1 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
4 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
6 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
10 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
12 (armv2, armv2a, armv3, armv3m): Delete architectures.
13 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
14 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
15 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
17 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
18 (*mulsidi3adddi): Likewise.
20 (*mulsidi3_nov6): Likewise.
21 (umulsidi3): Likewise.
22 (umulsidi3_nov6): Likewise.
23 (umaddsidi4): Likewise.
24 (*umulsidi3adddi): Likewise.
25 (smulsi3_highpart): Likewise.
26 (*smulsi3_highpart_nov6): Likewise.
27 (umulsi3_highpart): Likewise.
28 (*umulsi3_highpart_nov6): Likewise.
29 * config/arm/arm.h (arm_arch3m): Delete.
30 * config/arm/arm.c (arm_arch3m): Delete.
31 (arm_option_override_internal): Update armv3-related comment.
32 (arm_configure_build_target): Delete use of isa_bit_mode32.
33 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
34 (arm_rtx_costs_internal): Delete check of arm_arch3m.
35 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
38 * config/arm/arm-protos.h (arm_arch3m): Delete.
39 * config/arm/arm-tables.opt: Regenerate.
40 * config/arm/arm-tune.md: Likewise.
41 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
42 deleted architectures.
44 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
46 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
47 (armv5t, armv5te): New features.
48 (ARMv5, ARMv5e): Delete fgroups.
49 (ARMv5t, ARMv5te): Adjust for above changes.
51 (armv5, armv5e): Delete arches.
52 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
54 (*call_reg_arm): Likewise.
55 (*call_value_reg_armv5): Likewise.
56 (*call_value_reg_arm): Likewise.
57 (*call_symbol): Likewise.
58 (*call_value_symbol): Likewise.
59 (*sibcall_insn): Likewise.
60 (*sibcall_value_insn): Likewise.
63 (define_split and define_peephole2 dependent on arm_arch5):
65 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
67 (TARGET_ARM_QBIT): Likewise.
68 (TARGET_DSP_MULTIPLY): Likewise.
69 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
70 (arm_arch5, arm_arch5e): Delete.
71 (arm_arch5t, arm_arch5te): Declare.
72 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
73 (arm_arch5t): Declare.
74 (arm_option_reconfigure_globals): Update for the above.
75 (arm_options_perform_arch_sanity_checks): Update comment, replace
76 use of arm_arch5 with arm_arch5t.
77 (use_return_insn): Likewise.
78 (arm_emit_call_insn): Likewise.
79 (output_return_instruction): Likewise.
80 (arm_final_prescan_insn): Likewise.
81 (arm_coproc_builtin_available): Likewise.
82 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
83 arm_arch5e with arm_arch5t and arm_arch5te.
84 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
85 (arm_arch5t, arm_arch5te): Declare.
86 * config/arm/arm-tables.opt: Regenerate.
87 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
88 * config/arm/t-multilib: Likewise.
89 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
91 (*call_reg_thumb1): Likewise.
92 (*call_value_reg_thumb1_v5): Likewise.
93 (*call_value_reg_thumb1): Likewise.
94 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
96 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
98 2018-05-18 Martin Liska <mliska@suse.cz>
100 PR gcov-profile/84846
101 * doc/gcov.texi: Document -t option of gcov tool.
103 2018-05-18 Martin Liska <mliska@suse.cz>
105 PR gcov-profile/84846
106 * gcov.c (print_usage): Add new -t option.
107 (process_args): Handle the option.
108 (generate_results): Use stdout as output when requested by
111 2018-05-18 Martin Liska <mliska@suse.cz>
113 PR gcov-profile/84846
114 * coverage.c (coverage_init): Write PWD to .gcno file.
115 * doc/gcov.texi: Document how working directory is printed.
116 * gcov-dump.c (dump_gcov_file): Print PWD.
117 * gcov.c (output_intermediate_file): Likewise.
118 (read_graph_file): Read PWD string.
119 (output_lines): Print PWD.
121 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
124 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
125 for retval and return false if all args to phi are zero.
127 2018-05-18 Richard Biener <rguenther@suse.de>
129 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
131 (evrp_dom_walker::before_dom_children): Call it.
133 2018-05-18 Richard Biener <rguenther@suse.de>
135 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
136 results when processing array refs with variable index.
138 2018-05-18 Toon Moene <toon@moene.org>
140 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
141 directly after that of -floop-interchange. Indicate that both
142 options are enabled by default when specifying -O3.
144 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
146 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
147 iterator. Delete separate integer-mode vec_set<mode> expander.
148 (aarch64_simd_vec_setv2di): Delete.
149 (vec_setv2di): Delete.
150 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
151 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
152 the "w, r" alternative.
154 2018-05-18 Martin Liska <mliska@suse.cz>
156 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
157 * tree-pass.h (make_pass_lower_switch_O0): New function.
158 * tree-switch-conversion.c (node_has_low_bound): Remove.
159 (node_has_high_bound): Likewise.
160 (node_is_bounded): Likewise.
161 (class pass_lower_switch): Make it a template type and create
163 (pass_lower_switch::execute): Add template argument.
164 (make_pass_lower_switch): New function.
165 (make_pass_lower_switch_O0): New function.
166 (do_jump_if_equal): Remove.
167 (emit_case_nodes): Simplify to just handle all 3 cases and leave
168 all the hard work to tree optimization passes.
170 2018-05-18 Martin Liska <mliska@suse.cz>
172 * dbgcnt.c (limit_low): Renamed from limit.
173 (limit_high): New variable.
174 (dbg_cnt_is_enabled): Check for upper limit.
175 (dbg_cnt): Adjust dumping.
176 (dbg_cnt_set_limit_by_index): Add new argument for high
178 (dbg_cnt_set_limit_by_name): Likewise.
179 (dbg_cnt_process_single_pair): Parse new format.
180 (dbg_cnt_process_opt): Use strtok.
181 (dbg_cnt_list_all_counters): Remove 'value' and add
183 * doc/invoke.texi: Document changes.
185 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
187 * doc/sourcebuild.texi (scalar_all_fma): Document.
188 * tree.def (FMA_EXPR): Delete.
189 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
190 * internal-fn.c (ternary_direct): New macro.
191 (expand_ternary_optab_fn): Likewise.
192 (direct_ternary_optab_supported_p): Likewise.
193 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
194 * builtins.c (fold_builtin_fma): Delete.
195 (fold_builtin_3): Don't call it.
196 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
197 * expr.c (expand_expr_real_2): Likewise.
198 * fold-const.c (operand_equal_p): Likewise.
199 (fold_ternary_loc): Likewise.
200 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
201 * gimple.c (DEFTREECODE): Likewise.
202 * gimplify.c (gimplify_expr): Likewise.
203 * optabs-tree.c (optab_for_tree_code): Likewise.
204 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
205 * tree-eh.c (operation_could_trap_p): Likewise.
206 (stmt_could_throw_1_p): Likewise.
207 * tree-inline.c (estimate_operator_cost): Likewise.
208 * tree-pretty-print.c (dump_generic_node): Likewise.
209 (op_code_prio): Likewise.
210 * tree-ssa-loop-im.c (stmt_cost): Likewise.
211 * tree-ssa-operands.c (get_expr_operands): Likewise.
212 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
213 * fold-const-call.h (fold_fma): Delete.
214 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
215 CFN_FNMA and CFN_FNMS.
217 * genmatch.c (combined_fn): New enum.
218 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
219 (commutative_op): New function.
220 (commutate): Use it. Handle more than 2 operands.
221 (dt_operand::gen_gimple_expr): Use commutative_op.
222 (parser::parse_expr): Allow :c to be used with non-binary
223 operators if the commutative operand is known.
224 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
225 CFN_FMS, CFN_FNMA and CFN_FNMS.
226 (backprop::process_assign_use): Remove FMA_EXPR handling.
227 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
228 (gen_hsa_fma): New function.
229 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
230 IFN_FNMA and IFN_FNMS.
231 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
232 * gimple-fold.h (follow_all_ssa_edges): Declare.
233 * gimple-fold.c (follow_all_ssa_edges): New function.
234 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
235 gimple_build interface and use follow_all_ssa_edges to fold the result.
236 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
237 instead of checking for optabs directly.
238 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
239 rather than FMA_EXPRs.
240 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
241 call to IFN_FMA instead of an FMA_EXPR.
243 2018-05-17 Jim Wilson <jimw@sifive.com>
245 * expr.c (do_tablejump): When converting index to Pmode, if we have a
246 sign extended promoted subreg, and the range does not have the sign bit
247 set, then do a sign extend.
249 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
250 test, check for sign extended subreg and/or constant operands, and
251 do a sign extend in that case.
253 2018-05-17 Steve Ellcey <sellcey@cavium.com>
255 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
256 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
258 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
259 Change logics_shift_reg to logics_shift_imm.
260 (thunderx2t99_fp_loadpair_basic): Delete.
261 (thunderx2t99_fp_storepair_basic): Delete.
262 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
263 (thunderx2t99_asimd_polynomial): Delete.
264 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
265 and neon_fp_mul_d_scalar_q.
266 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
267 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
268 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
269 (thunderx2t99_asimd_lut): Add missing tbl types.
270 (thunderx2t99_asimd_ext): Delete.
271 (thunderx2t99_asimd_load1_1_mult): Delete.
272 (thunderx2t99_asimd_load1_2_mult): Delete.
273 (thunderx2t99_asimd_load1_ldp): New.
274 (thunderx2t99_asimd_load1): New.
275 (thunderx2t99_asimd_load2): Add missing *load2* types.
276 (thunderx2t99_asimd_load3): New.
277 (thunderx2t99_asimd_load4): New.
278 (thunderx2t99_asimd_store1_1_mult): Delete.
279 (thunderx2t99_asimd_store1_2_mult): Delete.
280 (thunderx2t99_asimd_store2_mult): Delete.
281 (thunderx2t99_asimd_store2_onelane): Delete.
282 (thunderx2t99_asimd_store_stp): New.
283 (thunderx2t99_asimd_store1): New.
284 (thunderx2t99_asimd_store2): New.
285 (thunderx2t99_asimd_store3): New.
286 (thunderx2t99_asimd_store4): New.
288 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
290 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
291 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
293 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
294 Segher Boessenkool <segher@kernel.crashing.org>
297 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
300 2018-05-17 Richard Biener <rguenther@suse.de>
302 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
303 for pruning loop and prune defs feeding only already visited PHIs.
305 2018-05-17 Richard Biener <rguenther@suse.de>
307 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
309 2018-05-17 Bin Cheng <bin.cheng@arm.com>
310 Richard Biener <rguenther@suse.de>
312 PR tree-optimization/85793
313 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
314 for VMAT_ELEMENTWISE.
316 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
318 * internal-fn.h (lookup_internal_fn): Declare
319 * internal-fn.c (lookup_internal_fn): New function.
320 * gimple.c (gimple_build_call_from_tree): Handle calls to
322 * gimple-pretty-print.c (dump_gimple_call): Print "." before
323 internal function names.
324 * tree-pretty-print.c (dump_generic_node): Likewise.
325 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
327 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
329 * gimple-fold.h (gimple_build): Make the function forms take
330 combined_fn rather than built_in_function.
331 (gimple_simplify): Likewise.
332 * gimple-match-head.c (gimple_simplify): Likewise.
333 * gimple-fold.c (gimple_build): Likewise.
334 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
335 rather than gimple_build_call_internal.
336 (get_initial_defs_for_reduction): Likewise.
337 (vect_create_epilog_for_reduction): Likewise.
338 (vectorizable_live_operation): Likewise.
340 2018-05-17 Martin Liska <mliska@suse.cz>
342 * gimple-ssa-sprintf.c (format_directive): Do not use
343 space in between 'G_' and '('.
345 2018-05-17 Jakub Jelinek <jakub@redhat.com>
348 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
349 even if the mask is not all ones.
352 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
354 (ix86_gimple_fold_builtin): Likewise.
357 * config/i386/i386.c: Include tree-vector-builder.h.
358 (ix86_vector_shift_count): New function.
359 (ix86_fold_builtin): Fold shift builtins by scalar count.
360 (ix86_gimple_fold_builtin): Likewise.
362 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
363 _mm512_setzero): New intrinsics.
365 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
366 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
368 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
369 code generation for cases where splatting a value is not useful.
370 * simplify-rtx.c (simplify_ternary_operation): Simplify
371 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
372 mode to a vec_concat.
374 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
376 * config.gcc: Support "goldmont-plus".
377 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
379 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
380 PROCESSOR_GOLDMONT_PLUS.
381 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
382 (processor_target_table): Add "goldmont-plus".
383 (PTA_GOLDMONT_PLUS): Define.
384 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
385 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
386 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
387 (fold_builtin_cpu): Add "goldmont-plus".
388 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
389 (ix86_option_override_internal): Add "goldmont-plus".
390 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
391 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
392 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
393 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
395 2018-05-17 Richard Biener <rguenther@suse.de>
397 PR tree-optimization/85757
398 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
399 remove defs that only feed that PHI from further processing.
401 2018-05-16 Jim Wilson <jimw@sifive.com>
403 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
405 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
407 2018-05-16 Mark Wielaard <mark@klomp.org>
409 * dwarf2out.c (count_index_strings): New function.
410 (output_indirect_strings): Call count_index_strings and generate
411 header for dwarf_version >= 5.
413 2018-05-16 Mark Wielaard <mark@klomp.org>
415 * dwarf2out.c (dwarf_FORM): New function.
416 (set_indirect_string): Use dwarf_FORM.
417 (reset_indirect_string): Likewise.
418 (size_of_die): Likewise.
419 (value_format): Likewise.
420 (output_die): Likewise.
421 (add_skeleton_AT_string): Likewise.
422 (output_macinfo_op): Likewise.
423 (index_string): Likewise.
424 (output_index_string_offset): Likewise.
425 (output_index_string): Likewise.
426 (count_index_strings): Likewise.
428 2018-05-16 Carl Love <cel@us.ibm.com>
430 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
431 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
433 2018-05-16 Martin Jambor <mjambor@suse.cz>
435 * ipa-prop.c (ipa_free_all_edge_args): Remove.
436 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
438 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
440 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
441 (fnma<mode>4): Likewise.
442 (fms<mode>4): Likewise.
443 (fnms<mode>4): Likewise.
444 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
445 (aarch64_fnma<mode>4): Likewise.
446 (aarch64_fms<mode>4): Likewise.
447 (aarch64_fnms<mode>4): Likewise.
448 (aarch64_fnmadd<mode>4): Likewise.
450 2018-05-16 Jason Merrill <jason@redhat.com>
452 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
454 2018-05-16 Richard Biener <rguenther@suse.de>
456 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
457 (dump_stmt_cost): Declare.
458 (add_stmt_cost): Dump cost we add.
459 (add_stmt_costs): New function.
460 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
462 (vect_analyze_stmt): Adjust prototype.
463 (vectorizable_condition): Likewise.
464 (vectorizable_live_operation): Likewise.
465 (vectorizable_reduction): Likewise.
466 (vectorizable_induction): Likewise.
467 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
468 cost vector to pass to vectorizable_ and record afterwards.
469 (vect_model_reduction_cost): Take cost vector argument and adjust.
470 (vect_model_induction_cost): Likewise.
471 (vectorizable_reduction): Likewise.
472 (vectorizable_induction): Likewise.
473 (vectorizable_live_operation): Likewise.
474 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
475 SLP_TREE_NUMBER_OF_VEC_STMTS.
476 (vect_analyze_slp_cost_1): Remove.
477 (vect_analyze_slp_cost): Likewise.
478 (vect_slp_analyze_node_operations): Take visited args and
479 a target cost vector. Avoid processing already visited stmt sets.
480 (vect_slp_analyze_operations): Use a local cost vector to gather
481 costs and register those of non-discarded instances.
482 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
483 (vect_schedule_slp_instance): Remove copying of
484 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
486 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
487 adding cost. Record cost entry location.
488 (vect_prologue_cost_for_slp_op): Function to compute cost of
489 a constant or invariant generated for SLP vect in the prologue,
490 split out from vect_analyze_slp_cost_1.
491 (vect_model_simple_cost): Make static. Adjust for SLP costing.
492 (vect_model_promotion_demotion_cost): Likewise.
493 (vect_model_store_cost): Likewise, make static.
494 (vect_model_load_cost): Likewise.
495 (vectorizable_bswap): Add cost vector arg and adjust.
496 (vectorizable_call): Likewise.
497 (vectorizable_simd_clone_call): Likewise.
498 (vectorizable_conversion): Likewise.
499 (vectorizable_assignment): Likewise.
500 (vectorizable_shift): Likewise.
501 (vectorizable_operation): Likewise.
502 (vectorizable_store): Likewise.
503 (vectorizable_load): Likewise.
504 (vectorizable_condition): Likewise.
505 (vectorizable_comparison): Likewise.
506 (can_vectorize_live_stmts): Likewise.
507 (vect_analyze_stmt): Likewise.
508 (vect_transform_stmt): Adjust calls to vectorizable_*.
509 * tree-vectorizer.c: Include gimple-pretty-print.h.
510 (dump_stmt_cost): New function.
512 2018-05-16 Richard Biener <rguenther@suse.de>
514 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
515 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
516 * tree-ssa-dse.c: Include tree-ssa-loop.h.
517 (check_name): New callback.
518 (dse_classify_store): Track cycles via a visited bitmap of PHI
519 defs and simplify handling of in-loop and across loop dead stores
520 and properly fail for loop-variant refs. Handle byte-tracking with
521 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
524 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
526 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
527 (vect_get_mask_type_for_stmt): Likewise.
528 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
530 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
531 to determine the statement's vector type and the vector type that
532 should be used for calculating nunits. Deal with cases in which
533 the type has to be deferred.
534 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
535 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
536 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
537 (vect_determine_vf_for_stmt): New functions, split out from...
538 (vect_determine_vectorization_factor): ...here.
539 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
540 (vect_get_mask_type_for_stmt): New functions, split out from
541 vect_determine_vectorization_factor.
543 2018-05-16 Richard Biener <rguenther@suse.de>
545 * tree-cfg.c (verify_gimple_assign_ternary): Properly
546 verify the [VEC_]COND_EXPR embedded comparison.
548 2018-05-15 Martin Sebor <msebor@redhat.com>
550 PR tree-optimization/85753
551 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
552 RECORD_TYPE in addition to ARRAY_TYPE.
554 2018-05-15 Martin Sebor <msebor@redhat.com>
557 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
559 2018-05-15 Richard Biener <rguenther@suse.de>
561 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
562 add by_clobber_p one. Change algorithm to collect all defs
563 representing uses we need to walk and try reducing them to
564 a single one before failing.
565 (dse_dom_walker::dse_optimize_stmt): Adjust.
567 2018-05-13 Mark Wielaard <mark@klomp.org>
569 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
570 (size_of_loc_descr): Likewise.
571 (output_loc_operands): Likewise.
572 (output_loc_operands_raw): Likewise.
573 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
574 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
575 (hash_loc_operands): Likewise.
576 (compare_loc_operands): Likewise.
578 2018-05-14 Mark Wielaard <mark@klomp.org>
580 * dwarf2out.c (count_index_addrs): New function.
581 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
583 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
585 PR tree-optimization/83648
586 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
587 return value as malloc candidate.
589 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
592 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
593 as true in call to suggest_attribute.
595 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
597 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
600 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
602 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
604 (aarch64_crypto_aesd_fused): Likewise.
606 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
608 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
609 (movsi_aarch64): Likewise.
610 (load_pairsi): Likewise.
611 (load_pairdi): Likewise.
612 (store_pairsi): Likewise.
613 (store_pairdi): Likewise.
614 (load_pairsf): Likewise.
615 (load_pairdf): Likewise.
616 (store_pairsf): Likewise.
617 (store_pairdf): Likewise.
618 (zero_extend): Likewise.
619 (trunc): Swap alternatives.
620 (fcvt_target): Add '?' to prefer w over r.
622 2018-05-14 Jakub Jelinek <jakub@redhat.com>
625 * config/i386/i386.md: Disallow non-commutative arithmetics in
626 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
627 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
628 in the peephole2 before it.
630 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
632 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
633 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
634 (ix86_handle_option): Handle -mcldemote.
635 * config.gcc: New header.
636 * config/i386/cldemoteintrin.h: New file.
637 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
638 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
640 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
641 OPTION_MASK_ISA_CLDEMOTE.
642 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
643 (ix86_valid_target_attribute_inner_p): Ditto.
644 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
645 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
646 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
647 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
648 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
650 * config/i386/i386.opt: Add -mcldemote.
651 * config/i386/x86intrin.h: New header.
652 * doc/invoke.texi: Add -mcldemote.
654 2018-05-14 Richard Biener <rguenther@suse.de>
656 * doc/match-and-simplify.texi: Adjust :s documentation.
658 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
660 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
661 intended memcpy size.
662 (REORDER_45): Likewise.
664 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
667 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
668 * vec.c (qsort_chk): Use gcc_qsort.
669 * Makefile.in (OBJS-libcommon): Add sort.o.
670 (build/sort.o): New target. Use it...
671 (BUILD_RTL): ... here, and...
672 (build/gencfn-macros): ... here, and...
673 (build/genmatch): ... here.
675 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
676 Chung-Ju Wu <jasonwucj@gmail.com>
678 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
679 * config/nds32/nds32-graywolf.md: New file.
680 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
681 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
683 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
684 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
685 * config/nds32/nds32.md (pipeline_model): Add graywolf.
686 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
687 * config/nds32/pipelines.md: Include n15 settings.
689 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
690 Chung-Ju Wu <jasonwucj@gmail.com>
692 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
693 * config/nds32/nds32-n13.md: New file.
694 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
695 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
697 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
698 * config/nds32/nds32.md (pipeline_model): Add n13.
699 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
700 * config/nds32/pipelines.md: Include n13 settings.
702 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
703 Chung-Ju Wu <jasonwucj@gmail.com>
705 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
706 * config/nds32/nds32-n10.md: New file.
707 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
708 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
710 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
711 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
712 * config/nds32/nds32.md (pipeline_model): Add n10.
713 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
714 * config/nds32/pipelines.md: Include n10 settings.
716 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
717 Kito Cheng <kito.cheng@gmail.com>
718 Chung-Ju Wu <jasonwucj@gmail.com>
720 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
721 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
722 Add enum values for DSP extension instructions.
723 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
725 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
726 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
728 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
729 * config/nds32/nds32-dspext.md: New file for DSP implementation.
730 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
731 * config/nds32/nds32-intrinsic.md: Likewise.
732 * config/nds32/nds32_intrinsic.h: Likewise.
733 * config/nds32/nds32-md-auxiliary.c: Likewise.
734 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
735 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
736 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
737 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
738 * config/nds32/nds32-protos.h: New declarations for DSP extension.
739 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
740 TYPE_DMAC in switch statement.
741 * config/nds32/nds32.c: New checking and implementation for DSP
742 extension instructions.
743 * config/nds32/nds32.h: Likewise.
744 * config/nds32/nds32.md: Likewise.
745 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
746 * config/nds32/predicates.md: Implement new predicates for DSP
749 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
751 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
752 Reformat alternatives and attributes so it is easier to identify
753 which constraints/attributes go with which instruction.
754 (mov<mode>_hardfloat32, FMOVE64): Likewise.
755 (mov<mode>_softfloat32, FMOVE64): Likewise.
756 (mov<mode>_hardfloat64, FMOVE64): Likewise.
757 (mov<mode>_softfloat64, FMOVE64): Likewise.
759 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
761 * doc/extend.texi (PowerPC Built-in Functions): Rename this
763 (Basic PowerPC Built-in Functions): The new name of the
764 subsection previously known as "PowerPC Built-in Functions".
765 (Basic PowerPC Built-in Functions Available on all Configurations):
767 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
768 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
769 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
770 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
772 2018-05-11 Martin Jambor <mjambor@suse.cz>
775 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
778 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
781 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
783 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
785 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
786 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
787 (ix86_handle_option): Handle -mwaitpkg.
788 * config.gcc: New header.
789 * config/i386/cpuid.h (bit_WAITPKG): New bit.
790 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
791 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
793 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
794 OPTION_MASK_ISA_WAITPKG.
795 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
796 (ix86_option_override_internal): Add PTA_WAITPKG.
797 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
798 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
800 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
801 __builtin_ia32_umwait and __builtin_ia32_tpause.
802 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
803 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
804 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
805 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
806 UNSPECV_TPAUSE): New.
807 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
808 * config/i386/i386.opt: Add -mwaitpkg.
809 * config/i386/waitpkgintrin.h: New file.
810 * config/i386/x86intrin.h: New header.
811 * doc/invoke.texi: Add -mwaitpkg.
813 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
816 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
818 (cortex-m0): Use armv6s-m isa.
819 (cortex-m0plus): Likewise.
820 (cortex-m1): Likewise.
821 (cortex-m0.small-multiply): Likewise.
822 (cortex-m0plus.small-multiply): Likewise.
823 (cortex-m1.small-multiply): Likewise.
825 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
826 Jakub Jelinek <jakub@redhat.com>
828 PR tree-optimization/85692
829 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
830 source permute as well.
832 2018-05-11 Martin Liska <mliska@suse.cz>
835 * doc/extend.texi: Document LLVM style format for no_sanitize
838 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
840 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
841 mode_supports_vsx_dform_quad to mode_supports_dq_form.
842 (mode_supports_vsx_dform_quad): Likewise.
843 (mode_supports_vmx_dform): Move these functions to be next to the
844 other mode_supports functions.
845 (mode_supports_dq_form): Likewise.
846 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
847 mode_supports_dq_form.
848 (reg_offset_addressing_ok_p): Likewise.
849 (offsettable_ok_by_alignment): Likewise.
850 (rs6000_legitimate_offset_address_p): Likewise.
851 (legitimate_lo_sum_address_p): Likewise.
852 (rs6000_legitimize_address): Likewise.
853 (rs6000_legitimize_reload_address): Likewise.
854 (rs6000_secondary_reload_inner): Likewise.
855 (rs6000_preferred_reload_class): Likewise.
856 (rs6000_output_move_128bit): Likewise.
858 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
860 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
861 Generate SImode target register for null target.
862 <case IX86_BUILTIN_XGETBV>: Ditto.
863 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
864 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
866 2018-05-10 Carl Love <cel@us.ibm.com>
868 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
869 dcbtt and dcbtstt if operands[2] is 0.
871 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
874 * config/i386/sse.md (usadv64qi): New expander.
876 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
878 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
879 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
880 -maltivec=be support.
881 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
882 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
883 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
884 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
885 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
886 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
887 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
888 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
889 altivec_vsumsws): Adjust.
890 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
891 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
892 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
894 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
895 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
896 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
897 (altivec_lve<VI_char>x): Delete expand.
898 (*altivec_lve<VI_char>x_internal): Rename to...
899 (altivec_lve<VI_char>x): ... this.
900 (altivec_lvxl_<mode>): Delete expand.
901 (*altivec_lvxl_<mode>_internal): Rename to ...
902 (altivec_lvxl_<mode>): ... this.
903 (altivec_stvxl_<mode>): Delete expand.
904 (*altivec_stvxl_<mode>_internal): Rename to ...
905 (altivec_stvxl_<mode>): ... this.
906 (altivec_stve<VI_char>x): Delete expand.
907 (*altivec_stve<VI_char>x_internal): Rename to ...
908 (altivec_stve<VI_char>x): ... this.
909 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
910 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
911 reduc_plus_scal_<mode>): Adjust.
912 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
914 (rs6000_cpu_cpp_builtins): Adjust.
915 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
916 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
917 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
918 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
919 -maltivec=be support.
920 (rs6000_split_vec_extract_var): Adjust.
921 (rs6000_split_v4si_init): Adjust.
922 (swap_selector_for_mode): Delete.
923 (altivec_expand_lvx_be, altivec_expand_stvx_be,
924 altivec_expand_stvex_be): Delete.
925 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
926 -maltivec=be support.
927 (rs6000_gimple_fold_builtin): Ditto.
928 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
930 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
931 (TARGET_DIRECT_MOVE_64BIT): Adjust.
932 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
933 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
934 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
935 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
936 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
937 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
938 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
939 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
940 anonymous split): Adjust.
941 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
942 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
944 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
946 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
947 when --with-gxx-include-dir is also specified.
948 * configure: Regenerate.
950 2018-05-09 Jim Wilson <jimw@sifive.com>
953 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
954 * config/riscv/t-withmultilib: New.
955 * config/riscv/withmultilib.h: New.
956 * doc/install.texi: Document RISC-V --with-multilib-list support.
958 2018-05-09 Richard Biener <rguenther@suse.de>
960 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
962 (vect_bb_vectorization_profitable_p): Adjust. Compute
963 actual scalar cost using the cost vector and the add_stmt_cost
966 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
968 PR rtl-optimization/85645
969 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
970 in the REG_CFA_REGISTER note for LR, don't leave it empty.
972 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
974 PR rtl-optimization/85645
975 * shrink-wrap.c (spread_components): Return a boolean saying if
976 anything was changed.
977 (try_shrink_wrapping_separate): Iterate spread_components until
978 nothing changes anymore.
980 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
982 PR rtl-optimization/85645
983 * regrename.c (build_def_use): Also kill the chains that include the
984 destination of a REG_CFA_REGISTER note.
986 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
988 PR rtl-optimization/85645
989 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
990 insn that has a REG_CFA_REGISTER note.
992 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
994 * cfgexpand.c (expand_clobber): New function.
995 (expand_gimple_stmt_1): Use it.
996 * tree-vect-stmts.c (vect_clobber_variable): New function,
998 (vectorizable_simd_clone_call): ...here.
999 (vectorizable_store): Emit a clobber either side of an
1000 IFN_STORE_LANES sequence.
1001 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
1003 2018-05-09 Tom de Vries <tom@codesourcery.com>
1006 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
1007 (define_insn "trap_if_false"): Add exit after trap.
1009 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
1011 PR rtl-optimization/85638
1012 * bb-reorder.c: Include common/common-target.h.
1013 (create_forwarder_block): New function extracted from...
1014 (fix_up_crossing_landing_pad): ...here. Rename into...
1015 (dw2_fix_up_crossing_landing_pad): ...this.
1016 (sjlj_fix_up_crossing_landing_pad): New function.
1017 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
1018 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
1019 from both partitions and exit the loop after one iteration.
1021 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1024 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1026 (Basic PowerPC Built-in Functions): The new name of the
1027 subsection previously known as "PowerPC Built-in Functions".
1028 (Basic PowerPC Built-in Functions Available on all Configurations):
1030 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1032 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1034 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1036 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1039 2018-05-08 Jim Wilson <jimw@sifive.com>
1041 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
1042 (LD_EMUL_SUFFIX): New.
1043 (LINK_SPEC): Use it.
1045 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1047 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1049 (Basic PowerPC Built-in Functions): The new name of the
1050 subsection previously known as "PowerPC Built-in Functions".
1051 (Basic PowerPC Built-in Functions Available on all Configurations):
1053 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1055 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1057 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1059 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1062 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1065 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
1066 after cmpelim optimization.
1068 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
1070 * config.gcc: Support "goldmont".
1071 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
1072 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1074 * config/i386/i386.c (m_GOLDMONT): Define.
1075 (processor_target_table): Add "goldmont".
1076 (PTA_GOLDMONT): Define.
1077 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
1078 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
1079 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
1080 (fold_builtin_cpu): Add "goldmont".
1081 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
1082 (ix86_option_override_internal): Add "goldmont".
1083 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
1084 (processor_type): Add PROCESSOR_GOLDMONT.
1085 * config/i386/i386.md: Add CPU "glm".
1086 * config/i386/glm.md: New file.
1087 * config/i386/x86-tune.def: Add m_GOLDMONT.
1088 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
1090 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1093 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
1095 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
1096 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
1097 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
1100 * config/i386/i386.c (ix86_fold_builtin): Handle
1101 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
1104 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
1105 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
1107 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
1110 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
1111 (check_arch): Likewise.
1112 (check_fpu): Return the result rather than printing it.
1113 (end arch): Fix operator precedence.
1114 (end cpu): Likewise.
1115 (END): Print the result from check_fpu.
1117 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1118 Alan Hayward <alan.hayward@arm.com>
1119 David Sherwood <david.sherwood@arm.com>
1121 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
1122 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
1123 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
1124 (*fcmuo<mode>_and): New patterns.
1126 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1128 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
1129 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
1130 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
1131 (cmp_op, sve_imm_con): New code attributes.
1132 (SVE_COND_INT_CMP, imm_con): Delete.
1133 (cmp_op): Remove above unspecs from int attribute.
1134 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
1136 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
1137 comparison-specific unspecs.
1138 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
1139 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
1140 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
1141 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
1142 (*vec_fcm<cmp_op><mode>): Rename to...
1143 (*fcm<cmp_op><mode>): ...this and adjust likewise.
1144 (*vec_fcmuo<mode>): Rename to...
1145 (*fcmuo<mode>): ...this and adjust likewise.
1146 (*pred_fcm<cmp_op><mode>): New pattern.
1147 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
1148 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
1150 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
1152 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
1153 (aarch64_emit_sve_predicated_cond): New function.
1154 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
1155 (aarch64_emit_unspec_cond_or): Replace with...
1156 (aarch64_emit_sve_or_conds): ...this new function. Use
1157 aarch64_emit_sve_ptrue_op for the individual comparisons and
1158 aarch64_emit_binop to OR them together.
1159 (aarch64_emit_inverted_unspec_cond): Replace with...
1160 (aarch64_emit_sve_inverted_cond): ...this new function. Use
1161 aarch64_emit_sve_ptrue_op for the comparison and
1162 aarch64_emit_unop to invert the result.
1163 (aarch64_expand_sve_vec_cmp_float): Update after the above
1164 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
1166 2018-05-07 Nathan Sidwell <nathan@acm.org>
1168 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
1169 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
1170 (Backwards Compatibility): Likewise.
1172 2018-05-07 Luis Machado <luis.machado@linaro.org>
1176 2018-05-07 Luis Machado <luis.machado@linaro.org>
1178 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1179 <prefetch_dynamic_strides>: New const bool field.
1180 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1181 prefetch_dynamic_strides.
1182 (exynosm1_prefetch_tune): Likewise.
1183 (thunderxt88_prefetch_tune): Likewise.
1184 (thunderx_prefetch_tune): Likewise.
1185 (thunderx2t99_prefetch_tune): Likewise.
1186 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1187 (aarch64_override_options_internal): Update to set
1188 PARAM_PREFETCH_DYNAMIC_STRIDES.
1189 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1190 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1191 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1192 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1193 prefetch-dynamic-strides setting.
1195 2018-05-07 Luis Machado <luis.machado@linaro.org>
1197 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1198 <minimum_stride>: New const int field.
1199 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1200 minimum_stride field.
1201 (exynosm1_prefetch_tune): Likewise.
1202 (thunderxt88_prefetch_tune): Likewise.
1203 (thunderx_prefetch_tune): Likewise.
1204 (thunderx2t99_prefetch_tune): Likewise.
1205 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1206 (aarch64_override_options_internal): Update to set
1207 PARAM_PREFETCH_MINIMUM_STRIDE.
1208 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1209 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1210 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1211 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1212 stride is constant and is below the minimum stride threshold.
1214 2018-05-07 Luis Machado <luis.machado@linaro.org>
1216 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
1219 2018-05-07 Luis Machado <luis.machado@linaro.org>
1221 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1222 <prefetch_dynamic_strides>: New const bool field.
1223 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1224 prefetch_dynamic_strides.
1225 (exynosm1_prefetch_tune): Likewise.
1226 (thunderxt88_prefetch_tune): Likewise.
1227 (thunderx_prefetch_tune): Likewise.
1228 (thunderx2t99_prefetch_tune): Likewise.
1229 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1230 (aarch64_override_options_internal): Update to set
1231 PARAM_PREFETCH_DYNAMIC_STRIDES.
1232 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1233 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1234 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1235 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1236 prefetch-dynamic-strides setting.
1238 2018-05-07 Luis Machado <luis.machado@linaro.org>
1240 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1241 <minimum_stride>: New const int field.
1242 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1243 minimum_stride field.
1244 (exynosm1_prefetch_tune): Likewise.
1245 (thunderxt88_prefetch_tune): Likewise.
1246 (thunderx_prefetch_tune): Likewise.
1247 (thunderx2t99_prefetch_tune): Likewise.
1248 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1249 (aarch64_override_options_internal): Update to set
1250 PARAM_PREFETCH_MINIMUM_STRIDE.
1251 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1252 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1253 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1254 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1255 stride is constant and is below the minimum stride threshold.
1257 2018-05-06 Jakub Jelinek <jakub@redhat.com>
1260 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
1261 the type is addressable. Don't force op into register if it has
1264 2018-05-05 Roland McGrath <mcgrathr@google.com>
1267 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
1268 any section for which we don't know a specific type it should have,
1269 regardless of name. Previously this was done only for the exact
1270 names ".init_array", ".fini_array", and ".preinit_array".
1271 (default_elf_asm_named_section): Add comment about
1272 relationship with default_section_type_flags and SECTION_NOTYPE.
1273 (get_section): Don't consider it a type conflict if one side has
1274 SECTION_NOTYPE and the other doesn't, as long as neither has the
1275 SECTION_BSS et al used in the default_section_type_flags logic.
1277 2018-05-05 Tom de Vries <tom@codesourcery.com>
1280 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
1281 (workaround_barsyncs): New function.
1282 (nvptx_reorg): Use workaround_barsyncs.
1283 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1284 (define_expand "nvptx_membar_cta"): New define_expand.
1285 (define_insn "*nvptx_membar_cta"): New insn.
1287 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
1289 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
1290 To improve optimization opportunities.
1291 * builtin-types.def: The new needed builtin types for the above.
1293 2018-05-04 Richard Biener <rguenther@suse.de>
1295 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
1296 * gimple-ssa-store-merging.c
1297 (imm_store_chain_info::output_merged_store): Remove redundant create,
1298 release split_store vector contents on failure.
1299 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
1300 scalar stmt vector on cache hit.
1302 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
1304 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
1306 * config.gcc (powerpc-xilinx-eabi*): Remove.
