump_pred_info): ...and here.
11394 (can_one_predicate_be_invalidated_p): Add debugging printfs.
11395 (can_chain_union_be_invalidated_p): Improve check for invalidation
11397 (uninit_uses_cannot_happen): Avoid unnecessary if
11398 convert_control_dep_chain_into_preds yielded nothing.
11400 2018-01-06 Martin Sebor <msebor@redhat.com>
11402 PR tree-optimization/83640
11403 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11404 subtracting negative offset from size.
11405 (builtin_access::overlap): Adjust offset bounds of the access to fall
11406 within the size of the object if possible.
11408 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
11410 PR rtl-optimization/83699
11411 * expmed.c (extract_bit_field_1): Restrict the vector usage of
11412 extract_bit_field_as_subreg to cases in which the extracted
11413 value is also a vector.
11415 * lra-constraints.c (process_alt_operands): Test for the equivalence
11416 substitutions when detecting a possible reload cycle.
11418 2018-01-06 Jakub Jelinek <jakub@redhat.com>
11421 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11422 by default if flag_selective_schedling{,2}. Formatting fixes.
11424 PR rtl-optimization/83682
11425 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11426 if it has non-VECTOR_MODE element mode.
11427 (vec_duplicate_p): Likewise.
11429 PR middle-end/83694
11430 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11431 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11433 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11436 * config/i386/i386-builtin.def
11437 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11438 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11439 Require also OPTION_MASK_ISA_AVX512F in addition to
11440 OPTION_MASK_ISA_GFNI.
11441 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11442 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11443 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11444 to OPTION_MASK_ISA_GFNI.
11445 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11446 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11447 OPTION_MASK_ISA_AVX512BW.
11448 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11449 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11450 addition to OPTION_MASK_ISA_GFNI.
11451 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11452 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11453 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11454 to OPTION_MASK_ISA_GFNI.
11455 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11456 a requirement for all ISAs rather than any of them with a few
11458 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11460 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11461 bitmasks to be enabled with 3 exceptions, instead of requiring any
11462 enabled ISA with lots of exceptions.
11463 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11464 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11465 Change avx512bw in isa attribute to avx512f.
11466 * config/i386/sgxintrin.h: Add license boilerplate.
11467 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11468 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11469 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11470 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11472 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11473 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11474 temporarily sse2 rather than sse if not enabled already.
11477 * config/i386/sse.md (VI248_VLBW): Rename to ...
11478 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11479 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11480 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11481 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11482 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11483 mode iterator instead of VI248_VLBW.
11485 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11487 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11488 (record_modified): Skip clobbers; add debug output.
11489 (param_change_prob): Use sreal frequencies.
11491 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11493 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11494 punt for user-aligned variables.
11496 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11498 * tree-chrec.c (chrec_contains_symbols): Return true for
11501 2018-01-05 Sudakshina Das <sudi.das@arm.com>
11504 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11505 of (x|y) == x for BICS pattern.
11507 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11509 PR tree-optimization/83605
11510 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11511 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11514 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
11516 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11517 * config/epiphany/rtems.h: New file.
11519 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11520 Uros Bizjak <ubizjak@gmail.com>
11523 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
11524 QIreg_operand instead of register_operand predicate.
11525 * config/i386/i386.c (ix86_rop_should_change_byte_p,
11526 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
11527 comments instead of -fmitigate[-_]rop.
11529 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11532 * cgraphunit.c (symbol_table::compile): Switch to text_section
11533 before calling assembly_start debug hook.
11534 * run-rtl-passes.c (run_rtl_passes): Likewise.
11537 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11539 * tree-vrp.c (extract_range_from_binary_expr_1): Check
11540 range_int_cst_p rather than !symbolic_range_p before calling
11541 extract_range_from_multiplicative_op_1.
11543 2018-01-04 Jeff Law <law@redhat.com>
11545 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11546 redundant test in assertion.
11548 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11550 * doc/rtl.texi: Document machine_mode wrapper classes.
11552 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11554 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11555 using tree_to_uhwi.
11557 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11559 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11560 the VEC_PERM_EXPR fold to fail.
11562 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11565 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
11566 to switched_sections.
11568 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11571 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
11572 test for d.testing.
11574 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
11577 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
11578 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
11580 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11583 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
11584 is BLKmode and bitpos not zero or mode change is needed.
11586 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11589 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
11592 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
11595 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
11596 instead of MULT rtx. Update all corresponding splitters.
11597 (*saddl_se): Ditto.
11598 (*ssub<modesuffix>): Ditto.
11599 (*ssubl_se): Ditto.
11600 (*cmp_sadd_di): Update split patterns.
11601 (*cmp_sadd_si): Ditto.
11602 (*cmp_sadd_sidi): Ditto.
11603 (*cmp_ssub_di): Ditto.
11604 (*cmp_ssub_si): Ditto.
11605 (*cmp_ssub_sidi): Ditto.
11606 * config/alpha/predicates.md (const23_operand): New predicate.
11607 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
11608 Look for ASHIFT, not MULT inner operand.
11609 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
11611 2018-01-04 Martin Liska <mliska@suse.cz>
11613 PR gcov-profile/83669
11614 * gcov.c (output_intermediate_file): Add version to intermediate
11616 * doc/gcov.texi: Document new field 'version' in intermediate
11617 file format. Fix location of '-k' option of gcov command.
11619 2018-01-04 Martin Liska <mliska@suse.cz>
11622 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
11624 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11626 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
11628 2018-01-03 Martin Sebor <msebor@redhat.com>
11630 PR tree-optimization/83655
11631 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
11632 checking calls with invalid arguments.
11634 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11636 * tree-vect-stmts.c (vect_get_store_rhs): New function.
11637 (vectorizable_mask_load_store): Delete.
11638 (vectorizable_call): Return false for masked loads and stores.
11639 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
11640 instead of gimple_assign_rhs1.
11641 (vectorizable_load): Handle IFN_MASK_LOAD.
11642 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
11644 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11646 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
11648 (vectorizable_mask_load_store): ...here.
11649 (vectorizable_load): ...and here.
11651 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11653 * tree-vect-stmts.c (vect_build_all_ones_mask)
11654 (vect_build_zero_merge_argument): New functions, split out from...
11655 (vectorizable_load): ...here.
11657 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11659 * tree-vect-stmts.c (vect_check_store_rhs): New function,
11661 (vectorizable_mask_load_store): ...here.
11662 (vectorizable_store): ...and here.
11664 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11666 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
11668 (vectorizable_mask_load_store): ...here.
11670 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11672 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
11673 (vect_model_store_cost): Take a vec_load_store_type instead of a
11675 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
11676 (vect_model_store_cost): Take a vec_load_store_type instead of a
11678 (vectorizable_mask_load_store): Update accordingly.
11679 (vectorizable_store): Likewise.
11680 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
11682 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11684 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
11685 IFN_MASK_LOAD calls here rather than...
11686 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
11688 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11689 Alan Hayward <alan.hayward@arm.com>
11690 David Sherwood <david.sherwood@arm.com>
11692 * expmed.c (extract_bit_field_1): For vector extracts,
11693 fall back to extract_bit_field_as_subreg if vec_extract
11696 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11697 Alan Hayward <alan.hayward@arm.com>
11698 David Sherwood <david.sherwood@arm.com>
11700 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
11701 they are variable or constant sized.
11702 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
11703 slots for constant-sized data.
11705 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11706 Alan Hayward <alan.hayward@arm.com>
11707 David Sherwood <david.sherwood@arm.com>
11709 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
11710 handling COND_EXPRs with boolean comparisons, try to find a better
11711 basis for the mask type than the boolean itself.