1307 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
1309 (fusion_addis_mem_combo_load): Ditto.
1310 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
1312 (rs6000_cpu_cpp_builtins): Ditto.
1313 * config/rs6000/rs6000-linux.c
1314 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
1315 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
1316 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
1318 (rs6000_setup_reg_addr_masks): Ditto.
1319 (rs6000_init_hard_regno_mode_ok): Ditto.
1320 (rs6000_option_override_internal): Ditto.
1321 (legitimate_lo_sum_address_p): Ditto.
1322 (rs6000_legitimize_address): Ditto.
1323 (rs6000_legitimize_reload_address): Ditto.
1324 (rs6000_legitimate_address_p): Ditto.
1325 (abi_v4_pass_in_fpr): Ditto.
1326 (setup_incoming_varargs): Ditto.
1327 (rs6000_gimplify_va_arg): Ditto.
1328 (rs6000_split_multireg_move): Ditto.
1329 (rs6000_savres_strategy): Ditto.
1330 (rs6000_emit_prologue_components): Ditto.
1331 (rs6000_emit_epilogue_components): Ditto.
1332 (rs6000_emit_prologue): Ditto.
1333 (rs6000_emit_epilogue): Ditto.
1334 (rs6000_elf_file_end): Ditto.
1335 (rs6000_function_value): Ditto.
1336 (rs6000_libcall_value): Ditto.
1337 * config/rs6000/rs6000.h: Ditto.
1338 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
1339 (TARGET_MINMAX): ... this. New.
1340 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
1341 * config/rs6000/rs6000.md: Remove Xilinx FP support.
1342 (*movsi_internal1_single): Delete.
1343 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
1344 mfpu=, mxilinx-fpu): Delete.
1345 * config/rs6000/singlefp.h: Delete.
1346 * config/rs6000/sysv4.h: Remove Xilinx FP support.
1347 * config/rs6000/t-rs6000: Ditto.
1348 * config/rs6000/t-xilinx: Delete.
1349 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
1350 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
1351 (VStype_simple): Delete.
1352 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
1353 * config/rs6000/xfpu.h: Delete.
1354 * config/rs6000/xfpu.md: Delete.
1355 * config/rs6000/xilinx.h: Delete.
1356 * config/rs6000/xilinx.opt: Delete.
1357 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
1358 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
1360 2018-05-04 Tom de Vries <tom@codesourcery.com>
1363 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
1366 2018-05-04 Richard Biener <rguenther@suse.de>
1369 * tree-complex.c (update_complex_assignment): We are always in SSA form.
1370 (expand_complex_div_wide): Likewise.
1371 (expand_complex_operations_1): Likewise.
1372 (expand_complex_libcall): Preserve EH info of the original stmt.
1373 (tree_lower_complex): Handle removed blocks.
1374 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
1375 on complex multiplication and division libcall builtins.
1377 2018-05-04 Richard Biener <rguenther@suse.de>
1380 * fold-const.c (negate_expr_p): Restrict negation of operand
1381 zero of a division to when we know that can happen without
1383 (fold_negate_expr_1): Likewise.
1385 2018-05-04 Jakub Jelinek <jakub@redhat.com>
1388 * real.h (real_nextafter): Declare.
1389 * real.c (real_nextafter): New function.
1390 * fold-const-call.c (fold_const_nextafter): New function.
1391 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
1392 CASE_CFN_NEXTTOWARD.
1393 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
1394 even when arg1_mode is different from arg0_mode.
1396 2018-05-03 Nathan Sidwell <nathan@acm.org>
1398 * doc/extend.texi (Deprecated Features): Remove
1400 (Backwards Compatibility): Likewise.
1401 * doc/invoke.texi (C++ Language Options): Likewise.
1402 (C++ Dialect Options): Likewise.
1404 2018-05-03 Jakub Jelinek <jakub@redhat.com>
1407 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1408 _mm512_mask_mullox_epi64): New intrinsics.
1410 2018-05-03 Tom de Vries <tom@codesourcery.com>
1413 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1414 dump files): Add offload-tree.
1416 2018-05-03 Richard Biener <rguenther@suse.de>
1418 PR tree-optimization/85615
1419 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1420 to loops not nested in BBs loop father to avoid creating multi-entry
1423 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1425 PR tree-optimization/70291
1426 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1427 arguments. Change return type to tree. Emit libcall as a new
1428 statement rather than replacing existing one when inplace_p is true.
1429 (expand_complex_multiplication_components): New function.
1430 (expand_complex_multiplication): Expand floating-point complex
1431 multiplication using the above.
1432 (expand_complex_division): Rename inner_type parameter to type.
1433 Update expand_complex_libcall call-site.
1434 (expand_complex_operations_1): Update expand_complex_multiplication
1435 and expand_complex_division call-sites.
1437 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1440 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1441 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1442 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1443 the highest significant bit of the shift count mask is clear. In
1444 check whether and[sq]i3 is needed verify that all significant bits
1445 of the shift count other than the highest are set.
1447 2018-05-02 Tom de Vries <tom@codesourcery.com>
1450 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1451 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1452 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1453 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1454 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1455 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1456 __builtin_goacc_parlevel_size.
1458 2018-05-02 Richard Biener <rguenther@suse.de>
1460 PR tree-optimization/85597
1461 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1462 do not use split vect_get_vec_defs call but call vect_get_slp_defs
1465 2018-05-02 Tom de Vries <tom@codesourcery.com>
1468 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1469 dump files): Add ltrans-tree.
1471 2018-05-02 Tom de Vries <tom@codesourcery.com>
1474 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1475 dump files): Add wpa-ipa.
1477 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
1479 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
1480 powerpc*-*-linux*paired* target.
1481 * config/rs6000/750cl.h: Delete.
1482 * config/rs6000/paired.h: Delete.
1483 * config/rs6000/paired.md: Delete.
1484 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1486 * config/rs6000/rs6000-builtin.def: Remove paired float support.
1487 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1488 comment. Remove paired float support.
1489 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1490 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1492 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1493 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1495 * config/rs6000/rs6000.c: Remove paired float support.
1496 (paired_expand_vector_init, paired_expand_vector_move,
1497 paired_emit_vector_compare, paired_emit_vector_cond_expr,
1498 (paired_expand_lv_builtin, paired_expand_stv_builtin,
1499 paired_expand_builtin, paired_expand_predicate_builtin,
1500 paired_init_builtins): Delete.
1501 * config/rs6000/rs6000.h: Remove paired float support.
1502 * config/rs6000/rs6000.md: Remove paired float support.
1503 (move_from_CR_ov_bit): Delete.
1504 * config/rs6000/rs6000.opt (mpaired): Delete.
1505 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1506 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1508 2018-05-02 Richard Biener <rguenther@suse.de>
1511 * gimplify.c (gimplify_save_expr): When in SSA form allow
1512 SAVE_EXPRs to compute to SSA vars.
1514 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1517 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1518 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1519 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1520 clobber operands[2], instead use a new pseudo. Formatting fixes.
1522 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1524 PR tree-optimization/85586
1525 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1526 exit early for statements in the same group if the accesses are
1529 2018-05-02 Tom de Vries <tom@codesourcery.com>
1532 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1535 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
1537 PR tree-optimization/85143
1538 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1540 2018-05-01 Tom de Vries <tom@codesourcery.com>
1543 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1544 not found" error message.
1546 2018-05-01 Tom de Vries <tom@codesourcery.com>
1549 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1550 * vec.c (test_ordered_remove_if): New function.
1551 (vec_c_tests): Call test_ordered_remove_if.
1552 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1553 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1554 * tree-vect-patterns.c (vect_pattern_recog_1): Use
1555 VEC_ORDERED_REMOVE_IF.
1557 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1559 PR tree-optimization/82665
1560 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1561 pointer subtraction where arguments come from a memchr call.
1563 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1565 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1566 --push-state --as-needed and --pop-state instead of --as-needed and
1567 --no-as-needed if ld supports it.
1568 * configure: Regenerated.
1571 * doc/install.texi2html: Replace _002d with - and _002a with * in
1572 generated html files using sed.
1574 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1577 * gcc-rich-location.c (blank_line_before_p): New function.
1578 (use_new_line): New function.
1579 (gcc_rich_location::add_fixit_insert_formatted): New function.
1580 * gcc-rich-location.h
1581 (gcc_rich_location::add_fixit_insert_formatted): New function.
1583 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1585 * selftest.c (assert_streq): Rename "expected" and "actual" to
1586 "val1" and "val2". Extend NULL-handling to cover both inputs
1587 symmetrically, while still requiring both to be non-NULL for a pass.
1588 * selftest.h (assert_streq): Rename "expected" and "actual" to
1590 (ASSERT_EQ): Likewise.
1591 (ASSERT_EQ_AT): Likewise.
1592 (ASSERT_KNOWN_EQ): Likewise.
1593 (ASSERT_KNOWN_EQ_AT): Likewise.
1594 (ASSERT_NE): Likewise.
1595 (ASSERT_MAYBE_NE): Likewise.
1596 (ASSERT_MAYBE_NE_AT): Likewise.
1597 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1598 the assertion to pass.
1599 (ASSERT_STREQ_AT): Likewise.
1601 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1603 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1604 interaction with -pie.
1606 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1608 * selftest.h: Fix alphabetization of per-source-file selftest
1611 2018-04-30 Jason Merrill <jason@redhat.com>
1613 PR c++/61982 - dead stores to destroyed objects.
1614 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1617 2018-04-30 Jason Merrill <jason@redhat.com>
1619 * tree.c (build_clobber): New.
1620 * tree.h: Declare it.
1621 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1623 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1625 * diagnostic-show-locus.c (layout::layout): Update for
1626 location_get_source_line returning a char_span.
1627 (struct char_span): Move to input.h.
1628 (struct correction): Update for fields in char_span becoming
1630 (struct source_line): Update for location_get_source_line
1631 returning a char_span.
1632 (layout::print_line): Likewise.
1633 * edit-context.c (edited_file::print_content): Likewise.
1634 (edited_file::print_diff_hunk): Likewise.
1635 (edited_file::print_run_of_changed_lines): Likewise.
1636 (edited_file::get_num_lines): Likewise.
1637 (edited_line::edited_line): Likewise.
1638 * final.c (asm_show_source): Likewise.
1639 * input.c (location_get_source_line): Convert return type
1640 from const char * to char_span, losing the final "line_len"
1642 (dump_location_info): Update for the above.
1643 (get_substring_ranges_for_loc): Likewise. Use a char_span
1644 when handling the literal within the line.
1645 (test_reading_source_line): Update for location_get_source_line
1646 returning a char_span.
1647 * input.h (class char_span): Move here from
1648 diagnostic-show-locus.c, converting from a struct to a class.
1649 Make data members private.
1650 (char_span::operator bool): New.
1651 (char_span::length): New.
1652 (char_span::get_buffer): New.
1653 (char_span::operator[]): New.
1654 (char_span::subspan): Make const.
1655 (char_span::xstrdup): New.
1656 (location_get_source_line): Convert return type from const char *
1657 to char_span, losing the final "line_size" param.
1659 2018-04-30 Jan Hubicka <jh@suse.cz>
1661 * lto-wrapper.c (ltrans_priorities): New static var.
1662 (cmp_priority): New.
1663 (run_gcc): Read priorities and if doing parallel build order
1664 the Makefile by them.
1666 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1668 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
1670 2018-04-30 Richard Biener <rguenther@suse.de>
1672 * tree-cfg.c (verify_address): Remove base argument, add
1673 flag whether to check TREE_ADDRESSABLE and do that.
1674 (verify_expr): Remove.
1675 (verify_types_in_gimple_reference): Add pieces from verify_expr.
1676 (verify_gimple_assign_single): Likewise.
1677 (verify_gimple_switch): Likewise.
1678 (verify_expr_location_1): Dereference tp once. Add (disabled)
1679 piece from verify_expr.
1680 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1682 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1684 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1686 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1688 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1689 (small_data_pattern): Likewise.
1690 (arc_rewrite_small_data): Likewise.
1691 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1692 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1693 (get_symbol_alignment): New function.
1694 (legitimate_small_data_address_p): Likewise.
1695 (legitimate_scaled_address): Update, call
1696 legitimate_small_data_address_p.
1697 (output_sdata): New static variable.
1698 (arc_print_operand): Update how we handle small data operands.
1699 (arc_print_operand_address): Likewise.
1700 (arc_legitimate_address_p): Update, use
1701 legitimate_small_data_address_p.
1702 (arc_rewrite_small_data_p): Remove.
1703 (arc_rewrite_small_data_1): Likewise.
1704 (arc_rewrite_small_data): Likewise.
1705 (small_data_pattern): Likewise.
1706 (compact_sda_memory_operand): Update to use
1707 legitimate_small_data_address_p and get_symbol_alignment.
1708 (prepare_move_operands): Don't rewite sdata pattern.
1709 (prepare_extend_operands): Remove.
1710 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
1712 (zero_extendqisi2): Likewise.
1713 (zero_extendhisi2): Likewise.
1714 (extendqihi2): Likewise.
1715 (extendqisi2): Likewise.
1716 (extendhisi2): Likewise.
1720 * config/arc/constraints.md (Usd): Change it to memory constraint.
1722 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1724 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
1725 as source of std instructions.
1726 * config/arc/arc.md (movsi_insn): Update pattern predicate to
1727 allow 6-bit constants as source for store instructions.
1728 (movdi_insn): Update instruction pattern to allow 6-bit constants
1729 as source for store instructions.
1731 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1733 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
1735 2018-04-30 Nathan Sidwell <nathan@acm.org>
1736 Sandra Loosemore <sandra@codesourcery.com>
1738 * dumpfile.c (dump_open): Allow '-' for stdout.
1739 * doc/invoke.texi (Developer Options): Document dump filename
1740 determination early. Document stdin/stdout selection.
1742 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
1744 Microblaze Target: PIC data text relative
1746 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
1747 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
1749 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
1750 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
1751 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
1752 New addressing mode for data-text relative position indepenedent code.
1753 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
1754 'ADDRESS_SYMBOLIC_TXT_REL'.
1755 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
1756 (microblaze_legitimate_pic_operand): Exclude function calls from
1757 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
1758 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
1760 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1761 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1762 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
1763 for 'address + offset'.
1764 (microblaze_expand_prologue): Add new function prologue call for
1766 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
1767 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
1768 table in case of TARGET_PIC_DATA_TEXT_REL.
1769 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
1770 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
1771 Add new macros 'UNSPEC_TEXT',
1772 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
1773 + exclude function calls from 'UNSPEC_PLT' in case of data text
1775 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1776 new target hook for generating address diff vector tables in case of
1778 * doc/tm.texi : Regenerate.
1779 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
1780 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
1781 of addr diff vector generation.
1782 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1783 target hook definition.
1784 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
1785 Add default function for generate_pic_addr_diff_vec -> flag_pic.
1786 * doc/invoke.texi (Add new pic option): Add new microblaze pic
1787 option for data text relative.
1789 2018-04-30 Richard Biener <rguenther@suse.de>
1791 * tree-chrec.h (evolution_function_is_constant_p): Remove
1793 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
1795 2018-04-30 Richard Biener <rguenther@suse.de>
1798 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
1800 2018-04-30 Richard Biener <rguenther@suse.de>
1802 PR tree-optimization/28364
1803 PR tree-optimization/85275
1804 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
1805 copying first exit test.
1807 2018-04-28 Mark Wielaard <mark@klomp.org>
1809 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
1811 (dwarf_AT): Handle DW_AT_addr_base.
1812 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
1814 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
1817 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
1818 (*ashl<dwi>3_doubleword_mask_1): Ditto.
1819 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
1820 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
1822 2018-04-28 Richard Biener <rguenther@suse.de>
1824 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
1825 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
1826 to reflect use. Only add interesting stmts.
1828 2018-04-27 Martin Jambor <mjambor@suse.cz>
1831 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
1832 the jump function allows for passing through aggregate values.
1834 2018-04-27 David Malcolm <dmalcolm@redhat.com>
1836 * input.h (in_system_header_at): Convert from macro to inline
1838 (from_macro_expansion_at): Likewise.
1839 (from_macro_definition_at): Likewise.
1841 2018-04-27 Jeff Law <law@redhat.com>
1843 * config.gcc: Mark tile* targets as deprecated/obsolete.
1845 2018-04-27 Richard Biener <rguenther@suse.de>
1847 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
1850 2018-04-27 Richard Biener <rguenther@suse.de>
1852 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
1854 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
1856 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
1857 with Yd constraint. Set "preferred_for_speed" attribute from
1858 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
1860 (*movdi_internal): Ditto.
1861 (movti_interunit splitters): Remove
1862 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
1863 (movdi_interunit splitters): Ditto.
1864 * config/i386/constraints.md (Ye): Remove.
1865 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
1867 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1870 * config/aarch64/constraints.md (Usg): Limit to 31.
1873 2018-04-27 Jakub Jelinek <jakub@redhat.com>
1875 PR tree-optimization/85529
1876 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
1877 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
1878 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
1879 zero extension or masking of the MSB bit.
1880 (optimize_range_tests): Add FIRST_BB argument, pass it through
1881 to optimize_range_tests_var_bound.
1882 (maybe_optimize_range_tests, reassociate_bb): Adjust
1883 optimize_range_tests callers.
1885 2018-04-26 Richard Biener <rguenther@suse.de>
1886 Jakub Jelinek <jakub@redhat.com>
1888 * cgraph.h (symbol_table): Just declare debug method here.
1889 * symtab.c (symbol_table::debug): Define.
1891 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
1893 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
1895 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
1897 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
1898 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
1899 (*movdi_internal): Substitute Yi and Yj constraint with x
1900 and Ym and Yn constraint with y constraint. Update "isa"
1901 attribute and set "preferred_for_speed" attribute from
1902 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
1903 (*movsi_internal): Ditto.
1904 (*movdf_internal): Ditto.
1905 (*movsf_internal): Ditto.
1906 (*zero_extendsidi2): Ditto.
1907 * config/i386/sse.md (vec_set<mode>_0): Ditto.
1908 (sse2_loadld): Ditto.
1909 (*vec_extract<ssevecmodelower>_0): Ditto.
1910 (*vec_extractv4si_0_zext_sse4): Ditto.
1911 (vec_concatv2di): Ditto.
1912 (*vec_dup<mode>): Ditto.
1913 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
1914 * config/i386/constraints.md (Yi): Remove.
1919 2018-04-26 Nathan Sidwell <nathan@acm.org>
1921 * dumpfile.c (dump_open): New.
1922 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
1923 (dump_finish): Detect stdio/stderr by value not name.
1925 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
1927 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
1929 2018-04-26 Tom de Vries <tom@codesourcery.com>
1932 * config/nvptx/nvptx.c (verify_neutering_jumps)
1933 (verify_neutering_labels): New function
1934 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
1936 2018-04-26 Tom de Vries <tom@codesourcery.com>
1939 * config/nvptx/nvptx.c (needs_neutering_p): New function.
1940 (nvptx_single): Use needs_neutering_p to skip over insns that do not
1943 2018-04-26 Richard Biener <rguenther@suse.de>
1944 Tom de Vries <tom@codesourcery.com>
1947 * lto-streamer-out.c (output_function): Fixup loops if required to match
1948 discovery done in the reader.
1950 2018-04-26 Richard Biener <rguenther@suse.de>
1952 PR tree-optimization/85116
1953 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
1954 have a loop exit from the single latch predecessor. Remove
1955 case of header with just condition.
1956 (ch_base::copy_headers): Exclude infinite loops from any
1958 (pass_ch::execute): Record exits.
1960 2018-04-26 Richard Biener <rguenther@suse.de>
1962 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
1963 prologue cost vector and pass it to vect_get_load_cost.
1964 (vect_get_peeling_costs_all_drs): Likewise.
1965 (vect_peeling_hash_get_lowest_cost): Likewise.
1966 (vect_enhance_data_refs_alignment): Likewise.
1968 2018-04-26 Richard Biener <rguenther@suse.de>
1971 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
1972 checking of integer<->pointer conversions.
1973 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
1974 sign-/zero-extending pointer types.
1975 (expand_omp_for_static_chunk): Likewise.
1977 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
1978 Jean Lee <xiaoyur347@gmail.com>
1980 * config/mips/mips.c (mips_asan_shadow_offset): New function.
1981 (TARGET_ASAN_SHADOW_OFFSET): Define.
1982 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
1983 true for -fsanitize=address.
1985 2018-04-25 Mark Wielaard <mark@klomp.org>
1987 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
1990 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1992 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
1993 than "alu", remove explicit "memory" and "imm_disp" attributes.
1994 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
1997 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
1998 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
1999 gen_lowpart_no_emit.
2001 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
2004 * config/i386/i386.c (ix86_expand_builtin): Change memory
2005 operand to XI, extend p0 to Pmode.
2006 * config/i386/i386.md: Change unspec volatile and operand
2007 1 mode to XI, change operand 0 mode to P.
2009 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2011 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
2012 GET_MODE_MASK before any checking.
2013 (nds32_can_use_bset_p): Likewise.
2014 (nds32_can_use_btgl_p): Likewise.
2016 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2018 * config/nds32/nds32-doubleword.md: New define_split pattern for
2019 illegal register number.
2021 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2023 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
2025 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2027 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
2029 2018-04-25 Richard Biener <rguenther@suse.de>
2031 * lto-streamer.h (LTO_major_version): Bump to 8.
2033 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2035 * BASE-VER: Set to 9.0.0.
2037 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
2039 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
2040 in __abskf2 and __powikf2.
2042 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2045 * config/aarch64/constraints.md (Usg, Usj): New constraints.
2046 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
2047 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
2048 Use the above on operand 2. Reindent.
2049 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
2051 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2054 * common/config/i386/i386-common.c (ix86_handle_option): Don't
2056 * config/i386/i386.opt (mcet): Removed.
2057 * doc/install.texi: Remove -mcet documentation.
2058 * doc/invoke.texi: Likewise.
2060 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2063 * doc/install.texi: Remove -mcet from bootstrap-cet.
2065 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2068 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
2069 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
2073 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
2074 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
2075 containing a CONST_VECTOR.
2077 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
2079 * doc/install.texi: Update newlib dependency for nvptx.
2081 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2084 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
2085 instead of INTVAL when shifting x left.
2087 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
2089 PR tree-optimization/85478
2090 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
2091 vect_grouped_store_supported for single element vectors.
2093 2018-04-24 Richard Biener <rguenther@suse.de>
2096 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
2097 load cost increase to the case of non-constant step.
2099 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2102 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
2103 destination if any_malformed_asm.
2105 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
2108 * expr.c (store_field): In the bitfield case, if the value comes from
2109 a function call and is returned in registers by means of a PARALLEL,
2110 do not change the mode of the temporary unless BLKmode and VOIDmode.
2112 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
2114 PR rtl-optimization/85423
2115 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
2116 dependencies to debug insns when the previous insn is non-debug.
2118 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2120 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
2121 enums into a single definition.
2122 (fls): Fix predicates and printing.
2125 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2127 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
2128 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
2129 and short u6 immediate.
2130 (check_if_valid_sleep_operand): Remove.
2131 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
2133 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2135 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
2136 flag_always_save_lp condition.
2137 * config/nds32/nds32.opt (malways-save-lp): New option.
2139 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2141 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
2142 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
2143 * config/nds32/nds32.h
2144 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
2145 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
2147 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2149 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
2150 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
2152 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2153 Chung-Ju Wu <jasonwucj@gmail.com>
2155 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
2157 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
2158 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
2160 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2162 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
2164 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2166 * config/nds32/nds32-protos.h (nds32_data_alignment,
2167 nds32_local_alignment): Declare.
2168 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
2169 nds32_local_alignment): New functions.
2170 (TARGET_CONSTANT_ALIGNMENT): Define.
2171 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
2173 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2175 * config/nds32/nds32.c
2176 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
2177 (TARGET_MODES_TIEABLE_P): Likewise.
2179 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2181 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
2184 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
2185 Chung-Ju Wu <jasonwucj@gmail.com>
2187 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2188 for unaligned access.
2189 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2191 * config/nds32/nds32-intrinsic.md: Likewise.
2192 * config/nds32/nds32_intrinsic.h: Likewise.
2193 * config/nds32/nds32.h (nds32_builtins): Likewise.
2194 * config/nds32/nds32.opt (munaligned-access): New option.
2195 * config/nds32/nds32.c (nds32_asm_file_start): Display
2196 flag_unaligned_access status.
2198 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2200 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
2201 -mno-relax is present.
2202 * config/riscv/linux.h (LINK_SPEC): Ditto.
2204 2018-04-20 Martin Sebor <msebor@redhat.com>
2207 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
2209 (gimple_fold_builtin_stxcpy_chk): Same.
2210 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
2212 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
2215 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
2216 __powikf2 when long double is IEEE 128-bit.
2218 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2220 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
2221 step to make sure stack always aligned.
2223 2018-04-20 Carl Love <cel@us.ibm.com>
2226 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
2227 size check for arg0.
2229 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
2230 Tom de Vries <tom@codesourcery.com>
2233 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
2234 Emit insns for calls too.
2235 (nvptx_find_par): Always look for worker-level predecessor insn.
2236 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
2238 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
2239 (nvptx_process_pars): Propagate frames for calls.
2241 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
2244 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
2246 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
2247 (ix86_handle_option): Don't handle OPT_mibt.
2248 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
2250 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
2252 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
2253 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
2254 (ix86_target_macros): Define __CET__ with flag_cf_protection
2255 for -fcf-protection.
2256 * config/i386/i386.c (isa2_opts): Remove -mibt.
2257 * config/i386/i386.h (TARGET_IBT): Removed.
2258 (TARGET_IBT_P): Likewise.
2259 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
2260 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
2261 * config/i386/i386.opt (mcet): Update help message.
2264 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
2265 -mcet as an alias for -mshstk.
2267 2018-04-20 Richard Biener <rguenther@suse.de>
2270 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
2271 complexity by forcing a single use of the multiply operand.
2273 2018-04-20 Martin Jambor <mjambor@suse.cz>
2276 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
2277 recursion dependency to only apply to non-clones.
2279 2018-04-20 Martin Jambor <mjambor@suse.cz>
2282 * ipa-cp.c (create_specialized_node): Check that clones of
2283 self-recursive edges exist during IPA-CP.
2285 2018-04-19 Toon Moene <toon@moene.org>
2287 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
2290 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2292 PR tree-optimization/85467
2293 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
2294 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
2295 VECTOR_CST element to type.
2297 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2300 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
2301 * config/i386/i386.md (builtin_setjmp_setup): Removed.
2302 (builtin_longjmp): Likewise.
2303 (save_stack_nonlocal): New pattern.
2304 (restore_stack_nonlocal): Likewise.
2306 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2309 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2310 Replace ASM_OUTPUT_LABEL with fprintf.
2312 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2315 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2316 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
2317 * config/i386/i386-c.c (ix86_target_macros_internal): Also
2318 define __IBT__ and __SHSTK__ for -fcf-protection.
2319 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
2321 (ix86_trampoline_init): Likewise.
2322 (x86_output_mi_thunk): Likewise.
2323 (ix86_notrack_prefixed_insn_p): Likewise.
2324 (ix86_option_override_internal): Don't disallow -fcf-protection.
2325 * config/i386/i386.md (rdssp<mode>): Also enable for
2327 (incssp<mode>): Likewise.
2328 (nop_endbr): Likewise.
2329 * config/i386/i386.opt (mcet): Change help message to built-in
2333 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
2334 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
2335 enable CET built-in functions.
2337 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
2339 * common/config/i386/i386-common.c
2340 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
2341 OPTION_MASK_ISA_MOVDIRI_UNSET,
2342 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
2343 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
2344 * config.gcc (movdirintrin.h): New header.
2345 * config/i386/cpuid.h (bit_MOVDIRI,
2346 bit_MOVDIR64B): New bits.
2347 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
2349 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
2350 (VOID, PVOID, PCVOID)): New function types.
2351 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
2352 __builtin_ia32_directstoreu_u64,
2353 __builtin_ia32_movdir64b): New builtins.
2354 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
2355 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
2357 (ix86_valid_target_attribute_inner_p): Ditto.
2358 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
2359 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
2360 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
2361 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
2362 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
2363 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
2364 (movdiri<mode>, movdir64b_<mode>): New.
2365 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
2366 * config/i386/immintrin.h: Include movdirintrin.h.
2367 * config/i386/movdirintrin.h: New file.
2368 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
2370 2018-04-19 Richard Biener <rguenther@suse.de>
2373 * cfg.c (clear_bb_flags): When loop state says we have
2374 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
2376 2018-04-19 Richard Biener <rguenther@suse.de>
2378 PR tree-optimization/84737
2379 * tree-vect-data-refs.c (vect_copy_ref_info): New function
2380 copying restrict info.
2381 (vect_setup_realignment): Use it.
2382 * tree-vectorizer.h (vect_copy_ref_info): Declare.
2383 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
2384 the first DR to all generated stores.
2385 (vectorizable_load): Likewise for loads.
2387 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2389 PR tree-optimization/85446
2390 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
2391 the integral and pointer types to have the same precision.
2393 * doc/install.texi: Document --disable-cet being the default and
2396 2018-04-18 Martin Liska <mliska@suse.cz>
2398 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
2401 2018-04-18 Martin Liska <mliska@suse.cz>
2404 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2407 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2408 arguments if they are comparable.
2410 2018-04-18 Martin Liska <mliska@suse.cz>
2413 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2416 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2419 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
2422 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2423 ENDBR after calling __morestack.
2425 2018-04-18 David Malcolm <dmalcolm@redhat.com>
2428 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2429 by using gcc_base_ver to generate a gcc_driver_version, and use
2430 it when generating GCC_DRIVER_NAME.
2431 * configure: Regenerate.
2433 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2436 * config.gcc: Obsolete powerpc*-*-*spe*.
2438 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2441 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2442 (stabstr_D): Change type of unum from unsigned int to
2443 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
2446 2018-04-17 Jim Wilson <jimw@sifive.com>
2449 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2450 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2451 Set arg_pointer_offset after using pretend_args_size.
2453 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2455 PR rtl-optimization/85431
2456 * dse.c (record_store): Ignore zero width stores.
2459 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
2460 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2461 __builtin_stack_restore rather than after it.
2462 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2463 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2464 argument instead of virtual_dynamic_stack_rtx.
2466 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
2468 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2470 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2471 Add note to error message to explain internal mapping of overloaded
2472 built-in function name to non-overloaded built-in function name.
2473 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2476 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2479 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2480 where the inputs overlap with the output.
2482 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2484 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2485 (=v, v) alternative and explicit "memory" attribute.
2486 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
2487 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2489 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2490 "sselog1" type instead of "sselog".
2491 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2492 "sselog". Remove explicit "memory" attribute.
2493 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2494 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2496 (vec_extract_hi_v32hi): Merge all alternatives into one, use
2497 "sselog1" type instead of "sselog". Remove explicit "memory"
2499 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2500 use "sselog1" type instead of "sselog". Remove explicit "memory"
2502 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2503 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2505 (vec_extract_hi_v64qi): Merge all alternatives into one, use
2506 "sselog1" type instead of "sselog". Remove explicit "memory"
2508 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2509 use "sselog1" type instead of "sselog". Remove explicit "memory"
2513 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2516 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2519 2018-04-17 Martin Jambor <mjambor@suse.cz>
2522 * ipa-cp.c (create_specialized_node): Call
2523 expand_all_artificial_thunks if necessary.
2525 2018-04-17 Martin Liska <mliska@suse.cz>
2528 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2529 in message, remote space in between '_G' and '('.
2531 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2534 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2535 avx512f_vmcmp<mode>3<round_saeonly_name>,
2536 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2537 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2538 avx512f_rndscale<mode><round_saeonly_name>,
2539 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2540 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2541 Use %<iptr>2 instead of %2 for -masm=intel.
2542 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2543 avx512f_vcvttss2usi<round_saeonly_name>,
2544 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2546 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2547 avx512f_vcvttsd2usi<round_saeonly_name>,
2548 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2549 Use %q1 instead of %1 for -masm=intel.
2550 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2551 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2552 of %3 for -masm=intel.
2553 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2555 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2557 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2559 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2560 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2562 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2564 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2565 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2566 %g1 and one with %0 and %1.
2567 (avx512er_vmrcp28<mode><round_saeonly_name>,
2568 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2570 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2571 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2572 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2573 of %0 and %{%4%} for -masm=intel.
2574 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2575 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2576 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2577 order of %0 and %{%5%}%{z%} for -masm=intel.
2579 2018-04-17 Jan Hubicka <jh@suse.cz>
2582 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2584 2018-04-17 Martin Liska <mliska@suse.cz>
2587 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2588 for target_clone error message. Make default implementation
2589 clone to be a local declaration.
2590 (separate_attrs): Add new argument and check for an empty
2592 (expand_target_clones): Handle it.
2593 (ipa_target_clone): Make redirection just for target_clones
2596 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2597 Tom de Vries <tom@codesourcery.com>
2600 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2601 tiled basic blocks without omp continue statements.
2603 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2606 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2607 vec_extract expression as having side effects to make sure it gets
2610 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2613 * config/i386/i386.c (get_builtin_code_for_version): Check
2616 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2619 * gcc/config.gcc: Support "skylake".
2620 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2622 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2623 (processor_target_table): Add "skylake".
2624 (ix86_option_override_internal): Add "skylake".
2625 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2626 PROCESSOR_CANNONLAKE.
2627 (get_builtin_code_for_version): Fix priority for
2628 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2629 PROCESSOR_SKYLAKE-AVX512.
2630 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2631 (processor_type): Add PROCESSOR_SKYLAKE.
2633 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
2634 Jason Merrill <jason@redhat.com>
2637 * convert.c (convert_to_integer_1): Use direct recursion for
2638 enumeral types and types with a precision less than the number
2639 of bits in their mode.
2641 2018-04-16 Julia Koval <julia.koval@intel.com>
2644 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2645 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2647 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
2650 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2651 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2652 and -mno-direct-move.
2654 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
2657 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
2658 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
2659 Ensure negative shifts result in {0}.
2661 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
2663 PR rtl-optimization/79916
2664 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
2665 regs (if any) to define how to gnerate SD moves when LRA is in
2668 2018-04-13 Jakub Jelinek <jakub@redhat.com>
2670 PR rtl-optimization/85393
2671 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
2672 * except.c (expand_dw2_landing_pad_for_region): Make static.
2673 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2674 a label and unconditional jump to old_bb, rather than
2675 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2678 PR rtl-optimization/85376
2679 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2680 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2681 instead of a specific value.
2683 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2684 Bin Cheng <bin.cheng@arm.com>
2686 PR tree-optimization/82965
2687 PR tree-optimization/83991
2688 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2689 by_profile_only parameter.
2690 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2691 information if the loop was predicted to iterate too many times.
2692 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2694 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2697 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
2700 2018-04-13 Martin Liska <mliska@suse.cz>
2701 Jakub Jelinek <jakub@redhat.com>
2704 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
2705 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
2706 * builtins.c (expand_builtin_memory_copy_args): Use
2707 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
2708 handle dest_addr == pc_rtx.
2710 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
2713 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
2714 asked to not generate direct moves.
2715 (fix_trunc<mode>si2_stfiwx): Similar.
2716 (fix_trunc<mode>si2_internal): Similar.
2718 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2721 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
2722 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
2723 lookup if dest in some wider mode is known to be const0_rtx and
2724 if so, record permanent equivalence for it to be ZERO_EXTEND of
2725 the narrower mode destination.
2727 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2729 * lto-streamer-out.c (output_function): Revert 259346.
2730 * omp-expand.c (expand_oacc_for): Likewise.
2732 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
2734 PR rtl-optimization/85354
2735 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
2736 * sel-sched.c (sel_global_init): ... here.
2738 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
2741 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
2742 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2743 mode for PE-COFF targets.
2744 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
2745 (i386_pe_asm_lto_end): Likewise.
2746 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
2747 (TARGET_ASM_LTO_END): Likewise.
2748 * config/i386/winnt.c (saved_debug_info_level): New static variable.
2749 (i386_pe_asm_lto_start): New function.
2750 (i386_pe_asm_lto_end): Likewise.
2752 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2753 Richard Biener <rguenther@suse.de>
2756 * lto-streamer-out.c (output_function): Fix CFG loop state before
2758 * omp-expand.c (expand_oacc_for): Handle calls to internal
2759 functions like regular functions.
2761 2018-04-12 Richard Biener <rguenther@suse.de>
2764 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
2765 for the early LTO debug to properly generate references to it
2766 during DIE emission. Do not re-use that for the skeleton for
2768 (dwarf2out_early_finish): Likewise.
2770 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2773 * config/i386/sse.md
2774 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
2775 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
2776 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
2777 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
2778 and output is a reg, avoid creating invalid lowpart subreg, but
2779 instead split into a 512-bit move. Don't split if not AVX512VL,
2780 input is xmm16+ reg and output is a mem.
2781 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
2782 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
2783 xmm16+ reg and output is a mem.
2785 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2787 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
2788 also for flag_dwarf2_cfi_asm.
2790 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2792 PR rtl-optimization/85342
2793 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
2794 a bool scalar var inside of the loop instead. Don't try to update
2795 recog_data.operand after failed apply_change_group.
2797 2018-04-12 Tom de Vries <tom@codesourcery.com>
2800 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
2801 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
2802 array with flexible array member as array without given dimension.
2803 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
2804 argument for undefined param to true.
2806 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
2809 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2810 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
2811 from PowerPC section.
2812 * config/rs6000/sysv4.opt (mcall-): Improve help text.
2813 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
2814 help text that is too long.
2815 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
2816 help text that is too long.
2817 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
2818 help text that is too long.
2820 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
2822 * config/alpha/alpha.md (stack_probe_internal): Rename
2823 from "probe_stack". Update all callers.
2825 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2827 PR rtl-optimization/84566
2828 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
2829 sched_macro_fuse_insns.
2831 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2834 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
2835 (compute_block_dependences): ... from here.
2837 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2839 PR tree-optimization/85331
2840 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
2841 from int to HOST_WIDE_INT.
2843 2018-04-11 Martin Jambor <mjambor@suse.cz>
2846 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
2847 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
2848 not the same as the source val.