11713 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11715 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
11716 is calculated and how it can be overridden.
11717 * genmodes.c (max_bitsize_mode_any_mode): New variable.
11718 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
11720 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
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 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
11728 Remove the mode argument.
11729 (aarch64_simd_valid_immediate): Remove the mode and inverse
11731 * config/aarch64/iterators.md (bitsize): New iterator.
11732 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
11733 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
11734 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
11735 aarch64_simd_valid_immediate.
11736 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
11737 (aarch64_reg_or_bic_imm): Likewise.
11738 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
11739 with an insn_type enum and msl with a modifier_type enum.
11740 Replace element_width with a scalar_mode. Change the shift
11741 to unsigned int. Add constructors for scalar_float_mode and
11742 scalar_int_mode elements.
11743 (aarch64_vect_float_const_representable_p): Delete.
11744 (aarch64_can_const_movi_rtx_p)
11745 (aarch64_simd_scalar_immediate_valid_for_move)
11746 (aarch64_simd_make_constant): Update call to
11747 aarch64_simd_valid_immediate.
11748 (aarch64_advsimd_valid_immediate_hs): New function.
11749 (aarch64_advsimd_valid_immediate): Likewise.
11750 (aarch64_simd_valid_immediate): Remove mode and inverse
11751 arguments. Rewrite to use the above. Use const_vec_duplicate_p
11752 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
11753 and aarch64_float_const_representable_p on the result.
11754 (aarch64_output_simd_mov_immediate): Remove mode argument.
11755 Update call to aarch64_simd_valid_immediate and use of
11756 simd_immediate_info.
11757 (aarch64_output_scalar_simd_mov_immediate): Update call
11760 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11761 Alan Hayward <alan.hayward@arm.com>
11762 David Sherwood <david.sherwood@arm.com>
11764 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
11765 (mode_nunits): Likewise CONST_MODE_NUNITS.
11766 * machmode.def (ADJUST_NUNITS): Document.
11767 * genmodes.c (mode_data::need_nunits_adj): New field.
11768 (blank_mode): Update accordingly.
11769 (adj_nunits): New variable.
11770 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
11772 (emit_mode_size_inline): Set need_bytesize_adj for all modes
11773 listed in adj_nunits.
11774 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
11775 listed in adj_nunits. Don't emit case statements for such modes.
11776 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
11777 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
11778 nothing if adj_nunits is nonnull.
11779 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
11780 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
11781 (emit_mode_fbit): Update use of print_maybe_const_decl.
11782 (emit_move_size): Likewise. Treat the array as non-const
11784 (emit_mode_adjustments): Handle adj_nunits.
11786 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11788 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
11789 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
11790 (VECTOR_MODES): Use it.
11791 (make_vector_modes): Take the prefix as an argument.
11793 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11794 Alan Hayward <alan.hayward@arm.com>
11795 David Sherwood <david.sherwood@arm.com>
11797 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
11798 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
11799 for MODE_VECTOR_BOOL.
11800 * machmode.def (VECTOR_BOOL_MODE): Document.
11801 * genmodes.c (VECTOR_BOOL_MODE): New macro.
11802 (make_vector_bool_mode): New function.
11803 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
11805 * lto-streamer-in.c (lto_input_mode_table): Likewise.
11806 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
11808 * stor-layout.c (int_mode_for_mode): Likewise.
11809 * tree.c (build_vector_type_for_mode): Likewise.
11810 * varasm.c (output_constant_pool_2): Likewise.
11811 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
11812 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
11813 for MODE_VECTOR_BOOL.
11814 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
11815 of mode class checks.
11816 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
11817 instead of a list of mode class checks.
11818 (expand_vector_scalar_condition): Likewise.
11819 (type_for_widest_vector_mode): Handle BImode as an inner mode.
11821 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11822 Alan Hayward <alan.hayward@arm.com>
11823 David Sherwood <david.sherwood@arm.com>
11825 * machmode.h (mode_size): Change from unsigned short to
11827 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
11828 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11829 or if measurement_type is not polynomial.
11830 (fixed_size_mode::includes_p): Check for constant-sized modes.
11831 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
11832 return a poly_uint16 rather than an unsigned short.
11833 (emit_mode_size): Change the type of mode_size from unsigned short
11834 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
11835 (emit_mode_adjustments): Cope with polynomial vector sizes.
11836 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11838 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11840 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
11841 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
11842 * caller-save.c (setup_save_areas): Likewise.
11843 (replace_reg_with_saved_mem): Likewise.
11844 * calls.c (emit_library_call_value_1): Likewise.
11845 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
11846 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
11847 (gen_lowpart_for_combine): Likewise.
11848 * convert.c (convert_to_integer_1): Likewise.
11849 * cse.c (equiv_constant, cse_insn): Likewise.
11850 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
11851 (cselib_subst_to_values): Likewise.
11852 * dce.c (word_dce_process_block): Likewise.
11853 * df-problems.c (df_word_lr_mark_ref): Likewise.
11854 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
11855 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
11856 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
11857 (rtl_for_decl_location): Likewise.
11858 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
11859 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
11860 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
11861 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
11862 (expand_expr_real_1): Likewise.
11863 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
11864 (pad_below): Likewise.
11865 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11866 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
11867 * ira.c (get_subreg_tracking_sizes): Likewise.
11868 * ira-build.c (ira_create_allocno_objects): Likewise.
11869 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
11870 (ira_sort_regnos_for_alter_reg): Likewise.
11871 * ira-costs.c (record_operand_costs): Likewise.
11872 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
11873 (resolve_simple_move): Likewise.
11874 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
11875 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
11876 (lra_constraints): Likewise.
11877 (CONST_POOL_OK_P): Reject variable-sized modes.
11878 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
11879 (add_pseudo_to_slot, lra_spill): Likewise.
11880 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11881 * optabs-query.c (get_best_extraction_insn): Likewise.
11882 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11883 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
11884 (expand_mult_highpart, valid_multiword_target_p): Likewise.
11885 * recog.c (offsettable_address_addr_space_p): Likewise.
11886 * regcprop.c (maybe_mode_change): Likewise.
11887 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
11888 * regrename.c (build_def_use): Likewise.
11889 * regstat.c (dump_reg_info): Likewise.
11890 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
11891 (find_reloads, find_reloads_subreg_address): Likewise.
11892 * reload1.c (eliminate_regs_1): Likewise.
11893 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
11894 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
11895 (simplify_binary_operation_1, simplify_subreg): Likewise.
11896 * targhooks.c (default_function_arg_padding): Likewise.
11897 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
11898 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
11899 (verify_gimple_assign_ternary): Likewise.
11900 * tree-inline.c (estimate_move_cost): Likewise.
11901 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11902 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
11903 (get_address_cost_ainc): Likewise.
11904 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
11905 (vect_supportable_dr_alignment): Likewise.
11906 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11907 (vectorizable_reduction): Likewise.
11908 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
11909 (vectorizable_operation, vectorizable_load): Likewise.
11910 * tree.c (build_same_sized_truth_vector_type): Likewise.
11911 * valtrack.c (cleanup_auto_inc_dec): Likewise.
11912 * var-tracking.c (emit_note_insn_var_location): Likewise.
11913 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
11914 (ADDR_VEC_ALIGN): Likewise.
11916 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11917 Alan Hayward <alan.hayward@arm.com>
11918 David Sherwood <david.sherwood@arm.com>
11920 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
11922 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11923 or if measurement_type is polynomial.
11924 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
11925 * combine.c (make_extraction): Likewise.