2849 (cgraph_edge_brings_value_p): New parameter.
2850 (gather_edges_for_value): Pass destination value to
2851 cgraph_edge_brings_value_p.
2852 (perhaps_add_new_callers): Likewise.
2853 (get_info_about_necessary_edges): Likewise and exclude values brought
2854 only by self-recursive edges.
2855 (create_specialized_node): Redirect only clones of self-calling edges.
2856 (+self_recursive_pass_through_p): New function.
2857 (find_more_scalar_values_for_callers_subset): Use it.
2858 (find_aggregate_values_for_callers_subset): Likewise.
2859 (known_aggs_to_agg_replacement_list): Removed.
2860 (decide_whether_version_node): Re-calculate known constants for all
2861 remaining context clones.
2863 2018-04-11 Richard Biener <rguenther@suse.de>
2866 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
2867 from early DWARF output.
2868 (dwarf2out_early_finish): Output line info unconditionally into
2869 early DWARF and add reference to it.
2871 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2874 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
2875 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
2876 other than V2DFmode using iptr mode attribute.
2877 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
2879 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2881 PR rtl-optimization/84659
2882 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
2884 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2887 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
2889 (output_loc_list): Pass address of a dummy size variable even in the
2890 locview handling loop.
2891 (index_location_lists): Add comment on why skip_loc_list_entry can't
2894 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
2897 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
2900 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
2903 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2904 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
2905 and -mstring-compare-inline-limit.
2907 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2910 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
2911 for stack clash protection in a register whenever we need it to be in
2914 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2916 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
2917 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
2919 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2922 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
2924 (mlong-double-): Ditto.
2925 * config/rs6000/sysv4.opt (msdata=): Ditto.
2926 (mtls-size=): Ditto.
2928 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
2930 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2931 erroneous entries for
2932 "vector int vec_ldl (int, long int *)", and
2933 "vector unsigned int vec_ldl (int, unsigned long int *)".
2934 Add comments and entries for
2935 "vector bool char vec_ldl (int, bool char *)",
2936 "vector bool short vec_ldl (int, bool short *)",
2937 "vector bool int vec_ldl (int, bool int *)",
2938 "vector bool long long vec_ldl (int, bool long long *)",
2939 "vector pixel vec_ldl (int, pixel *)",
2940 "vector long long vec_ldl (int, long long *)",
2941 "vector unsigned long long vec_ldl (int, unsigned long long *)".
2942 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
2943 type tree bool_long_long_type_node and correct definition of
2944 bool_V2DI_type_node to make reference to this new type tree.
2945 (rs6000_mangle_type): Replace erroneous reference to
2946 bool_long_type_node with bool_long_long_type_node.
2947 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
2948 comments to emphasize sign distinctions for char and int types and
2949 replace RS6000_BTI_bool_long constant with
2950 RS6000_BTI_bool_long_long constant. Also add comment to restrict
2951 use of RS6000_BTI_pixel.
2952 (bool_long_type_node): Remove this macro definition.
2953 (bool_long_long_type_node): New macro definition
2955 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2957 PR rtl-optimization/85300
2958 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
2959 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
2960 simplify_unary_operation fails.
2962 2018-04-10 Martin Liska <mliska@suse.cz>
2964 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
2965 cgraph_edge and ipa_ref.
2967 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2971 * config/i386/sse.md
2972 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
2973 computation of the VEC_MERGE selector from mask.
2974 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
2975 Fix decoding of the VEC_MERGE selector into mask.
2977 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2979 PR tree-optimization/85286
2980 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
2982 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2984 * final.c (final_1): Set insn_last_address as well as
2985 insn_current_address.
2987 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2990 * explow.c (emit_stack_probe): Call validize_mem on memory location
2991 before passing it to gen_probe_stack. Create address operand and
2992 legitimize it for the probe_stack_address case.
2994 2018-04-09 Jan Hubicka <jh@suse.cz>
2997 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
2998 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
2999 * tree.c (free_lang_data_in_type): Fix handling of binfos;
3001 (free_lang_data): Rebuild type inheritance graph.
3003 2018-04-09 Martin Sebor <msebor@redhat.com>
3005 * invoke.texi (-finline-small-functions): Mention other optimization
3007 (-findirect-inlining, -fpartial-inlining): Same.
3008 (-finline-functions-called-once): Same.
3009 (-freorder-blocks-and-partition): Same.
3011 2018-04-09 Jan Hubicka <jh@suse.cz>
3014 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
3015 jumps; choose last target that matches the criteria (i.e.
3016 no partition changes for non-crossing jumps).
3017 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
3018 support for redirecting crossing jumps to non-crossing.
3020 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
3022 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
3023 also for naked functions.
3025 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
3027 * config/arc/arc.md (add_shift): New pattern.
3028 (add_shift2): Likewise.
3029 (sub_shift): Likewise.
3030 (sub_shift_cmp0_noout): Likewise.
3031 (compare_si_ashiftsi): Likewise.
3032 (xbfu_cmp0_noout): New combine pattern.
3033 (xbfu_cmp0"): Likewise.
3034 (movsi_set_cc_insn): Place the predicable variant first.
3035 (commutative_binary_cmp0_noout): Remove clobber.
3036 (commutative_binary_cmp0): New pattern.
3037 (noncommutative_binary_cmp0): Likewise.
3038 (noncommutative_binary_cmp0_noout): Likewise.
3039 (noncommutative_binary_comparison_result_used): Removed.
3040 (rsub_cmp0): New pattern.
3041 (rsub_cmp0_noout): Likewise.
3042 (extzvsi): Changed, keep only meaningful variants.
3043 (SQH, SEZ): New iterators.
3044 (SQH_postfix): New mode attribute.
3045 (SEZ_prefix): New code attribute.
3046 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
3047 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
3048 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
3050 (noncommutative_operator): Check the availability of barrel
3053 2018-04-09 Richard Biener <rguenther@suse.de>
3055 PR tree-optimization/85284
3056 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
3057 Only use the niter constraining form of simple_iv when the exit
3060 2018-04-09 Tom de Vries <tom@codesourcery.com>
3063 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3064 (define_expand "*memory_barrier"): New define_expand.
3065 (define_insn "memory_barrier"): New insn.
3067 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3069 PR rtl-optimization/80463
3070 PR rtl-optimization/83972
3071 PR rtl-optimization/83480
3073 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
3074 correct producer for the insn.
3075 (tidy_control_flow): Fixup seqnos in case of debug insns.
3077 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3079 PR rtl-optimization/83913
3081 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
3082 different sched-times when merging exprs.
3084 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3086 PR rtl-optimization/83962
3088 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
3089 tidy_fallthru_edge and tidy_control_flow.
3091 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3093 PR rtl-optimization/83530
3095 * sel-sched.c (force_next_insn): New global variable.
3096 (remove_insn_for_debug): When force_next_insn is true, also leave only
3097 next insn in the ready list.
3098 (sel_sched_region): When the region wasn't scheduled, make another pass
3099 over it with force_next_insn set to 1.
3101 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3103 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
3105 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3106 for interrupt control.
3107 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3108 functions for interrupt control.
3109 * config/nds32/nds32-intrinsic.md: Likewise.
3110 * config/nds32/nds32_intrinsic.h: Likewise.
3111 * config/nds32/nds32.h (nds32_builtins): Likewise.
3113 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3115 * config/nds32/nds32.c (nds32_init_machine_status,
3116 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
3117 strict_aligned_p field.
3118 (nds32_expand_to_rtl_hook): New function.
3119 (TARGET_EXPAND_TO_RTL_HOOK): Define.
3120 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
3122 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3123 Chung-Ju Wu <jasonwucj@gmail.com>
3125 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
3126 * config/nds32/nds32-n7.md: New file.
3127 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
3128 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
3130 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
3131 * config/nds32/nds32.md (pipeline_model): Add n7.
3132 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
3133 * config/nds32/pipelines.md: Include n7 settings.
3135 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3136 Chung-Ju Wu <jasonwucj@gmail.com>
3138 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
3139 * config/nds32/nds32-e8.md: New file.
3140 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
3141 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
3143 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
3144 * config/nds32/nds32.md (pipeline_model): Add e8.
3145 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
3146 * config/nds32/pipelines.md: Include e8 settings.
3148 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3149 Chung-Ju Wu <jasonwucj@gmail.com>
3151 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
3152 * config/nds32/nds32-n8.md: New file.
3153 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
3154 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
3156 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
3157 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
3158 * config/nds32/nds32.md (pipeline_model): Add n8.
3159 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
3160 * config/nds32/pipelines.md: Include n8 settings.
3162 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3163 Chung-Ju Wu <jasonwucj@gmail.com>
3165 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
3166 * config/nds32/nds32-n9-2r1w.md: New file.
3167 * config/nds32/nds32-n9-3r2w.md: New file.
3168 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
3169 nds32_register_ports): New or modify for cpu n9.
3170 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
3172 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
3173 * config/nds32/nds32-utils.c: New file.
3174 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
3175 TARGET_MUL_SLOW): Define.
3176 * config/nds32/nds32.md (pipeline_model): New attribute.
3177 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
3178 New options that support cpu n9.
3179 * config/nds32/pipelines.md: Include n9 settings.
3180 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
3182 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3184 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
3185 information if necessary.
3186 (output_cond_branch_compare_zero): Likewise.
3187 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
3188 (nds32_target_alignment): Refine for alignment.
3189 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
3190 (FUNCTION_BOUNDARY): Modify.
3191 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
3193 * config/nds32/nds32.opt (malways-align, malign-functions): New.
3195 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3197 * config/nds32/constants.md (unspec_volatile_element): Add values for
3198 TLB operation and data prefetch.
3199 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3200 functions for TLB operation and data prefetch.
3201 * config/nds32/nds32-intrinsic.md: Likewise.
3202 * config/nds32/nds32_intrinsic.h: Likewise.
3203 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
3204 (nds32_print_operand): Likewise.
3205 * config/nds32/nds32.h (nds32_builtins): Likewise.
3207 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
3208 Andrew Pinski <pinsika@gcc.gnu.org>
3211 * match.pd: Use constant_boolean_node of correct type instead of
3212 boolean_true_node or boolean_false_node for simplifying
3213 pointer comparisons to zero.
3215 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3217 PR tree-optimization/80021
3218 * tree.c (verify_type_variant): Make error call in verify_variant_match
3219 translatable and remove final full stop.
3221 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3223 * config/nds32/constants.md (unspec_volatile_element): Add
3224 UNSPEC_VOLATILE_EH_RETURN.
3225 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
3226 nds32_output_stack_pop): Support dwarf exception handling process.
3227 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
3228 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
3229 exception handling process.
3230 (nds32_compute_stack_frame): Likewise.
3231 (nds32_return_addr_rtx): Likewise.
3232 (nds32_initial_elimination_offset): Likewise.
3233 (nds32_expand_prologue): Likewise.
3234 (nds32_expand_epilogue): Likewise.
3235 (nds32_dynamic_chain_address): New function.
3236 * config/nds32/nds32.h (machine_function): Add fields for dwarf
3238 (DYNAMIC_CHAIN_ADDRESS): Define.
3239 (EH_RETURN_DATA_REGNO): Define.
3240 (EH_RETURN_STACKADJ_RTX): Define.
3241 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
3242 patterns for dwarf exception handling.
3244 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3246 * config/nds32/nds32.h: Clean up obsolete macros.
3248 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3250 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3251 Add enum values for particular instructions.
3252 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3253 particular intrinsic functions.
3254 * config/nds32/nds32-intrinsic.md: Likewise.
3255 * config/nds32/nds32_intrinsic.h: Likewise.
3256 * config/nds32/nds32.h (nds32_builtins): Likewise.
3257 * config/nds32/nds32.md (type): Add pbsad and pbsada.
3258 (btst, ave): New patterns for particular instructions.
3260 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3262 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3263 Add enum values for atomic load/store and memory sync.
3264 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
3266 * config/nds32/nds32-intrinsic.md: Likewise.
3267 * config/nds32/nds32_intrinsic.h: Likewise.
3268 * config/nds32/nds32.h (nds32_builtins): Likewise.
3270 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3272 PR tree-optimization/85257
3273 * fold-const.c (native_encode_vector): If not all elts could fit
3274 and off is -1, return 0 rather than offset.
3275 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
3276 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
3277 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
3278 adjust buffer in native_interpret_expr call.
3280 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3282 * config/nds32/constants.md (unspec_volatile_element): Add cache
3283 control enum values.
3284 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
3285 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
3286 * config/nds32/nds32.c (nds32_cctl_names): New.
3287 (nds32_print_operand): Handle cache control register names.
3288 * config/nds32/nds32.h (nds32_builtins): New enum values.
3289 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
3291 * config/nds32/nds32.md (type): Add mmu.
3292 * config/nds32/pipelines.md (simple_insn): Add mmu.
3294 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3296 * config/nds32/nds32.md (type): Remove call.
3297 * config/nds32/pipelines.md (simple_insn): Likewise.
3299 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3301 * config/nds32/constants.md (unspec_volatile_element): Add
3302 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
3303 UNSPEC_VOLATILE_FMFCFG.
3304 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
3305 description for fmfcfg and fmfcsr.
3306 (bdesc_1arg): Add fmtcsr.
3307 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
3308 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
3309 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
3310 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
3311 unspec_fmfcfg): New patterns.
3312 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
3313 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
3314 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
3315 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
3316 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
3317 __nds32__fmfcfg): Define.
3319 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3321 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
3322 intrinsic register names.
3323 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
3324 intrinsic register enum values and macros.
3326 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3328 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
3329 for load/store addressing form.
3330 (nds32_print_operand_address): Likewise.
3332 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
3335 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
3336 based on LABEL_REF. Remove useless assertion.
3337 (pic_address_needs_scratch): Fix formatting.
3338 (sparc_legitimize_pic_address): Minor tweaks.
3339 (sparc_delegitimize_address): Adjust assertion accordingly.
3340 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
3341 into symbolic_operand.
3342 (movsi_high_pic_label_ref): Likewise.
3343 (movsi_lo_sum_pic_label_ref): Likewise.
3344 (movdi_pic_label_ref): Likewise.
3345 (movdi_high_pic_label_ref): Likewise.
3346 (movdi_lo_sum_pic_label_ref): Likewise.
3348 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
3350 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
3351 custom LIB_SPEC setup.
3353 2018-04-06 Ruslan Bukin <br@bsdpad.com>
3354 Kito Cheng <kito.cheng@gmail.com>
3356 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
3357 * config/riscv/freebsd.h: New.
3359 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3361 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
3362 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
3365 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3366 Kito Cheng <kito.cheng@gmail.com>
3368 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
3369 nds32_output_call, nds32_symbol_binds_local_p): New functions.
3370 * config/nds32/nds32-protos.h (nds32_output_call,
3371 nds32_output_return): Declare.
3372 * config/nds32/nds32.md: Refine all the call and return patterns.
3374 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3377 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
3378 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
3380 PR rtl-optimization/84872
3381 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
3382 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
3385 2018-04-06 Tamar Christina <tamar.christina@arm.com>
3387 * expr.c (copy_blkmode_to_reg): Revert 254862.
3388 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
3390 2018-04-06 Richard Biener <rguenther@suse.de>
3393 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
3394 after seeing a component reference with an adjacent field. Treat
3395 refs to arrays at struct end of external decls similar to
3396 refs to unconstrained commons.
3398 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3401 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
3402 look through SAVE_EXPRs with non-side-effects argument. Adjust
3404 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3407 2018-04-06 Richard Biener <rguenther@suse.de>
3410 * alias.c (find_base_term): New wrapper around find_base_term
3411 unwinding CSELIB_VAL_PTR changes.
3412 (find_base_term): Do not restore CSELIB_VAL_PTR during the
3415 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3417 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3419 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3420 constant definitions.
3421 ("nop"): lr 0,0 -> nopr r0
3422 ("nop_lr0", "nop_lr1"): New insn definitions.
3424 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3426 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3427 NDS32_V3PUSH_AVAILABLE_P macro.
3429 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
3430 Chung-Ju Wu <jasonwucj@gmail.com>
3432 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3433 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3434 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3435 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3436 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3437 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3438 * config/nds32/constraints.md: New constraints and checking for hard
3439 float configuration.
3440 * config/nds32/iterators.md: New mode iterator and attribute for hard
3441 float configuration.
3442 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3444 * config/nds32/nds32-fpu.md: New file.
3445 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3446 deal with hard float code generation.
3447 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3449 (abi_type, float_reg_number): New enum type.
3450 * config/nds32/nds32-predicates.c: New predicates for hard float.
3451 * config/nds32/nds32-protos.h: Declare functions for hard float.
3452 * config/nds32/nds32.c: Implementation for hard float configuration.
3453 * config/nds32/nds32.h: Definitions for hard float configuration.
3454 * config/nds32/nds32.md: Include hard float machine description and
3455 modify patterns for hard float configuration.
3456 * config/nds32/nds32.opt: New options for hard float configuration.
3457 * config/nds32/predicates.md: New predicates for hard float
3460 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3462 * common/config/nds32/nds32-common.c
3463 (nds32_option_optimization_table): Enable -mreleax-hint by default.
3465 2018-04-05 Jakub Jelinek <jakub@redhat.com>
3468 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3469 CONSTRUCTOR_ELT (ctor, ...)->value.
3471 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
3474 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3476 2018-04-05 Tom de Vries <tom@codesourcery.com>
3479 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3482 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3483 Kito Cheng <kito.cheng@gmail.com>
3485 * config/nds32/constraints.md (U33): Fine-tune checking condition.
3486 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3487 * config/nds32/nds32.h (nds32_16bit_address_type): Add
3488 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3490 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3491 Kito Cheng <kito.cheng@gmail.com>
3493 * config/nds32/constraints.md (Ufe): New memory constraint.
3494 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3495 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3496 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3498 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3499 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3501 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3503 * config/nds32/nds32.md: Use optimize_size in the condition for
3504 alu-shift instructions.
3506 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3508 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3510 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3512 * config/nds32/nds32.md (negsi2): Refine pattern.
3514 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3515 Chung-Ju Wu <jasonwucj@gmail.com>
3517 * config/nds32/iterators.md (shift_rotate): New code iterator.
3518 (shift): New code attribute.
3519 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3520 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3521 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3522 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3523 bit-wise operations.
3524 (andsi3, *andsi3): Ditto.
3525 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3526 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3527 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3528 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3529 nds32_ior_operand, nds32_xor_operand): New predicates.
3531 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3533 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3534 (addsi3, subsi3): ... this.
3536 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3538 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3540 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3542 * config/nds32/nds32.md: Adjust indention.
3544 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3546 * config/nds32/nds32.md (feature): New attribute.
3548 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3550 * config/nds32/nds32.md (subtype): New attribute.
3552 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
3555 * config/arm/arm-builtins.c (arm_expand_builtin): Change
3556 expansion to perform a bitwise AND of the argument followed by a
3557 boolean negation of the result.
3559 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3561 PR rtl-optimization/84878
3562 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3563 the basic block. Assert the use reference is not artificial and that
3564 it has an associated insn.
3566 2018-04-04 Michael Matz <matz@suse.de>
3568 * builtins.c (compute_objsize): Pass correct operand
3569 to array_at_struct_end_p.
3571 2018-04-04 Richard Biener <rguenther@suse.de>
3574 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3575 from contexts for DINFO_LEVEL_TERSE and below.
3577 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3579 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3580 resiter_operand condition.
3581 * config/nds32/nds32.md (*move<mode>): Ditto.
3583 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3584 Monk Chiang <sh.chiang04@gmail.com>
3586 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3588 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3590 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3592 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3593 Kito Cheng <kito.cheng@gmail.com>
3595 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3596 nds32_cond_code_str, output_cond_branch,
3597 output_cond_branch_compare_zero, nds32_expand_cbranch,
3598 nds32_expand_cstore, nds32_expand_movcc,
3599 nds32_output_cbranchsi4_equality_zero,
3600 nds32_output_cbranchsi4_equality_reg,
3601 nds32_output_cbranchsi4_equality_reg_or_const_int,
3602 nds32_output_cbranchsi4_greater_less_zero: New functions.
3603 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3604 nds32_expand_cstore, nds32_expand_movcc,
3605 nds32_output_cbranchsi4_equality_zero,
3606 nds32_output_cbranchsi4_equality_reg,
3607 nds32_output_cbranchsi4_equality_reg_or_const_int,
3608 nds32_output_cbranchsi4_greater_less_zero): Declare.
3609 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3610 nds32_rimm11s_operand): New predicates.
3611 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3612 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3615 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3617 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3618 * config/nds32/nds32.md: Ditto.
3619 * config/nds32/pipelines.md: Ditto.
3621 2018-04-04 Richard Biener <rguenther@suse.de>
3623 PR tree-optimization/85168
3624 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3625 propagating abnormals.
3627 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3629 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3631 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3632 Kito Cheng <kito.cheng@gmail.com>
3634 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3635 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3636 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3637 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3638 * config/nds32/nds32.md (sibcall_internal): New.
3639 (sibcall_register): Remove.
3640 (sibcall_immediate): Remove.
3641 (sibcall_value_internal): New.
3642 (sibcall_value_register): Remove.
3643 (sibcall_value_immediate): Remove.
3644 * config/nds32/predicates.md (nds32_general_register_operand): New.
3645 (nds32_call_address_operand): New.
3647 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3649 PR rtl-optimization/85167
3650 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3651 bb_defs if *split_p, instead preinitialize it to NULL.
3653 PR tree-optimization/85156
3654 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3655 evaluating the argument multiple times.
3657 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
3659 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
3661 (_mm_cvtpd_ps): Likewise.
3662 (_mm_cvttpd_epi32): Likewise.
3663 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
3664 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
3665 vector, pixel, and bool following altivec.h include.
3667 2018-04-03 Martin Sebor <msebor@redhat.com>
3669 * doc/extend.texi (Common Function Attributes): Clarify.
3670 (const attribute): Likewise.
3671 (pure attribute): Likewise.
3673 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3676 * config/i386/i386.c (ix86_expand_vector_set): Use
3677 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
3679 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
3681 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3682 instructions when changing rounding bits to preserve precision bits
3683 in the x87 control word.
3685 2018-04-03 Martin Liska <mliska@suse.cz>
3687 PR tree-optimization/82491
3688 * rtl.h (strip_offset_and_add): Replace += suboffset with
3689 poly_uint64 () + suboffset.
3691 2018-03-29 Martin Liska <mliska@suse.cz>
3692 Martin Jambor <mjambor@suse.cz>
3695 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
3696 param_type is not an integral or pointer type.
3698 2018-04-03 Richard Biener <rguenther@suse.de>
3700 * sese.h (recompute_all_dominators): Remove.
3702 2018-04-02 Martin Sebor <msebor@redhat.com>
3704 * doc/invoke.texi (-Wrestrict): Fix typos.
3706 2018-04-02 Jim Wilson <jimw@sifive.com>
3708 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
3709 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
3710 (<optab>di3, <optab>si3_extend): Likewise.
3711 (<optab>si3_mask, <optab>si3_mask_1): New.
3712 (<optab>di3_mask, <optab>di3_mask_1): New.
3713 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
3714 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
3715 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
3717 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
3719 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
3722 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
3724 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
3725 (nds32_canonicalize_comparison): New function.
3727 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3728 Kito Cheng <kito.cheng@gmail.com>
3729 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3731 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
3732 * config/nds32/constants.md (unspec_volatile_element): Add
3733 UNSPEC_VOLATILE_RELAX_GROUP.
3734 * config/nds32/nds32-relax-opt.c: New file.
3735 * config/nds32/nds32-predicates.c
3736 (nds32_symbol_load_store_p): New function.
3737 * config/nds32/nds32-protos.h
3738 (nds32_symbol_load_store_p): Declare function.
3739 (make_pass_nds32_relax_opt): Declare new rtl pass function.
3740 * config/nds32/nds32.c
3741 (nds32_register_pass): New function to register pass.
3742 (nds32_register_passes): New function to register passes.
3743 * config/nds32/nds32.md (relax_group): New pattern.
3744 * config/nds32/nds32.opt (mrelax-hint): New option.
3745 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
3747 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
3749 * config/nds32/t-nds32: Modify files dependency.
3751 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3753 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
3754 (PROFILE_HOOK): Define its implementation.
3756 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3758 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
3759 type and 32-bit size.
3761 2018-04-01 Jakub Jelinek <jakub@redhat.com>
3764 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
3765 (V_128_256): New mode iterator.
3766 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
3767 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
3768 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
3770 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
3771 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
3773 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
3776 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
3777 NaN inputs correctly.
3779 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
3782 * config/rs6000/vsx.md (??r): New mode attribute.
3783 (*vsx_mov<mode>_64bit): Use it.
3784 (*vsx_mov<mode>_32bit): Likewise.
3786 2018-03-30 Martin Sebor <msebor@redhat.com>
3788 PR tree-optimization/84818
3789 * builtins.c (check_access): Use warning_n.
3791 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3794 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
3796 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
3799 2018-03-30 Julia Koval <julia.koval@intel.com>
3802 * x86-tune.def (movx, partial_reg_dependency): Enable for
3805 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
3808 * lra-constraints.c (process_alt_operands): Move setting
3809 this_alternative_matches below.
3811 2018-03-29 Martin Liska <mliska@suse.cz>
3814 * doc/invoke.texi: Document how LTO works with debug info.
3815 Describe auto-load support of binutils. Mention 'x86-64'
3816 as valid option value of -march option.
3818 2018-03-29 Jakub Jelinek <jakub@redhat.com>
3820 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
3823 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
3824 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
3825 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
3828 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3831 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
3832 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
3833 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
3834 for RS6000_BTM_POWERPC64.
3835 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
3836 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
3837 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
3842 2018-03-28 Carl Love <cel@us.ibm.com>
3845 2017-09-27 Carl Love <cel@us.ibm.com>
3847 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
3848 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
3849 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
3852 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3854 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
3855 instead of __vector bool.
3856 (_mm_max_pu8): Likewise.
3857 (_mm_min_pi16): Likewise.
3859 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3862 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
3863 (DIVWEUO): Likewise.
3865 (DIVDEUO): Likewise.
3866 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
3867 DIVWEUO and DIVDEUO.
3868 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
3869 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
3870 (div_extend): Likewise.
3871 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
3873 (__builtin_divweuo): Likewise.
3874 (__builtin_divdeo): Likewise.
3875 (__builtin_divdeuo): Likewise.
3877 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3880 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
3881 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
3883 PR tree-optimization/82004
3884 * gimple-match-head.c (optimize_pow_to_exp): New function.
3885 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
3886 Don't fold to exp if optimize_pow_to_exp is false.
3888 2018-03-28 Martin Liska <mliska@suse.cz>
3891 * calls.c (initialize_argument_information): Fix trailing space.
3892 * common.opt: Fix typo and provide better explanation for
3893 -fsanitize-coverage option.
3894 * config/i386/i386.opt: Fix typo.
3896 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3897 Martin Liska <mliska@suse.cz>
3900 * gimplify.c (asan_poison_variable): Don't do the check for
3901 gimplify_omp_ctxp here.
3902 (gimplify_decl_expr): Do it here.
3903 (gimplify_target_expr): Likewise.
3905 2018-03-28 Martin Liska <mliska@suse.cz>
3908 * config/i386/i386.c (ix86_function_arg_advance): Do not call
3909 chkp_type_bounds_count if MPX is not enabled.
3911 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
3913 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
3915 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
3918 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
3919 function to create the function decl for complex long double
3920 multiply and divide for -mabi=ieeelongdouble.
3921 (init_float128_ieee): Call it.
3923 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
3926 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
3927 -fcf-protection=branch -mibt.
3928 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
3930 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3933 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
3935 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
3938 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
3939 extern array declarations.
3941 2018-03-27 Richard Biener <rguenther@suse.de>
3944 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
3945 explicit single_use checks.
3947 2018-03-27 Richard Biener <rguenther@suse.de>
3949 PR tree-optimization/85082
3950 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
3953 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3955 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
3956 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
3957 Turn on fasynchronous-unwind-tables and funwind-tables.
3959 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
3962 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
3963 (*bmi_blsr_<mode>_ccz): Ditto.
3965 2018-03-26 Tom de Vries <tom@codesourcery.com>
3967 PR tree-optimization/85063
3968 * omp-general.c (offloading_function_p): New function. Factor out
3970 * omp-offload.c (pass_omp_target_link::gate): ... here.
3971 * omp-general.h (offloading_function_p): Declare.
3972 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
3973 with attribute omp declare target for offloading functions.
3975 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
3977 PR tree-optimization/84005
3978 * tree-data-ref.h (get_base_for_alignment): Declare.
3979 * tree-data-ref.c (get_base_for_alignment_1): New function.
3980 (get_base_for_alignment): Likewise.
3981 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
3982 get_base_for_alignment to find a suitable base object, instead
3983 of always using drb->base_address.
3985 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3988 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
3989 known size by default.
3991 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
3994 * lra-constraints.c (process_alt_operands): Don't match BLKmode
3995 and non BLKmode operands.
3997 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4000 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
4001 Clean up attributes.
4003 2018-03-23 Richard Biener <rguenther@suse.de>
4006 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
4007 we are going to emit early debug for LTO.
4009 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4012 * function.c (match_asm_constraints_1): Don't optimize if input
4013 doesn't satisfy general_operand predicate for output's mode.
4016 * alias.c (write_dependence_p): Don't require for x_canonicalized
4017 non-VOIDmode if x has VOIDmode.
4020 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
4021 just don't try to optimize it rather than assert it never happens.
4023 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4025 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
4026 macro expansions for definition of ST_INTERNAL_<mode> and
4027 LD_INTERNAL_<mode> builtins.
4028 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
4030 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
4032 (altivec_expand_st_builtin): Likewise.
4033 (altivec_expand_builtin): Remove calls to deleted functions.
4034 (rs6000_address_for_altivec): Delete this function.
4035 * config/rs6000/vector.md: Remove expands for
4036 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
4038 2018-03-22 Sudakshina Das <sudi.das@arm.com>
4041 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
4042 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
4043 re-computing once computed.
4044 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
4045 (arm_init_machine_status): Initialize
4046 machine->static_chain_stack_bytes.
4048 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4051 * doc/extend.texi: Add four new prototypes for vec_ld.
4052 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
4053 definitions for more logical presentation.
4054 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
4055 entries for V1TI variants of __builtin_altivec_ld builtin.
4056 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
4057 handling of V1TI variant of LVX icode pattern.
4058 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
4059 (rs6000_gimple_fold_builtin): Likewise.
4060 (altivec_init_builtins): Add code to define
4061 __builtin_altivec_lvx_v1ti function.
4063 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4066 * function.c (match_asm_constraints_1): Don't do the optimization
4067 if input isn't a REG, SUBREG, MEM or constant.
4069 2018-03-22 Tom de Vries <tom@codesourcery.com>
4071 PR tree-optimization/84956
4072 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
4073 bb_has_abnormal_pred.
4075 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4078 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
4079 DECL_INITIAL (decl) to decl at the end.
4080 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
4083 2018-03-21 Joseph Myers <joseph@codesourcery.com>
4085 * doc/extend.texi (__builtin_tgmath): Document when complex
4086 integer types are treated as _Complex _Float64.
4088 2018-03-21 Tom de Vries <tom@codesourcery.com>
4090 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
4092 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4094 PR tree-optimization/84960
4095 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
4096 if it is ENTRY block, move them into single succ of ENTRY in that case.
4098 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
4100 PR tree-optimization/84811
4101 * poly-int.h (poly_span_traits): Remove the T3 parameter and
4102 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
4103 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
4104 (known_subrange_p): Update accordingly. Cast each value involved
4105 in the size comparison, rather than casting the result of the
4108 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4110 PR tree-optimization/84982
4111 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
4112 by flipping the least significant bit rather than all bits from
4113 bitpos to bitpos + bitsize - 1.
4115 2018-03-21 Nathan Sidwell <nathan@acm.org>
4117 * doc/extend.texi (Deprecated Features): Remove mention of
4118 long-deleted deprecations.
4120 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4123 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
4124 * configure: Regenerate.
4126 2018-03-21 Tom de Vries <tom@codesourcery.com>
4128 PR tree-optimization/83126
4129 * tree-parloops.c (num_phis): New function.
4130 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
4132 2018-03-21 Nathan Sidwell <nathan@acm.org>
4134 * doc/extend.texi (Deprecated Features): Update deprecated flags,
4135 mention anon-struct/union members and trailing attributes.
4137 2018-03-21 Bin Cheng <bin.cheng@arm.com>
4139 PR tree-optimization/84969
4140 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
4141 builtin memset partitions if they set different rhs values.
4143 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4145 PR rtl-optimization/84989
4146 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
4147 VEC_DUPLICATE with scalar result mode.
4149 2018-03-21 Martin Liska <mliska@suse.cz>
4152 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
4153 not intended return statement.
4155 2018-03-21 Martin Liska <mliska@suse.cz>
4158 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
4159 (chkp_find_bound_slots_1): Limit number of iterations.
4161 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
4164 * Minor grammar fixes for x86 options.
4166 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4169 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
4170 holding REG_CFA_RESTORE notes, instead turn them into a USE.
4172 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
4175 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
4176 (altivec_lvx_<mode>_1op): Likewise.
4177 (altivec_stvx_<mode>_2op): Likewise.
4178 (altivec_stvx_<mode>_1op): Likewise.
4179 (altivec_lvx_<VM2:mode>): New define_expand.
4180 (altivec_stvx_<VM2:mode>): Likewise.
4181 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
4182 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4183 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
4184 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4185 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
4186 (rs6000_gen_lvx): Likewise.
4187 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
4188 (altivec_expand_stv_builtin): Likewise.
4189 (altivec_expand_builtin): Likewise.
4190 * config/rs6000/vector.md: Likewise.
4192 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4195 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
4198 2018-03-20 Richard Biener <rguenther@suse.de>
4201 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
4202 sign-conversions as zero, fall back to standard scalar_stmt
4205 2018-03-20 Martin Liska <mliska@suse.cz>
4208 * predict.c (rebuild_frequencies): Handle case when we have
4209 PROFILE_ABSENT, but flag_guess_branch_prob is false.
4211 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4214 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
4215 flag_section_anchors.
4216 * varasm.c (use_blocks_for_decl_p): Remove hack for
4217 dw2_force_const_mem.
4220 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
4222 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
4223 be created, use lowpart_subreg of operands[0] rather than operands[0]
4225 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
4226 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
4227 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
4228 and n constraint instead of aarch64_shift_imm_di and Usd.
4229 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
4230 (*aarch64_<optab>_reg_minus<mode>3): ... this.
4232 2018-03-20 Sudakshina Das <sudi.das@arm.com>
4235 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
4236 to favor GPR over NEON registers.
4237 (<shift>di3_neon): Likewise.
4239 2018-03-20 Tom de Vries <tom@codesourcery.com>
4242 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
4243 (nvptx_process_pars): Emit bar.sync asap and alap.
4245 2018-03-20 Tom de Vries <tom@codesourcery.com>
4248 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
4249 seen_label if seen_label is already set.
4251 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4254 * config/i386/i386.c (fold_builtin_cpu): For features above 31
4255 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
4256 Use 1U instead of 1. Formatting fixes.
4259 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
4260 instead of TREE_TYPE (s1) for the return value.
4262 2018-03-19 Jakub Jelinek <jakub@redhat.com>
4264 PR tree-optimization/84946
4265 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
4266 bitsize + bitsize in poly_uint64 rather than poly_int64.
4269 * dwarf2asm.c: Include fold-const.c.
4270 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
4271 of decl rather than decl itself.
4273 PR rtl-optimization/84643
4274 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
4276 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
4279 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
4280 calling assemble_variable.
4282 2018-03-19 Sudakshina Das <sudi.das@arm.com>
4285 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
4286 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
4288 2018-03-19 Jim Wilson <jimw@sifive.com>
4291 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
4292 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
4293 (riscv_first_stack_step): Likewise.
4294 (riscv_option_override): Use STACK_BOUNDARY instead of
4296 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
4298 (BIGGEST_ALIGNMENT): Set to 128.
4299 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
4300 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
4303 2018-03-19 Richard Biener <rguenther@suse.de>
4305 PR tree-optimization/84933
4306 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
4307 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
4309 2018-03-19 Richard Biener <rguenther@suse.de>
4311 PR tree-optimization/84859
4312 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
4313 (cond_if_else_store_replacement): Perform sinking operation on
4314 single-store BBs regardless of MAX_STORES_TO_SINK setting.
4315 Generalize what a BB with a single eligible store is.
4317 2018-03-19 Richard Biener <rguenther@suse.de>
4319 PR tree-optimization/84929
4320 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
4321 chrec_is_positive against non-chrec arg.
4323 2018-03-19 Tamar Christina <tamar.christina@arm.com>
4326 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
4328 2018-03-18 Martin Liska <mliska@suse.cz>
4330 PR rtl-optimization/84635
4331 * regrename.c (build_def_use): Use matches_mode only when
4334 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
4336 PR tree-optimization/84913
4337 * tree-vect-loop.c (vectorizable_reduction): Don't try to
4338 vectorize chains of COND_EXPRs.
4340 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4342 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
4344 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4346 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
4348 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4350 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
4352 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4353 Kito Cheng <kito.cheng@gmail.com>
4355 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
4356 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
4357 (nds32_adjust_reg_alloc_order): New function.
4358 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
4360 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
4362 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
4363 nds32_print_operand, nds32_print_operand_address): Use
4364 HOST_WIDE_INT_PRINT_DEC instead.
4366 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4368 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
4370 2018-03-17 Jakub Jelinek <jakub@redhat.com>
4373 * config/i386/i386.c (initial_ix86_tune_features,
4374 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
4376 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
4377 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
4378 rather than 1u << ix86_tune. Formatting fix.
4379 (ix86_option_override_internal): Change ix86_arch_mask from
4380 unsigned int to unsigned HOST_WIDE_INT, initialize to
4381 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
4382 (ix86_function_specific_restore): Likewise.
4384 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4387 * postreload.c (reload_combine_recognize_pattern): Perform
4388 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
4389 truncate_int_for_mode the result for the destination's mode.
4392 * hsa-gen.c (mem_type_for_type): Fix comment typo.
4393 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
4395 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4398 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
4401 * lra-assigns.c (lra_split_hard_reg_for): Don't use
4402 regno_allocno_class_array and sorted_pseudos.
4403 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
4404 insns where regno is used.
4406 2018-03-16 Martin Liska <mliska@suse.cz>
4409 * multiple_target.c (create_dispatcher_calls): Redirect
4410 reference in the symbol table.
4412 2018-03-16 Martin Liska <mliska@suse.cz>
4415 * multiple_target.c (create_dispatcher_calls): Redirect also
4418 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4422 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4423 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4424 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4426 2018-03-16 Julia Koval <julia.koval@intel.com>
4428 * doc/invoke.texi (Skylake Server): Add CLWB.
4429 Cannonlake): Remove CLWB.
4431 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4433 PR tree-optimization/84841
4434 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4436 (FLOAT_ONE_CONST_TYPE): Define.
4437 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4438 (sort_by_operand_rank): Put entries with higher constant_type last
4439 rather than first to match comments.
4441 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
4443 * config/nios2/nios2.md (movsi_internal): Fix thinko in
4446 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4449 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4450 check and use address of target always.
4452 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
4455 * config/i386/i386.c (indirect_thunk_needed): Update comments.
4456 (indirect_thunk_bnd_needed): Likewise.
4457 (indirect_thunks_used): Likewise.
4458 (indirect_thunks_bnd_used): Likewise.
4459 (indirect_return_needed): New.
4460 (indirect_return_bnd_needed): Likewise.
4461 (output_indirect_thunk_function): Add a bool argument for
4463 (output_indirect_thunk_function): Don't generate alias for
4464 function return thunk.
4465 (ix86_code_end): Call output_indirect_thunk_function to generate
4466 function return thunks.
4467 (ix86_output_function_return): Set indirect_return_bnd_needed
4468 and indirect_return_needed instead of indirect_thunk_bnd_needed
4469 and indirect_thunk_needed.
4471 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
4473 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4474 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4475 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4477 2018-03-15 David Malcolm <dmalcolm@redhat.com>
4478 Paul Hua <paul.hua.gm@gmail.com>
4481 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4483 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
4485 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4486 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4489 2018-03-15 Tamar Christina <tamar.christina@arm.com>
4492 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4493 instead of GET_MODE_SIZE when comparing Units.
4495 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4498 * varasm.c (hash_section): Return an unchangeble hash value
4499 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4500 Return !aarch64_can_use_per_function_literal_pools_p ().
4502 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4505 * optabs.c (emit_conditional_move): Pass address of cmode's copy
4506 rather than address of cmode as last argument to prepare_cmp_insn.
4508 2018-03-15 Julia Koval <julia.koval@intel.com>
4510 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4511 F_AVX512VNNI, F_AVX512BITALG): New.
4513 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4516 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
4517 insn for floating-point loads and stores.
4519 2018-03-14 Carl Love <cel@us.ibm.com>
4521 * config/rs6000/rs6000-c.c: Add macro definitions for
4522 ALTIVEC_BUILTIN_VEC_PERMXOR.
4523 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4524 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4525 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4526 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4528 * config/doc/extend.texi: Add prototypes for vec_permxor.
4530 2018-03-14 David Malcolm <dmalcolm@redhat.com>
4533 * diagnostic-show-locus.c (class layout_point): Convert m_line
4534 from int to linenum_type.
4535 (line_span::comparator): Use linenum "compare" function when
4536 comparing line numbers.
4537 (test_line_span): New function.
4538 (layout_range::contains_point): Convert param "row" from int to
4540 (layout_range::intersects_line_p): Likewise.
4541 (layout::will_show_line_p): Likewise.
4542 (layout::print_source_line): Likewise.
4543 (layout::should_print_annotation_line_p): Likewise.
4544 (layout::print_annotation_line): Likewise.
4545 (layout::print_leading_fixits): Likewise.
4546 (layout::annotation_line_showed_range_p): Likewise.
4547 (struct line_corrections): Likewise for field m_row.
4548 (line_corrections::line_corrections): Likewise for param "row".
4549 (layout::print_trailing_fixits): Likewise.
4550 (layout::get_state_at_point): Likewise.
4551 (layout::get_x_bound_for_row): Likewise.
4552 (layout::print_line): Likewise.
4553 (diagnostic_show_locus): Likewise for locals "last_line" and
4555 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4556 * input.c (selftest::test_linenum_comparisons): New function.
4557 (selftest::input_c_tests): Call it.
4558 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4559 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4560 * selftest.h (ASSERT_GT): New macro.
4561 (ASSERT_GT_AT): New macro.
4562 (ASSERT_LT): New macro.
4563 (ASSERT_LT_AT): New macro.
4565 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4567 PR rtl-optimization/84780
4568 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4570 2018-03-14 Martin Liska <mliska@suse.cz>
4572 * tree.c (record_node_allocation_statistics): Use
4573 get_stats_node_kind.
4574 (get_stats_node_kind): New function extracted from
4575 record_node_allocation_statistics.
4576 (free_node): Use get_stats_node_kind.
4578 2018-03-14 Richard Biener <rguenther@suse.de>
4580 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4581 that the value-set of ANTIC_IN doesn't grow.
4584 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4586 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4587 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4589 2018-03-14 Julia Koval <julia.koval@intel.com>
4591 * config.gcc (icelake-client, icelake-server): New.
4593 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4594 (initial_ix86_arch_features): Ditto.
4595 (PTA_SKYLAKE): Add SGX.
4596 (PTA_ICELAKE): Remove.
4597 (PTA_ICELAKE_CLIENT): New.
4598 (PTA_ICELAKE_SERVER): New.
4599 (ix86_option_override_internal): Split up icelake on icelake client and
4601 (get_builtin_code_for_version): Ditto.
4602 (fold_builtin_cpu): Ditto.
4603 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4604 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4605 * config/i386/i386.h (processor_type): Ditto.
4606 * doc/invoke.texi: Ditto.
4608 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4611 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4612 INTEGER_CST offset, add it together with bitpos / 8 and
4613 sign extend based on POINTER_SIZE.
4617 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4620 * config/i386/constraints.md (Yc): New register constraint.
4621 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4622 Use Yc constraint for alternative 2 of operand 0. Remove
4623 preferred_for_speed attribute.
4625 2018-03-14 Richard Biener <rguenther@suse.de>
4627 PR tree-optimization/84830
4628 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4629 with the old one to avoid oscillations.
4631 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
4634 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4636 (assign_by_spills): Return a flag of reload assignment failure.
4637 Do not process the reload assignment failures. Do not spill other
4638 reload pseudos if they has the same reg class. Update n if
4640 (lra_assign): Add a return arg. Set up from the result of
4641 assign_by_spills call.
4642 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4643 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4644 usage_insns if it is not NULL.
4645 (spill_hard_reg_in_range): New function.
4646 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4647 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4648 function prototypes.
4649 (lra_assign): Change prototype.
4650 * lra.c (lra): Add code to deal with fails by splitting hard reg
4653 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
4655 * config/riscv/riscv.opt (mrelax): New option.
4656 * config/riscv/riscv.c (riscv_file_start): Emit ".option
4657 "norelax" when riscv_mrelax is disabled.
4658 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
4660 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4663 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
4664 reassociation for int modes.
4666 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4668 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
4669 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
4671 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
4672 * config/aarch64/aarch64-sve.md
4673 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4674 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4675 (*extend<mode><Vwide>2): Rename to...
4676 (aarch64_sve_extend<mode><Vwide>2): ...this.
4677 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4678 renaming the old pattern to...
4679 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
4681 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4682 define_expand, renaming the old pattern to...
4683 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4684 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4685 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4686 account when deciding which SVE instruction the optab should use.
4687 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4689 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4691 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4692 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4693 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4694 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4695 (tlsdesc_small_<mode>): Turn a define_expand and use
4696 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
4697 (tlsdesc_small_advsimd_<mode>): ...this.
4698 (tlsdesc_small_sve_<mode>): New pattern.
4700 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4702 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
4703 (UNSPEC_UMUL_HIGHPART): New constants.
4704 (MUL_HIGHPART): New int iteraor.
4705 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
4706 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
4708 (*<su>mul<mode>3_highpart): New define_insn.
4710 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4713 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4716 2018-03-13 Martin Liska <mliska@suse.cz>
4719 * (sem_item_optimizer::sem_item_optimizer): Initialize new
4721 (sem_item_optimizer::~sem_item_optimizer): Release it.
4722 (sem_item_optimizer::merge_classes): Register variable aliases.
4723 (sem_item_optimizer::fixup_pt_set): New function.
4724 (sem_item_optimizer::fixup_points_to_sets): Likewise.
4725 * ipa-icf.h: Declare new variables and functions.
4727 2018-03-13 Jakub Jelinek <jakub@redhat.com>
4730 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
4731 integer_pow2p@2 and test integer_pow2p in condition.
4732 (A < 0 ? C : 0): Similarly for @1.
4735 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
4736 characters starting at p contain '\0' character, don't look beyond
4740 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
4741 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
4744 * reg-stack.c (change_stack): Change update_end var from int to
4745 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
4746 also call set_block_for_insn on the newly added insns and rescan.
4749 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
4750 on the last operand.
4753 * tree.c (stabilize_reference_1): Return save_expr (e) for
4754 STATEMENT_LIST even if it doesn't have side-effects.
4756 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
4758 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
4760 2018-03-12 Renlin Li <renlin.li@arm.com>
4762 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
4763 aarch64_output_scalar_simd_mov_immediate.
4765 2018-03-12 Martin Sebor <msebor@redhat.com>
4767 PR tree-optimization/83456
4768 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
4769 for perfectly overlapping calls to memcpy.
4770 (gimple_fold_builtin_memory_chk): Same.
4771 (gimple_fold_builtin_strcpy): Handle no-warning.
4772 (gimple_fold_builtin_stxcpy_chk): Same.
4773 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
4775 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4777 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
4778 parameter. Use it for SFmode.
4779 (rs6000_function_arg_advance_1): Adjust.
4780 (rs6000_function_arg): Adjust.
4781 (rs6000_gimplify_va_arg): Pass false for that new parameter.
4783 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4785 PR rtl-optimization/84169
4786 PR rtl-optimization/84780
4787 * combine.c (can_combine_p): Check for a 2-insn combination whether
4788 the destination register is used between the two insns, too.
4790 2018-03-12 Richard Biener <rguenther@suse.de>
4792 PR tree-optimization/84803
4793 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
4794 for refs DR analysis didn't process.
4796 2018-03-12 Richard Biener <rguenther@suse.de>
4798 PR tree-optimization/84777
4799 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
4800 force-vectorize loops ignore whether we are optimizing for size.
4802 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
4804 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
4805 (TARGET_MD_ASM_ADJUST): Define.
4807 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
4808 Kito Cheng <kito.cheng@gmail.com>
4809 Chung-Ju Wu <jasonwucj@gmail.com>
4811 * config/nds32/nds32.c (nds32_compute_stack_frame,
4812 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
4813 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
4814 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
4815 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
4816 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
4817 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
4818 * config/nds32/nds32.md (prologue, epilogue): Use macro
4819 NDS32_V3PUSH_AVAILABLE_P to do checking.
4821 2018-03-11 Jakub Jelinek <jakub@redhat.com>
4824 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
4825 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
4826 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
4827 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
4828 addition of most attributes on !orig_type_die or the attribute not
4829 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
4831 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4832 Chung-Ju Wu <jasonwucj@gmail.com>
4834 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
4836 * config/nds32/nds32.opt (mvh): New option.
4838 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4839 Chung-Ju Wu <jasonwucj@gmail.com>
4841 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
4843 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
4844 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
4847 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4848 Chung-Ju Wu <jasonwucj@gmail.com>
4850 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
4852 * config/nds32/nds32-multiple.md (strlensi): New pattern.
4853 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
4855 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
4856 Kito Cheng <kito.cheng@gmail.com>
4857 Chung-Ju Wu <jasonwucj@gmail.com>
4859 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
4860 UNSPEC_FFMISM and UNSPEC_FLMISM.
4861 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
4862 for ffb, ffmism and flmism.
4863 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
4864 (unspec_ffmism): Ditto.
4865 (unspec_flmism): Ditto.
4866 (nds32_expand_builtin_impl): Check if string extension is available.
4867 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
4868 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
4870 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
4873 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4876 * lra-assigns.c (assign_by_spills): Return a flag of reload
4877 assignment failure. Do not process the reload assignment
4878 failures. Do not spill other reload pseudos if they has the same
4880 (lra_assign): Add a return arg. Set up from the result of
4881 assign_by_spills call.
4882 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4883 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4884 usage_insns if it is not NULL.
4885 (spill_hard_reg_in_range): New function.
4886 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4887 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4888 function prototypes.
4889 (lra_assign): Change prototype.
4890 * lra.c (lra): Add code to deal with fails by splitting hard reg
4893 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
4896 * config/i386/i386.opt: Replace Enforcment with Enforcement.
4898 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
4901 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
4902 (dw_val_node): Add val_symbolic_view.
4903 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
4904 (symview_upper_bound): New.
4905 (new_line_info_table): Initialize symviews_since_reset.
4906 (dwarf2out_source_line): Count symviews_since_reset and set
4907 symview_upper_bound.
4908 (dw_val_equal_p): Handle symview.
4909 (add_AT_symview): New.
4910 (print_dw_val): Handle symview.
4911 (attr_checksum, attr_checksum_ordered): Likewise.
4912 (same_dw_val_p, size_of_die): Likewise.
4913 (value_format, output_die): Likewise.
4914 (add_high_low_attributes): Use add_AT_symview for entry_view.
4915 (dwarf2out_finish): Reset symview_upper_bound, clear
4918 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
4921 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
4922 Add strict argument and use it.
4923 (rs6000_split_multireg_move): Update for new strict argument.
4924 (mem_operand_gpr): Disallow all non-offsettable addresses.
4925 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
4927 2018-03-09 Jakub Jelinek <jakub@redhat.com>
4930 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
4931 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
4932 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
4935 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
4936 decl, use remap_type if we want to use the type.
4938 2018-03-09 Martin Sebor <msebor@redhat.com>
4940 PR tree-optimization/84526
4941 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4943 (builtin_access::generic_overlap): Be prepared to handle non-array
4946 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
4948 PR rtl-optimization/84682
4949 * lra-constraints.c (process_address_1): Check is_address flag
4950 for address constraints.
4951 (process_alt_operands): Likewise.
4952 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
4953 preprocess_constraints.
4954 * recog.h (preprocess_constraints): Add oploc parameter.
4956 * recog.c (preprocess_constraints): Test address_operand for
4957 CT_ADDRESS constraints.
4959 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4962 * lra-assigns.c (assign_by_spills): Return a flag of reload
4963 assignment failure. Do not process the reload assignment
4964 failures. Do not spill other reload pseudos if they has the same
4966 (lra_assign): Add a return arg. Set up from the result of
4967 assign_by_spills call.
4968 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4969 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4970 usage_insns if it is not NULL.
4971 (spill_hard_reg_in_range): New function.
4972 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4973 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4974 function prototypes.
4975 (lra_assign): Change prototype.
4976 * lra.c (lra): Add code to deal with fails by splitting hard reg
4979 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4982 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
4983 Accept complain bool parameter. Only emit errors if it is true.
4984 (arm_parse_cpu_option_name): Likewise.
4985 (arm_target_thumb_only): Adjust callers of the above.
4986 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
4987 prototype to take a default true bool parameter.
4988 (arm_parse_arch_option_name): Likewise.
4990 2018-03-09 David Malcolm <dmalcolm@redhat.com>
4991 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
4995 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
4996 * configure: Regenerate.
4997 * configure.ac ("linker --version-script option"): New.
4998 ("linker soname option"): New.
5000 2018-03-09 Richard Biener <rguenther@suse.de>
5002 PR tree-optimization/84775
5003 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
5004 immediate uses of predicate stmts and mark them modified.
5007 PR tree-optimization/84178
5008 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5010 (version_loop_for_if_conversion): Delay update_ssa call.
5011 (tree_if_conversion): Delay update_ssa until after predicate
5014 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
5017 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
5018 when the function accesses prior frames.
5020 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5023 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
5024 gen_llsym, otherwise call maybe_gen_llsym.
5027 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
5028 has ',' character inside of it.
5030 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5033 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
5034 as clobbering CC_REGNUM.
5036 2018-03-08 Richard Biener <rguenther@suse.de>
5039 * tree-scalar-evolution.c: Include tree-into-ssa.h.
5040 (follow_copies_to_constant): Do not follow SSA names registered
5043 2018-03-08 Richard Biener <rguenther@suse.de>
5045 PR tree-optimization/84178
5046 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5048 (version_loop_for_if_conversion): Delay update_ssa call.
5049 (tree_if_conversion): Delay update_ssa until after predicate
5052 2018-03-08 David Malcolm <dmalcolm@redhat.com>
5054 PR tree-optimization/84178
5055 * tree-if-conv.c (release_bb_predicate): Remove the
5056 the assertion that the stmts have NULL use_ops.
5057 Discard the statements, asserting that they haven't
5058 yet been added to a BB.
5060 2018-03-08 Richard Biener <rguenther@suse.de>
5062 PR tree-optimization/84746
5063 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
5064 (phi_translate): Pass in destination ANTIC_OUT set.
5065 (phi_translate_1): Likewise. For a simplified result lookup
5066 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
5067 (phi_translate_set): Adjust.
5068 (do_pre_regular_insertion): Likewise.
5069 (do_pre_partial_partial_insertion): Likewise.
5071 2018-03-08 Martin Liska <mliska@suse.cz>
5073 PR gcov-profile/84735
5074 * doc/gcov.texi: Document usage of profile files.
5075 * gcov-io.h: Document changes in the format.
5077 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
5081 * dwarf2out.c (struct dw_line_info_table): Update comments for
5083 (FORCE_RESET_NEXT_VIEW): New.
5084 (FORCE_RESETTING_VIEW_P): New.
5085 (RESETTING_VIEW_P): Check for -1 too.
5086 (ZERO_VIEW_P): Likewise.
5087 (new_line_info_table): Force-reset next view.
5088 (dwarf2out_begin_function): Likewise.
5089 (dwarf2out_source_line): Simplify zero_view_p initialization.
5090 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
5091 view directly. Omit view when omitting .loc at line 0.
5093 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5095 PR tree-optimization/84740
5096 * tree-switch-conversion.c (process_switch): Call build_constructors
5097 only if info.phi_count is non-zero.
5099 PR tree-optimization/84739
5100 * tree-tailcall.c (find_tail_calls): Check call arguments against
5101 DECL_ARGUMENTS (current_function_decl) rather than
5102 DECL_ARGUMENTS (func) when checking for tail recursion.
5104 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5106 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
5107 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
5108 Volker Reichelt's entry and add entries for people that perform
5109 GCC fuzzy testing and report numerous bugs.
5111 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
5114 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
5115 readonly data in sdata, if that is disabled.
5116 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
5117 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
5118 -mreadonly-in-sdata option.
5120 2018-03-07 Martin Sebor <msebor@redhat.com>
5122 PR tree-optimization/84468
5123 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
5124 basic block when looking for nul assignment.
5126 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
5129 * except.h (output_function_exception_table): Adjust prototype.
5130 * except.c (output_function_exception_table): Remove FNNAME parameter
5131 and add SECTION parameter. Ouput one part of the table at a time.
5132 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
5133 the first part of the exception table and emit unwind directives.
5134 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
5135 (i386_pe_seh_cold_init): Likewise.
5136 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
5137 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
5138 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
5139 (ix86_output_call_insn): Emit a nop in one more case for SEH.
5140 * config/i386/winnt.c: Include except.h.
5141 (struct seh_frame_state): Add reg_offset, after_prologue and
5142 in_cold_section fields.
5143 (i386_pe_seh_end_prologue): Set seh->after_prologue.
5144 (i386_pe_seh_cold_init): New function.
5145 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
5146 to seh->in_cold_section.
5147 (seh_emit_push): Record the offset of the push.
5148 (seh_emit_save): Record the offet of the save.
5149 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
5150 Test seh->after_prologue to disregard the epilogue.
5151 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
5152 (i386_pe_end_cold_function): New function.
5154 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5157 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
5158 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
5161 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
5163 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
5164 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
5167 * multiple_target.c: Include tree-inline.h and intl.h.
5168 (expand_target_clones): Diagnose and fail if node->definition and
5169 !tree_versionable_function_p (node->decl).
5171 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
5173 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
5175 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
5176 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
5177 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
5179 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5182 * combine.c (try_combine): Use reg_or_subregno instead of handling
5183 just paradoxical SUBREGs and REGs.
5185 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
5187 * config/arc/arc.c (arc_finalize_pic): Remove function.
5188 (arc_must_save_register): We use single base PIC register, remove
5189 checks to save/restore the PIC register.
5190 (arc_expand_prologue): Likewise.
5191 * config/arc/arc-protos.h (arc_set_default_type_attributes):
5193 (arc_verify_short): Likewise.
5194 (arc_attr_type): Likewise.
5195 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
5196 (walk_stores): Likewise.
5197 (arc_address_cost): Make it static.
5198 (arc_verify_short): Likewise.
5199 (branch_dest): Likewise.
5200 (arc_attr_type): Likewise.
5201 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
5202 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
5203 (arc_final_prescan_insn): Remove inserting the nops due to
5204 hardware hazards. It is done in reorg step.
5205 (insn_length_variant_t): Remove.
5206 (insn_length_parameters_t): Likewise.
5207 (arc_insn_length_parameters): Likewise.
5208 (arc_get_insn_variants): Likewise.
5209 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
5211 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5214 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
5217 PR tree-optimization/84687
5218 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
5220 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
5222 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5224 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
5225 Rename to ppc_speculation_barrier.
5226 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
5227 __builtin_ppc_speculation_barrier.
5229 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5232 * combine.c (combine_simplify_rtx): Don't try to simplify if
5233 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
5236 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
5238 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
5239 to 32 bytes when compiling for POWER9.
5241 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5244 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
5245 regparm >= 3 with no arg reg available also for calls with
5246 flag_force_indirect_call. Pass decl to ix86_function_regparm.
5249 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
5251 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
5253 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
5256 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
5258 2018-03-05 Richard Biener <rguenther@suse.de>
5260 PR tree-optimization/84486
5261 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
5262 When inserting a __builtin_assume_aligned call set the LHS
5263 SSA name alignment info accordingly.
5265 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
5267 PR tree-optimization/84114
5268 * config/aarch64/aarch64.c (aarch64_reassociation_width)
5269 Avoid reassociation of FLOAT_MODE addition.
5271 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
5273 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
5274 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
5275 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
5276 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
5277 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
5278 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
5279 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
5281 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
5282 __builtin_ia32_wbinvd): New builtins.
5283 (SPECIAL_ARGS2): New.
5284 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
5285 (SPECIAL_ARGS2): New.
5286 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
5287 (ix86_valid_target_attribute_inner_p): Ditto.
5288 (ix86_init_mmx_sse_builtins): Add special_args2.
5289 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
5290 TARGET_WBNOINVD_P): New.
5291 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
5292 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
5293 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
5294 * config/i386/immintrin.h (_wbinvd): New intrinsic.
5295 * config/i386/pconfigintrin.h: New file.
5296 * config/i386/wbnoinvdintrin.h: Ditto.
5297 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
5298 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
5300 2018-03-05 Richard Biener <rguenther@suse.de>
5302 PR tree-optimization/84670
5303 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5305 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5306 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5307 (compute_antic_aux): Only assert the number of values in ANTIC_IN
5308 doesn't grow if all successors (recursively) were visited at least
5311 2018-03-05 Richard Biener <rguenther@suse.de>
5313 PR tree-optimization/84650
5314 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
5315 if executed in the loop pipeline.
5317 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
5319 * doc/configfiles.texi (Configuration Files): Move info about
5320 conditionalizing $target-protos.h to...
5321 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
5322 differs from $target-protos.h.
5324 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
5325 Chung-Ju Wu <jasonwucj@gmail.com>
5327 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
5328 * config/nds32/nds32-multiple.md (setmemsi): Define.
5329 * config/nds32/nds32-memory-manipulation.c
5330 (nds32_gen_dup_4_byte_to_word_value): New.
5331 (emit_setmem_word_loop): New.
5332 (emit_setmem_byte_loop): New.
5333 (nds32_expand_setmem_loop): New.
5334 (nds32_expand_setmem_loop_v3m): New.
5335 (nds32_expand_setmem_unroll): New.
5336 (nds32_expand_setmem): New.
5338 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5339 Chung-Ju Wu <jasonwucj@gmail.com>
5341 * config/nds32/nds32-memory-manipulation.c
5342 (nds32_emit_load_store): New.
5343 (nds32_emit_post_inc_load_store): New.
5344 (nds32_emit_mem_move): New.
5345 (nds32_emit_mem_move_block): New.
5346 (nds32_expand_movmemsi_loop_unknown_size): New.
5347 (nds32_expand_movmemsi_loop_known_size): New.
5348 (nds32_expand_movmemsi_loop): New.
5349 (nds32_expand_movmemsi_unroll): New.
5350 (nds32_expand_movmemqi): Rename ...
5351 (nds32_expand_movmemsi): ... to this.
5352 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
5353 (movmemsi): ... to this.
5354 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
5355 (nds32_expand_movmemsi): ... to this.
5357 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5358 Monk Chiang <sh.chiang04@gmail.com>
5359 Chung-Ju Wu <jasonwucj@gmail.com>
5361 * config/nds32/nds32-protos.h
5362 (nds32_expand_load_multiple): New arguments.
5363 (nds32_expand_store_multiple): Ditto.
5364 (nds32_valid_multiple_load_store): Rename ...
5365 (nds32_valid_multiple_load_store_p): ... to this.
5366 * config/nds32/nds32-memory-manipulation.c
5367 (nds32_expand_load_multiple): Refine implementation.
5368 (nds32_expand_store_multiple): Ditto.
5369 * config/nds32/nds32-multiple.md
5370 (load_multiple): Update nds32_expand_load_multiple interface.
5371 (store_multiple): Update nds32_expand_store_multiple interface.
5372 * config/nds32/nds32-predicates.c
5373 (nds32_valid_multiple_load_store): Rename ...
5374 (nds32_valid_multiple_load_store_p): ... to this and refine
5376 * config/nds32/predicates.md
5377 (nds32_load_multiple_and_update_address_operation): New predicate.
5378 (nds32_store_multiple_and_update_address_operation): New predicate.
5380 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5381 Chung-Ju Wu <jasonwucj@gmail.com>
5383 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
5384 (combo): New attribute.
5385 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
5387 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
5389 * config/nds32/nds32.opt: Change -mcmodel= default value.
5391 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
5392 Monk Chiang <sh.chiang04@gmail.com>
5393 Chung-Ju Wu <jasonwucj@gmail.com>
5395 * config/nds32/constants.md (unspec_element): New enum.
5396 * config/nds32/constraints.md (Umw): New constraint.
5397 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
5398 * config/nds32/nds32-intrinsic.md: Likewise.
5399 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
5400 (nds32_valid_smw_lwm_base_p): New.
5401 (nds32_output_smw_single_word): New.
5402 (nds32_output_lmw_single_word): New.
5403 (nds32_expand_unaligned_load): New.
5404 (nds32_expand_unaligned_store): New.
5405 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5406 (nds32_output_smw_single_word): Declare.
5407 (nds32_output_lmw_single_word): Declare.
5408 (nds32_expand_unaligned_load): Declare.
5409 (nds32_expand_unaligned_store): Declare.
5410 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5411 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5412 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5413 NDS32_BUILTIN_UASTORE_DW.
5414 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5417 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
5418 Kito Cheng <kito.cheng@gmail.com>
5419 Chung-Ju Wu <jasonwucj@gmail.com>
5421 * config/nds32/nds32-intrinsic.c
5422 (nds32_expand_builtin_null_ftype_reg): Delete.
5423 (nds32_expand_builtin_reg_ftype_imm): Ditto.
5424 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5425 (nds32_read_argument): New.
5426 (nds32_legitimize_target): Ditto.
5427 (nds32_legitimize_argument): Ditto.
5428 (nds32_check_constant_argument): Ditto.
5429 (nds32_expand_unop_builtin): Ditto.
5430 (nds32_expand_unopimm_builtin): Ditto.
5431 (nds32_expand_binop_builtin): Ditto.
5432 (nds32_builtin_decl_impl): Ditto.
5433 (builtin_description): Ditto.
5434 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5435 (nds32_init_builtins_impl): Ditto.
5436 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5437 (nds32_builtin_decl): New.
5438 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5439 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5441 2018-03-02 Jeff Law <law@redhat.com>
5443 * reorg.c (stop_search_p): Handle DEBUG_INSN.
5444 (redundant_insn, fill_simple_delay_slots): Likewise.
5445 (fill_slots_from_thread): Likewise.
5446 * resource.c (mark_referenced_resources): Likewise.
5447 (mark_set_resources, find_dead_or_set_registers): Likewise.
5449 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5451 * substring-locations.h (format_warning_va): Formatting fix for
5453 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5455 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5456 * substring-locations.c: Include intl.h.
5457 (format_warning_va): Turned into small wrapper around
5458 format_warning_n_va, renamed to ...
5459 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5460 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5462 (format_warning_at_substring_n): New function.
5463 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5464 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
5465 format_warning_at_substring with just a shorter name instead of
5466 const function pointer.
5467 (fmtwarn_n): New function.
5468 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5469 appropriate, get rid of all the fmtstr temporaries, move conditionals
5470 with G_() wrapped string literals directly into fmtwarn arguments,
5471 cast dir.len to (int), formatting fixes.
5473 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
5475 * doc/invoke.texi: Remove "Cilk Plus" references.
5477 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5478 Richard Biener <rguenther@suse.de>
5481 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5482 for error or warning attributes if CALL_FROM_THUNK_P is set.
5485 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5488 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5489 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5492 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5493 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5495 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5496 instead of -1U in last predictors element's probability member.
5498 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5501 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5502 arguments if they are comparable.
5504 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5506 PR tree-optimization/84634
5507 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5508 masks and masked_loop_p with a single loop_masks, making sure it's
5509 null for bb vectorization.
5511 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5513 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5514 (vect_analyze_data_ref_access): Use loop->safe_len rather than
5515 loop->force_vectorize to check whether there is no alias.
5517 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5520 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5522 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5524 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5525 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5526 instead of a loop around prev_real_insn.
5527 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5531 * config/i386/i386.c (ix86_print_operand): Use conditional
5532 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5535 2018-03-02 Richard Biener <rguenther@suse.de>
5537 PR tree-optimization/84427
5538 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5539 (bitmap_set_subtract_values): Rewrite to handle multiple
5542 (prune_clobbered_mems): Likewise.
5543 (phi_translate): Take edge instead of pred/phiblock.
5544 (phi_translate_1): Likewise.
5545 (phi_translate_set): Likewise. Insert all translated
5546 exprs for a value into the set, keeping possibly multiple
5547 expressions per value.
5548 (compute_antic_aux): Adjust for phi_translate changes.
5549 When intersecting union the expressions and prune those
5550 not in the final value set, keeping possibly multiple
5551 expressions per value. Do not use value-insertion
5552 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5553 all expressions. Add verification that the value-sets
5554 only shrink during iteration.
5555 (compute_partial_antic_aux): Adjust for the phi_translate changes.
5556 (do_pre_regular_insertion): Likewise.
5557 (do_pre_partial_partial_insertion): Likewise.
5559 2018-03-02 Richard Biener <rguenther@suse.de>
5562 * config/darwin.c (saved_debug_info_level): New static global.
5563 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5564 (darwin_asm_lto_end): Restore debug info generation settings.
5566 2018-03-01 Martin Liska <mliska@suse.cz>
5569 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5572 2018-03-01 Richard Biener <rguenther@suse.de>
5575 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5576 type annotation with LTO when debug was disabled at compile-time.
5578 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5580 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5582 (mips_final_postscan_insn): Likewise.
5584 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5586 PR rtl-optimization/84528
5587 * alias.c (init_alias_target): Add commentary.
5588 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5589 a unique base value if the frame pointer is not eliminated
5590 to the stack pointer.
5592 2018-03-01 Tom de Vries <tom@codesourcery.com>
5594 PR rtl-optimization/83327
5595 * lra-int.h (hard_regs_spilled_into): Declare.
5596 * lra.c (hard_regs_spilled_into): Define.
5597 (init_reg_info): Init hard_regs_spilled_into.
5598 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5599 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5600 (process_bb_lives): Handle hard_regs_spilled_into.
5601 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5602 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5604 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5606 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5607 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5608 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5609 * config/rs6000/aix72.h: New file.
5611 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5613 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5614 instead of warning_at with conditional singular and plural messages
5618 * stor-layout.c (update_alignment_for_field): For
5619 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5620 && !DECL_PACKED (field), do the alignment update, just use
5621 only desired_align instead of MAX (type_align, desired_align)
5623 (place_field): Don't do known_align < desired_align handling
5624 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5625 is non-NULL, instead do it after rli->prev_field handling and
5626 only if not within a bitfield word. For DECL_PACKED (field)
5627 use type_align of BITS_PER_UNIT.
5629 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5631 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5632 superfluous parentheses and trailing spaces.
5634 2018-02-28 Richard Biener <rguenther@suse.de>
5636 PR tree-optimization/84584
5637 * graphite-scop-detection.c (scop_detection::add_scop): Discard
5638 SCoPs with fake exit edge.
5640 2018-02-28 Martin Liska <mliska@suse.cz>
5643 * timevar.c (timer::print): Fix format to properly print 100%
5646 2018-02-28 Richard Biener <rguenther@suse.de>
5649 * genmatch.c (capture_info::walk_match): Do not mark
5650 captured expressions without operands as expr_p given
5651 they act more like predicates and should be subject to
5652 "lost tail" side-effect preserving.
5654 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
5656 PR rtl-optimization/81611
5657 * auto-inc-dec.c (attempt_change): Move dead note from
5658 mem_insn if it's the next use of regno
5659 (find_address): Take address use of reg holding
5660 non-incremented value. Add parm to limit search to the named
5662 (merge_in_block): Attempt to use a mem insn that is the next
5663 use of the original regno.
5665 2018-02-27 Martin Sebor <msebor@redhat.com>
5668 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
5669 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
5671 2018-02-27 Martin Sebor <msebor@redhat.com>
5673 PR translation/84207
5674 * diagnostic-core.h (warning_n, error_n, inform_n): Change
5675 n argument to unsigned HOST_WIDE_INT.
5676 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5677 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
5678 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5679 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5681 2018-02-27 Richard Biener <rguenther@suse.de>
5683 PR tree-optimization/84512
5684 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5685 Do not use the estimate returned from record_stmt_cost for
5686 the scalar iteration cost but sum properly using add_stmt_cost.
5688 2018-02-27 Richard Biener <rguenther@suse.de>
5690 PR tree-optimization/84466
5691 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5692 Adjust last change to less strictly validate use operands.
5694 2018-02-27 Martin Liska <mliska@suse.cz>
5696 PR gcov-profile/84548
5697 * gcov.c (process_file): Allow partial overlap and consider it
5698 also as group functions.
5699 (output_lines): Properly calculate range of lines for a group.
5701 2018-02-27 Martin Liska <mliska@suse.cz>
5703 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
5704 'ggc' suffixes. Change first column width.
5705 (timer::print): Fix formatting of the column.
5707 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
5709 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
5710 preserve inline entry blocks for the sake of debug inline
5711 entry point markers alone.
5712 (remove_unused_locals): Suggest in comments a better place to
5713 force the preservation of inline entry blocks that are
5714 otherwise unused, but do not preserve them.
5716 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5718 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
5720 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5723 * config/i386/constraints.md (Bs): Replace
5724 ix86_indirect_branch_register with
5725 TARGET_INDIRECT_BRANCH_REGISTER.
5727 * config/i386/i386.md (indirect_jump): Likewise.
5728 (tablejump): Likewise.
5729 (*sibcall_memory): Likewise.
5730 (*sibcall_value_memory): Likewise.
5731 Peepholes of indirect call and jump via memory: Likewise.
5732 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
5733 (*sibcall_value_GOT_32): Likewise.
5734 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5735 (GOT_memory_operand): Likewise.
5736 (call_insn_operand): Likewise.
5737 (sibcall_insn_operand): Likewise.
5738 (GOT32_symbol_operand): Likewise.
5739 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
5741 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5743 PR rtl-optimization/83496
5744 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
5745 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
5746 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
5747 redundant insn, if any.
5748 (relax_delay_slots): Likewise.
5749 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
5751 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
5753 PR tree-optimization/83965
5754 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
5755 that grouped statements are part of a reduction chain. Return
5756 true if the statement is not marked as a reduction itself but
5758 (vect_recog_dot_prod_pattern): Don't check whether the statement
5759 is part of a group here.
5760 (vect_recog_sad_pattern): Likewise.
5761 (vect_recog_widen_sum_pattern): Likewise.
5763 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5766 * final.c (rest_of_clean_state): Also look for calls inside sequences.
5768 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5771 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
5773 (ix86_output_indirect_function_return): New prototype.
5774 (ix86_split_simple_return_pop_internal): Likewise.
5775 * config/i386/i386.c (indirect_return_via_cx): New.
5776 (indirect_return_via_cx_bnd): Likewise.
5777 (indirect_thunk_name): Handle return va CX_REG.
5778 (output_indirect_thunk_function): Create alias for
5779 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
5780 (ix86_output_indirect_jmp): Remove the bool argument.
5781 (ix86_output_indirect_function_return): New function.
5782 (ix86_split_simple_return_pop_internal): Likewise.
5783 * config/i386/i386.md (*indirect_jump): Don't pass false
5784 to ix86_output_indirect_jmp.
5785 (*tablejump_1): Likewise.
5786 (simple_return_pop_internal): Change it to define_insn_and_split.
5787 Call ix86_split_simple_return_pop_internal to split it for
5789 (simple_return_indirect_internal): Call
5790 ix86_output_indirect_function_return instead of
5791 ix86_output_indirect_jmp.