11926 * dse.c (find_shift_sequence): Likewise.
11927 * dwarf2out.c (mem_loc_descriptor): Likewise.
11928 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
11929 (extract_bit_field, extract_low_bits): Likewise.
11930 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
11931 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
11932 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
11933 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
11934 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11935 * reload.c (find_reloads): Likewise.
11936 * reload1.c (alter_reg): Likewise.
11937 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
11938 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
11939 * tree-if-conv.c (predicate_mem_writes): Likewise.
11940 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
11941 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
11942 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
11943 * valtrack.c (dead_debug_insert_temp): Likewise.
11944 * varasm.c (mergeable_constant_section): Likewise.
11945 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
11947 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11948 Alan Hayward <alan.hayward@arm.com>
11949 David Sherwood <david.sherwood@arm.com>
11951 * expr.c (expand_assignment): Cope with polynomial mode sizes
11952 when assigning to a CONCAT.
11954 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11955 Alan Hayward <alan.hayward@arm.com>
11956 David Sherwood <david.sherwood@arm.com>
11958 * machmode.h (mode_precision): Change from unsigned short to
11960 (mode_to_precision): Return a poly_uint16 rather than an unsigned
11962 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
11963 or if measurement_type is not polynomial.
11964 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
11965 in which the mode is already known to be a scalar_int_mode.
11966 * genmodes.c (emit_mode_precision): Change the type of mode_precision
11967 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
11969 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11970 for GET_MODE_PRECISION.
11971 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11972 for GET_MODE_PRECISION.
11973 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
11975 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
11976 (expand_field_assignment, make_extraction): Likewise.
11977 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
11978 (get_last_value): Likewise.
11979 * convert.c (convert_to_integer_1): Likewise.
11980 * cse.c (cse_insn): Likewise.
11981 * expr.c (expand_expr_real_1): Likewise.
11982 * lra-constraints.c (simplify_operand_subreg): Likewise.
11983 * optabs-query.c (can_atomic_load_p): Likewise.
11984 * optabs.c (expand_atomic_load): Likewise.
11985 (expand_atomic_store): Likewise.
11986 * ree.c (combine_reaching_defs): Likewise.
11987 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
11988 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
11989 * tree.h (type_has_mode_precision_p): Likewise.
11990 * ubsan.c (instrument_si_overflow): Likewise.
11992 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11993 Alan Hayward <alan.hayward@arm.com>
11994 David Sherwood <david.sherwood@arm.com>
11996 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
11997 polynomial numbers of units.
11998 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
11999 (valid_vector_subparts_p): New function.
12000 (build_vector_type): Remove temporary shim and take the number
12001 of units as a poly_uint64 rather than an int.
12002 (build_opaque_vector_type): Take the number of units as a
12003 poly_uint64 rather than an int.
12004 * tree.c (build_vector_from_ctor): Handle polynomial
12005 TYPE_VECTOR_SUBPARTS.
12006 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
12007 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
12008 (build_vector_from_val): If the number of units is variable,
12009 use build_vec_duplicate_cst for constant operands and
12010 VEC_DUPLICATE_EXPR otherwise.
12011 (make_vector_type): Remove temporary is_constant ().
12012 (build_vector_type, build_opaque_vector_type): Take the number of
12013 units as a poly_uint64 rather than an int.
12014 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
12016 * cfgexpand.c (expand_debug_expr): Likewise.
12017 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
12018 (store_constructor, expand_expr_real_1): Likewise.
12019 (const_scalar_mask_from_tree): Likewise.
12020 * fold-const-call.c (fold_const_reduction): Likewise.
12021 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
12022 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
12023 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
12024 (fold_relational_const): Likewise.
12025 (native_interpret_vector): Likewise. Change the size from an
12026 int to an unsigned int.
12027 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
12028 TYPE_VECTOR_SUBPARTS.
12029 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
12030 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
12031 duplicating a non-constant operand into a variable-length vector.
12032 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
12033 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
12034 * ipa-icf.c (sem_variable::equals): Likewise.
12035 * match.pd: Likewise.
12036 * omp-simd-clone.c (simd_clone_subparts): Likewise.
12037 * print-tree.c (print_node): Likewise.
12038 * stor-layout.c (layout_type): Likewise.
12039 * targhooks.c (default_builtin_vectorization_cost): Likewise.
12040 * tree-cfg.c (verify_gimple_comparison): Likewise.
12041 (verify_gimple_assign_binary): Likewise.
12042 (verify_gimple_assign_ternary): Likewise.
12043 (verify_gimple_assign_single): Likewise.
12044 * tree-pretty-print.c (dump_generic_node): Likewise.
12045 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12046 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
12047 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
12048 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
12049 (vect_shift_permute_load_chain): Likewise.
12050 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
12051 (expand_vector_condition, optimize_vector_constructor): Likewise.
12052 (lower_vec_perm, get_compute_type): Likewise.
12053 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12054 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
12055 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
12056 (vect_recog_mask_conversion_pattern): Likewise.
12057 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
12058 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
12059 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12060 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
12061 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
12062 (vectorizable_shift, vectorizable_operation, vectorizable_store)
12063 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
12064 (supportable_widening_operation): Likewise.
12065 (supportable_narrowing_operation): Likewise.
12066 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
12068 * varasm.c (output_constant): Likewise.
12070 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12071 Alan Hayward <alan.hayward@arm.com>
12072 David Sherwood <david.sherwood@arm.com>
12074 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
12075 so that both the length == 3 and length != 3 cases set up their
12076 own permute vectors. Add comments explaining why we know the
12077 number of elements is constant.
12078 (vect_permute_load_chain): Likewise.
12080 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12081 Alan Hayward <alan.hayward@arm.com>
12082 David Sherwood <david.sherwood@arm.com>
12084 * machmode.h (mode_nunits): Change from unsigned char to
12086 (ONLY_FIXED_SIZE_MODES): New macro.
12087 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
12088 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
12089 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
12091 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
12092 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
12093 or if measurement_type is not polynomial.
12094 * genmodes.c (ZERO_COEFFS): New macro.
12095 (emit_mode_nunits_inline): Make mode_nunits_inline return a
12097 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
12098 Use ZERO_COEFFS when emitting initializers.
12099 * data-streamer.h (bp_pack_poly_value): New function.
12100 (bp_unpack_poly_value): Likewise.
12101 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12102 for GET_MODE_NUNITS.
12103 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12104 for GET_MODE_NUNITS.
12105 * tree.c (make_vector_type): Remove temporary shim and make
12106 the real function take the number of units as a poly_uint64
12107 rather than an int.
12108 (build_vector_type_for_mode): Handle polynomial nunits.
12109 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
12110 * emit-rtl.c (const_vec_series_p_1): Likewise.
12111 (gen_rtx_CONST_VECTOR): Likewise.
12112 * fold-const.c (test_vec_duplicate_folding): Likewise.
12113 * genrecog.c (validate_pattern): Likewise.
12114 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
12115 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12116 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
12117 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
12118 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
12119 * rtlanal.c (subreg_get_info): Likewise.
12120 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12121 (vect_grouped_load_supported): Likewise.
12122 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
12123 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
12124 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12125 (simplify_const_unary_operation, simplify_binary_operation_1)
12126 (simplify_const_binary_operation, simplify_ternary_operation)
12127 (test_vector_ops_duplicate, test_vector_ops): Likewise.
12128 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
12129 instead of CONST_VECTOR_NUNITS.
12130 * varasm.c (output_constant_pool_2): Likewise.
12131 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
12132 explicit-encoded elements in the XVEC for variable-length vectors.