5793 2018-02-26 Jakub Jelinek <jakub@redhat.com>
5796 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
5797 memset and value initialization afterwards.
5799 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
5801 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
5803 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5806 * common/config/aarch64/aarch64-common.c
5807 (aarch_option_optimization_table[]): Switch
5808 off fomit-frame-pointer
5810 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5811 Chung-Ju Wu <jasonwucj@gmail.com>
5813 * config/nds32/nds32-multiple.md (load_multiple): Disallow
5815 (store_multiple): Ditto.
5817 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5819 * config.gcc: Add --with-cpu support for nds32 target.
5820 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
5821 * config/nds32/nds32.opt: Add -mcpu= option.
5823 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
5825 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
5826 isel=yes): Warn for these deprecated options.
5828 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
5830 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
5831 ISA_2_5_MASKS_EMBEDDED.
5833 2018-02-23 Jakub Jelinek <jakub@redhat.com>
5835 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
5836 p->max as pointers rather than using iterative_hash_expr.
5838 2018-02-23 Carl Love <cel@us.ibm.com>
5840 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
5841 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
5843 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
5844 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
5845 P8V_BUILTIN_VEC_VUNSIGNED2.
5847 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
5850 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
5851 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
5853 * lra-constraints.c (curr_insn_transform): Set up
5854 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
5856 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
5857 * lra-eliminations.c (spill_pseudos): Ditto.
5858 (process_insn_for_elimination): Ditto.
5859 * lra-lives.c (reg_early_clobber_p): Use the new macros.
5860 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
5861 LRA_NON_CLOBBERED_ALT.
5863 2018-02-22 Martin Sebor <msebor@redhat.com>
5865 PR tree-optimization/84480
5866 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
5867 to maybe_diag_stxncpy_trunc. Call it.
5868 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
5869 from gimple_fold_builtin_strcpy. Print inlining stack.
5870 (handle_builtin_stxncpy): Print inlining stack.
5871 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
5873 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
5876 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
5877 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
5878 and -fcheck-pointer-bounds are used together.
5879 (indirect_thunk_prefix): New enum.
5880 (indirect_thunk_need_prefix): New function.
5881 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
5882 "_nt" instead of "_bnd" for NOTRACK prefix.
5883 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
5884 (output_indirect_thunk_function): Likewise.
5886 (ix86_code_end): Update output_indirect_thunk_function calls.
5887 (ix86_output_indirect_branch_via_reg): Replace
5888 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
5889 (ix86_output_indirect_branch_via_push): Likewise.
5890 (ix86_output_function_return): Likewise.
5891 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
5892 incompatible with -fcf-protection=branch and
5893 -fcheck-pointer-bounds.
5895 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5898 * config/aarch64/aarch64.c (aarch64_print_address_internal):
5899 Change gcc_assert call to output_operand_lossage.
5901 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5903 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
5905 2018-02-22 DJ Delorie <dj@redhat.com>
5906 Sebastian Perta <sebastian.perta@renesas.com>
5907 Oleg Endo <olegendo@gcc.gnu.org>
5909 * config/rx/rx.c (rx_rtx_costs): New function.
5910 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
5912 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
5914 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
5916 2018-02-22 Martin Liska <mliska@suse.cz>
5919 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
5920 Add "native" as a possible value.
5922 2018-02-22 Martin Liska <mliska@suse.cz>
5925 * config/i386/i386.c (ix86_option_override_internal):
5926 Add "native" as a possible value for -march and -mtune.
5928 2018-02-22 Jakub Jelinek <jakub@redhat.com>
5931 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
5932 to all type variants.
5934 PR tree-optimization/84503
5935 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
5936 width as info->bitpos + info->bitsize - start.
5937 (merged_store_group::merge_overlapping): Simplify width computation.
5938 (check_no_overlap): New function.
5939 (imm_store_chain_info::try_coalesce_bswap): Compute expected
5940 start + width and last_order of the group, fail if check_no_overlap
5942 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
5943 to group if check_no_overlap fails.
5945 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5947 * config/rs6000/altivec.md: Delete contraint arguments to
5948 define_expand, define_split, and define_peephole2, and in
5949 define_insn_and_split if always unused.
5950 * config/rs6000/darwin.md: Ditto.
5951 * config/rs6000/dfp.md: Ditto.
5952 * config/rs6000/rs6000.md: Ditto.
5953 * config/rs6000/sync.md: Ditto.
5954 * config/rs6000/vector.md: Ditto.
5955 * config/rs6000/vsx.md: Ditto.
5957 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5959 * config/rs6000/altivec.md: Write output control strings as braced
5960 blocks instead of double-quoted strings.
5961 * config/rs6000/darwin.md: Ditto.
5962 * config/rs6000/rs6000.md: Ditto.
5963 * config/rs6000/vector.md: Ditto.
5964 * config/rs6000/vsx.md: Ditto.
5966 2018-02-21 Jason Merrill <jason@redhat.com>
5968 PR c++/84314 - ICE with templates and fastcall attribute.
5969 * attribs.c (build_type_attribute_qual_variant): Remove assert.
5971 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5973 * ipa-cp.c (determine_versionability): Fix comment typos.
5975 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5978 * ipa-cp.c (determine_versionability): Do not version functions caling
5981 2018-02-21 Martin Liska <mliska@suse.cz>
5984 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
5985 Add "native" as a possible value.
5986 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
5987 the macro when native cpu detection is available.
5989 2018-02-21 Martin Liska <mliska@suse.cz>
5992 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
5993 Add "native" as a possible value.
5994 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
5995 when native cpu detection is available.
5997 2018-02-21 Jakub Jelinek <jakub@redhat.com>
5998 Martin Sebor <msebor@redhat.com>
6000 PR tree-optimization/84478
6001 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
6003 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
6004 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
6005 support which is conservatively correct, for 2 only stay conservative
6006 for maxlen. Formatting and comment capitalization fixes. Add STRICT
6007 argument to the 2 argument get_range_strlen, adjust 6 arg
6008 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
6010 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
6011 (gimple_fold_builtin_strlen): Pass true as last argument to
6014 2018-02-20 Martin Sebor <msebor@redhat.com>
6017 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
6018 (builtin_memref::set_base_and_offset): Same. Handle inner references.
6019 (builtin_memref::builtin_memref): Factor out parts into
6020 set_base_and_offset and call it.
6022 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6025 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
6026 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
6027 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
6028 search at the associated MODE_INT.
6030 2018-02-20 Jeff Law <law@redhat.com>
6033 PR tree-optimization/81592
6035 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
6036 for range data rather than using global data.
6037 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
6038 range data rather than using global data.
6039 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
6040 pass it to children as needed.
6041 (struct directive::fmtresult): Similarly.
6042 (struct directive::set_width): Similarly.
6043 (struct directive::set_precision): Similarly.
6044 (format_integer, format_directive, parse_directive): Similarly.
6045 (format_none): Accept unnamed vr_values parameter.
6046 (format_percent, format_floating, format_character): Similarly.
6047 (format_string, format_plain): Similarly.
6048 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
6049 the EVRP range analyzer for range data rather than using global data.
6050 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
6051 gimple-ssa-evrp-analyze.h
6052 (class sprintf_dom_walker): Add after_dom_children member function.
6053 Add evrp_range_analyzer member.
6054 (sprintf_dom_walker::before_dom_children): Call into the EVRP
6055 range analyzer as needed.
6056 (sprintf_dom_walker::after_dom_children): New member function.
6057 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
6059 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6060 (evrp_range_analyzer::pop_to_marker): Likewise.
6062 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6064 PR tree-optimization/84419
6065 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
6066 with the required type if its current type is compatible but
6069 2018-02-20 Jakub Jelinek <jakub@redhat.com>
6072 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
6073 after vectorization.
6075 2018-02-20 Martin Liska <mliska@suse.cz>
6078 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
6079 possible values if we don't have a hint.
6081 2018-02-20 Martin Liska <mliska@suse.cz>
6085 * final.c (shorten_branches): Build align_tab array with one
6087 * opts.c (finish_options): Add alignment option limit check.
6088 (MAX_CODE_ALIGN): Likewise.
6089 (MAX_CODE_ALIGN_VALUE): Likewise.
6090 * doc/invoke.texi: Document maximum allowed option value for
6091 all -falign-* options.
6093 2018-02-19 Jakub Jelinek <jakub@redhat.com>
6096 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
6097 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
6098 * var-tracking.c (emit_note_insn_var_location): Remove all references
6099 to NOTE_INSN_CALL_ARG_LOCATION.
6100 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
6101 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
6102 Use copy_rtx_if_shared.
6103 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
6104 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
6105 (dwarf2out_var_location): Remove handling of
6106 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
6108 * final.c (final_scan_insn): Remove all references to
6109 NOTE_INSN_CALL_ARG_LOCATION.
6110 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
6111 before dumping final insns.
6112 * except.c (emit_note_eh_region_end): Remove all references to
6113 NOTE_INSN_CALL_ARG_LOCATION.
6114 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
6115 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
6116 * config/arc/arc.c (hwloop_optimize): Likewise.
6117 * config/arm/arm.c (create_fix_barrier): Likewise.
6118 * config/s390/s390.c (s390_chunkify_start): Likewise.
6119 * config/sh/sh.c (find_barrier): Likewise.
6120 * config/i386/i386.c (rest_of_insert_endbranch,
6121 ix86_seh_fixup_eh_fallthru): Likewise.
6122 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
6123 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
6124 * config/frv/frv.c (frv_function_prologue): Likewise.
6125 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
6127 (note_outside_basic_block_p): Remove all references to
6128 NOTE_INSN_CALL_ARG_LOCATION.
6129 * gengtype.c (adjust_field_rtx_def): Likewise.
6130 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
6132 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
6133 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
6136 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
6139 PR tree-optimization/84452
6140 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
6141 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
6144 2018-02-19 Martin Liska <mliska@suse.cz>
6147 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
6149 2018-02-19 Martin Liska <mliska@suse.cz>
6150 Richard Sandiford <richard.sandiford@linaro.org>
6152 PR tree-optimization/82491
6153 * gimple-fold.c (get_base_constructor): Make earlier bail out
6156 2018-02-19 Carl Love <cel@us.ibm.com>
6158 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
6159 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
6161 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
6162 P8V_BUILTIN_VEC_NEG.
6164 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
6166 * config/rl78/rl78.md (movdf): New define expand.
6168 2018-02-19 Martin Liska <mliska@suse.cz>
6171 * doc/invoke.texi: Fix typo.
6172 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
6174 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
6176 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
6177 handle rs6000_single_float and rs6000_double_float specially for
6180 2018-02-16 Jeff Law <law@redhat.com>
6182 * config/rx/rx.c (add_pop_cfi_notes): New function.;
6185 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6188 * ipa-inline.c (inline_small_functions): Fix a typo.
6190 2018-02-16 Nathan Sidwell <nathan@acm.org>
6192 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
6194 2018-02-16 Carl Love <cel@us.ibm.com>
6196 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
6197 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
6198 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
6199 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
6200 expansion to P8V_BUILTIN_VEC_FLOAT2.
6202 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
6204 PR rtl-optimization/70023
6205 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
6206 src_regno into account.
6208 2018-02-16 Carl Love <cel@us.ibm.com>
6210 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
6211 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
6212 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
6213 * config/rs6000/rs6000.c: Remove case statements for
6214 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
6215 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
6216 and P9V_BUILTIN_VEC_VINSERT4B.
6217 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
6218 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
6219 * config/rs6000/vsx.md:
6220 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
6223 2018-02-16 Carl Love <cel@us.ibm.com>
6225 * config/rs6000/altivec.h: Add builtin names vec_extract4b
6227 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
6229 * config/rs6000/rs6000-c.c: Add the definitions for
6230 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
6231 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
6232 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
6233 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
6234 definition for insert4b and define insn *insert3b_internal.
6235 * doc/extend.texi: Add documentation for vec_extract4b.
6237 2018-02-16 Nathan Sidwell <nathan@acm.org>
6239 * doc/extend.texi (Backwards Compatibility): Mention friend
6240 injection. Note for-scope is deprecated.
6241 * doc/invoke.texi (-ffriend-injection): Deprecate.
6243 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
6245 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
6246 that moved to I2, also allow destinations that are a paradoxical
6247 subreg (instead of a normal reg).
6249 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
6252 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
6255 2018-02-16 Richard Biener <rguenther@suse.de>
6257 PR tree-optimization/84037
6258 PR tree-optimization/84016
6260 * config/i386/i386.c (ix86_builtin_vectorization_cost):
6261 Adjust vec_construct for the fact we need additional higher latency
6262 128bit inserts for AVX256 and AVX512 vector builds.
6263 (ix86_add_stmt_cost): Scale vector construction cost for
6266 2018-02-16 Richard Biener <rguenther@suse.de>
6268 PR tree-optimization/84417
6269 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
6270 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
6271 (non_rewritable_lvalue_p): Likewise, use poly-ints.
6273 2018-02-16 Martin Liska <mliska@suse.cz>
6276 * internal-fn.def (ASAN_CHECK): Set proper flags.
6277 (ASAN_MARK): Likewise.
6279 2018-02-16 Julia Koval <julia.koval@intel.com>
6281 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
6282 from PTA_CANNONLAKE.
6284 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6287 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
6288 Use ++iter rather than iter++ for std::list iterators.
6289 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
6290 defer deleting them until all nodes in the forest are processed. Do
6291 free even leaf nodes. Change to_process into auto_vec.
6294 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
6295 * vec.h (vec_default_construct): Use memset instead of placement new
6296 if BROKEN_VALUE_INITIALIZATION is defined.
6297 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
6298 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
6301 PR rtl-optimization/83723
6302 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
6303 * lra.c (lra_substitute_pseudo): Likewise. If true, use
6304 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
6306 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
6308 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
6310 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
6312 PR rtl-optimization/81443
6313 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
6314 from inner REGs to paradoxical SUBREGs.
6316 2018-02-16 Richard Biener <rguenther@suse.de>
6318 PR tree-optimization/84399
6319 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6320 For operands we can analyze at their definition make sure we can
6321 analyze them at each use as well.
6323 2018-02-16 Richard Biener <rguenther@suse.de>
6325 PR tree-optimization/84190
6326 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
6327 volatile accesses if the decl isn't volatile.
6329 2018-02-15 Jason Merrill <jason@redhat.com>
6331 PR c++/84314 - ICE with templates and fastcall attribute.
6332 * attribs.c (build_type_attribute_qual_variant): Don't clobber
6333 TYPE_CANONICAL on an existing type.
6335 2018-02-15 Jakub Jelinek <jakub@redhat.com>
6337 PR tree-optimization/84383
6338 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
6339 dstoff nor call operand_equal_p if dstbase is NULL.
6341 PR tree-optimization/84334
6342 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
6343 also a CONSTANT_CLASS_P, punt.
6345 2018-02-14 Jim Wilson <jimw@sifive.com>
6347 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
6348 first SMALL_OPERAND check. New local min_second_step. Move assert
6349 to where locals are set. Add TARGET_RVC support.
6350 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
6352 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
6354 * doc/invoke.texi: Correct -Wformat-overflow code sample.
6356 2018-02-14 Martin Sebor <msebor@redhat.com>
6358 PR tree-optimization/83698
6359 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
6360 arrays constrain the offset range to their bounds.
6361 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
6362 (builtin_access::overlap): Avoid setting the size of overlap if it's
6364 (maybe_diag_overlap): Also consider arrays when deciding what values
6365 of offsets to include in diagnostics.
6367 2018-02-14 Martin Sebor <msebor@redhat.com>
6370 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
6371 that correspond to the kind of a declaration.
6373 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
6376 * config/pa/pa.md: Load address of PIC label using the linkage table
6377 if the label is nonlocal.
6379 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
6381 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
6382 warning message if user requests -maltivec=be.
6383 * doc/invoke.texi: Document deprecation of -maltivec=be.
6385 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
6388 * config/rs6000/rs6000-c.c: Update definitions for
6389 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
6390 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
6392 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6395 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
6396 add _get_ssp intrinsics. Remove argument from
6397 __builtin_ia32_rdssp[d|q].
6398 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
6399 * config/i386/i386-builtin.def: Remove argument from
6400 __builtin_ia32_rdssp[d|q].
6401 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
6402 ix86_expand_special_args_builtin for _rdssp[d|q].
6403 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
6404 Clear register before usage.
6405 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6406 Add documentation for new _get_ssp and _inc_ssp intrinsics.
6408 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
6410 PR tree-optimization/84357
6411 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6412 operand 1 of an ARRAY_REF too.
6414 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
6417 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6418 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6420 (set_of_reg): New struct.
6421 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6422 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6423 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6425 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6426 Split into bitclr, bitset, bitinvert patterns if appropriate.
6427 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6428 use rx_fuse_in_memory_bitop.
6429 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6430 to named insn, correct maximum insn length.
6432 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6435 * machmode.def: Define a complex mode for PARTIAL_INT.
6436 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6438 * doc/rtl.texi: Document CSPImode.
6439 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6441 (msp430_hard_regno_nregs_with_padding): Likewise.
6443 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
6446 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6448 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
6450 PR rtl-optimization/84169
6451 * combine.c (try_combine): New variable split_i2i3. Set it to true if
6452 we generated a parallel as new i3 and we split that to new i2 and i3
6453 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
6454 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
6455 those to i2, not i1. Partially rewrite this scan code.
6457 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6460 * stor-layout.c (place_field): For variable length fields, adjust
6461 offset_align afterwards not just based on the field's alignment,
6462 but also on the size.
6465 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6466 of exps and logs in the use_exp2 case.
6468 2018-02-13 Jeff Law <law@redhat.com>
6470 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6473 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6476 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
6480 * common.opt (gas-loc-support, gas-locview-support): New.
6481 (ginline-points, ginternal-reset-location-views): New.
6482 * doc/invoke.texi: Document them. Use @itemx where intended.
6483 (gvariable-location-views): Adjust.
6484 * target.def (reset_location_view): New.
6485 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6486 (TARGET_RESET_LOCATION_VIEW): New.
6487 * doc/tm.texi: Rebuilt.
6488 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6489 (dwarf2out_default_as_locview_support): New.
6490 (output_asm_line_debug_info): Use option variables.
6491 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6492 (output_loc_list): Likewise.
6493 (add_high_low_attributes): Check option variables.
6494 Don't output entry view attribute in strict mode.
6495 (gen_inlined_subroutine_die): Check option variables.
6496 (dwarf2out_inline_entry): Likewise.
6497 (init_sections_and_labels): Likewise.
6498 (dwarf2out_early_finish): Likewise.
6499 (maybe_reset_location_view): New, from...
6500 (dwarf2out_var_location): ... here. Call it.
6501 * debug.h (dwarf2out_default_as_loc_support): Declare.
6502 (dwarf2out_default_as_locview_support): Declare.
6503 * hooks.c (hook_int_rtx_insn_0): New.
6504 * hooks.h (hook_int_rtx_insn_0): Declare.
6505 * toplev.c (process_options): Take -gas-loc-support and
6506 -gas-locview-support from dwarf2out. Enable
6507 -gvariable-location-views by default only with locview
6508 assembler support. Enable -ginternal-reset-location-views by
6509 default only if the target defines the corresponding hook.
6510 Enable -ginline-points by default if location views are
6511 enabled; force it disabled if statement frontiers are
6513 * tree-inline.c (expand_call_inline): Check option variables.
6514 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6516 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
6518 PR tree-optimization/84321
6519 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6520 handling. Also check whether the anti-range contains any values
6521 that satisfy the mask; switch to a VR_RANGE if not.
6523 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
6526 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6528 2018-02-13 Martin Jambor <mjambor@suse.cz>
6531 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6532 of call statements, also set location of a load to a temporary.
6534 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
6536 * config/rl78/rl78.c (add_vector_labels): New function.
6537 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6538 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6539 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6540 which checks that no arguments are passed.
6541 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6542 * doc/extend.texi: Documentation for the new attribute.
6544 2018-02-13 Andreas Schwab <schwab@suse.de>
6546 * config/riscv/linux.h (CPP_SPEC): Define.
6548 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6551 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6552 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6553 OPTION_MASK_ISA_AES as first argument to def_builtin_const
6554 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6555 instead of OPTION_MASK_ISA_PCLMUL as first argument to
6556 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6557 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6558 temporarily for AES and PCLMUL builtins.
6560 PR tree-optimization/84339
6561 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6562 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6566 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6567 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6568 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6570 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6572 * omp-simd-clone.h: New file.
6573 * omp-simd-clone.c: Include omp-simd-clone.h.
6574 (expand_simd_clones): No longer static.
6575 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6576 cgraph.h and omp-simd-clone.h.
6577 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6578 (vect_recog_widen_shift_pattern): Formatting fix.
6579 (vect_pattern_recog_1): Don't check optab for calls.
6582 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6583 operands[2] into a REG before using gen_lowpart on it.
6585 2018-02-12 Jeff Law <law@redhat.com>
6588 * config/sh/sh.c (find_barrier): Consider a sibling call
6591 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6592 successfully back substituting a reg.
6594 2018-02-12 Richard Biener <rguenther@suse.de>
6596 PR tree-optimization/84037
6597 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6598 parameter, move visited init to caller.
6599 (vect_slp_analyze_operations): Separate cost from validity
6600 check, initialize visited once for all instances.
6601 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6603 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6605 (vect_model_promotion_demotion_cost): Likewise.
6606 (vectorizable_bswap): Guard cost modeling with !slp_node
6607 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6609 (vectorizable_call): Likewise.
6610 (vectorizable_conversion): Likewise.
6611 (vectorizable_assignment): Likewise.
6612 (vectorizable_shift): Likewise.
6613 (vectorizable_operation): Likewise.
6614 (vectorizable_store): Likewise.
6615 (vectorizable_load): Likewise.
6616 (vectorizable_condition): Likewise.
6617 (vectorizable_comparison): Likewise.
6619 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6622 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6623 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6624 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6626 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6629 * params.def (inline-min-speedup): Increase from 8 to 15.
6630 (max-inline-insns-auto): Decrease from 40 to 30.
6631 * ipa-split.c (consider_split): Add some buffer for function to
6632 be considered inlining candidate.
6633 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6636 2018-02-12 Richard Biener <rguenther@suse.de>
6638 PR tree-optimization/84037
6639 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6640 matched stmts if we cannot swap the non-matched ones.
6642 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6644 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6645 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6646 _mm_maskz_scalef_round_ss): New intrinsics.
6647 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6648 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6649 __builtin_ia32_scalefss_round): Remove.
6650 (__builtin_ia32_scalefsd_mask_round,
6651 __builtin_ia32_scalefss_mask_round): New intrinsics.
6652 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6653 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6654 ((match_operand:VF_128 2 "<round_nimm_predicate>"
6655 "<round_constraint>")): Changed to ...
6656 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
6657 "<round_scalar_constraint>")): ... this.
6658 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
6659 %0, %1, %2<round_op3>}"): Changed to ...
6660 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
6661 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
6662 %2<round_scalar_mask_op3>}"): ... this.
6663 * config/i386/subst.md (round_scalar_nimm_predicate): New.
6665 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6667 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
6668 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
6669 (_mm_maskz_sqrt_round_ss): New intrinsics.
6670 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
6671 (__builtin_ia32_sqrtsd_mask_round)
6672 (__builtin_ia32_sqrtss_mask_round): New builtins.
6673 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6674 (__builtin_ia32_sqrtss_round): Remove.
6675 (__builtin_ia32_sqrtsd_mask_round)
6676 (__builtin_ia32_sqrtss_mask_round): New builtins.
6677 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6678 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6679 ((match_operand:VF_128 1 "vector_operand"
6680 "xBm,<round_constraint>")): Changed to ...
6681 ((match_operand:VF_128 1 "vector_operand"
6682 "xBm,<round_scalar_constraint>")): ... this.
6683 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6684 %0, %2, %<iptr>1<round_op3>}): Changed to ...
6685 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6686 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6687 %<iptr>1<round_scalar_mask_op3>}): ... this.
6688 ((set_attr "prefix" "<round_prefix>")): Changed to ...
6689 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6691 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
6694 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6695 Cast vec_cmpeq result to correct type.
6696 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
6697 Cast vec_cmpgt result to correct type.
6699 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
6701 * final.c (final_scan_insn_1): Renamed from...
6702 (final_scan_insn): ... this. New wrapper, to recover
6703 seen from the outermost call in recursive ones.
6704 * config/sparc/sparc.c (output_return): Drop seen from call.
6705 (output_sibcall): Likewise.
6706 * config/visium/visium.c (output_branch): Likewise.
6708 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
6710 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
6713 2018-02-10 Alan Modra <amodra@gmail.com>
6716 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
6717 Specify LR as an input.
6719 2018-02-10 Jakub Jelinek <jakub@redhat.com>
6722 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
6723 remove_member_access_dummy_vars): New functions.
6724 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
6725 lower_omp_1, execute_lower_omp): Use them.
6727 PR rtl-optimization/84308
6728 * shrink-wrap.c (spread_components): Release todo vector.
6730 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
6732 PR rtl-optimization/57193
6733 * ira-color.c (struct allocno_color_data): Add member
6734 conflict_allocno_hard_prefs.
6735 (update_conflict_allocno_hard_prefs): New.
6736 (bucket_allocno_compare_func): Add a preference based on
6737 conflict_allocno_hard_prefs.
6738 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
6739 (color_allocnos): Remove a dead code. Initiate
6740 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
6742 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6745 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
6746 constraint from =wa to wa. Avoid a subreg on the output operand,
6747 instead use a pseudo and subreg it in a move.
6748 (p9_xxbrd_<mode>): Changed to ...
6749 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
6750 (p9_xxbrd_v2df): New expander.
6751 (p9_xxbrw_<mode>): Changed to ...
6752 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
6753 (p9_xxbrw_v4sf): New expander.
6755 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6757 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
6759 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
6762 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
6763 multiply in 32-bit mode.
6764 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
6765 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
6768 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6770 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
6771 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
6772 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
6773 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
6775 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
6778 * dwarf2out.c (is_trivial_indirect_ref): New function.
6779 (dwarf2out_late_global_decl): Do not generate a location
6780 attribute for variables that have a non-trivial DECL_VALUE_EXPR
6781 and that are not defined in the current unit.
6783 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6785 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
6786 instead of a libcall for UNORDERED.
6788 2018-02-09 Tamar Christina <tamar.christina@arm.com>
6791 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
6792 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
6794 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6797 * config/s390/s390.c (s390_set_current_function): Invoke
6798 s390_indirect_branch_settings also if fndecl didn't change.
6800 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6802 * config/rs6000/rs6000.md (blockage): Set length to zero.
6804 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6806 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
6808 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6811 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
6812 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
6816 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
6817 PARALLEL incoming that failed vt_get_decl_and_offset check.
6820 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
6821 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
6823 (get_variable_section): For decls in named .bss* sections pass true as
6824 second argument to bss_initializer_p.
6826 2018-02-09 Marek Polacek <polacek@redhat.com>
6827 Jakub Jelinek <jakub@redhat.com>
6830 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
6831 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
6832 Sync some changes from cxx_fold_indirect_ref.
6834 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6836 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
6838 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
6839 (BLOCK_INLINE_ENTRY_LABEL): New.
6840 (dwarf2out_var_location): Disregard inline entry markers.
6841 (inline_entry_data): New struct.
6842 (inline_entry_data_hasher): New hashtable type.
6843 (inline_entry_data_hasher::hash): New.
6844 (inline_entry_data_hasher::equal): New.
6845 (inline_entry_data_table): New variable.
6846 (add_high_low_attributes): Add DW_AT_entry_pc and
6847 DW_AT_GNU_entry_view attributes if a pending entry is found
6848 in inline_entry_data_table. Add old entry_pc attribute only
6849 if debug nonbinding markers are disabled.
6850 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
6851 markers are enabled.
6852 (block_within_block_p, dwarf2out_inline_entry): New.
6853 (dwarf2out_finish): Check that no entries remained in
6854 inline_entry_data_table.
6855 * final.c (reemit_insn_block_notes): Handle inline entry notes.
6856 (final_scan_insn, notice_source_line): Likewise.
6857 (rest_of_clean_state): Skip inline entry markers.
6858 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
6860 * gimple.c (gimple_build_debug_inline_entry): New.
6861 * gimple.h (enum gimple_debug_subcode): Add
6862 GIMPLE_DEBUG_INLINE_ENTRY.
6863 (gimple_build_debug_inline_entry): Declare.
6864 (gimple_debug_inline_entry_p): New.
6865 (gimple_debug_nonbind_marker_p): Adjust.
6866 * insn-notes.def (INLINE_ENTRY): New.
6867 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
6868 inline entry marker notes.
6869 (print_insn): Likewise.
6870 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
6871 (INSN_DEBUG_MARKER_KIND): Likewise.
6872 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
6873 * tree-inline.c (expand_call_inline): Build and insert
6874 debug_inline_entry stmt.
6875 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
6876 inline entry blocks early, if nonbind markers are enabled.
6877 (dump_scope_block): Dump fragment info.
6878 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
6879 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
6880 (gimple_build_debug_inline_entry): New.
6881 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
6882 Enable/disable inline entry points too.
6883 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
6884 (DEBUG_INSN): Describe inline entry markers.
6886 * common.opt (gvariable-location-views): New.
6887 (gvariable-location-views=incompat5): New.
6888 * config.in: Rebuilt.
6889 * configure: Rebuilt.
6890 * configure.ac: Test assembler for view support.
6891 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
6892 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
6893 * dwarf2out.c (var_loc_view): New typedef.
6894 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
6895 (dwarf2out_locviews_in_attribute): New.
6896 (dwarf2out_locviews_in_loclist): New.
6897 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
6898 (enum dw_line_info_opcode): Add LI_adv_address.
6899 (struct dw_line_info_table): Add view.
6900 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
6901 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
6902 (zero_view_p): New variable.
6903 (ZERO_VIEW_P): New macro.
6904 (output_asm_line_debug_info): New.
6905 (struct var_loc_node): Add view.
6906 (add_AT_view_list, AT_loc_list): New.
6907 (add_var_loc_to_decl): Add view param. Test it against last.
6908 (new_loc_list): Add view params. Record them.
6909 (AT_loc_list_ptr): Handle loc and view lists.
6910 (view_list_to_loc_list_val_node): New.
6911 (print_dw_val): Handle dw_val_class_view_list.
6912 (size_of_die): Likewise.
6913 (value_format): Likewise.
6914 (loc_list_has_views): New.
6915 (gen_llsym): Set vl_symbol too.
6916 (maybe_gen_llsym, skip_loc_list_entry): New.
6917 (dwarf2out_maybe_output_loclist_view_pair): New.
6918 (output_loc_list): Output view list or entries too.
6919 (output_view_list_offset): New.
6920 (output_die): Handle dw_val_class_view_list.
6921 (output_dwarf_version): New.
6922 (output_compilation_unit_header): Use it.
6923 (output_skeleton_debug_sections): Likewise.
6924 (output_rnglists, output_line_info): Likewise.
6925 (output_pubnames, output_aranges): Update version comments.
6926 (output_one_line_info_table): Output view numbers in asm comments.
6927 (dw_loc_list): Determine current endview, pass it to new_loc_list.
6928 Call maybe_gen_llsym.
6929 (loc_list_from_tree_1): Adjust.
6930 (add_AT_location_description): Create view list attribute if
6931 needed, check it's absent otherwise.
6932 (convert_cfa_to_fb_loc_list): Adjust.
6933 (maybe_emit_file): Call output_asm_line_debug_info for test.
6934 (dwarf2out_var_location): Reset views as needed. Precompute
6935 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
6936 attribute. Set view.
6937 (new_line_info_table): Reset next view.
6938 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
6939 (dwarf2out_source_line): Likewise. Output view resets and labels to
6940 the assembler, or select appropriate line info opcodes.
6941 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
6942 (optimize_string_length): Catch it. Adjust.
6943 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
6944 dw_val_class_view_list, and remove it if no longer needed.
6945 (hash_loc_list): Hash view numbers.
6946 (loc_list_hasher::equal): Compare them.
6947 (optimize_location_lists): Check whether a view list symbol is
6948 needed, and whether the locview attribute is present, and
6949 whether they match. Remove the locview attribute if no longer
6951 (index_location_lists): Call skip_loc_list_entry for test.
6952 (dwarf2out_finish): Call output_asm_line_debug_info for test.
6953 Use output_dwarf_version.
6954 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
6955 (struct dw_val_node): Add val_view_list.
6956 * final.c (SEEN_NEXT_VIEW): New.
6957 (set_next_view_needed): New.
6958 (clear_next_view_needed): New.
6959 (maybe_output_next_view): New.
6960 (final_start_function): Rename to...
6961 (final_start_function_1): ... this. Take pointer to FIRST,
6962 add SEEN parameter. Emit param bindings in the initial view.
6963 (final_start_function): Reintroduce SEEN-less interface.
6964 (final): Rename to...
6965 (final_1): ... this. Take SEEN parameter. Output final pending
6966 next view at the end.
6967 (final): Reintroduce seen-less interface.
6968 (final_scan_insn): Output pending next view before switching
6969 sections or ending a block. Mark the next view as needed when
6970 outputting variable locations. Notify debug backend of section
6971 changes, and of location view changes.
6972 (rest_of_handle_final): Adjust.
6973 * toplev.c (process_options): Autodetect value for debug variable
6974 location views option. Warn on incompat5 without -gdwarf-5.
6975 * doc/invoke.texi (gvariable-location-views): New.
6976 (gvariable-location-views=incompat5): New.
6977 (gno-variable-location-views): New.
6979 2018-02-08 David Malcolm <dmalcolm@redhat.com>
6981 PR tree-optimization/84136
6982 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
6983 that the result of find_edge is non-NULL.
6985 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6988 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
6989 storing integer register in SImode. Fix cost of 256 and 512
6990 byte aligned SSE register store.
6992 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6994 * config/i386/i386.c (ix86_multiplication_cost): Fix
6995 multiplication cost for TARGET_AVX512DQ.
6997 2018-02-08 Marek Polacek <polacek@redhat.com>
6999 PR tree-optimization/84238
7000 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
7003 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7005 PR tree-optimization/84265
7006 * tree-vect-stmts.c (vectorizable_store): Don't treat
7007 VMAT_CONTIGUOUS accesses as grouped.
7008 (vectorizable_load): Likewise.
7010 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7012 PR tree-optimization/81635
7013 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
7014 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
7015 (test_round_for_mask): New functions.
7016 (wide_int_cc_tests): Call test_round_for_mask.
7017 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
7018 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
7019 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
7020 range returned by get_range_info.
7022 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7025 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
7026 * symtab.c: Include builtins.h
7027 (symtab_node::output_to_lto_symbol_table_p): Move here
7028 from lto-streamer-out.c:output_symbol_p.
7029 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
7030 (output_symbol_p): Move all logic to symtab.c
7031 (produce_symtab): Update.
7033 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7035 * config/s390/s390-opts.h (enum indirect_branch): Define.
7036 * config/s390/s390-protos.h (s390_return_addr_from_memory)
7037 (s390_indirect_branch_via_thunk)
7038 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
7039 (enum s390_indirect_branch_type): Define.
7040 * config/s390/s390.c (struct s390_frame_layout, struct
7041 machine_function): Remove.
7042 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
7043 (indirect_branch_table_label_no, indirect_branch_table_name):
7045 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
7046 (enum s390_indirect_branch_option): Define.
7047 (s390_return_addr_from_memory): New function.
7048 (s390_handle_string_attribute): New function.
7049 (s390_attribute_table): Add new attribute handler.
7050 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
7051 (s390_indirect_branch_via_thunk): New function.
7052 (s390_indirect_branch_via_inline_thunk): New function.
7053 (s390_function_ok_for_sibcall): When jumping via thunk disallow
7054 sibling call optimization for non z10 compiles.
7055 (s390_emit_call): Force indirect branch target to be a single
7056 register. Add r1 clobber for non-z10 compiles.
7057 (s390_emit_epilogue): Emit return jump via return_use expander.
7058 (s390_reorg): Handle JUMP_INSNs as execute targets.
7059 (s390_option_override_internal): Perform validity checks for the
7060 new command line options.
7061 (s390_indirect_branch_attrvalue): New function.
7062 (s390_indirect_branch_settings): New function.
7063 (s390_set_current_function): Invoke s390_indirect_branch_settings.
7064 (s390_output_indirect_thunk_function): New function.
7065 (s390_code_end): Implement target hook.
7066 (s390_case_values_threshold): Implement target hook.
7067 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
7069 * config/s390/s390.h (struct s390_frame_layout)
7070 (struct machine_function): Move here from s390.c.
7071 (TARGET_INDIRECT_BRANCH_NOBP_RET)
7072 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
7073 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
7074 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
7075 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
7076 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
7077 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
7078 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
7079 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
7080 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
7081 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
7082 (mnemonic attribute): Add values which aren't recognized
7084 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
7085 pattern for branch conversion. Fix mnemonic attribute.
7086 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
7087 indirect branch via thunk if requested.
7088 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
7089 ("*indirect_jump"): Disable for branch conversion using out of
7091 ("indirect_jump_via_thunk<mode>_z10")
7092 ("indirect_jump_via_thunk<mode>")
7093 ("indirect_jump_via_inlinethunk<mode>_z10")
7094 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
7095 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
7096 ("casesi_jump_via_inlinethunk<mode>_z10")
7097 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
7098 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
7099 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
7100 ("*indirect2_jump"): Disable for branch conversion.
7101 ("casesi_jump"): Turn into expander and expand patterns for branch
7103 ("return_use"): New expander.
7104 ("*return"): Emit return via thunk and rename it to ...
7105 ("*return<mode>"): ... this one.
7106 * config/s390/s390.opt: Add new options and and enum for the
7109 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7111 * lra-constraints.c (match_reload): Unconditionally use
7112 gen_lowpart_SUBREG, rather than selecting between that
7113 and equivalent gen_rtx_SUBREG code.
7115 2018-02-08 Richard Biener <rguenther@suse.de>
7117 PR tree-optimization/84233
7118 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
7119 changed flag instead of boguously re-using phi_inserted.
7121 2018-02-08 Martin Jambor <mjambor@suse.cz>
7123 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
7124 static local variables.