12134 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12136 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
12138 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12139 Alan Hayward <alan.hayward@arm.com>
12140 David Sherwood <david.sherwood@arm.com>
12142 * coretypes.h (fixed_size_mode): Declare.
12143 (fixed_size_mode_pod): New typedef.
12144 * builtins.h (target_builtins::x_apply_args_mode)
12145 (target_builtins::x_apply_result_mode): Change type to
12146 fixed_size_mode_pod.
12147 * builtins.c (apply_args_size, apply_result_size, result_vector)
12148 (expand_builtin_apply_args_1, expand_builtin_apply)
12149 (expand_builtin_return): Update accordingly.
12151 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12153 * cse.c (hash_rtx_cb): Hash only the encoded elements.
12154 * cselib.c (cselib_hash_rtx): Likewise.
12155 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
12156 CONST_VECTOR encoding.
12158 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12159 Jeff Law <law@redhat.com>
12162 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
12163 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
12164 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
12165 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
12168 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
12169 explicitly probe *sp in a noreturn function if there were any callee
12170 register saves or frame pointer is needed.
12172 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12175 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
12176 BLKmode for ternary, binary or unary expressions.
12179 * var-tracking.c (delete_vta_debug_insn): New inline function.
12180 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
12181 insns from get_insns () to NULL instead of each bb separately.
12182 Use delete_vta_debug_insn. No longer static.
12183 (vt_debug_insns_local, variable_tracking_main_1): Adjust
12184 delete_vta_debug_insns callers.
12185 * rtl.h (delete_vta_debug_insns): Declare.
12186 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
12187 instead of variable_tracking_main.
12189 2018-01-03 Martin Sebor <msebor@redhat.com>
12191 PR tree-optimization/83603
12192 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
12193 arguments past the endof the argument list in functions declared
12194 without a prototype.
12195 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
12196 Avoid checking when arguments are null.
12198 2018-01-03 Martin Sebor <msebor@redhat.com>
12201 * doc/extend.texi (attribute const): Fix a typo.
12202 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
12203 issuing -Wsuggest-attribute for void functions.
12205 2018-01-03 Martin Sebor <msebor@redhat.com>
12207 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
12208 offset_int::from instead of wide_int::to_shwi.
12209 (maybe_diag_overlap): Remove assertion.
12210 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
12211 * gimple-ssa-sprintf.c (format_directive): Same.
12212 (parse_directive): Same.
12213 (sprintf_dom_walker::compute_format_length): Same.
12214 (try_substitute_return_value): Same.
12216 2018-01-03 Jeff Law <law@redhat.com>
12218 PR middle-end/83654
12219 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
12220 non-constant residual for zero at runtime and avoid probing in
12221 that case. Reorganize code for trailing problem to mirror handling
12224 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12226 PR tree-optimization/83501
12227 * tree-ssa-strlen.c (get_string_cst): New.
12228 (handle_char_store): Call get_string_cst.
12230 2018-01-03 Martin Liska <mliska@suse.cz>
12232 PR tree-optimization/83593
12233 * tree-ssa-strlen.c: Include tree-cfg.h.
12234 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
12235 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
12236 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
12238 (strlen_dom_walker::before_dom_children): Call
12239 gimple_purge_dead_eh_edges. Dump tranformation with details
12241 (strlen_dom_walker::before_dom_children): Update call by adding
12242 new argument cleanup_eh.
12243 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
12245 2018-01-03 Martin Liska <mliska@suse.cz>
12248 * cif-code.def (VARIADIC_THUNK): New enum value.
12249 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
12252 2018-01-03 Jan Beulich <jbeulich@suse.com>
12254 * sse.md (mov<mode>_internal): Tighten condition for when to use
12255 vmovdqu<ssescalarsize> for TI and OI modes.
12257 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12259 Update copyright years.
12261 2018-01-03 Martin Liska <mliska@suse.cz>
12264 * ipa-visibility.c (function_and_variable_visibility): Skip
12265 functions with noipa attribure.
12267 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12269 * gcc.c (process_command): Update copyright notice dates.
12270 * gcov-dump.c (print_version): Ditto.
12271 * gcov.c (print_version): Ditto.
12272 * gcov-tool.c (print_version): Ditto.
12273 * gengtype.c (create_file): Ditto.
12274 * doc/cpp.texi: Bump @copying's copyright year.
12275 * doc/cppinternals.texi: Ditto.
12276 * doc/gcc.texi: Ditto.
12277 * doc/gccint.texi: Ditto.
12278 * doc/gcov.texi: Ditto.
12279 * doc/install.texi: Ditto.
12280 * doc/invoke.texi: Ditto.
12282 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12284 * vector-builder.h (vector_builder::m_full_nelts): Change from
12285 unsigned int to poly_uint64.
12286 (vector_builder::full_nelts): Update prototype accordingly.
12287 (vector_builder::new_vector): Likewise.
12288 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12289 (vector_builder::operator ==): Likewise.
12290 (vector_builder::finalize): Likewise.
12291 * int-vector-builder.h (int_vector_builder::int_vector_builder):
12292 Take the number of elements as a poly_uint64 rather than an
12294 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12295 from unsigned int to poly_uint64.
12296 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12297 (vec_perm_indices::new_vector): Likewise.
12298 (vec_perm_indices::length): Likewise.
12299 (vec_perm_indices::nelts_per_input): Likewise.
12300 (vec_perm_indices::input_nelts): Likewise.
12301 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12302 number of elements per input as a poly_uint64 rather than an
12303 unsigned int. Use the original encoding for variable-length
12304 vectors, rather than clamping each individual element.
12305 For the second and subsequent elements in each pattern,
12306 clamp the step and base before clamping their sum.
12307 (vec_perm_indices::series_p): Handle polynomial element counts.
12308 (vec_perm_indices::all_in_range_p): Likewise.
12309 (vec_perm_indices_to_tree): Likewise.
12310 (vec_perm_indices_to_rtx): Likewise.
12311 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12312 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12313 (tree_vector_builder::new_binary_operation): Handle polynomial
12314 element counts. Return false if we need to know the number
12315 of elements at compile time.
12316 * fold-const.c (fold_vec_perm): Punt if the number of elements
12317 isn't known at compile time.
12319 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12321 * vec-perm-indices.h (vec_perm_builder): Change element type
12322 from HOST_WIDE_INT to poly_int64.
12323 (vec_perm_indices::element_type): Update accordingly.
12324 (vec_perm_indices::clamp): Handle polynomial element_types.
12325 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12326 (vec_perm_indices::all_in_range_p): Likewise.
12327 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12329 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12330 polynomial vec_perm_indices element types.
12331 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12332 * fold-const.c (fold_vec_perm): Likewise.
12333 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12334 * tree-vect-generic.c (lower_vec_perm): Likewise.
12335 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12336 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12337 element type to HOST_WIDE_INT.
12339 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12340 Alan Hayward <alan.hayward@arm.com>
12341 David Sherwood <david.sherwood@arm.com>
12343 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12344 rather than an int. Use plus_constant.
12345 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12346 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12348 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12349 Alan Hayward <alan.hayward@arm.com>
12350 David Sherwood <david.sherwood@arm.com>
12352 * calls.c (emit_call_1, expand_call): Change struct_value_size from
12353 a HOST_WIDE_INT to a poly_int64.
12355 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12356 Alan Hayward <alan.hayward@arm.com>
12357 David Sherwood <david.sherwood@arm.com>
12359 * calls.c (load_register_parameters): Cope with polynomial
12360 mode sizes. Require a constant size for BLKmode parameters
12361 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
12362 forces a parameter to be padded at the lsb end in order to
12363 fill a complete number of words, require the parameter size
12364 to be ordered wrt UNITS_PER_WORD.