7126 2018-02-08 Richard Biener <rguenther@suse.de>
7128 PR tree-optimization/84278
7129 * tree-vect-stmts.c (vectorizable_store): When looking for
7130 smaller vector types to perform grouped strided loads/stores
7131 make sure the mode is supported by the target.
7132 (vectorizable_load): Likewise.
7134 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7136 * config/aarch64/aarch64.c (aarch64_components_for_bb):
7137 Increase LDP/STP opportunities by adding adjacent callee-saves.
7139 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7141 PR rtl-optimization/84068
7142 PR rtl-optimization/83459
7143 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
7145 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
7147 PR tree-optimization/84224
7148 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
7149 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
7152 2018-02-07 Iain Sandoe <iain@codesourcery.com>
7155 * config/rs6000/altivec.md (*restore_world): Remove LR use.
7156 * config/rs6000/predicates.md (restore_world_operation): Adjust op
7157 count, remove one USE.
7159 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
7161 * doc/install.texi (Configuration): Document the
7162 --with-long-double-format={ibm,ieee} PowerPC configuration
7166 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
7167 Convert from define_expand to be define_insn_and_split. Rework
7168 float/double/_Float128 conversions to QI/HI/SImode to work with
7169 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
7170 conversions to QI/HImode types did a store and then a load to
7171 truncate the value. For conversions to VSX registers, don't split
7172 the insn, instead emit the code directly. Use the code iterator
7173 any_fix to combine signed and unsigned conversions.
7174 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
7175 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
7176 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7177 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
7178 (fix_<mode>di2_hw): Likewise.
7179 (fixuns_<mode>di2_hw): Likewise.
7180 (fix_<mode>si2_hw): Likewise.
7181 (fixuns_<mode>si2_hw): Likewise.
7182 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
7183 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7184 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
7185 fix<uns>_trunc<SFDF:mode>si2_p8.
7186 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
7188 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
7189 (fix<uns>_<mode>_mem): Likewise.
7190 (fctiw<u>z_<mode>_mem): Likewise.
7191 (fix<uns>_<mode>_mem): Likewise.
7192 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
7193 the register allocator from doing a direct move to the GPRs to do
7194 a store, and instead use the ISA 3.0 store byte/half-word from
7195 vector register instruction. For IEEE 128-bit floating point,
7196 also optimize stores of 32-bit ints.
7197 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
7199 2018-02-07 Alan Hayward <alan.hayward@arm.com>
7201 * genextract.c (push_pathstr_operand): New function to support
7203 (walk_rtx): Call push_pathstr_operand.
7204 (print_path): Support [a-zA-Z].
7206 2018-02-07 Richard Biener <rguenther@suse.de>
7208 PR tree-optimization/84037
7209 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
7210 (cse_and_gimplify_to_preheader): Declare.
7211 (vect_get_place_in_interleaving_chain): Likewise.
7212 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7214 (_loop_vec_info::~_loop_vec_info): Delete it.
7215 (cse_and_gimplify_to_preheader): New function.
7216 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
7217 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
7218 (vectorizable_load): Likewise. For grouped stores always base
7219 the IV on the first element.
7220 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
7221 condition before gimplifying.
7223 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7225 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
7226 *DIV_EXPR and *MOD_EXPR.
7228 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
7231 * config/i386/i386.c (ix86_option_override_internal): Mask out
7232 the CF_SET bit when checking -fcf-protection.
7234 2018-02-07 Tom de Vries <tom@codesourcery.com>
7237 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
7240 2018-02-07 Richard Biener <rguenther@suse.de>
7242 PR tree-optimization/84204
7243 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
7246 PR tree-optimization/84205
7247 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
7248 special-case isl_ast_op_zdiv_r.
7250 PR tree-optimization/84223
7251 * graphite-scop-detection.c (gather_bbs::before_dom_children):
7252 Only add conditions from within the region.
7253 (gather_bbs::after_dom_children): Adjust.
7255 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
7258 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
7259 * config/avr/avr.md: Only post-reload split REG-REG moves if
7260 either register is GENERAL_REG_P.
7262 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7264 PR tree-optimization/84235
7265 * tree-ssa-scopedtables.c
7266 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
7267 if the subtraction is performed in floating point type where NaNs are
7268 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
7269 build 1. Formatting fix.
7271 2018-02-06 Jakub Jelinek <jakub@redhat.com>
7274 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
7275 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
7276 and skip it regardless of bb boundaries. Use CALL_P macro,
7277 don't test INSN_P (insn) together with CALL_P or JUMP_P check
7278 unnecessarily, formatting fix.
7280 2018-02-06 Michael Collison <michael.collison@arm.com>
7282 * config/arm/thumb2.md:
7283 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
7284 (*thumb_mov_notscc): Ditto.
7286 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
7289 * config/rs6000/rs6000.md (su code attribute): Use "u" for
7290 unsigned_fix, not "s".
7292 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7294 * configure.ac (gcc_fn_eh_frame_ro): New function.
7295 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
7296 correct .eh_frame permissions.
7297 * configure: Regenerate.
7299 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
7301 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
7304 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7306 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7307 Display warning message for -mno-speculate-indirect-jumps.
7309 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
7311 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
7313 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
7315 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
7317 PR tree-optimization/84225
7318 * tree-eh.c (find_trapping_overflow): Only call
7319 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
7321 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7324 * config/i386/i386.c: Reimplement the check of possible options
7325 -mibt/-mshstk conbination. Change error messages.
7326 * doc/invoke.texi: Fix a typo: remove extra '='.
7328 2018-02-06 Marek Polacek <polacek@redhat.com>
7330 PR tree-optimization/84228
7331 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
7333 2018-02-06 Tamar Christina <tamar.christina@arm.com>
7336 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
7337 emitted arch directives.
7338 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
7339 __ARM_FEATURE_COPROC before changing architectures.
7341 2018-02-06 Richard Biener <rguenther@suse.de>
7343 * config/i386/i386.c (print_reg): Fix typo.
7344 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
7346 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
7348 * configure: Regenerate.
7350 2018-02-05 Martin Sebor <msebor@redhat.com>
7352 PR tree-optimization/83369
7353 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
7356 2018-02-05 Martin Liska <mliska@suse.cz>
7358 * doc/invoke.texi: Cherry-pick upstream r323995.
7360 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
7362 * ira.c (ira_init_register_move_cost): Adjust comment.
7364 2018-02-05 Martin Liska <mliska@suse.cz>
7366 PR gcov-profile/84137
7367 * doc/gcov.texi: Fix typo in documentation.
7369 2018-02-05 Martin Liska <mliska@suse.cz>
7371 PR gcov-profile/83879
7372 * doc/gcov.texi: Document necessity of --dynamic-list-data when
7373 using dlopen functionality.
7375 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
7377 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
7378 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
7379 _mm_maskz_range_ss, _mm_mask_range_round_ss,
7380 _mm_maskz_range_round_ss): New intrinsics.
7381 (__builtin_ia32_rangesd128_round)
7382 (__builtin_ia32_rangess128_round): Remove.
7383 (__builtin_ia32_rangesd128_mask_round,
7384 __builtin_ia32_rangess128_mask_round): New builtins.
7385 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
7386 __builtin_ia32_rangess128_round): Remove.
7387 (__builtin_ia32_rangesd128_mask_round,
7388 __builtin_ia32_rangess128_mask_round): New builtins.
7389 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
7390 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
7391 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
7392 "<round_saeonly_constraint>")): Changed to ...
7393 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
7394 "<round_saeonly_scalar_constraint>")): ... this.
7395 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
7396 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
7397 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
7398 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
7399 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
7401 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
7403 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
7405 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7406 Remove all values except native, 8540 and 8548.
7408 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
7410 * config/i386/i386.c (ix86_output_function_return): Pass
7411 INVALID_REGNUM, instead of -1, as invalid register number to
7412 indirect_thunk_name and output_indirect_thunk.
7414 2018-02-02 Julia Koval <julia.koval@intel.com>
7416 * config.gcc: Add -march=icelake.
7417 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7418 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7419 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7420 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7421 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7422 (processor_target_table): Add icelake.
7423 (ix86_option_override_internal): Handle new PTAs.
7424 (get_builtin_code_for_version): Handle icelake.
7425 (M_INTEL_COREI7_ICELAKE): New.
7426 (fold_builtin_cpu): Handle icelake.
7427 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7428 * doc/invoke.texi: Add -march=icelake.
7430 2018-02-02 Julia Koval <julia.koval@intel.com>
7432 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7433 to wide_int_bitmask.
7434 * wide-int-bitmask.h: New.
7436 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7439 * config/i386/i386.md: Replace Pmode with word_mode in
7440 builtin_setjmp_setup and builtin_longjmp to support x32.
7442 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
7446 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7448 (rs6000_supported_cpu_names): New static variable.
7449 (linux_cpu_translation_table): Likewise.
7450 (elf_platform) <cpu>: Define new static variable and use it.
7451 Translate kernel AT_PLATFORM name to canonical name if needed.
7452 Error if platform name is unknown.
7454 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
7457 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7459 2018-02-01 Jeff Law <law@redhat.com>
7462 * config/i386/i386.c (release_scratch_register_on_entry): Add new
7463 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
7464 the scratch if RELEASE_VIA_POP is false.
7465 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7466 If we have to save a temporary register, decrement SIZE appropriately.
7467 Pass new arguments to release_scratch_register_on_entry.
7468 (ix86_adjust_stack_and_probe): Likewise.
7469 (ix86_emit_probe_stack_range): Pass new arguments to
7470 release_scratch_register_on_entry.
7472 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
7474 PR rtl-optimization/84157
7475 * combine.c (change_zero_ext): Use REG_P predicate in
7476 front of HARD_REGISTER_P predicate.
7478 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
7480 * config/avr/avr.c (avr_option_override): Move disabling of
7481 -fdelete-null-pointer-checks to...
7482 * common/config/avr/avr-common.c (avr_option_optimization_table):
7485 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7487 PR tree-optimization/81635
7488 * tree-data-ref.c (split_constant_offset_1): For types that
7489 wrap on overflow, try to use range info to prove that wrapping
7492 2018-02-01 Renlin Li <renlin.li@arm.com>
7495 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7497 (aarch64_register_move_cost): Likewise.
7498 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7500 (REG_CLASS_NAMES): Likewise.
7501 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7502 TAILCALL_ADDR_REGS. Remove IP registers.
7503 * config/aarch64/aarch64.md (Ucs): Update register constraint.
7505 2018-02-01 Richard Biener <rguenther@suse.de>
7507 * domwalk.h (dom_walker::dom_walker): Add additional constructor
7508 for specifying RPO order and allow NULL for that.
7509 * domwalk.c (dom_walker::dom_walker): Likewise.
7510 (dom_walker::walk): Handle NULL RPO order.
7511 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7513 (rewrite_update_dom_walker): Likewise.
7514 (mark_def_dom_walker): Likewise.
7516 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7518 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7519 (aarch64_maybe_expand_sve_subreg_move): Declare.
7520 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7521 * config/aarch64/predicates.md (aarch64_any_register_operand): New
7523 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7524 that are semantically a reverse operation.
7525 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7526 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7527 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7529 (aarch64_can_change_mode_class): For big-endian, forbid changes
7530 between two SVE modes if they have different element sizes.
7532 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7534 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7535 the TImode handling for big-endian targets.
7537 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7539 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7540 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
7542 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7543 Remove BSWAP handing for big-endian targets and use the form of
7544 LD1RQ appropariate for the mode.
7546 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7548 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7549 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7552 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7555 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7556 check for operands that need to go through aarch64_sve_reload_be.
7558 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7560 PR tree-optimization/81661
7561 PR tree-optimization/84117
7562 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7563 * tree-eh.c: Include gimplify.h.
7564 (find_trapping_overflow, replace_trapping_overflow,
7565 rewrite_to_non_trapping_overflow): New functions.
7566 * tree-vect-loop.c: Include tree-eh.h.
7567 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7568 * tree-data-ref.c: Include tree-eh.h.
7569 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7571 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7573 PR rtl-optimization/84123
7574 * combine.c (change_zero_ext): Check if hard register satisfies
7575 can_change_dest_mode before calling gen_lowpart_SUBREG.
7577 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7580 * ira.c (ira_init_register_move_cost): Remove assert.
7582 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7584 PR rtl-optimization/84071
7585 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7586 * doc/tm.texi: Regenerate.
7588 2018-01-31 Richard Biener <rguenther@suse.de>
7590 PR tree-optimization/84132
7591 * tree-data-ref.c (analyze_miv_subscript): Properly
7592 check whether evolution_function_is_affine_multivariate_p
7593 before calling gcd_of_steps_may_divide_p.
7595 2018-01-31 Julia Koval <julia.koval@intel.com>
7598 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7599 * config/i386/i386.md (rdpid_rex64) New.
7600 (rdpid): Make 32bit only.
7602 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7605 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7606 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7608 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7611 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7613 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7615 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7617 PR rtl-optimization/84071
7618 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7619 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7621 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7623 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7624 (arc_attribute_table): Add 'aux' attribute.
7625 (arc_in_small_data_p): Consider aux like variables.
7626 (arc_is_aux_reg_p): New function.
7627 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7628 (arc_get_aux_arg): New function.
7629 (prepare_move_operands): Handle aux-register access.
7630 (arc_handle_aux_attribute): New function.
7631 * doc/extend.texi (ARC Variable attributes): Add subsection.
7633 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7635 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7636 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7637 (arc_attribute_table): Add 'uncached' attribute.
7638 (arc_print_operand): Print '.di' flag for uncached memory
7640 (arc_in_small_data_p): Do not consider for small data the uncached
7642 (arc_is_uncached_mem_p): New function.
7643 * config/arc/predicates.md (compact_store_memory_operand): Check
7644 for uncached memory accesses.
7645 (nonvol_nonimm_operand): Likewise.
7646 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7648 2018-01-31 Jakub Jelinek <jakub@redhat.com>
7651 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7652 falign-loops=): Add Optimization flag.
7654 2018-01-30 Jeff Law <law@redhat.com>
7657 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
7658 INT_REGISTERS_SAVED. Check it prior to calling
7659 get_scratch_register_on_entry.
7660 (ix86_adjust_stack_and_probe): Similarly.
7661 (ix86_emit_probe_stack_range): Similarly.
7662 (ix86_expand_prologue): Corresponding changes.
7664 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7667 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
7668 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
7670 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
7673 * lra-constraints.c (curr_insn_transform): Process AND in the
7676 2018-01-30 Jakub Jelinek <jakub@redhat.com>
7678 PR rtl-optimization/83986
7679 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7680 dependence against last_pending_memory_flush in addition to
7683 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
7685 PR tree-optimization/81611
7686 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7689 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7692 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7695 2018-01-30 Richard Biener <rguenther@suse.de>
7696 Jakub Jelinek <jakub@redhat.com>
7698 PR tree-optimization/84111
7699 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
7700 inner loops added during recursion, as they don't have up-to-date
7703 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7706 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
7707 (can_inline_edge_by_limits_p): ... here.
7708 (can_early_inline_edge_p, check_callers,
7709 update_caller_keys, update_callee_keys, recursive_inlining,
7710 add_new_edges_to_heap, speculation_useful_p,
7711 inline_small_functions,
7712 inline_small_functions, flatten_function,
7713 inline_to_all_callers_1): Update.
7715 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7717 * profile-count.c (profile_count::combine_with_ipa_count): Handle
7720 2018-01-30 Richard Biener <rguenther@suse.de>
7722 PR tree-optimization/83008
7723 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
7724 invariant and constant vector uses in stmts when they need
7727 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7730 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
7731 * configure: Regenerate.
7733 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7735 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
7737 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
7738 Use gen_rtx_REG rather than gen_lowpart.
7740 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7742 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
7743 rather than 0 when creating partial subregs.
7745 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7747 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
7750 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
7753 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
7754 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
7755 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
7756 flags. This restores the settings used before the 2017-07-24.
7757 Turning off pre increment/decrement/modify allows IVOPTS to
7758 optimize DF/SF loops where the index is an int.
7760 2018-01-29 Richard Biener <rguenther@suse.de>
7761 Kelvin Nilsen <kelvin@gcc.gnu.org>
7764 * tree-vect-stmts.c (vectorizable_call): Don't call
7765 targetm.vectorize_builtin_md_vectorized_function if callee is
7768 2018-01-22 Carl Love <cel@us.ibm.com>
7770 * doc/extend.tex: Fix typo in second arg in
7771 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
7773 2018-01-29 Richard Biener <rguenther@suse.de>
7775 PR tree-optimization/84086
7776 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
7777 (flush_ssaname_freelist): When SSA names were released reset
7778 the SCEV hash table.
7780 2018-01-29 Richard Biener <rguenther@suse.de>
7782 PR tree-optimization/84057
7783 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
7784 removed paths when removing edges.
7786 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
7788 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
7789 -mfunction-return=@var{choice}.
7791 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
7794 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
7796 (layout::print_source_line): Likewise.
7797 (test_get_line_width_without_trailing_whitespace): Add test cases.
7799 2018-01-27 Jakub Jelinek <jakub@redhat.com>
7802 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
7805 2018-01-26 Jim Wilson <jimw@sifive.com>
7807 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
7809 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
7812 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7814 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
7815 and CMP + SUB-immediate -> SUBS.
7817 2018-01-26 Martin Sebor <msebor@redhat.com>
7819 PR tree-optimization/83896
7820 * tree-ssa-strlen.c (get_string_len): Rename...
7821 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
7822 Avoid assuming length is constant.
7823 (handle_char_store): Use HOST_WIDE_INT for string length.
7825 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
7828 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
7829 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
7831 2018-01-26 Richard Biener <rguenther@suse.de>
7833 PR rtl-optimization/84003
7834 * dse.c (record_store): Only record redundant stores when
7835 the earlier store aliases at least all accesses the later one does.
7837 2018-01-26 Jakub Jelinek <jakub@redhat.com>
7839 PR rtl-optimization/83985
7840 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
7841 REG_CFA_RESTORE insns.
7842 (delete_unmarked_insns): Don't ignore separate shrink wrapping
7843 REG_CFA_RESTORE insns here.
7846 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
7847 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
7849 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7851 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
7852 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
7853 (arc_init): Likewise.
7854 (arc_override_options): Likewise.
7855 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
7857 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
7859 * config/arc/arc.h (TARGET_DBNZ): Define.
7860 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
7861 properly set the tune attribute.
7862 (dbnz): Use TARGET_DBNZ guard.
7863 * config/arc/arc.opt (mtune): Add core3 option.
7865 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7867 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
7868 recognize new pic like addresses.
7869 (arc_delegitimize_address): Clean up.
7871 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7873 * config/arc/arc-arches.def: Option mrf16 valid for all
7875 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
7876 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
7877 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
7878 * config/arc/arc-tables.opt: Regenerate.
7879 * config/arc/arc.c (arc_conditional_register_usage): Handle
7880 reduced register file case.
7881 (arc_file_start): Set must have build attributes.
7882 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
7884 * config/arc/arc.opt (mrf16): Add new option.
7885 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
7886 * config/arc/genmultilib.awk: Handle new mrf16 option.
7887 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
7888 * config/arc/t-multilib: Regenerate.
7889 * doc/invoke.texi (ARC Options): Document mrf16 option.
7891 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7893 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
7894 * config/arc/arc.c (arc_handle_secure_attribute): New function.
7895 (arc_attribute_table): Add 'secure_call' attribute.
7896 (arc_print_operand): Print secure call operand.
7897 (arc_function_ok_for_sibcall): Don't optimize tail calls when
7899 (arc_is_secure_call_p): New function. * config/arc/arc.md
7900 (call_i): Add support for sjli instruction.
7901 (call_value_i): Likewise.
7902 * config/arc/constraints.md (Csc): New constraint.
7904 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7905 John Eric Martin <John.Martin@emmicro-us.com>
7907 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
7908 * config/arc/arc.c (_arc_jli_section): New struct.
7909 (arc_jli_section): New type.
7910 (rc_jli_sections): New static variable.
7911 (arc_handle_jli_attribute): New function.
7912 (arc_attribute_table): Add jli_always and jli_fixed attribute.
7913 (arc_file_end): New function.
7914 (TARGET_ASM_FILE_END): Define.
7915 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
7916 (arc_add_jli_section): New function.
7917 (jli_call_scan): Likewise.
7918 (arc_reorg): Call jli_call_scan.
7919 (arc_output_addsi): Remove 'S' from printing asm operand.
7920 (arc_is_jli_call_p): New function.
7921 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
7923 (movhi_insn): Likewise.
7924 (movsi_insn): Likewise.
7925 (movsi_set_cc_insn): Likewise.
7926 (loadqi_update): Likewise.
7927 (load_zeroextendqisi_update): Likewise.
7928 (load_signextendqisi_update): Likewise.
7929 (loadhi_update): Likewise.
7930 (load_zeroextendhisi_update): Likewise.
7931 (load_signextendhisi_update): Likewise.
7932 (loadsi_update): Likewise.
7933 (loadsf_update): Likewise.
7934 (movsicc_insn): Likewise.
7935 (bset_insn): Likewise.
7936 (bxor_insn): Likewise.
7937 (bclr_insn): Likewise.
7938 (bmsk_insn): Likewise.
7939 (bicsi3_insn): Likewise.
7940 (cmpsi_cc_c_insn): Likewise.
7941 (movsi_ne): Likewise.
7942 (movsi_cond_exec): Likewise.
7943 (clrsbsi2): Likewise.
7954 (call_i): Remove 'S' asm letter, add jli instruction.
7955 (call_value_i): Likewise.
7956 * config/arc/arc.op (mjli-always): New option.
7957 * config/arc/constraints.md (Cji): New constraint.
7958 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
7960 (subsf3_fpx): Likewise.
7961 (mulsf3_fpx): Likewise.
7962 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
7964 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
7966 * doc/invoke.texi (ARC): Document mjli-always option.
7968 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
7970 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
7971 avoid addition with 0 and use incw and decw where possible.
7973 2018-01-26 Richard Biener <rguenther@suse.de>
7975 PR tree-optimization/81082
7976 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
7977 association if it requires casting to unsigned.
7978 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
7979 from fold_plusminus_mult_expr to catch important cases late when
7980 range info is available.
7982 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7984 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
7985 * configure.ac (hidden_linkonce): New test.
7986 * configure: Regenerate.
7987 * config.in: Regenerate.
7989 2018-01-26 Julia Koval <julia.koval@intel.com>
7991 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
7992 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
7993 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
7994 _mm_mask_bitshuffle_epi64_mask): Fix type.
7995 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
7996 USI_FTYPE_V4DI_V4DI_USI): Remove.
7997 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
7998 __builtin_ia32_vpshufbitqmb256_mask,
7999 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
8000 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
8001 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
8003 2018-01-26 Alan Modra <amodra@gmail.com>
8006 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
8007 UNSPEC_VBPERMQ. Sort other unspecs.
8009 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
8011 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
8013 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
8016 * predict.c (drop_profile): Do not push/pop cfun; update also
8018 (handle_missing_profiles): Fix logic looking for zero profiles.
8020 2018-01-25 Jakub Jelinek <jakub@redhat.com>
8023 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
8024 on functions with #pragma omp declare simd or functions with simd
8026 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
8027 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
8028 Remove trailing \n from warning_at calls.
8030 2018-01-25 Tom de Vries <tom@codesourcery.com>
8033 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8034 for neutered workers.
8036 2018-01-24 Joseph Myers <joseph@codesourcery.com>
8039 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
8040 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
8042 2018-01-24 Jeff Law <law@redhat.com>
8045 * i386.c (get_probe_interval): Move to earlier point.
8046 (ix86_compute_frame_layout): If -fstack-clash-protection and
8047 the frame is larger than the probe interval, then use pushes
8048 to save registers rather than reg->mem moves.
8049 (ix86_expand_prologue): Remove conditional for int_registers_saved
8052 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
8055 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
8056 min/max for never referenced object.
8058 2018-01-24 Jakub Jelinek <jakub@redhat.com>
8061 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
8063 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
8064 attributes from DECL_ATTRIBUTES (decl) without affecting
8065 DECL_ATTRIBUTES (current_function_decl).
8066 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
8067 functions with non-NULL DECL_ABSTRACT_ORIGIN.
8069 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
8071 PR tree-optimization/83979
8072 * fold-const.c (fold_comparison): Use constant_boolean_node
8073 instead of boolean_{true,false}_node.
8075 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
8077 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
8080 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8082 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8083 Simplify the clause that sets the length attribute.
8084 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8085 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
8086 clause that sets the length attribute.
8087 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8089 2018-01-24 Tom de Vries <tom@codesourcery.com>
8092 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
8093 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
8094 Add strict parameter.
8095 (prevent_branch_around_nothing): Insert dummy insn between branch to
8096 label and label with no ptx insn inbetween.
8097 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
8099 2018-01-24 Tom de Vries <tom@codesourcery.com>
8102 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8103 for neutered threads in warp.
8104 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
8106 2018-01-24 Richard Biener <rguenther@suse.de>
8108 PR tree-optimization/83176
8109 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
8112 2018-01-24 Richard Biener <rguenther@suse.de>
8114 PR tree-optimization/82819
8115 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
8116 code generating pluses that are no-ops in the target precision.
8118 2018-01-24 Richard Biener <rguenther@suse.de>
8121 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
8123 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8125 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
8126 to merge probabilities.
8127 * predict.c (probably_never_executed): Also mark as cold functions
8128 with global 0 profile and guessed local profile.
8129 * profile-count.c (profile_probability::combine_with_count): New
8131 * profile-count.h (profile_probability::operator*,
8132 profile_probability::operator*=, profile_probability::operator/,
8133 profile_probability::operator/=): Reduce precision to adjusted
8134 and set value to guessed on contradictory divisions.
8135 (profile_probability::combine_with_freq): Remove.
8136 (profile_probability::combine_wiht_count): Declare.
8137 (profile_count::force_nonzero):: Set to adjusted.
8138 (profile_count::probability_in):: Set quality to adjusted.
8139 * tree-ssa-tail-merge.c (replace_block_by): Use
8142 2018-01-23 Andrew Waterman <andrew@sifive.com>
8143 Jim Wilson <jimw@sifive.com>
8145 * config/riscv/riscv.c (riscv_stack_boundary): New.
8146 (riscv_option_override): Set riscv_stack_boundary. Handle
8147 riscv_preferred_stack_boundary_arg.
8148 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
8149 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
8150 (STACK_BOUNDARY): Set to riscv_stack_boundary.
8151 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
8152 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
8153 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
8155 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
8158 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
8159 of struct ix86_frame.
8160 (ix86_expand_epilogue): Likewise. Add a local variable for
8161 the reg_save_offset field in struct ix86_frame.
8163 2018-01-23 Bin Cheng <bin.cheng@arm.com>
8165 PR tree-optimization/82604
8166 * tree-loop-distribution.c (enum partition_kind): New enum item
8167 PKIND_PARTIAL_MEMSET.
8168 (partition_builtin_p): Support above new enum item.
8169 (generate_code_for_partition): Ditto.
8170 (compute_access_range): Differentiate cases that equality can be
8171 proven at all loops, the innermost loops or no loops.
8172 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
8173 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
8174 (finalize_partitions, distribute_loop): Don't fuse partition of
8175 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
8176 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
8179 2018-01-23 Martin Liska <mliska@suse.cz>
8181 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
8182 order to ignore the predictor.
8183 (PRED_POLYMORPHIC_CALL): Likewise.
8184 (PRED_RECURSIVE_CALL): Likewise.
8186 2018-01-23 Martin Liska <mliska@suse.cz>
8188 * tree-profile.c (tree_profiling): Print function header to
8189 aware reader which function we are working on.
8190 * value-prof.c (gimple_find_values_to_profile): Do not print
8191 not interesting value histograms.
8193 2018-01-23 Martin Liska <mliska@suse.cz>
8195 * profile-count.h (enum profile_quality): Add
8196 profile_uninitialized as the first value. Do not number values
8197 as they are zero based.
8198 (profile_count::verify): Update sanity check.
8199 (profile_probability::verify): Likewise.
8201 2018-01-23 Nathan Sidwell <nathan@acm.org>
8203 * doc/invoke.texi (ffor-scope): Deprecate.
8205 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8207 PR tree-optimization/83510
8208 * domwalk.c (set_all_edges_as_executable): New function.
8209 (dom_walker::dom_walker): Convert bool param
8210 "skip_unreachable_blocks" to enum reachability. Move setup of
8211 edge flags to set_all_edges_as_executable and only do it when
8212 reachability is REACHABLE_BLOCKS.
8213 * domwalk.h (enum dom_walker::reachability): New enum.
8214 (dom_walker::dom_walker): Convert bool param
8215 "skip_unreachable_blocks" to enum reachability.
8216 (set_all_edges_as_executable): New decl.
8217 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
8218 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
8220 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
8221 but converting true to REACHABLE_BLOCKS.
8222 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
8224 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
8225 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
8226 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
8228 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
8229 if check_all_array_refs will be called.
8231 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8233 * tree.c (selftest::test_location_wrappers): Add more test
8236 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8238 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
8239 (selftest::test_bit_in_range): Likewise.
8241 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8244 * doc/sourcebuild.texi (vect_float): Say that the selector
8245 only describes the situation when -funsafe-math-optimizations is on.
8246 (vect_float_strict): Document.
8248 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8250 PR tree-optimization/83965
8251 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
8252 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
8253 instead of checking only for a reduction.
8254 (vect_recog_widen_sum_pattern): Likewise.
8256 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8258 * predict.c (probably_never_executed): Only use precise profile info.
8259 (compute_function_frequency): Skip after inlining hack since we now
8260 have quality checking.
8262 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8264 * profile-count.h (profile_probability::very_unlikely,
8265 profile_probability::unlikely, profile_probability::even): Set
8266 precision to guessed.
8268 2018-01-23 Richard Biener <rguenther@suse.de>
8270 PR tree-optimization/83963
8271 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
8272 Properly terminate dominator walk when crossing the exit edge not
8273 when visiting its source block.
8275 2018-01-23 Jakub Jelinek <jakub@redhat.com>
8278 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
8279 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
8281 2018-01-22 Jakub Jelinek <jakub@redhat.com>
8283 PR tree-optimization/83957
8284 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
8285 semicolon after for body surrounded by braces.
8287 PR tree-optimization/83081
8288 * profile-count.h (profile_probability::split): New method.
8289 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
8290 Use profile_probability::split.
8291 (do_compare_rtx_and_jump): Fix adjustment of probabilities
8292 when splitting a single conditional jump into 2.
8294 2018-01-22 David Malcolm <dmalcolm@redhat.com>
8296 PR tree-optimization/69452
8297 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
8300 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8302 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
8303 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
8304 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
8306 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8308 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
8309 * config/rl78/rl78.md (movdi): New define_expand.
8310 * config/rl78/rl78.c (rl78_split_movdi): New function.
8312 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
8315 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
8317 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
8318 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
8319 128-bit to produce an UNSPEC move to get the double word with the
8320 signbit and then a shift directly to do signbit.
8321 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
8322 implementation with a new version that just does either a direct
8323 move or a regular move. Move memory interface to separate insns.
8324 Move insns so they are next to the expander.
8325 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
8326 with signbit move. Split big and little endian case.
8327 (signbit<mode>2_dm_mem_le): Likewise.
8328 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
8329 (signbit<mode>2_dm2): Likewise.
8331 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8333 * config/rl78/rl78.md (anddi3): New define_expand.
8335 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8337 * config/rl78/rl78.md (umindi3): New define_expand.
8339 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8341 * config/rl78/rl78.md (smindi3): New define_expand.
8343 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8345 * config/rl78/rl78.md (smaxdi3): New define_expand.
8347 2018-01-22 Carl Love <cel@us.ibm.com>
8349 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
8350 LVX_V1TI): Add macro expansion.
8351 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
8352 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
8353 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
8354 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
8355 Change check to determine if the instruction is a byte reversing
8356 entry. Fix typo in comment.
8357 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
8358 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
8359 Add def_builtin calls for new builtins.
8360 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
8361 Add define_insn expansion.
8363 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8365 * config/rl78/rl78.md (umaxdi3): New define_expand.
8367 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8369 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
8370 for non-QImode registers.
8372 2018-01-22 Richard Biener <rguenther@suse.de>
8374 PR tree-optimization/83963
8375 * graphite-scop-detection.c (scop_detection::get_sese): Delay
8376 including the loop exit block.
8377 (scop_detection::merge_sese): Likewise.
8378 (scop_detection::add_scop): Do it here instead.
8380 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8382 * doc/sourcebuild.texi (arm_softfloat): Document.
8384 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
8387 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
8388 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
8389 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
8391 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8392 David Edelsohn <dje.gcc@gmail.com>
8395 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8396 Change "crset eq" to "crset 2".
8397 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8398 (*call_indirect_aix<mode>_nospec): Likewise.
8399 (*call_value_indirect_aix<mode>_nospec): Likewise.
8400 (*call_indirect_elfv2<mode>_nospec): Likewise.
8401 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
8402 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
8403 change assembly output from . to $.
8404 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8405 (indirect_jump<mode>_nospec): Change assembly output from . to $.
8406 (*tablejump<mode>_internal1_nospec): Likewise.
8408 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
8411 * config/sh/sh_optimize_sett_clrt.cc:
8412 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8414 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
8416 PR tree-optimization/83940
8417 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8418 offset_dt to vect_constant_def rather than vect_unknown_def_type.
8419 (vect_check_load_store_mask): Add a mask_dt_out parameter and
8420 use it to pass back the definition type.
8421 (vect_check_store_rhs): Likewise rhs_dt_out.
8422 (vect_build_gather_load_calls): Add a mask_dt argument and use
8423 it instead of a call to vect_is_simple_use.
8424 (vectorizable_store): Update calls to vect_check_load_store_mask
8425 and vect_check_store_rhs. Use the dt returned by the latter instead
8426 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
8427 instead of calls to vect_is_simple_use. Pass the scalar rather
8428 than the vector operand to vect_is_simple_use when handling
8429 second and subsequent copies of an rhs value.
8430 (vectorizable_load): Update calls to vect_check_load_store_mask
8431 and vect_build_gather_load_calls. Use the cached mask_dt and
8432 gs_info.offset_dt instead of calls to vect_is_simple_use.
8434 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8437 * tree-emutls.c: Include gimplify.h.
8438 (lower_emutls_2): New function.
8439 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8440 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8441 it before further processing.
8444 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8445 UINTVAL (trueop1) instead of INTVAL (op1).
8447 2018-01-19 Jakub Jelinek <jakub@redhat.com>
8451 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8452 INCOMING_FRAME_SP_OFFSET if not defined.
8453 (scan_trace): Add ENTRY argument. If true and
8454 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8455 emit a note to adjust the CFA offset.
8456 (create_cfi_notes): Adjust scan_trace callers.
8457 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8458 INCOMING_FRAME_SP_OFFSET in the CIE.
8459 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8460 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8462 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8463 * doc/tm.texi: Regenerated.
8465 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8467 PR rtl-optimization/83147
8468 * lra-constraints.c (remove_inheritance_pseudos): Use
8469 lra_substitute_pseudo_within_insn.
8471 2018-01-19 Tom de Vries <tom@codesourcery.com>
8472 Cesar Philippidis <cesar@codesourcery.com>
8475 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8477 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
8480 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8481 spaces for function labels.
8483 2018-01-19 Martin Liska <mliska@suse.cz>
8485 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8486 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8487 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8488 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8489 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8490 (PRED_CONST_RETURN): Change from 69 to 65.
8491 (PRED_NULL_RETURN): Change from 91 to 71.
8492 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8493 (PRED_LOOP_GUARD): Change from 66 to 73.
8495 2018-01-19 Martin Liska <mliska@suse.cz>
8497 * predict.c (predict_insn_def): Add new assert.
8498 (struct branch_predictor): Change type to signed integer.
8499 (test_prediction_value_range): Amend test to cover
8501 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8502 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8503 (PRED_LOOP_ITERATIONS_MAX): Likewise.
8504 (PRED_LOOP_IV_COMPARE): Likewise.
8505 * predict.h (PROB_UNINITIALIZED): Define new constant.
8507 2018-01-19 Martin Liska <mliska@suse.cz>
8509 * predict.c (dump_prediction): Add new format for
8510 analyze_brprob.py script which is enabled with -details
8512 * profile-count.h (precise_p): New function.
8514 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8516 PR tree-optimization/83922
8517 * tree-vect-loop.c (vect_verify_full_masking): Return false if
8518 there are no statements that need masking.
8519 (vect_active_double_reduction_p): New function.
8520 (vect_analyze_loop_operations): Use it when handling phis that
8521 are not in the loop header.
8523 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8525 PR tree-optimization/83914
8526 * tree-vect-loop.c (vectorizable_induction): Don't convert
8527 init_expr or apply the peeling adjustment for inductions
8528 that are nested within the vectorized loop.
8530 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8532 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8535 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8539 * function.h (gimplify_parameters): Add gimple_seq * argument.
8540 * function.c: Include gimple.h and options.h.
8541 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8542 for the added local temporaries if needed.
8543 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8544 if there are any parameter cleanups, wrap whole body into a
8545 try/finally with the cleanups.
8547 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8550 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8551 Use GET_MODE_CLASS for scalar floating point.
8553 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8557 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8558 Fix call of call_cgraph_insertion_hooks.
8560 2018-01-18 Martin Sebor <msebor@redhat.com>
8562 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8564 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8567 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8570 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8573 * common.opt: (-ffile-prefix-map): New option.
8574 * opts.c (common_handle_option): Defer it.
8575 * opts-global.c (handle_common_deferred_options): Handle it.
8576 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8577 * file-prefix-map.h: New file.
8578 (remap_debug_filename, add_debug_prefix_map): ...here.
8579 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8580 * final.c (debug_prefix_map, add_debug_prefix_map
8581 remap_debug_filename): Move to...
8582 * file-prefix-map.c: New file.
8583 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8584 generalize, get rid of alloca(), use strrchr() instead of strchr().
8585 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8586 Implement in terms of add_prefix_map().
8587 (remap_macro_filename, remap_debug_filename): Implement in term of
8589 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8590 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8591 * dbxout.c: Include file-prefix-map.h.