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 * reload1.c (spill_stack_slot_width): Change element type
12371 from unsigned int to poly_uint64_pod.
12372 (alter_reg): Treat mode sizes as polynomial.
12374 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12375 Alan Hayward <alan.hayward@arm.com>
12376 David Sherwood <david.sherwood@arm.com>
12378 * reload.c (complex_word_subreg_p): New function.
12379 (reload_inner_reg_of_subreg, push_reload): Use it.
12381 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12382 Alan Hayward <alan.hayward@arm.com>
12383 David Sherwood <david.sherwood@arm.com>
12385 * lra-constraints.c (process_alt_operands): Reject matched
12386 operands whose sizes aren't ordered.
12387 (match_reload): Refer to this check here.
12389 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12390 Alan Hayward <alan.hayward@arm.com>
12391 David Sherwood <david.sherwood@arm.com>
12393 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12394 that the mode size is in the set {1, 2, 4, 8, 16}.
12396 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12397 Alan Hayward <alan.hayward@arm.com>
12398 David Sherwood <david.sherwood@arm.com>
12400 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12401 Use plus_constant instead of gen_rtx_PLUS.
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 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12408 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12409 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12410 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12411 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12412 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12413 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12414 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12415 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12416 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12418 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12419 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12420 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12421 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12422 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12423 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12424 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12425 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12426 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12427 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12429 * expr.c (emit_move_resolve_push): Treat the input and result
12430 of PUSH_ROUNDING as a poly_int64.
12431 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12432 (emit_push_insn): Likewise.
12433 * lra-eliminations.c (mark_not_eliminable): Likewise.
12434 * recog.c (push_operand): Likewise.
12435 * reload1.c (elimination_effects): Likewise.
12436 * rtlanal.c (nonzero_bits1): Likewise.
12437 * calls.c (store_one_arg): Likewise. Require the padding to be
12438 known at compile time.
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 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12445 Use plus_constant instead of gen_rtx_PLUS.
12447 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12448 Alan Hayward <alan.hayward@arm.com>
12449 David Sherwood <david.sherwood@arm.com>
12451 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12452 rather than an int.
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 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12459 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12460 via stack temporaries. Treat the mode size as polynomial too.
12462 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12463 Alan Hayward <alan.hayward@arm.com>
12464 David Sherwood <david.sherwood@arm.com>
12466 * expr.c (expand_expr_real_2): When handling conversions involving
12467 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12468 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12469 as a poly_uint64 too.
12471 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12472 Alan Hayward <alan.hayward@arm.com>
12473 David Sherwood <david.sherwood@arm.com>
12475 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
12477 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12478 Alan Hayward <alan.hayward@arm.com>
12479 David Sherwood <david.sherwood@arm.com>
12481 * combine.c (can_change_dest_mode): Handle polynomial
12482 REGMODE_NATURAL_SIZE.
12483 * expmed.c (store_bit_field_1): Likewise.
12484 * expr.c (store_constructor): Likewise.
12485 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12486 and polynomial REGMODE_NATURAL_SIZE.
12487 (gen_lowpart_common): Likewise.
12488 * reginfo.c (record_subregs_of_mode): Likewise.
12489 * rtlanal.c (read_modify_subreg_p): Likewise.
12491 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12492 Alan Hayward <alan.hayward@arm.com>
12493 David Sherwood <david.sherwood@arm.com>
12495 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12496 numbers of elements.
12498 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12499 Alan Hayward <alan.hayward@arm.com>
12500 David Sherwood <david.sherwood@arm.com>
12502 * match.pd: Cope with polynomial numbers of vector elements.
12504 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12505 Alan Hayward <alan.hayward@arm.com>
12506 David Sherwood <david.sherwood@arm.com>
12508 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12509 in a POINTER_PLUS_EXPR.
12511 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12512 Alan Hayward <alan.hayward@arm.com>
12513 David Sherwood <david.sherwood@arm.com>
12515 * omp-simd-clone.c (simd_clone_subparts): New function.
12516 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12517 (ipa_simd_modify_function_body): Likewise.
12519 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12520 Alan Hayward <alan.hayward@arm.com>
12521 David Sherwood <david.sherwood@arm.com>
12523 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
12524 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
12525 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
12526 (expand_vector_condition, vector_element): Likewise.
12527 (subparts_gt): New function.
12528 (get_compute_type): Use subparts_gt.
12529 (count_type_subparts): Delete.
12530 (expand_vector_operations_1): Use subparts_gt instead of
12531 count_type_subparts.
12533 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12534 Alan Hayward <alan.hayward@arm.com>
12535 David Sherwood <david.sherwood@arm.com>
12537 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
12538 (vect_compile_time_alias): ...this new function. Do the calculation
12539 on poly_ints rather than trees.
12540 (vect_prune_runtime_alias_test_list): Update call accordingly.
12542 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12543 Alan Hayward <alan.hayward@arm.com>
12544 David Sherwood <david.sherwood@arm.com>
12546 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
12548 (vect_schedule_slp_instance): Likewise.
12550 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12551 Alan Hayward <alan.hayward@arm.com>
12552 David Sherwood <david.sherwood@arm.com>
12554 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12555 constant and extern definitions for variable-length vectors.
12556 (vect_get_constant_vectors): Note that the number of units
12557 is known to be constant.
12559 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12560 Alan Hayward <alan.hayward@arm.com>
12561 David Sherwood <david.sherwood@arm.com>
12563 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
12564 of units as polynomial. Choose between WIDE and NARROW based
12567 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12568 Alan Hayward <alan.hayward@arm.com>
12569 David Sherwood <david.sherwood@arm.com>
12571 * tree-vect-stmts.c (simd_clone_subparts): New function.
12572 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
12574 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12575 Alan Hayward <alan.hayward@arm.com>
12576 David Sherwood <david.sherwood@arm.com>
12578 * tree-vect-stmts.c (vectorizable_call): Treat the number of
12579 vectors as polynomial. Use build_index_vector for
12580 IFN_GOMP_SIMD_LANE.
12582 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12583 Alan Hayward <alan.hayward@arm.com>
12584 David Sherwood <david.sherwood@arm.com>
12586 * tree-vect-stmts.c (get_load_store_type): Treat the number of
12587 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
12588 for variable-length vectors.
12589 (vectorizable_mask_load_store): Treat the number of units as
12590 polynomial, asserting that it is constant if the condition has
12591 already been enforced.
12592 (vectorizable_store, vectorizable_load): Likewise.
12594 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12595 Alan Hayward <alan.hayward@arm.com>
12596 David Sherwood <david.sherwood@arm.com>
12598 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
12599 of units as polynomial. Punt if we can't tell at compile time
12600 which vector contains the final result.
12602 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12603 Alan Hayward <alan.hayward@arm.com>
12604 David Sherwood <david.sherwood@arm.com>
12606 * tree-vect-loop.c (vectorizable_induction): Treat the number
12607 of units as polynomial. Punt on SLP inductions. Use an integer
12608 VEC_SERIES_EXPR for variable-length integer reductions. Use a
12609 cast of such a series for variable-length floating-point
12612 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12613 Alan Hayward <alan.hayward@arm.com>
12614 David Sherwood <david.sherwood@arm.com>
12616 * tree.h (build_index_vector): Declare.
12617 * tree.c (build_index_vector): New function.
12618 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
12619 of units as polynomial, forcibly converting it to a constant if
12620 vectorizable_reduction has already enforced the condition.
12621 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
12622 to create a {1,2,3,...} vector.
12623 (vectorizable_reduction): Treat the number of units as polynomial.
12624 Choose vectype_in based on the largest scalar element size rather
12625 than the smallest number of units. Enforce the restrictions
12628 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12629 Alan Hayward <alan.hayward@arm.com>
12630 David Sherwood <david.sherwood@arm.com>
12632 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
12633 number of units as polynomial.
12635 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12636 Alan Hayward <alan.hayward@arm.com>
12637 David Sherwood <david.sherwood@arm.com>
12639 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
12640 * target.def (autovectorize_vector_sizes): Return the vector sizes
12641 by pointer, using vector_sizes rather than a bitmask.
12642 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
12643 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
12644 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
12646 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
12647 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
12648 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
12649 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
12650 * omp-general.c (omp_max_vf): Likewise.
12651 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12652 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
12653 * tree-vect-loop.c (vect_analyze_loop): Likewise.
12654 * tree-vect-slp.c (vect_slp_bb): Likewise.
12655 * doc/tm.texi: Regenerate.
12656 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
12658 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
12659 the vector size as a poly_uint64 rather than an unsigned int.
12660 (current_vector_size): Change from an unsigned int to a poly_uint64.
12661 (get_vectype_for_scalar_type): Update accordingly.
12662 * tree.h (build_truth_vector_type): Take the size and number of
12663 units as a poly_uint64 rather than an unsigned int.
12664 (build_vector_type): Add a temporary overload that takes
12665 the number of units as a poly_uint64 rather than an unsigned int.
12666 * tree.c (make_vector_type): Likewise.
12667 (build_truth_vector_type): Take the number of units as a poly_uint64
12668 rather than an unsigned int.
12670 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12671 Alan Hayward <alan.hayward@arm.com>
12672 David Sherwood <david.sherwood@arm.com>
12674 * target.def (get_mask_mode): Take the number of units and length
12675 as poly_uint64s rather than unsigned ints.
12676 * targhooks.h (default_get_mask_mode): Update accordingly.
12677 * targhooks.c (default_get_mask_mode): Likewise.
12678 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
12679 * doc/tm.texi: Regenerate.
12681 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12682 Alan Hayward <alan.hayward@arm.com>
12683 David Sherwood <david.sherwood@arm.com>
12685 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
12686 * omp-general.c (omp_max_vf): Likewise.
12687 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
12688 (expand_omp_simd): Handle polynomial safelen.
12689 * omp-low.c (omplow_simd_context): Add a default constructor.
12690 (omplow_simd_context::max_vf): Change from int to poly_uint64.
12691 (lower_rec_simd_input_clauses): Update accordingly.
12692 (lower_rec_input_clauses): Likewise.
12694 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12695 Alan Hayward <alan.hayward@arm.com>
12696 David Sherwood <david.sherwood@arm.com>
12698 * tree-vectorizer.h (vect_nunits_for_cost): New function.
12699 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
12700 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
12701 (vect_analyze_slp_cost): Likewise.
12702 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
12703 (vect_model_load_cost): Likewise.
12705 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12706 Alan Hayward <alan.hayward@arm.com>
12707 David Sherwood <david.sherwood@arm.com>
12709 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
12710 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
12711 from an unsigned int * to a poly_uint64_pod *.
12712 (calculate_unrolling_factor): New function.
12713 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
12715 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12716 Alan Hayward <alan.hayward@arm.com>
12717 David Sherwood <david.sherwood@arm.com>
12719 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
12720 from an unsigned int to a poly_uint64.
12721 (_loop_vec_info::slp_unrolling_factor): Likewise.
12722 (_loop_vec_info::vectorization_factor): Change from an int
12724 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
12725 (vect_get_num_vectors): New function.
12726 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
12727 (vect_get_num_copies): Use vect_get_num_vectors.
12728 (vect_analyze_data_ref_dependences): Change max_vf from an int *
12729 to an unsigned int *.
12730 (vect_analyze_data_refs): Change min_vf from an int * to a
12732 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12733 than an unsigned HOST_WIDE_INT.
12734 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
12735 (vect_analyze_data_ref_dependence): Change max_vf from an int *
12736 to an unsigned int *.
12737 (vect_analyze_data_ref_dependences): Likewise.
12738 (vect_compute_data_ref_alignment): Handle polynomial vf.
12739 (vect_enhance_data_refs_alignment): Likewise.
12740 (vect_prune_runtime_alias_test_list): Likewise.
12741 (vect_shift_permute_load_chain): Likewise.
12742 (vect_supportable_dr_alignment): Likewise.
12743 (dependence_distance_ge_vf): Take the vectorization factor as a
12744 poly_uint64 rather than an unsigned HOST_WIDE_INT.
12745 (vect_analyze_data_refs): Change min_vf from an int * to a
12747 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
12748 vfm1 as a poly_uint64 rather than an int. Make the same change
12749 for the returned bound_scalar.
12750 (vect_gen_vector_loop_niters): Handle polynomial vf.
12751 (vect_do_peeling): Likewise. Update call to
12752 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
12753 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
12755 * tree-vect-loop.c (vect_determine_vectorization_factor)
12756 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
12757 (vect_get_known_peeling_cost): Likewise.
12758 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
12759 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
12760 (vect_transform_loop): Likewise. Use the lowest possible VF when
12761 updating the upper bounds of the loop.
12762 (vect_min_worthwhile_factor): Make static. Return an unsigned int
12763 rather than an int.
12764 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
12765 polynomial unroll factors.
12766 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
12767 (vect_make_slp_decision): Likewise.
12768 (vect_supported_load_permutation_p): Likewise, and polynomial
12770 (vect_analyze_slp_cost): Handle polynomial vf.
12771 (vect_slp_analyze_node_operations): Likewise.
12772 (vect_slp_analyze_bb_1): Likewise.
12773 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12774 than an unsigned HOST_WIDE_INT.
12775 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
12776 (vectorizable_load): Handle polynomial vf.
12777 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
12779 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
12781 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12782 Alan Hayward <alan.hayward@arm.com>
12783 David Sherwood <david.sherwood@arm.com>
12785 * match.pd: Handle bit operations involving three constants
12786 and try to fold one pair.
12788 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12790 * tree-vect-loop-manip.c: Include gimple-fold.h.
12791 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
12792 niters_maybe_zero parameters. Handle other cases besides a step of 1.
12793 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
12794 Add a path that uses a step of VF instead of 1, but disable it
12796 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
12797 and niters_no_overflow parameters. Update calls to
12798 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
12799 Create a new SSA name if the latter choses to use a ste other
12800 than zero, and return it via niters_vector_mult_vf_var.
12801 * tree-vect-loop.c (vect_transform_loop): Update calls to
12802 vect_do_peeling, vect_gen_vector_loop_niters and
12803 slpeel_make_loop_iterate_ntimes.
12804 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
12805 (vect_gen_vector_loop_niters): Update declarations after above changes.
12807 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
12809 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
12810 128-bit round to integer instructions.
12811 (ceil<mode>2): Likewise.
12812 (btrunc<mode>2): Likewise.
12813 (round<mode>2): Likewise.
12815 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12817 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
12818 unaligned VSX load/store on P8/P9.
12819 (expand_block_clear): Allow the use of unaligned VSX
12820 load/store on P8/P9.
12822 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12824 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
12826 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
12827 swap associated with both a load and a store.
12829 2018-01-02 Andrew Waterman <andrew@sifive.com>
12831 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
12832 * config/riscv/riscv.md (clear_cache): Use it.