8592 * varasm.c: Likewise.
8593 * vmsdbgout.c: Likewise.
8594 * xcoffout.c: Likewise.
8595 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8596 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8597 * doc/invoke.texi (-ffile-prefix-map): Document.
8598 (-fdebug-prefix-map): Update description.
8600 2018-01-18 Martin Liska <mliska@suse.cz>
8602 * config/i386/i386.c (indirect_thunk_name): Document that also
8604 (output_indirect_thunk): Document why both instructions
8605 (pause and lfence) are generated.
8607 2018-01-18 Richard Biener <rguenther@suse.de>
8609 PR tree-optimization/83887
8610 * graphite-scop-detection.c
8611 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8612 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8613 (scop_detection::merge_sese): Re-implement with a flood-fill
8614 algorithm that properly finds a SESE region if it exists.
8616 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8619 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8620 pointer_diff optimizations use view_convert instead of convert.
8622 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8624 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8625 Generate different code for -mno-speculate-indirect-jumps.
8626 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8627 (*call_indirect_aix<mode>): Disable for
8628 -mno-speculate-indirect-jumps.
8629 (*call_indirect_aix<mode>_nospec): New define_insn.
8630 (*call_value_indirect_aix<mode>): Disable for
8631 -mno-speculate-indirect-jumps.
8632 (*call_value_indirect_aix<mode>_nospec): New define_insn.
8633 (*sibcall_nonlocal_sysv<mode>): Generate different code for
8634 -mno-speculate-indirect-jumps.
8635 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8637 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8639 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8640 long double type, set the flags for noting the default long double
8641 type, even if we don't pass or return a long double type.
8643 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
8646 * ipa-inline.c (flatten_function): Do not overwrite final inlining
8649 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
8651 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8652 support for merge[hl].
8653 (fold_mergehl_helper): New helper function.
8654 (tree-vector-builder.h): New #include for tree_vector_builder usage.
8655 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
8656 (altivec_vmrglw_direct): Add xxmrglw insn.
8658 2018-01-17 Andrew Waterman <andrew@sifive.com>
8660 * config/riscv/riscv.c (riscv_conditional_register_usage): If
8661 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
8663 2018-01-17 David Malcolm <dmalcolm@redhat.com>
8666 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
8667 call the lto_location_cache before reading the
8668 DECL_SOURCE_LOCATION of the types.
8670 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
8671 Richard Sandiford <richard.sandiford@linaro.org>
8673 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8674 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8675 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8676 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8677 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8679 * config/aarch64/constraints.md (aarch64_movti_operand):
8681 * config/aarch64/predicates.md (Uti): Add new constraint.
8683 2018-01-17 Carl Love <cel@us.ibm.com>
8685 * config/rs6000/vsx.md (define_expand xl_len_r,
8686 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8687 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8689 (define_expand, define_insn): Move the shift left from the
8690 define_insn to the define_expand for lxvl and stxvl instructions.
8691 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8692 and XL_LEN_R definitions to PURE.
8694 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
8696 * config/i386/i386.c (indirect_thunk_name): Declare regno
8697 as unsigned int. Compare regno with INVALID_REGNUM.
8698 (output_indirect_thunk): Ditto.
8699 (output_indirect_thunk_function): Ditto.
8700 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
8701 in the call to output_indirect_thunk_function.
8703 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
8706 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
8707 rather than the size of inner_type to determine the stack slot size
8708 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
8710 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
8713 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
8716 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
8718 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
8719 endian Linux systems to optionally enable multilibs for selecting
8720 the long double type if the user configured an explicit type.
8721 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
8722 have no long double multilibs if not defined.
8723 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
8724 warn if the user used -mabi={ieee,ibm}longdouble and we built
8725 multilibs for long double.
8726 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
8727 appropriate multilib option.
8728 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
8730 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
8731 for building long double multilibs.
8732 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
8734 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
8736 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
8739 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
8741 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
8744 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
8747 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
8750 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8752 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
8753 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
8755 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8757 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
8758 different rtl trees depending on TARGET_64BIT.
8759 (rs6000_gen_lvx): Likewise.
8761 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8763 * config/visium/visium.md (nop): Tweak comment.
8764 (hazard_nop): Likewise.
8766 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8768 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
8769 -mspeculate-indirect-jumps.
8770 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
8771 for -mno-speculate-indirect-jumps.
8772 (*call_indirect_elfv2<mode>_nospec): New define_insn.
8773 (*call_value_indirect_elfv2<mode>): Disable for
8774 -mno-speculate-indirect-jumps.
8775 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
8776 (indirect_jump): Emit different RTL for
8777 -mno-speculate-indirect-jumps.
8778 (*indirect_jump<mode>): Disable for
8779 -mno-speculate-indirect-jumps.
8780 (*indirect_jump<mode>_nospec): New define_insn.
8781 (tablejump): Emit different RTL for
8782 -mno-speculate-indirect-jumps.
8783 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
8784 (tablejumpsi_nospec): New define_expand.
8785 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
8786 (tablejumpdi_nospec): New define_expand.
8787 (*tablejump<mode>_internal1): Disable for
8788 -mno-speculate-indirect-jumps.
8789 (*tablejump<mode>_internal1_nospec): New define_insn.
8790 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
8793 2018-01-16 Artyom Skrobov tyomitch@gmail.com
8795 * caller-save.c (insert_save): Drop unnecessary parameter. All
8798 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8799 Richard Biener <rguenth@suse.de>
8802 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
8803 return early, inline manually is_gimple_sizepos. Make sure if we
8804 call gimplify_expr we don't end up with a gimple constant.
8805 * tree.c (variably_modified_type_p): Don't return true for
8806 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
8807 * gimplify.h (is_gimple_sizepos): Remove.
8809 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8811 PR tree-optimization/83857
8812 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
8813 vectorizable_live_operation for pure SLP statements.
8814 (vectorizable_live_operation): Handle PHIs.
8816 2018-01-16 Richard Biener <rguenther@suse.de>
8818 PR tree-optimization/83867
8819 * tree-vect-stmts.c (vect_transform_stmt): Precompute
8820 nested_in_vect_loop_p since the scalar stmt may get invalidated.
8822 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8825 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
8826 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
8827 If off is not INTEGER_CST, issue a may not be aligned warning
8828 rather than isn't aligned. Use isn%'t rather than isn't.
8829 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
8831 <case MULT_EXPR>: Improve the case when bottom and one of the
8832 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
8833 operand, in that case check if the other operand is multiple of
8834 bottom divided by the INTEGER_CST operand.
8836 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8839 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
8840 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
8841 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
8842 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
8843 * config/pa/pa.c (pa_function_arg_advance): Likewise.
8844 (pa_function_arg, pa_arg_partial_bytes): Likewise.
8845 (pa_function_arg_size): New function.
8847 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8849 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
8850 in a separate statement.
8852 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8854 PR tree-optimization/83847
8855 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
8856 group gathers and scatters.
8858 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8860 PR rtl-optimization/86620
8861 * params.def (max-sched-ready-insns): Bump minimum value to 1.
8863 PR rtl-optimization/83213
8864 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
8865 to last if both are JUMP_INSNs.
8867 PR tree-optimization/83843
8868 * gimple-ssa-store-merging.c
8869 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
8870 store_immediate_info for bswap/nop orig_stores.
8872 2018-01-15 Andrew Waterman <andrew@sifive.com>
8874 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
8876 <UDIV>: Increase cost if !TARGET_DIV.
8878 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
8880 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
8881 (define_attr "cr_logical_3op"): New.
8882 (cceq_ior_compare): Adjust.
8883 (cceq_ior_compare_complement): Adjust.
8884 (*cceq_rev_compare): Adjust.
8885 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
8886 (is_cracked_insn): Adjust.
8887 (insn_must_be_first_in_group): Adjust.
8888 * config/rs6000/40x.md: Adjust.
8889 * config/rs6000/440.md: Adjust.
8890 * config/rs6000/476.md: Adjust.
8891 * config/rs6000/601.md: Adjust.
8892 * config/rs6000/603.md: Adjust.
8893 * config/rs6000/6xx.md: Adjust.
8894 * config/rs6000/7450.md: Adjust.
8895 * config/rs6000/7xx.md: Adjust.
8896 * config/rs6000/8540.md: Adjust.
8897 * config/rs6000/cell.md: Adjust.
8898 * config/rs6000/e300c2c3.md: Adjust.
8899 * config/rs6000/e500mc.md: Adjust.
8900 * config/rs6000/e500mc64.md: Adjust.
8901 * config/rs6000/e5500.md: Adjust.
8902 * config/rs6000/e6500.md: Adjust.
8903 * config/rs6000/mpc.md: Adjust.
8904 * config/rs6000/power4.md: Adjust.
8905 * config/rs6000/power5.md: Adjust.
8906 * config/rs6000/power6.md: Adjust.
8907 * config/rs6000/power7.md: Adjust.
8908 * config/rs6000/power8.md: Adjust.
8909 * config/rs6000/power9.md: Adjust.
8910 * config/rs6000/rs64.md: Adjust.
8911 * config/rs6000/titan.md: Adjust.
8913 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8915 * config/i386/predicates.md (indirect_branch_operand): Rewrite
8916 ix86_indirect_branch_register logic.
8918 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8920 * config/i386/constraints.md (Bs): Update
8921 ix86_indirect_branch_register check. Don't check
8922 ix86_indirect_branch_register with GOT_memory_operand.
8924 * config/i386/predicates.md (GOT_memory_operand): Don't check
8925 ix86_indirect_branch_register here.
8926 (GOT32_symbol_operand): Likewise.
8928 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8930 * config/i386/predicates.md (constant_call_address_operand):
8931 Rewrite ix86_indirect_branch_register logic.
8932 (sibcall_insn_operand): Likewise.
8934 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8936 * config/i386/constraints.md (Bs): Replace
8937 ix86_indirect_branch_thunk_register with
8938 ix86_indirect_branch_register.
8940 * config/i386/i386.md (indirect_jump): Likewise.
8941 (tablejump): Likewise.
8942 (*sibcall_memory): Likewise.
8943 (*sibcall_value_memory): Likewise.
8944 Peepholes of indirect call and jump via memory: Likewise.
8945 * config/i386/i386.opt: Likewise.
8946 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8947 (GOT_memory_operand): Likewise.
8948 (call_insn_operand): Likewise.
8949 (sibcall_insn_operand): Likewise.
8950 (GOT32_symbol_operand): Likewise.
8952 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8955 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
8956 type rather than type addr's type points to.
8957 (expand_omp_atomic_mutex): Likewise.
8958 (expand_omp_atomic): Likewise.
8960 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8963 * config/i386/i386.c (output_indirect_thunk_function): Use
8964 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
8965 for __x86_return_thunk.
8967 2018-01-15 Richard Biener <rguenther@suse.de>
8970 * expmed.c (extract_bit_field_1): Fix typo.
8972 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8975 * config/arm/iterators.md (VF): New mode iterator.
8976 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
8977 Remove integer-related logic from pattern.
8978 (neon_vabd<mode>_3): Likewise.
8980 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8983 * common.opt (fstrict-overflow): No longer an alias.
8984 (fwrapv-pointer): New option.
8985 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
8986 also for pointer types based on flag_wrapv_pointer.
8987 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
8988 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
8989 opts->x_flag_wrapv got set.
8990 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
8991 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
8992 POINTER_TYPE_OVERFLOW_UNDEFINED.
8993 * match.pd: Likewise in address comparison pattern.
8994 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
8996 2018-01-15 Richard Biener <rguenther@suse.de>
8999 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
9000 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
9001 Reset type names to their identifier if their TYPE_DECL doesn't
9002 have linkage (and thus is used for ODR and devirt).
9003 (save_debug_info_for_decl): Remove.
9004 (save_debug_info_for_type): Likewise.
9005 (add_tree_to_fld_list): Adjust.
9006 * tree-pretty-print.c (dump_generic_node): Make dumping of
9007 type names more robust.
9009 2018-01-15 Richard Biener <rguenther@suse.de>
9011 * BASE-VER: Bump to 8.0.1.
9013 2018-01-14 Martin Sebor <msebor@redhat.com>
9016 * builtins.c (check_access): Avoid warning when the no-warning bit
9019 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
9021 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
9022 * ira-color (allocno_hard_regs_compare): Likewise.
9024 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
9027 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
9028 Use .pushsection/.popsection.
9030 2018-01-14 Martin Sebor <msebor@redhat.com>
9033 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
9035 2018-01-14 Jakub Jelinek <jakub@redhat.com>
9037 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
9038 entry from extra_headers.
9039 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
9040 extra_headers, make the list bitwise identical to the i?86-*-* one.
9042 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9044 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
9045 -mcmodel=large with -mindirect-branch=thunk,
9046 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
9047 -mfunction-return=thunk-extern.
9048 * doc/invoke.texi: Document -mcmodel=large is incompatible with
9049 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
9050 -mfunction-return=thunk and -mfunction-return=thunk-extern.
9052 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9054 * config/i386/i386.c (print_reg): Print the name of the full
9055 integer register without '%'.
9056 (ix86_print_operand): Handle 'V'.
9057 * doc/extend.texi: Document 'V' modifier.
9059 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9061 * config/i386/constraints.md (Bs): Disallow memory operand for
9062 -mindirect-branch-register.
9064 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9065 (GOT_memory_operand): Likewise.
9066 (call_insn_operand): Likewise.
9067 (sibcall_insn_operand): Likewise.
9068 (GOT32_symbol_operand): Likewise.
9069 * config/i386/i386.md (indirect_jump): Call convert_memory_address
9070 for -mindirect-branch-register.
9071 (tablejump): Likewise.
9072 (*sibcall_memory): Likewise.
9073 (*sibcall_value_memory): Likewise.
9074 Disallow peepholes of indirect call and jump via memory for
9075 -mindirect-branch-register.
9076 (*call_pop): Replace m with Bw.
9077 (*call_value_pop): Likewise.
9078 (*sibcall_pop_memory): Replace m with Bs.
9079 * config/i386/i386.opt (mindirect-branch-register): New option.
9080 * doc/invoke.texi: Document -mindirect-branch-register option.
9082 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9084 * config/i386/i386-protos.h (ix86_output_function_return): New.
9085 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
9086 set function_return_type.
9087 (indirect_thunk_name): Add ret_p to indicate thunk for function
9089 (output_indirect_thunk_function): Pass false to
9090 indirect_thunk_name.
9091 (ix86_output_indirect_branch_via_reg): Likewise.
9092 (ix86_output_indirect_branch_via_push): Likewise.
9093 (output_indirect_thunk_function): Create alias for function
9094 return thunk if regno < 0.
9095 (ix86_output_function_return): New function.
9096 (ix86_handle_fndecl_attribute): Handle function_return.
9097 (ix86_attribute_table): Add function_return.
9098 * config/i386/i386.h (machine_function): Add
9099 function_return_type.
9100 * config/i386/i386.md (simple_return_internal): Use
9101 ix86_output_function_return.
9102 (simple_return_internal_long): Likewise.
9103 * config/i386/i386.opt (mfunction-return=): New option.
9104 (indirect_branch): Mention -mfunction-return=.
9105 * doc/extend.texi: Document function_return function attribute.
9106 * doc/invoke.texi: Document -mfunction-return= option.
9108 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9110 * config/i386/i386-opts.h (indirect_branch): New.
9111 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
9112 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
9113 with local indirect jump when converting indirect call and jump.
9114 (ix86_set_indirect_branch_type): New.
9115 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
9116 (indirectlabelno): New.
9117 (indirect_thunk_needed): Likewise.
9118 (indirect_thunk_bnd_needed): Likewise.
9119 (indirect_thunks_used): Likewise.
9120 (indirect_thunks_bnd_used): Likewise.
9121 (INDIRECT_LABEL): Likewise.
9122 (indirect_thunk_name): Likewise.
9123 (output_indirect_thunk): Likewise.
9124 (output_indirect_thunk_function): Likewise.
9125 (ix86_output_indirect_branch_via_reg): Likewise.
9126 (ix86_output_indirect_branch_via_push): Likewise.
9127 (ix86_output_indirect_branch): Likewise.
9128 (ix86_output_indirect_jmp): Likewise.
9129 (ix86_code_end): Call output_indirect_thunk_function if needed.
9130 (ix86_output_call_insn): Call ix86_output_indirect_branch if
9132 (ix86_handle_fndecl_attribute): Handle indirect_branch.
9133 (ix86_attribute_table): Add indirect_branch.
9134 * config/i386/i386.h (machine_function): Add indirect_branch_type
9135 and has_local_indirect_jump.
9136 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
9138 (tablejump): Likewise.
9139 (*indirect_jump): Use ix86_output_indirect_jmp.
9140 (*tablejump_1): Likewise.
9141 (simple_return_indirect_internal): Likewise.
9142 * config/i386/i386.opt (mindirect-branch=): New option.
9143 (indirect_branch): New.
9146 (thunk-inline): Likewise.
9147 (thunk-extern): Likewise.
9148 * doc/extend.texi: Document indirect_branch function attribute.
9149 * doc/invoke.texi: Document -mindirect-branch= option.
9151 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
9154 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
9156 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
9158 * ipa-inline.c (want_inline_small_function_p): Return false if
9159 inlining has already failed with CIF_FINAL_ERROR.
9160 (update_caller_keys): Call want_inline_small_function_p before
9162 (update_callee_keys): Likewise.
9164 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9166 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
9168 (rs6000_quadword_masked_address_p): Likewise.
9169 (quad_aligned_load_p): Likewise.
9170 (quad_aligned_store_p): Likewise.
9171 (const_load_sequence_p): Add comment to describe the outer-most loop.
9172 (mimic_memory_attributes_and_flags): New function.
9173 (rs6000_gen_stvx): Likewise.
9174 (replace_swapped_aligned_store): Likewise.
9175 (rs6000_gen_lvx): Likewise.
9176 (replace_swapped_aligned_load): Likewise.
9177 (replace_swapped_load_constant): Capitalize argument name in
9178 comment describing this function.
9179 (rs6000_analyze_swaps): Add a third pass to search for vector loads
9180 and stores that access quad-word aligned addresses and replace
9181 with stvx or lvx instructions when appropriate.
9182 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
9183 New function prototype.
9184 (rs6000_quadword_masked_address_p): Likewise.
9185 (rs6000_gen_lvx): Likewise.
9186 (rs6000_gen_stvx): Likewise.
9187 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
9188 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
9189 when memory address is aligned.
9190 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
9191 this split to select lvx instruction when memory address is aligned.
9192 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
9193 instruction when memory address is aligned.
9194 (*vsx_le_perm_load_v16qi): Likewise.
9195 (four unnamed splitters): Modify to select the stvx instruction
9196 when memory is aligned.
9198 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
9200 * predict.c (determine_unlikely_bbs): Handle correctly BBs
9201 which appears in the queue multiple times.
9203 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9204 Alan Hayward <alan.hayward@arm.com>
9205 David Sherwood <david.sherwood@arm.com>
9207 * tree-vectorizer.h (vec_lower_bound): New structure.
9208 (_loop_vec_info): Add check_nonzero and lower_bounds.
9209 (LOOP_VINFO_CHECK_NONZERO): New macro.
9210 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
9211 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
9212 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
9213 fields. Make seg_len the distance travelled, not including the
9215 (dr_direction_indicator): Declare.
9216 (dr_zero_step_indicator): Likewise.
9217 (dr_known_forward_stride_p): Likewise.
9218 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
9220 (runtime_alias_check_p): Allow runtime alias checks with
9222 (operator ==): Compare access_size and align.
9223 (prune_runtime_alias_test_list): Rework for new distinction between
9224 the access_size and seg_len.
9225 (create_intersect_range_checks_index): Likewise. Cope with polynomial
9227 (get_segment_min_max): New function.
9228 (create_intersect_range_checks): Use it.
9229 (dr_step_indicator): New function.
9230 (dr_direction_indicator): Likewise.
9231 (dr_zero_step_indicator): Likewise.
9232 (dr_known_forward_stride_p): Likewise.
9233 * tree-loop-distribution.c (data_ref_segment_size): Return
9234 DR_STEP * (niters - 1).
9235 (compute_alias_check_pairs): Update call to the dr_with_seg_len
9237 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
9238 (vect_preserves_scalar_order_p): New function, split out from...
9239 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
9240 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
9241 (vect_vfa_access_size): New function.
9242 (vect_vfa_align): Likewise.
9243 (vect_compile_time_alias): Take access_size_a and access_b arguments.
9244 (dump_lower_bound): New function.
9245 (vect_check_lower_bound): Likewise.
9246 (vect_small_gap_p): Likewise.
9247 (vectorizable_with_step_bound_p): Likewise.
9248 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
9249 depencies if the vectorization factor is 1. Convert the checks
9250 for nonzero steps into checks on the bounds of DR_STEP. Try using
9251 a bunds check for variable steps if the minimum required step is
9252 relatively small. Update calls to the dr_with_seg_len
9253 constructor and to vect_compile_time_alias.
9254 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
9256 (vect_loop_versioning): Call it.
9257 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
9259 (vect_estimate_min_profitable_iters): Account for any bounds checks.
9261 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9262 Alan Hayward <alan.hayward@arm.com>
9263 David Sherwood <david.sherwood@arm.com>
9265 * doc/sourcebuild.texi (vect_scatter_store): Document.
9266 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
9268 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
9270 * genopinit.c (main): Add supports_vec_scatter_store and
9271 supports_vec_scatter_store_cached to target_optabs.
9272 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
9273 IFN_MASK_SCATTER_STORE.
9274 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
9276 * internal-fn.h (internal_store_fn_p): Declare.
9277 (internal_fn_stored_value_index): Likewise.
9278 * internal-fn.c (scatter_store_direct): New macro.
9279 (expand_scatter_store_optab_fn): New function.
9280 (direct_scatter_store_optab_supported_p): New macro.
9281 (internal_store_fn_p): New function.
9282 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
9283 IFN_MASK_SCATTER_STORE.
9284 (internal_fn_mask_index): Likewise.
9285 (internal_fn_stored_value_index): New function.
9286 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
9288 * optabs-query.h (supports_vec_scatter_store_p): Declare.
9289 * optabs-query.c (supports_vec_scatter_store_p): New function.
9290 * tree-vectorizer.h (vect_get_store_rhs): Declare.
9291 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
9292 true for scatter stores.
9293 (vect_gather_scatter_fn_p): Handle scatter stores too.
9294 (vect_check_gather_scatter): Consider using scatter stores if
9295 supports_vec_scatter_store_p.
9296 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
9298 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9299 internal_fn_stored_value_index.
9300 (check_load_store_masking): Handle scatter stores too.
9301 (vect_get_store_rhs): Make public.
9302 (vectorizable_call): Use internal_store_fn_p.
9303 (vectorizable_store): Handle scatter store internal functions.
9304 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
9305 when deciding whether the end of the group has been reached.
9306 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
9307 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
9308 (mask_scatter_store<mode>): New insns.
9310 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9311 Alan Hayward <alan.hayward@arm.com>
9312 David Sherwood <david.sherwood@arm.com>
9314 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
9315 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
9316 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
9318 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
9319 Use vect_truncate_gather_scatter_offset if we can't treat the
9320 operation as a normal gather load or scatter store.
9321 (get_group_load_store_type): Take the gather_scatter_info
9322 as argument. Try using a gather load or scatter store for
9323 single-element groups.
9324 (get_load_store_type): Update calls to get_group_load_store_type
9325 and vect_use_strided_gather_scatters_p.
9327 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9328 Alan Hayward <alan.hayward@arm.com>
9329 David Sherwood <david.sherwood@arm.com>
9331 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
9332 optional tree argument.
9333 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
9335 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
9336 but continue to use the current value as a fallback.
9337 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
9338 to compare the updates.
9339 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
9340 (get_load_store_type): Use it when handling a strided access.
9341 (vect_get_strided_load_store_ops): New function.
9342 (vect_get_data_ptr_increment): Likewise.
9343 (vectorizable_load): Handle strided gather loads. Always pass
9344 a step to vect_create_data_ref_ptr and bump_vector_ptr.
9346 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9347 Alan Hayward <alan.hayward@arm.com>
9348 David Sherwood <david.sherwood@arm.com>
9350 * doc/md.texi (gather_load@var{m}): Document.
9351 (mask_gather_load@var{m}): Likewise.
9352 * genopinit.c (main): Add supports_vec_gather_load and
9353 supports_vec_gather_load_cached to target_optabs.
9354 * optabs-tree.c (init_tree_optimization_optabs): Use
9355 ggc_cleared_alloc to allocate target_optabs.
9356 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
9357 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
9359 * internal-fn.h (internal_load_fn_p): Declare.
9360 (internal_gather_scatter_fn_p): Likewise.
9361 (internal_fn_mask_index): Likewise.
9362 (internal_gather_scatter_fn_supported_p): Likewise.
9363 * internal-fn.c (gather_load_direct): New macro.
9364 (expand_gather_load_optab_fn): New function.
9365 (direct_gather_load_optab_supported_p): New macro.
9366 (direct_internal_fn_optab): New function.
9367 (internal_load_fn_p): Likewise.
9368 (internal_gather_scatter_fn_p): Likewise.
9369 (internal_fn_mask_index): Likewise.
9370 (internal_gather_scatter_fn_supported_p): Likewise.
9371 * optabs-query.c (supports_at_least_one_mode_p): New function.
9372 (supports_vec_gather_load_p): Likewise.
9373 * optabs-query.h (supports_vec_gather_load_p): Declare.
9374 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
9375 and memory_type field.
9376 (NUM_PATTERNS): Bump to 15.
9377 * tree-vect-data-refs.c: Include internal-fn.h.
9378 (vect_gather_scatter_fn_p): New function.
9379 (vect_describe_gather_scatter_call): Likewise.
9380 (vect_check_gather_scatter): Try using internal functions for
9381 gather loads. Recognize existing calls to a gather load function.
9382 (vect_analyze_data_refs): Consider using gather loads if
9383 supports_vec_gather_load_p.
9384 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
9385 (vect_get_gather_scatter_offset_type): Likewise.
9386 (vect_convert_mask_for_vectype): Likewise.
9387 (vect_add_conversion_to_patterm): Likewise.
9388 (vect_try_gather_scatter_pattern): Likewise.
9389 (vect_recog_gather_scatter_pattern): New pattern recognizer.
9390 (vect_vect_recog_func_ptrs): Add it.
9391 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9392 internal_fn_mask_index and internal_gather_scatter_fn_p.
9393 (check_load_store_masking): Take the gather_scatter_info as an
9394 argument and handle gather loads.
9395 (vect_get_gather_scatter_ops): New function.
9396 (vectorizable_call): Check internal_load_fn_p.
9397 (vectorizable_load): Likewise. Handle gather load internal
9399 (vectorizable_store): Update call to check_load_store_masking.
9400 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
9401 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
9402 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
9403 (aarch64_gather_scale_operand_d): New predicates.
9404 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
9405 (mask_gather_load<mode>): New insns.
9407 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9408 Alan Hayward <alan.hayward@arm.com>
9409 David Sherwood <david.sherwood@arm.com>
9411 * optabs.def (fold_left_plus_optab): New optab.
9412 * doc/md.texi (fold_left_plus_@var{m}): Document.
9413 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9414 * internal-fn.c (fold_left_direct): Define.
9415 (expand_fold_left_optab_fn): Likewise.
9416 (direct_fold_left_optab_supported_p): Likewise.
9417 * fold-const-call.c (fold_const_fold_left): New function.
9418 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9419 * tree-parloops.c (valid_reduction_p): New function.
9420 (gather_scalar_reductions): Use it.
9421 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9422 (vect_finish_replace_stmt): Declare.
9423 * tree-vect-loop.c (fold_left_reduction_fn): New function.
9424 (needs_fold_left_reduction_p): New function, split out from...
9425 (vect_is_simple_reduction): ...here. Accept reductions that
9426 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9427 (vect_force_simple_reduction): Also store the reduction type in
9428 the assignment's STMT_VINFO_REDUC_TYPE.
9429 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9430 (merge_with_identity): New function.
9431 (vect_expand_fold_left): Likewise.
9432 (vectorize_fold_left_reduction): Likewise.
9433 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9434 scalar phi in place for it. Check for target support and reject
9435 cases that would reassociate the operation. Defer the transform
9436 phase to vectorize_fold_left_reduction.
9437 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9438 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9439 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9441 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9443 * tree-if-conv.c (predicate_mem_writes): Remove redundant
9444 call to ifc_temp_var.
9446 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9447 Alan Hayward <alan.hayward@arm.com>
9448 David Sherwood <david.sherwood@arm.com>
9450 * target.def (legitimize_address_displacement): Take the original
9451 offset as a poly_int.
9452 * targhooks.h (default_legitimize_address_displacement): Update
9454 * targhooks.c (default_legitimize_address_displacement): Likewise.
9455 * doc/tm.texi: Regenerate.
9456 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9457 as an argument, moving assert of ad->disp == ad->disp_term to...
9458 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
9459 Try calling targetm.legitimize_address_displacement before expanding
9460 the address rather than afterwards, and adjust for the new interface.
9461 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9462 Match the new hook interface. Handle SVE addresses.
9463 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9466 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9468 * Makefile.in (OBJS): Add early-remat.o.
9469 * target.def (select_early_remat_modes): New hook.
9470 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9471 * doc/tm.texi: Regenerate.
9472 * targhooks.h (default_select_early_remat_modes): Declare.
9473 * targhooks.c (default_select_early_remat_modes): New function.
9474 * timevar.def (TV_EARLY_REMAT): New timevar.
9475 * passes.def (pass_early_remat): New pass.
9476 * tree-pass.h (make_pass_early_remat): Declare.
9477 * early-remat.c: New file.
9478 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9480 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
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-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9487 vfm1 with a bound_epilog parameter.
9488 (vect_do_peeling): Update calls accordingly, and move the prologue
9489 call earlier in the function. Treat the base bound_epilog as 0 for
9490 fully-masked loops and retain vf - 1 for other loops. Add 1 to
9491 this base when peeling for gaps.
9492 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9493 with fully-masked loops.
9494 (vect_estimate_min_profitable_iters): Handle the single peeled
9495 iteration in that case.
9497 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9498 Alan Hayward <alan.hayward@arm.com>
9499 David Sherwood <david.sherwood@arm.com>
9501 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9502 single-element interleaving even if the size is not a power of 2.
9503 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9504 accesses for single-element interleaving if the group size is
9507 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9508 Alan Hayward <alan.hayward@arm.com>
9509 David Sherwood <david.sherwood@arm.com>
9511 * doc/md.texi (fold_extract_last_@var{m}): Document.
9512 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9513 * optabs.def (fold_extract_last_optab): New optab.
9514 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9515 * internal-fn.c (fold_extract_direct): New macro.
9516 (expand_fold_extract_optab_fn): Likewise.
9517 (direct_fold_extract_optab_supported_p): Likewise.
9518 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9519 * tree-vect-loop.c (vect_model_reduction_cost): Handle
9520 EXTRACT_LAST_REDUCTION.
9521 (get_initial_def_for_reduction): Do not create an initial vector
9522 for EXTRACT_LAST_REDUCTION reductions.
9523 (vectorizable_reduction): Leave the scalar phi in place for
9524 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
9525 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
9526 epilogue code for EXTRACT_LAST_REDUCTION and defer the
9527 transform phase to vectorizable_condition.
9528 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9530 (vect_finish_stmt_generation): ...here.
9531 (vect_finish_replace_stmt): New function.
9532 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9533 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9535 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9537 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9538 Alan Hayward <alan.hayward@arm.com>
9539 David Sherwood <david.sherwood@arm.com>
9541 * doc/md.texi (extract_last_@var{m}): Document.
9542 * optabs.def (extract_last_optab): New optab.
9543 * internal-fn.def (EXTRACT_LAST): New internal function.
9544 * internal-fn.c (cond_unary_direct): New macro.
9545 (expand_cond_unary_optab_fn): Likewise.
9546 (direct_cond_unary_optab_supported_p): Likewise.
9547 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9548 loops using EXTRACT_LAST.
9549 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9550 (extract_last_<mode>): ...this optab.
9551 (vec_extract<mode><Vel>): Update accordingly.
9553 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9554 Alan Hayward <alan.hayward@arm.com>
9555 David Sherwood <david.sherwood@arm.com>
9557 * target.def (empty_mask_is_expensive): New hook.
9558 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9559 * doc/tm.texi: Regenerate.
9560 * targhooks.h (default_empty_mask_is_expensive): Declare.
9561 * targhooks.c (default_empty_mask_is_expensive): New function.
9562 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9563 if the target says that empty masks are expensive.
9564 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9566 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9568 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9569 Alan Hayward <alan.hayward@arm.com>
9570 David Sherwood <david.sherwood@arm.com>
9572 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9573 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9574 (vect_use_loop_mask_for_alignment_p): New function.
9575 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9576 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9577 niters_skip argument. Make sure that the first niters_skip elements
9578 of the first iteration are inactive.
9579 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9580 Update call to vect_set_loop_masks_directly.
9581 (get_misalign_in_elems): New function, split out from...
9582 (vect_gen_prolog_loop_niters): ...here.
9583 (vect_update_init_of_dr): Take a code argument that specifies whether
9584 the adjustment should be added or subtracted.
9585 (vect_update_init_of_drs): Likewise.
9586 (vect_prepare_for_masked_peels): New function.
9587 (vect_do_peeling): Skip prologue peeling if we're using a mask
9588 instead. Update call to vect_update_inits_of_drs.
9589 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9591 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9592 alignment. Do not include the number of peeled iterations in
9593 the minimum threshold in that case.
9594 (vectorizable_induction): Adjust the start value down by
9595 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9596 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9597 Take the number of skipped iterations into account when calculating
9599 * tree-vect-stmts.c (vect_gen_while_not): New function.
9601 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9602 Alan Hayward <alan.hayward@arm.com>
9603 David Sherwood <david.sherwood@arm.com>
9605 * doc/sourcebuild.texi (vect_fully_masked): Document.
9606 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9608 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9610 (vect_analyze_loop_2): ...here. Don't check the vectorization
9611 factor against the number of loop iterations if the loop is
9614 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9615 Alan Hayward <alan.hayward@arm.com>
9616 David Sherwood <david.sherwood@arm.com>
9618 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9619 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9620 (dump_groups): Update accordingly.
9621 (iv_use::mem_type): New member variable.
9622 (address_p): New function.
9623 (record_use): Add a mem_type argument and initialize the new
9625 (record_group_use): Add a mem_type argument. Use address_p.
9626 Remove obsolete null checks of base_object. Update call to record_use.
9627 (find_interesting_uses_op): Update call to record_group_use.
9628 (find_interesting_uses_cond): Likewise.
9629 (find_interesting_uses_address): Likewise.
9630 (get_mem_type_for_internal_fn): New function.
9631 (find_address_like_use): Likewise.
9632 (find_interesting_uses_stmt): Try find_address_like_use before
9633 calling find_interesting_uses_op.
9634 (addr_offset_valid_p): Use the iv mem_type field as the type
9635 of the addressed memory.
9636 (add_autoinc_candidates): Likewise.
9637 (get_address_cost): Likewise.
9638 (split_small_address_groups_p): Use address_p.
9639 (split_address_groups): Likewise.
9640 (add_iv_candidate_for_use): Likewise.
9641 (autoinc_possible_for_pair): Likewise.
9642 (rewrite_groups): Likewise.
9643 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9644 (determine_group_iv_cost): Update after split of USE_ADDRESS.
9645 (get_alias_ptr_type_for_ptr_address): New function.
9646 (rewrite_use_address): Rewrite address uses in calls that were
9647 identified by find_address_like_use.
9649 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9650 Alan Hayward <alan.hayward@arm.com>
9651 David Sherwood <david.sherwood@arm.com>
9653 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9655 * gimple-expr.h (is_gimple_addressable: Likewise.
9656 * gimple-expr.c (is_gimple_address): Likewise.
9657 * internal-fn.c (expand_call_mem_ref): New function.
9658 (expand_mask_load_optab_fn): Use it.
9659 (expand_mask_store_optab_fn): Likewise.
9661 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9662 Alan Hayward <alan.hayward@arm.com>
9663 David Sherwood <david.sherwood@arm.com>
9665 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
9666 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
9667 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
9668 (cond_umax@var{mode}): Document.
9669 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
9670 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
9671 (cond_umin_optab, cond_umax_optab): New optabs.
9672 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
9673 (COND_IOR, COND_XOR): New internal functions.
9674 * internal-fn.h (get_conditional_internal_fn): Declare.
9675 * internal-fn.c (cond_binary_direct): New macro.
9676 (expand_cond_binary_optab_fn): Likewise.
9677 (direct_cond_binary_optab_supported_p): Likewise.
9678 (get_conditional_internal_fn): New function.
9679 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9680 Cope with reduction statements that are vectorized as calls rather
9682 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9683 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9684 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9685 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9686 (UNSPEC_COND_EOR): New unspecs.
9687 (optab): Add mappings for them.
9688 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9689 (sve_int_op, sve_fp_op): New int attributes.
9691 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9692 Alan Hayward <alan.hayward@arm.com>
9693 David Sherwood <david.sherwood@arm.com>
9695 * optabs.def (while_ult_optab): New optab.
9696 * doc/md.texi (while_ult@var{m}@var{n}): Document.
9697 * internal-fn.def (WHILE_ULT): New internal function.
9698 * internal-fn.h (direct_internal_fn_supported_p): New override
9699 that takes two types as argument.
9700 * internal-fn.c (while_direct): New macro.
9701 (expand_while_optab_fn): New function.
9702 (convert_optab_supported_p): Likewise.
9703 (direct_while_optab_supported_p): New macro.
9704 * wide-int.h (wi::udiv_ceil): New function.
9705 * tree-vectorizer.h (rgroup_masks): New structure.
9706 (vec_loop_masks): New typedef.
9707 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
9709 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
9710 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
9711 (vect_max_vf): New function.
9712 (slpeel_make_loop_iterate_ntimes): Delete.
9713 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
9714 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
9715 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
9716 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
9717 internal-fn.h, stor-layout.h and optabs-query.h.