12834 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
12836 * web.c: Remove out-of-date comment.
12838 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12840 * expr.c (fixup_args_size_notes): Check that any existing
12841 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
12842 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
12843 (emit_single_push_insn): ...here.
12845 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12847 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
12848 (const_vector_encoded_nelts): New function.
12849 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
12850 (const_vector_int_elt, const_vector_elt): Declare.
12851 * emit-rtl.c (const_vector_int_elt_1): New function.
12852 (const_vector_elt): Likewise.
12853 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
12854 of CONST_VECTOR_ELT.
12856 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12858 * expr.c: Include rtx-vector-builder.h.
12859 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
12860 directly on the tree encoding.
12861 (const_vector_from_tree): Likewise.
12862 * optabs.c: Include rtx-vector-builder.h.
12863 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
12864 sequence of "u" values.
12865 * vec-perm-indices.c: Include rtx-vector-builder.h.
12866 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
12867 directly on the vec_perm_indices encoding.
12869 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12871 * doc/rtl.texi (const_vector): Describe new encoding scheme.
12872 * Makefile.in (OBJS): Add rtx-vector-builder.o.
12873 * rtx-vector-builder.h: New file.
12874 * rtx-vector-builder.c: Likewise.
12875 * rtl.h (rtx_def::u2): Add a const_vector field.
12876 (CONST_VECTOR_NPATTERNS): New macro.
12877 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
12878 (CONST_VECTOR_DUPLICATE_P): Likewise.
12879 (CONST_VECTOR_STEPPED_P): Likewise.
12880 (CONST_VECTOR_ENCODED_ELT): Likewise.
12881 (const_vec_duplicate_p): Check for a duplicated vector encoding.
12882 (unwrap_const_vec_duplicate): Likewise.
12883 (const_vec_series_p): Check for a non-duplicated vector encoding.
12884 Say that the function only returns true for integer vectors.
12885 * emit-rtl.c: Include rtx-vector-builder.h.
12886 (gen_const_vec_duplicate_1): Delete.
12887 (gen_const_vector): Call gen_const_vec_duplicate instead of
12888 gen_const_vec_duplicate_1.
12889 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
12890 (gen_const_vec_duplicate): Use rtx_vector_builder.
12891 (gen_const_vec_series): Likewise.
12892 (gen_rtx_CONST_VECTOR): Likewise.
12893 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
12894 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12895 Build a new vector rather than modifying a CONST_VECTOR in-place.
12896 (handle_special_swappables): Update call accordingly.
12897 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
12898 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12899 Build a new vector rather than modifying a CONST_VECTOR in-place.
12900 (handle_special_swappables): Update call accordingly.
12902 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12904 * simplify-rtx.c (simplify_const_binary_operation): Use
12905 CONST_VECTOR_ELT instead of XVECEXP.
12907 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12909 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
12910 the selector elements to be different from the data elements
12911 if the selector is a VECTOR_CST.
12912 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
12913 ssizetype for the selector.
12915 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12917 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
12918 before testing each element individually.
12919 * tree-vect-generic.c (lower_vec_perm): Likewise.
12921 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12923 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
12924 * selftest-run-tests.c (selftest::run_tests): Call it.
12925 * vector-builder.h (vector_builder::operator ==): New function.
12926 (vector_builder::operator !=): Likewise.
12927 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
12928 (vec_perm_indices::all_from_input_p): New function.
12929 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12930 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
12931 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
12932 instead of reading the VECTOR_CST directly. Detect whether both
12933 vector inputs are the same before constructing the vec_perm_indices,
12934 and update the number of inputs argument accordingly. Use the
12935 utility functions added above. Only construct sel2 if we need to.
12937 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12939 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
12940 the broadcast of the low byte.
12941 (expand_mult_highpart): Use an explicit encoding for the permutes.
12942 * optabs-query.c (can_mult_highpart_p): Likewise.
12943 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
12944 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12945 (vectorizable_bswap): Likewise.
12946 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
12947 explicit encoding for the power-of-2 permutes.
12948 (vect_permute_store_chain): Likewise.
12949 (vect_grouped_load_supported): Likewise.
12950 (vect_permute_load_chain): Likewise.
12952 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12954 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
12955 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
12956 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
12957 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12958 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
12959 (vect_gen_perm_mask_any): Likewise.
12961 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12963 * int-vector-builder.h: New file.
12964 * vec-perm-indices.h: Include int-vector-builder.h.
12965 (vec_perm_indices): Redefine as an int_vector_builder.
12966 (auto_vec_perm_indices): Delete.
12967 (vec_perm_builder): Redefine as a stand-alone class.
12968 (vec_perm_indices::vec_perm_indices): New function.
12969 (vec_perm_indices::clamp): Likewise.
12970 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
12971 (vec_perm_indices::new_vector): New function.
12972 (vec_perm_indices::new_expanded_vector): Update for new
12973 vec_perm_indices class.
12974 (vec_perm_indices::rotate_inputs): New function.
12975 (vec_perm_indices::all_in_range_p): Operate directly on the
12976 encoded form, without computing elided elements.
12977 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
12978 encoding. Update for new vec_perm_indices class.
12979 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
12980 the given vec_perm_builder.
12981 (expand_vec_perm_var): Update vec_perm_builder constructor.
12982 (expand_mult_highpart): Use vec_perm_builder instead of
12983 auto_vec_perm_indices.
12984 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
12985 vec_perm_indices instead of auto_vec_perm_indices. Use a single
12986 or double series encoding as appropriate.
12987 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
12988 vec_perm_indices instead of auto_vec_perm_indices.
12989 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12990 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12991 (vect_permute_store_chain): Likewise.
12992 (vect_grouped_load_supported): Likewise.
12993 (vect_permute_load_chain): Likewise.
12994 (vect_shift_permute_load_chain): Likewise.
12995 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12996 (vect_transform_slp_perm_load): Likewise.
12997 (vect_schedule_slp_instance): Likewise.
12998 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12999 (vectorizable_mask_load_store): Likewise.
13000 (vectorizable_bswap): Likewise.
13001 (vectorizable_store): Likewise.
13002 (vectorizable_load): Likewise.
13003 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
13004 vec_perm_indices instead of auto_vec_perm_indices. Use
13005 tree_to_vec_perm_builder to read the vector from a tree.
13006 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
13007 vec_perm_builder instead of a vec_perm_indices.
13008 (have_whole_vector_shift): Use vec_perm_builder and
13009 vec_perm_indices instead of auto_vec_perm_indices. Leave the
13010 truncation to calc_vec_perm_mask_for_shift.
13011 (vect_create_epilog_for_reduction): Likewise.
13012 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
13013 from auto_vec_perm_indices to vec_perm_indices.
13014 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13015 instead of changing individual elements.
13016 (aarch64_vectorize_vec_perm_const): Use new_vector to install
13017 the vector in d.perm.
13018 * config/arm/arm.c (expand_vec_perm_d::perm): Change
13019 from auto_vec_perm_indices to vec_perm_indices.
13020 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13021 instead of changing individual elements.
13022 (arm_vectorize_vec_perm_const): Use new_vector to install
13023 the vector in d.perm.
13024 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
13025 Update vec_perm_builder constructor.
13026 (rs6000_expand_interleave): Likewise.
13027 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
13028 (rs6000_expand_interleave): Likewise.
13030 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13032 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
13033 to qimode could truncate the indices.
13034 * optabs.c (expand_vec_perm_var): Likewise.
13036 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13038 * Makefile.in (OBJS): Add vec-perm-indices.o.
13039 * vec-perm-indices.h: New file.
13040 * vec-perm-indices.c: Likewise.
13041 * target.h (vec_perm_indices): Replace with a forward class
13043 (auto_vec_perm_indices): Move to vec-perm-indices.h.
13044 * optabs.h: Include vec-perm-indices.h.
13045 (expand_vec_perm): Delete.
13046 (selector_fits_mode_p, expand_vec_perm_var): Declare.
13047 (expand_vec_perm_const): Declare.
13048 * target.def (vec_perm_const_ok): Replace with...
13049 (vec_perm_const): ...this new hook.
13050 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
13051 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
13052 * doc/tm.texi: Regenerate.
13053 * optabs.def (vec_perm_const): Delete.
13054 * doc/md.texi (vec_perm_const): Likewise.
13055 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
13056 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
13057 expand_vec_perm for constant permutation vectors. Assert that
13058 the mode of variable permutation vectors is the integer equivalent
13059 of the mode that is being permuted.
13060 * optabs-query.h (selector_fits_mode_p): Declare.
13061 * optabs-query.c: Include vec-perm-indices.h.
13062 (selector_fits_mode_p): New function.
13063 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
13064 is defined, instead of checking whether the vec_perm_const_optab
13065 exists. Use targetm.vectorize.vec_perm_const instead of
13066 targetm.vectorize.vec_perm_const_ok. Check whether the indices
13067 fit in the vector mode before using a variable permute.
13068 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
13069 vec_perm_indices instead of an rtx.
13070 (expand_vec_perm): Replace with...
13071 (expand_vec_perm_const): ...this new function. Take the selector
13072 as a vec_perm_indices rather than an rtx. Also take the mode of
13073 the selector. Update call to shift_amt_for_vec_perm_mask.
13074 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
13075 Use vec_perm_indices::new_expanded_vector to expand the original
13076 selector into bytes. Check whether the indices fit in the vector
13077 mode before using a variable permute.
13078 (expand_vec_perm_var): Make global.
13079 (expand_mult_highpart): Use expand_vec_perm_const.
13080 * fold-const.c: Includes vec-perm-indices.h.
13081 * tree-ssa-forwprop.c: Likewise.
13082 * tree-vect-data-refs.c: Likewise.
13083 * tree-vect-generic.c: Likewise.
13084 * tree-vect-loop.c: Likewise.
13085 * tree-vect-slp.c: Likewise.
13086 * tree-vect-stmts.c: Likewise.
13087 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
13089 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
13090 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
13091 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
13092 (aarch64_vectorize_vec_perm_const): ...this new function.
13093 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13094 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13095 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
13096 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
13097 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13098 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13099 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
13101 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
13102 check for NEON modes.
13103 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
13104 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
13105 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
13106 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
13108 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
13109 the old VEC_PERM_CONST conditions.
13110 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
13111 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
13112 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
13113 (ia64_vectorize_vec_perm_const_ok): Merge into...
13114 (ia64_vectorize_vec_perm_const): ...this new function.
13115 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
13116 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
13117 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
13118 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
13119 * config/mips/mips.c (mips_expand_vec_perm_const)
13120 (mips_vectorize_vec_perm_const_ok): Merge into...
13121 (mips_vectorize_vec_perm_const): ...this new function.
13122 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
13123 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
13124 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
13125 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
13126 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
13127 (rs6000_expand_vec_perm_const): Delete.
13128 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
13130 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13131 (altivec_expand_vec_perm_const_le): Take each operand individually.
13132 Operate on constant selectors rather than rtxes.
13133 (altivec_expand_vec_perm_const): Likewise. Update call to
13134 altivec_expand_vec_perm_const_le.
13135 (rs6000_expand_vec_perm_const): Delete.
13136 (rs6000_vectorize_vec_perm_const_ok): Delete.
13137 (rs6000_vectorize_vec_perm_const): New function.
13138 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13139 an element count and rtx array.
13140 (rs6000_expand_extract_even): Update call accordingly.
13141 (rs6000_expand_interleave): Likewise.
13142 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
13143 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
13144 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
13145 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
13146 (rs6000_expand_vec_perm_const): Delete.
13147 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13148 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13149 (altivec_expand_vec_perm_const_le): Take each operand individually.
13150 Operate on constant selectors rather than rtxes.
13151 (altivec_expand_vec_perm_const): Likewise. Update call to
13152 altivec_expand_vec_perm_const_le.
13153 (rs6000_expand_vec_perm_const): Delete.
13154 (rs6000_vectorize_vec_perm_const_ok): Delete.
13155 (rs6000_vectorize_vec_perm_const): New function. Remove stray
13156 reference to the SPE evmerge intructions.
13157 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13158 an element count and rtx array.
13159 (rs6000_expand_extract_even): Update call accordingly.
13160 (rs6000_expand_interleave): Likewise.
13161 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
13162 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
13164 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13166 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13168 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
13169 vector mode and that that mode matches the mode of the data
13171 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
13172 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
13173 directly using expand_vec_perm_1 when forcing selectors into
13175 (expand_vec_perm_var): New function, split out from expand_vec_perm.
13177 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13179 * optabs-query.h (can_vec_perm_p): Delete.
13180 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
13181 * optabs-query.c (can_vec_perm_p): Split into...
13182 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
13183 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
13184 particular selector is valid.
13185 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13186 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13187 (vect_grouped_load_supported): Likewise.
13188 (vect_shift_permute_load_chain): Likewise.
13189 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13190 (vect_transform_slp_perm_load): Likewise.
13191 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13192 (vectorizable_bswap): Likewise.
13193 (vect_gen_perm_mask_checked): Likewise.
13194 * fold-const.c (fold_ternary_loc): Likewise. Don't take
13195 implementations of variable permutation vectors into account
13196 when deciding which selector to use.
13197 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
13198 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
13199 with a false third argument.
13200 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
13201 to test whether the constant selector is valid and can_vec_perm_var_p
13202 to test whether a variable selector is valid.
13204 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13206 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
13207 * optabs-query.c (can_vec_perm_p): Likewise.
13208 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
13209 instead of vec_perm_indices.
13210 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
13211 (vect_gen_perm_mask_checked): Likewise,
13212 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
13213 (vect_gen_perm_mask_checked): Likewise,
13215 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13217 * optabs-query.h (qimode_for_vec_perm): Declare.
13218 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
13219 (qimode_for_vec_perm): ...this new function.
13220 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
13222 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13224 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
13225 does not have a conditional at the top.
13227 2018-01-02 Richard Biener <rguenther@suse.de>
13229 * ipa-inline.c (big_speedup_p): Fix expression.
13231 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13234 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
13237 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13241 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
13242 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
13243 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
13244 cond_taken_branch_cost 3->4.
13246 2018-01-01 Jakub Jelinek <jakub@redhat.com>
13248 PR tree-optimization/83581
13249 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
13250 TODO_cleanup_cfg if any changes have been made.
13252 PR middle-end/83608
13253 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
13254 convert_modes if target mode has the right side, but different mode
13257 PR middle-end/83609
13258 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
13259 last argument when extracting from CONCAT. If either from_real or
13260 from_imag is NULL, use expansion through memory. If result is not
13261 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
13262 the parts directly to inner mode, if even that fails, use expansion
13265 PR middle-end/83623
13266 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
13267 check for bswap in mode rather than HImode and use that in expand_unop
13270 Copyright (C) 2018 Free Software Foundation, Inc.
13272 Copying and distribution of this file, with or without modification,
13273 are permitted in any medium without royalty provided the copyright
13274 notice and this notice are preserved.