9718 (vect_set_loop_mask): New function.
9719 (add_preheader_seq): Likewise.
9720 (add_header_seq): Likewise.
9721 (interleave_supported_p): Likewise.
9722 (vect_maybe_permute_loop_masks): Likewise.
9723 (vect_set_loop_masks_directly): Likewise.
9724 (vect_set_loop_condition_masked): Likewise.
9725 (vect_set_loop_condition_unmasked): New function, split out from
9726 slpeel_make_loop_iterate_ntimes.
9727 (slpeel_make_loop_iterate_ntimes): Rename to..
9728 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
9729 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
9730 (vect_do_peeling): Update call accordingly.
9731 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
9733 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9734 mask_compare_type, can_fully_mask_p and fully_masked_p.
9735 (release_vec_loop_masks): New function.
9736 (_loop_vec_info): Use it to free the loop masks.
9737 (can_produce_all_loop_masks_p): New function.
9738 (vect_get_max_nscalars_per_iter): Likewise.
9739 (vect_verify_full_masking): Likewise.
9740 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
9741 retries, and free the mask rgroups before retrying. Check loop-wide
9742 reasons for disallowing fully-masked loops. Make the final decision
9743 about whether use a fully-masked loop or not.
9744 (vect_estimate_min_profitable_iters): Do not assume that peeling
9745 for the number of iterations will be needed for fully-masked loops.
9746 (vectorizable_reduction): Disable fully-masked loops.
9747 (vectorizable_live_operation): Likewise.
9748 (vect_halve_mask_nunits): New function.
9749 (vect_double_mask_nunits): Likewise.
9750 (vect_record_loop_mask): Likewise.
9751 (vect_get_loop_mask): Likewise.
9752 (vect_transform_loop): Handle the case in which the final loop
9753 iteration might handle a partial vector. Call vect_set_loop_condition
9754 instead of slpeel_make_loop_iterate_ntimes.
9755 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
9756 (check_load_store_masking): New function.
9757 (prepare_load_store_mask): Likewise.
9758 (vectorizable_store): Handle fully-masked loops.
9759 (vectorizable_load): Likewise.
9760 (supportable_widening_operation): Use vect_halve_mask_nunits for
9762 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
9763 (vect_gen_while): New function.
9764 * config/aarch64/aarch64.md (umax<mode>3): New expander.
9765 (aarch64_uqdec<mode>): New insn.
9767 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9768 Alan Hayward <alan.hayward@arm.com>
9769 David Sherwood <david.sherwood@arm.com>
9771 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
9772 (reduc_xor_scal_optab): New optabs.
9773 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
9774 (reduc_xor_scal_@var{m}): Document.
9775 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
9776 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
9778 * fold-const-call.c (fold_const_call): Handle them.
9779 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
9780 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
9781 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
9782 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
9783 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
9784 (UNSPEC_XORV): New unspecs.
9785 (optab): Add entries for them.
9786 (BITWISEV): New int iterator.
9787 (bit_reduc_op): New int attributes.
9789 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9790 Alan Hayward <alan.hayward@arm.com>
9791 David Sherwood <david.sherwood@arm.com>
9793 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
9794 * internal-fn.def (VEC_SHL_INSERT): New internal function.
9795 * optabs.def (vec_shl_insert_optab): New optab.
9796 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
9797 (duplicate_and_interleave): Likewise.
9798 * tree-vect-loop.c: Include internal-fn.h.
9799 (neutral_op_for_slp_reduction): New function, split out from
9800 get_initial_defs_for_reduction.
9801 (get_initial_def_for_reduction): Handle option 2 for variable-length
9802 vectors by loading the neutral value into a vector and then shifting
9803 the initial value into element 0.
9804 (get_initial_defs_for_reduction): Replace the code argument with
9805 the neutral value calculated by neutral_op_for_slp_reduction.
9806 Use gimple_build_vector for constant-length vectors.
9807 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
9808 but the first group_size elements have a neutral value.
9809 Use duplicate_and_interleave otherwise.
9810 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
9811 Update call to get_initial_defs_for_reduction. Handle SLP
9812 reductions for variable-length vectors by creating one vector
9813 result for each scalar result, with the elements associated
9814 with other scalar results stubbed out with the neutral value.
9815 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
9816 Require IFN_VEC_SHL_INSERT for double reductions on
9817 variable-length vectors, or SLP reductions that have
9818 a neutral value. Require can_duplicate_and_interleave_p
9819 support for variable-length unchained SLP reductions if there
9820 is no neutral value, such as for MIN/MAX reductions. Also require
9821 the number of vector elements to be a multiple of the number of
9822 SLP statements when doing variable-length unchained SLP reductions.
9823 Update call to vect_create_epilog_for_reduction.
9824 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
9825 and remove initial values.
9826 (duplicate_and_interleave): Make public.
9827 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
9828 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
9830 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9831 Alan Hayward <alan.hayward@arm.com>
9832 David Sherwood <david.sherwood@arm.com>
9834 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
9835 (can_duplicate_and_interleave_p): New function.
9836 (vect_get_and_check_slp_defs): Take the vector of statements
9837 rather than just the current one. Remove excess parentheses.
9838 Restriction rejectinon of vect_constant_def and vect_external_def
9839 for variable-length vectors to boolean types, or types for which
9840 can_duplicate_and_interleave_p is false.
9841 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
9842 (duplicate_and_interleave): New function.
9843 (vect_get_constant_vectors): Use gimple_build_vector for
9844 constant-length vectors and suitable variable-length constant
9845 vectors. Use duplicate_and_interleave for other variable-length
9846 vectors. Don't defer the update when inserting new statements.
9848 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9849 Alan Hayward <alan.hayward@arm.com>
9850 David Sherwood <david.sherwood@arm.com>
9852 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
9853 min_profitable_iters doesn't go negative.
9855 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9856 Alan Hayward <alan.hayward@arm.com>
9857 David Sherwood <david.sherwood@arm.com>
9859 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
9860 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
9861 * optabs.def (vec_mask_load_lanes_optab): New optab.
9862 (vec_mask_store_lanes_optab): Likewise.
9863 * internal-fn.def (MASK_LOAD_LANES): New internal function.
9864 (MASK_STORE_LANES): Likewise.
9865 * internal-fn.c (mask_load_lanes_direct): New macro.
9866 (mask_store_lanes_direct): Likewise.
9867 (expand_mask_load_optab_fn): Handle masked operations.
9868 (expand_mask_load_lanes_optab_fn): New macro.
9869 (expand_mask_store_optab_fn): Handle masked operations.
9870 (expand_mask_store_lanes_optab_fn): New macro.
9871 (direct_mask_load_lanes_optab_supported_p): Likewise.
9872 (direct_mask_store_lanes_optab_supported_p): Likewise.
9873 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
9875 (vect_load_lanes_supported): Likewise.
9876 * tree-vect-data-refs.c (strip_conversion): New function.
9877 (can_group_stmts_p): Likewise.
9878 (vect_analyze_data_ref_accesses): Use it instead of checking
9879 for a pair of assignments.
9880 (vect_store_lanes_supported): Take a masked_p parameter.
9881 (vect_load_lanes_supported): Likewise.
9882 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
9883 vect_store_lanes_supported and vect_load_lanes_supported.
9884 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9885 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
9886 parameter. Don't allow gaps for masked accesses.
9887 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
9888 and vect_load_lanes_supported.
9889 (get_load_store_type): Take a masked_p parameter and update
9890 call to get_group_load_store_type.
9891 (vectorizable_store): Update call to get_load_store_type.
9892 Handle IFN_MASK_STORE_LANES.
9893 (vectorizable_load): Update call to get_load_store_type.
9894 Handle IFN_MASK_LOAD_LANES.
9896 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9897 Alan Hayward <alan.hayward@arm.com>
9898 David Sherwood <david.sherwood@arm.com>
9900 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
9902 * config/aarch64/aarch64-protos.h
9903 (aarch64_sve_struct_memory_operand_p): Declare.
9904 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
9905 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
9906 (VPRED, vpred): Handle SVE structure modes.
9907 * config/aarch64/constraints.md (Utx): New constraint.
9908 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
9909 (aarch64_sve_struct_nonimmediate_operand): New predicates.
9910 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
9911 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
9912 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
9913 structure modes. Split into pieces after RA.
9914 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
9915 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
9917 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
9918 SVE structure modes.
9919 (aarch64_classify_address): Likewise.
9920 (sizetochar): Move earlier in file.
9921 (aarch64_print_operand): Handle SVE register lists.
9922 (aarch64_array_mode): New function.
9923 (aarch64_sve_struct_memory_operand_p): Likewise.
9924 (TARGET_ARRAY_MODE): Redefine.
9926 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9927 Alan Hayward <alan.hayward@arm.com>
9928 David Sherwood <david.sherwood@arm.com>
9930 * target.def (array_mode): New target hook.
9931 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
9932 * doc/tm.texi: Regenerate.
9933 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
9934 * hooks.c (hook_optmode_mode_uhwi_none): New function.
9935 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
9937 * stor-layout.c (mode_for_array): Likewise. Support polynomial
9940 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9941 Alan Hayward <alan.hayward@arm.com>
9942 David Sherwood <david.sherwood@arm.com>
9944 * fold-const.c (fold_binary_loc): Check the argument types
9945 rather than the result type when testing for a vector operation.
9947 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9949 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
9950 * doc/tm.texi: Regenerate.
9952 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9953 Alan Hayward <alan.hayward@arm.com>
9954 David Sherwood <david.sherwood@arm.com>
9956 * doc/invoke.texi (-msve-vector-bits=): Document new option.
9957 (sve): Document new AArch64 extension.
9958 * doc/md.texi (w): Extend the description of the AArch64
9959 constraint to include SVE vectors.
9960 (Upl, Upa): Document new AArch64 predicate constraints.
9961 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
9963 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
9964 (msve-vector-bits=): New option.
9965 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
9966 SVE when these are disabled.
9967 (sve): New extension.
9968 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
9969 modes. Adjust their number of units based on aarch64_sve_vg.
9970 (MAX_BITSIZE_MODE_ANY_MODE): Define.
9971 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
9972 aarch64_addr_query_type.
9973 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
9974 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
9975 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
9976 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
9977 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
9978 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
9979 (aarch64_simd_imm_zero_p): Delete.
9980 (aarch64_check_zero_based_sve_index_immediate): Declare.
9981 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9982 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9983 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9984 (aarch64_sve_float_mul_immediate_p): Likewise.
9985 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9987 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
9988 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
9989 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
9990 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
9991 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
9992 (aarch64_regmode_natural_size): Likewise.
9993 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
9994 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
9996 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
9997 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
9998 for VG and the SVE predicate registers.
9999 (V_ALIASES): Add a "z"-prefixed alias.
10000 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
10001 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
10002 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
10003 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
10004 (REG_CLASS_NAMES): Add entries for them.
10005 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
10006 and the predicate registers.
10007 (aarch64_sve_vg): Declare.
10008 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
10009 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
10010 (REGMODE_NATURAL_SIZE): Define.
10011 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
10013 * config/aarch64/aarch64.c: Include cfgrtl.h.
10014 (simd_immediate_info): Add a constructor for series vectors,
10015 and an associated step field.
10016 (aarch64_sve_vg): New variable.
10017 (aarch64_dbx_register_number): Handle VG and the predicate registers.
10018 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
10019 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
10020 (VEC_ANY_DATA, VEC_STRUCT): New constants.
10021 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
10022 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
10023 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
10024 (aarch64_get_mask_mode): New functions.
10025 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
10026 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
10027 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
10028 predicate modes and predicate registers. Explicitly restrict
10029 GPRs to modes of 16 bytes or smaller. Only allow FP registers
10030 to store a vector mode if it is recognized by
10031 aarch64_classify_vector_mode.
10032 (aarch64_regmode_natural_size): New function.
10033 (aarch64_hard_regno_caller_save_mode): Return the original mode
10035 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
10036 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
10037 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
10038 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
10040 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
10041 does not overlap dest if the function is frame-related. Handle
10043 (aarch64_split_add_offset): New function.
10044 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
10045 them aarch64_add_offset.
10046 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
10047 and update call to aarch64_sub_sp.
10048 (aarch64_add_cfa_expression): New function.
10049 (aarch64_expand_prologue): Pass extra temporary registers to the
10050 functions above. Handle the case in which we need to emit new
10051 DW_CFA_expressions for registers that were originally saved
10052 relative to the stack pointer, but now have to be expressed
10053 relative to the frame pointer.
10054 (aarch64_output_mi_thunk): Pass extra temporary registers to the
10056 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
10057 IP0 and IP1 values for SVE frames.
10058 (aarch64_expand_vec_series): New function.
10059 (aarch64_expand_sve_widened_duplicate): Likewise.
10060 (aarch64_expand_sve_const_vector): Likewise.
10061 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
10062 Handle SVE constants. Use emit_move_insn to move a force_const_mem
10063 into the register, rather than emitting a SET directly.
10064 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
10065 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
10066 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
10067 (offset_9bit_signed_scaled_p): New functions.
10068 (aarch64_replicate_bitmask_imm): New function.
10069 (aarch64_bitmask_imm): Use it.
10070 (aarch64_cannot_force_const_mem): Reject expressions involving
10071 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
10072 (aarch64_classify_index): Handle SVE indices, by requiring
10073 a plain register index with a scale that matches the element size.
10074 (aarch64_classify_address): Handle SVE addresses. Assert that
10075 the mode of the address is VOIDmode or an integer mode.
10076 Update call to aarch64_classify_symbol.
10077 (aarch64_classify_symbolic_expression): Update call to
10078 aarch64_classify_symbol.
10079 (aarch64_const_vec_all_in_range_p): New function.
10080 (aarch64_print_vector_float_operand): Likewise.
10081 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
10082 "vN" for FP registers with SVE modes. Handle (const ...) vectors
10083 and the FP immediates 1.0 and 0.5.
10084 (aarch64_print_address_internal): Handle SVE addresses.
10085 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
10086 (aarch64_regno_regclass): Handle predicate registers.
10087 (aarch64_secondary_reload): Handle big-endian reloads of SVE
10089 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
10090 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
10091 (aarch64_convert_sve_vector_bits): New function.
10092 (aarch64_override_options): Use it to handle -msve-vector-bits=.
10093 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10094 rather than an rtx.
10095 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
10096 Handle SVE vector and predicate modes. Accept VL-based constants
10097 that need only one temporary register, and VL offsets that require
10098 no temporary registers.
10099 (aarch64_conditional_register_usage): Mark the predicate registers
10100 as fixed if SVE isn't available.
10101 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
10102 Return true for SVE vector and predicate modes.
10103 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
10104 rather than an unsigned int. Handle SVE modes.
10105 (aarch64_preferred_simd_mode): Update call accordingly. Handle
10107 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
10109 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10110 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10111 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10112 (aarch64_sve_float_mul_immediate_p): New functions.
10113 (aarch64_sve_valid_immediate): New function.
10114 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
10115 Explicitly reject structure modes. Check for INDEX constants.
10116 Handle PTRUE and PFALSE constants.
10117 (aarch64_check_zero_based_sve_index_immediate): New function.
10118 (aarch64_simd_imm_zero_p): Delete.
10119 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
10120 vector modes. Accept constants in the range of CNT[BHWD].
10121 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
10122 ask for an Advanced SIMD mode.
10123 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
10124 (aarch64_simd_vector_alignment): Handle SVE predicates.
10125 (aarch64_vectorize_preferred_vector_alignment): New function.
10126 (aarch64_simd_vector_alignment_reachable): Use it instead of
10128 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
10129 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
10131 (MAX_VECT_LEN): Delete.
10132 (expand_vec_perm_d): Add a vec_flags field.
10133 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
10134 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
10135 (aarch64_evpc_ext): Don't apply a big-endian lane correction
10137 (aarch64_evpc_rev): Rename to...
10138 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
10139 (aarch64_evpc_rev_global): New function.
10140 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
10141 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
10143 (aarch64_evpc_sve_tbl): New function.
10144 (aarch64_expand_vec_perm_const_1): Update after rename of
10145 aarch64_evpc_rev. Handle SVE permutes too, trying
10146 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
10147 than aarch64_evpc_tbl.
10148 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
10149 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
10150 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
10151 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
10152 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
10153 (aarch64_expand_sve_vcond): New functions.
10154 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
10155 of aarch64_vector_mode_p.
10156 (aarch64_dwarf_poly_indeterminate_value): New function.
10157 (aarch64_compute_pressure_classes): Likewise.
10158 (aarch64_can_change_mode_class): Likewise.
10159 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
10160 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
10161 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
10162 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
10163 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
10164 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
10165 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
10166 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
10168 (Dn, Dl, Dr): Accept const as well as const_vector.
10169 (Dz): Likewise. Compare against CONST0_RTX.
10170 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
10171 of "vector" where appropriate.
10172 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
10173 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
10174 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
10175 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
10176 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
10177 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
10178 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
10179 (v_int_equiv): Extend to SVE modes.
10180 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
10182 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
10183 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
10184 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
10185 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
10186 (SVE_COND_FP_CMP): New int iterators.
10187 (perm_hilo): Handle the new unpack unspecs.
10188 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
10190 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
10191 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
10192 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
10193 (aarch64_equality_operator, aarch64_constant_vector_operand)
10194 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
10195 (aarch64_sve_nonimmediate_operand): Likewise.
10196 (aarch64_sve_general_operand): Likewise.
10197 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
10198 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
10199 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
10200 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
10201 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
10202 (aarch64_sve_float_arith_immediate): Likewise.
10203 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
10204 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
10205 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
10206 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
10207 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
10208 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
10209 (aarch64_sve_float_arith_operand): Likewise.
10210 (aarch64_sve_float_arith_with_sub_operand): Likewise.
10211 (aarch64_sve_float_mul_operand): Likewise.
10212 (aarch64_sve_vec_perm_operand): Likewise.
10213 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
10214 (aarch64_mov_operand): Accept const_poly_int and const_vector.
10215 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
10216 as well as const_vector.
10217 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
10218 in file. Use CONST0_RTX and CONSTM1_RTX.
10219 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
10220 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
10221 Use aarch64_simd_imm_zero.
10222 * config/aarch64/aarch64-sve.md: New file.
10223 * config/aarch64/aarch64.md: Include it.
10224 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
10225 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
10226 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
10227 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
10228 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
10229 (sve): New attribute.
10230 (enabled): Disable instructions with the sve attribute unless
10232 (movqi, movhi): Pass CONST_POLY_INT operaneds through
10233 aarch64_expand_mov_immediate.
10234 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
10235 CNT[BHSD] immediates.
10236 (movti): Split CONST_POLY_INT moves into two halves.
10237 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
10238 Split additions that need a temporary here if the destination
10239 is the stack pointer.
10240 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
10241 (*add<mode>3_poly_1): New instruction.
10242 (set_clobber_cc): New expander.
10244 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10246 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
10247 parameter and use it instead of GET_MODE_SIZE (innermode). Use
10248 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
10249 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
10250 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
10251 Change innermode from fixed_mode_size to machine_mode.
10252 (simplify_subreg): Update call accordingly. Handle a constant-sized
10253 subreg of a variable-length CONST_VECTOR.
10255 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10256 Alan Hayward <alan.hayward@arm.com>
10257 David Sherwood <david.sherwood@arm.com>
10259 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
10260 (add_offset_to_base): New function, split out from...
10261 (create_mem_ref): ...here. When handling a scale other than 1,
10262 check first whether the address is valid without the offset.
10263 Add it into the base if so, leaving the index and scale as-is.
10265 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10268 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
10269 fold_for_warn before checking if arg2 is INTEGER_CST.
10271 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
10273 * config/rs6000/predicates.md (load_multiple_operation): Delete.
10274 (store_multiple_operation): Delete.
10275 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
10276 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
10277 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
10278 guarded by TARGET_STRING.
10279 (rs6000_output_load_multiple): Delete.
10280 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
10281 OPTION_MASK_STRING / TARGET_STRING handling.
10282 (print_operand) <'N', 'O'>: Add comment that these are unused now.
10283 (const rs6000_opt_masks) <"string">: Change mask to 0.
10284 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
10285 (MASK_STRING): Delete.
10286 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
10288 (load_multiple): Delete.
10295 (store_multiple): Delete.
10302 (movmemsi_8reg): Delete.
10303 (corresponding unnamed define_insn): Delete.
10304 (movmemsi_6reg): Delete.
10305 (corresponding unnamed define_insn): Delete.
10306 (movmemsi_4reg): Delete.
10307 (corresponding unnamed define_insn): Delete.
10308 (movmemsi_2reg): Delete.
10309 (corresponding unnamed define_insn): Delete.
10310 (movmemsi_1reg): Delete.
10311 (corresponding unnamed define_insn): Delete.
10312 * config/rs6000/rs6000.opt (mno-string): New.
10313 (mstring): Replace by deprecation warning stub.
10314 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
10316 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10318 * regrename.c (regrename_do_replace): If replacing the same
10319 reg multiple times, try to reuse last created gen_raw_REG.
10322 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
10323 main to workaround a bug in GDB.
10325 2018-01-12 Tom de Vries <tom@codesourcery.com>
10328 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
10330 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
10332 PR rtl-optimization/80481
10333 * ira-color.c (get_cap_member): New function.
10334 (allocnos_conflict_by_live_ranges_p): Use it.
10335 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
10336 (setup_slot_coalesced_allocno_live_ranges): Ditto.
10338 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
10341 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
10342 (*saddl_se_1): Ditto.
10343 (*ssubsi_1): Ditto.
10344 (*ssubl_se_1): Ditto.
10346 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10348 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
10349 rather than wi::to_widest for DR_INITs.
10350 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
10351 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
10352 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
10354 (vect_analyze_group_access_1): Note that here.
10356 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10358 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
10359 polynomial type sizes.
10361 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10363 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
10364 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
10365 (gimple_add_tmp_var): Likewise.
10367 2018-01-12 Martin Liska <mliska@suse.cz>
10369 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
10370 (gimple_alloc_sizes): Likewise.
10371 (dump_gimple_statistics): Use PRIu64 in printf format.
10372 * gimple.h: Change uint64_t to int.
10374 2018-01-12 Martin Liska <mliska@suse.cz>
10376 * tree-core.h: Use uint64_t instead of int.
10377 * tree.c (tree_node_counts): Likewise.
10378 (tree_node_sizes): Likewise.
10379 (dump_tree_statistics): Use PRIu64 in printf format.
10381 2018-01-12 Martin Liska <mliska@suse.cz>
10383 * Makefile.in: As qsort_chk is implemented in vec.c, add
10384 vec.o to linkage of gencfn-macros.
10385 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
10386 passing the info to record_node_allocation_statistics.
10387 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
10389 * ggc-common.c (struct ggc_usage): Add operator== and use
10390 it in operator< and compare function.
10391 * mem-stats.h (struct mem_usage): Likewise.
10392 * vec.c (struct vec_usage): Remove operator< and compare
10393 function. Can be simply inherited.
10395 2018-01-12 Martin Jambor <mjambor@suse.cz>
10398 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
10399 * tree-ssa-math-opts.c: Include domwalk.h.
10400 (convert_mult_to_fma_1): New function.
10401 (fma_transformation_info): New type.
10402 (fma_deferring_state): Likewise.
10403 (cancel_fma_deferring): New function.
10404 (result_of_phi): Likewise.
10405 (last_fma_candidate_feeds_initial_phi): Likewise.
10406 (convert_mult_to_fma): Added deferring logic, split actual
10407 transformation to convert_mult_to_fma_1.
10408 (math_opts_dom_walker): New type.
10409 (math_opts_dom_walker::after_dom_children): New method, body moved
10410 here from pass_optimize_widening_mul::execute, added deferring logic
10412 (pass_optimize_widening_mul::execute): Moved most of code to
10413 math_opts_dom_walker::after_dom_children.
10414 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10415 * config/i386/i386.c (ix86_option_override_internal): Added
10416 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10418 2018-01-12 Richard Biener <rguenther@suse.de>
10421 * dwarf2out.c (gen_variable_die): Do not reset old_die for
10422 inline instance vars.
10424 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
10427 * config/rx/rx.c (rx_is_restricted_memory_address):
10428 Handle SUBREG case.
10430 2018-01-12 Richard Biener <rguenther@suse.de>
10432 PR tree-optimization/80846
10433 * target.def (split_reduction): New target hook.
10434 * targhooks.c (default_split_reduction): New function.
10435 * targhooks.h (default_split_reduction): Declare.
10436 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10437 target requests first reduce vectors by combining low and high
10439 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10440 (get_vectype_for_scalar_type_and_size): Export.
10441 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10442 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10443 * doc/tm.texi: Regenerate.
10444 * config/i386/i386.c (ix86_split_reduction): Implement
10445 TARGET_VECTORIZE_SPLIT_REDUCTION.
10447 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10450 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10451 in PIC mode except for TARGET_VXWORKS_RTP.
10452 * config/sparc/sparc.c: Include cfgrtl.h.
10453 (TARGET_INIT_PIC_REG): Define.
10454 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10455 (sparc_pic_register_p): New predicate.
10456 (sparc_legitimate_address_p): Use it.
10457 (sparc_legitimize_pic_address): Likewise.
10458 (sparc_delegitimize_address): Likewise.
10459 (sparc_mode_dependent_address_p): Likewise.
10460 (gen_load_pcrel_sym): Remove 4th parameter.
10461 (load_got_register): Adjust call to above. Remove obsolete stuff.
10462 (sparc_expand_prologue): Do not call load_got_register here.
10463 (sparc_flat_expand_prologue): Likewise.
10464 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10465 (sparc_use_pseudo_pic_reg): New function.
10466 (sparc_init_pic_reg): Likewise.
10467 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10468 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10470 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
10472 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10473 Add item for branch_cost.
10475 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10477 PR rtl-optimization/83565
10478 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10479 not extend the result to a larger mode for rotate operations.
10480 (num_sign_bit_copies1): Likewise.
10482 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10485 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10487 Use values-Xc.o for -pedantic.
10488 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10490 2018-01-12 Martin Liska <mliska@suse.cz>
10493 * ipa-devirt.c (final_warning_record::grow_type_warnings):
10495 (possible_polymorphic_call_targets): Use it.
10496 (ipa_devirt): Likewise.
10498 2018-01-12 Martin Liska <mliska@suse.cz>
10500 * profile-count.h (enum profile_quality): Use 0 as invalid
10501 enum value of profile_quality.
10503 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
10505 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10506 -mext-string options.
10508 2018-01-12 Richard Biener <rguenther@suse.de>
10510 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10511 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10512 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10514 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10516 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10518 * configure.ac (--with-long-double-format): Add support for the
10519 configuration option to change the default long double format on
10521 * config.gcc (powerpc*-linux*-*): Likewise.
10522 * configure: Regenerate.
10523 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10524 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10525 used without modification.
10527 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10529 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10530 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10531 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10532 MISC_BUILTIN_SPEC_BARRIER.
10533 (rs6000_init_builtins): Likewise.
10534 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10536 (speculation_barrier): New define_insn.
10537 * doc/extend.texi: Document __builtin_speculation_barrier.
10539 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10542 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10543 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10544 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10546 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10547 integral modes instead of "ss" and "sd".
10548 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10549 vectors with 32-bit and 64-bit elements.
10550 (vecdupssescalarmodesuffix): New mode attribute.
10551 (vec_dup<mode>): Use it.
10553 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
10556 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10557 frame if argument is passed on stack.
10559 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10562 * ree.c (combine_reaching_defs): Optimize also
10563 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10564 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10566 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10568 PR middle-end/83189
10569 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10571 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10573 PR middle-end/83718
10574 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10575 after they are computed.
10577 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10579 PR tree-optimization/83695
10580 * gimple-loop-linterchange.cc
10581 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10582 reset cached scev information after interchange.
10583 (pass_linterchange::execute): Remove call to scev_reset_htab.
10585 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10587 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10588 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10589 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10590 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10591 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10592 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10593 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10594 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10595 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10596 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10597 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10598 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10599 (V_lane_reg): Likewise.
10600 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10602 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10603 (vfmal_lane_low<mode>_intrinsic,
10604 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10605 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10606 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10607 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10608 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10609 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10611 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10613 * config/arm/arm-cpus.in (fp16fml): New feature.
10614 (ALL_SIMD): Add fp16fml.
10615 (armv8.2-a): Add fp16fml as an option.
10616 (armv8.3-a): Likewise.
10617 (armv8.4-a): Add fp16fml as part of fp16.
10618 * config/arm/arm.h (TARGET_FP16FML): Define.
10619 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10621 * config/arm/arm-modes.def (V2HF): Define.
10622 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10623 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10624 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10625 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10626 vfmsl_low, vfmsl_high): New set of builtins.
10627 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10628 (vfml_op): New code attribute.
10629 (VFMLHALVES): New int iterator.
10630 (VFML, VFMLSEL): New mode attributes.
10631 (V_reg): Define mapping for V2HF.
10632 (V_hi, V_lo): New mode attributes.
10633 (VF_constraint): Likewise.
10634 (vfml_half, vfml_half_selector): New int attributes.
10635 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10637 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10638 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10640 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10641 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10642 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10643 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
10645 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10646 Document new effective target and option set.
10648 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10650 * config/arm/arm-cpus.in (armv8_4): New feature.
10651 (ARMv8_4a): New fgroup.
10652 (armv8.4-a): New arch.
10653 * config/arm/arm-tables.opt: Regenerate.
10654 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10655 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
10656 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
10657 Add matching rules for -march=armv8.4-a and extensions.
10658 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
10660 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
10663 * config/rx/rx.md (BW): New mode attribute.
10664 (sync_lock_test_and_setsi): Add mode suffix to insn output.
10666 2018-01-11 Richard Biener <rguenther@suse.de>
10668 PR tree-optimization/83435
10669 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
10670 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
10671 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
10673 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10674 Alan Hayward <alan.hayward@arm.com>
10675 David Sherwood <david.sherwood@arm.com>
10677 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
10679 (aarch64_classify_address): Initialize it. Track polynomial offsets.
10680 (aarch64_print_address_internal): Use it to check for a zero offset.
10682 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10683 Alan Hayward <alan.hayward@arm.com>
10684 David Sherwood <david.sherwood@arm.com>
10686 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10687 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10688 Return a poly_int64 rather than a HOST_WIDE_INT.
10689 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10690 rather than a HOST_WIDE_INT.
10691 * config/aarch64/aarch64.h (aarch64_frame): Protect with
10692 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
10693 hard_fp_offset, frame_size, initial_adjust, callee_offset and
10694 final_offset from HOST_WIDE_INT to poly_int64.
10695 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
10696 to_constant when getting the number of units in an Advanced SIMD
10698 (aarch64_builtin_vectorized_function): Check for a constant number
10700 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
10702 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
10703 attribute instead of GET_MODE_NUNITS.
10704 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
10705 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
10706 GET_MODE_SIZE for fixed-size registers.
10707 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
10708 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
10709 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
10710 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
10711 (aarch64_print_operand, aarch64_print_address_internal)
10712 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
10713 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
10714 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
10715 Handle polynomial GET_MODE_SIZE.
10716 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
10717 wider than SImode without modification.
10718 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
10719 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
10720 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
10721 passing and returning SVE modes.
10722 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
10723 rather than GEN_INT.
10724 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
10725 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
10726 (aarch64_allocate_and_probe_stack_space): Likewise.
10727 (aarch64_layout_frame): Cope with polynomial offsets.
10728 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
10729 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
10730 polynomial offsets.
10731 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
10732 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
10733 poly_int64 rather than a HOST_WIDE_INT.
10734 (aarch64_get_separate_components, aarch64_process_components)
10735 (aarch64_expand_prologue, aarch64_expand_epilogue)
10736 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
10737 (aarch64_anchor_offset): New function, split out from...
10738 (aarch64_legitimize_address): ...here.
10739 (aarch64_builtin_vectorization_cost): Handle polynomial
10740 TYPE_VECTOR_SUBPARTS.
10741 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
10743 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
10744 number of elements from the PARALLEL rather than the mode.
10745 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
10746 rather than GET_MODE_BITSIZE.
10747 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
10748 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
10749 (aarch64_expand_vec_perm_const_1): Handle polynomial
10750 d->perm.length () and d->perm elements.
10751 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
10752 Apply to_constant to d->perm elements.
10753 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
10754 polynomial CONST_VECTOR_NUNITS.
10755 (aarch64_move_pointer): Take amount as a poly_int64 rather
10757 (aarch64_progress_pointer): Avoid temporary variable.
10758 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
10759 the mode attribute instead of GET_MODE.
10761 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10762 Alan Hayward <alan.hayward@arm.com>
10763 David Sherwood <david.sherwood@arm.com>
10765 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
10766 x exists before using it.
10767 (aarch64_add_constant_internal): Rename to...
10768 (aarch64_add_offset_1): ...this. Replace regnum with separate
10769 src and dest rtxes. Handle the case in which they're different,
10770 including when the offset is zero. Replace scratchreg with an rtx.
10771 Use 2 additions if there is no spare register into which we can
10772 move a 16-bit constant.
10773 (aarch64_add_constant): Delete.
10774 (aarch64_add_offset): Replace reg with separate src and dest
10775 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
10776 Use aarch64_add_offset_1.
10777 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
10778 an rtx rather than an int. Take the delta as a poly_int64
10779 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
10780 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
10781 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
10782 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
10783 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
10784 and aarch64_add_sp.
10785 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
10786 aarch64_add_constant.
10788 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10790 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
10791 Use scalar_float_mode.
10793 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10795 * config/aarch64/aarch64-simd.md
10796 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
10797 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
10798 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
10799 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
10800 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
10801 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
10802 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
10803 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
10804 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
10805 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
10807 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10810 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
10811 targ_options->x_arm_arch_string is non NULL.
10813 2018-01-11 Tamar Christina <tamar.christina@arm.com>
10815 * config/aarch64/aarch64.h
10816 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
10818 2018-01-11 Sudakshina Das <sudi.das@arm.com>
10821 * expmed.c (emit_store_flag_force): Swap if const op0
10822 and change VOIDmode to mode of op0.
10824 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10826 PR rtl-optimization/83761
10827 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
10828 than bytes to mode_for_size.
10830 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10832 PR middle-end/83189
10833 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
10834 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
10837 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10839 PR middle-end/83575
10840 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
10841 when in layout mode.
10842 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
10843 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
10846 2018-01-10 Michael Collison <michael.collison@arm.com>
10848 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
10849 * config/aarch64/aarch64-option-extension.def: Add
10850 AARCH64_OPT_EXTENSION of 'fp16fml'.
10851 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10852 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
10853 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
10854 * config/aarch64/constraints.md (Ui7): New constraint.
10855 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
10856 (VFMLA_SEL_W): Ditto.
10859 (VFMLA16_LOW): New int iterator.
10860 (VFMLA16_HIGH): Ditto.
10861 (UNSPEC_FMLAL): New unspec.
10862 (UNSPEC_FMLSL): Ditto.
10863 (UNSPEC_FMLAL2): Ditto.
10864 (UNSPEC_FMLSL2): Ditto.
10865 (f16mac): New code attribute.
10866 * config/aarch64/aarch64-simd-builtins.def
10867 (aarch64_fmlal_lowv2sf): Ditto.
10868 (aarch64_fmlsl_lowv2sf): Ditto.
10869 (aarch64_fmlalq_lowv4sf): Ditto.
10870 (aarch64_fmlslq_lowv4sf): Ditto.
10871 (aarch64_fmlal_highv2sf): Ditto.
10872 (aarch64_fmlsl_highv2sf): Ditto.
10873 (aarch64_fmlalq_highv4sf): Ditto.
10874 (aarch64_fmlslq_highv4sf): Ditto.
10875 (aarch64_fmlal_lane_lowv2sf): Ditto.
10876 (aarch64_fmlsl_lane_lowv2sf): Ditto.
10877 (aarch64_fmlal_laneq_lowv2sf): Ditto.
10878 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
10879 (aarch64_fmlalq_lane_lowv4sf): Ditto.
10880 (aarch64_fmlsl_lane_lowv4sf): Ditto.
10881 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
10882 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
10883 (aarch64_fmlal_lane_highv2sf): Ditto.
10884 (aarch64_fmlsl_lane_highv2sf): Ditto.
10885 (aarch64_fmlal_laneq_highv2sf): Ditto.
10886 (aarch64_fmlsl_laneq_highv2sf): Ditto.
10887 (aarch64_fmlalq_lane_highv4sf): Ditto.
10888 (aarch64_fmlsl_lane_highv4sf): Ditto.
10889 (aarch64_fmlalq_laneq_highv4sf): Ditto.
10890 (aarch64_fmlsl_laneq_highv4sf): Ditto.
10891 * config/aarch64/aarch64-simd.md:
10892 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
10893 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10894 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
10895 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10896 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
10897 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
10898 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
10899 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
10900 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
10901 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
10902 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
10903 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
10904 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
10905 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
10906 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
10907 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
10908 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
10909 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
10910 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
10911 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
10912 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
10913 (vfmlsl_low_u32): Ditto.
10914 (vfmlalq_low_u32): Ditto.
10915 (vfmlslq_low_u32): Ditto.
10916 (vfmlal_high_u32): Ditto.
10917 (vfmlsl_high_u32): Ditto.
10918 (vfmlalq_high_u32): Ditto.
10919 (vfmlslq_high_u32): Ditto.
10920 (vfmlal_lane_low_u32): Ditto.
10921 (vfmlsl_lane_low_u32): Ditto.
10922 (vfmlal_laneq_low_u32): Ditto.
10923 (vfmlsl_laneq_low_u32): Ditto.
10924 (vfmlalq_lane_low_u32): Ditto.
10925 (vfmlslq_lane_low_u32): Ditto.
10926 (vfmlalq_laneq_low_u32): Ditto.
10927 (vfmlslq_laneq_low_u32): Ditto.
10928 (vfmlal_lane_high_u32): Ditto.
10929 (vfmlsl_lane_high_u32): Ditto.
10930 (vfmlal_laneq_high_u32): Ditto.
10931 (vfmlsl_laneq_high_u32): Ditto.
10932 (vfmlalq_lane_high_u32): Ditto.
10933 (vfmlslq_lane_high_u32): Ditto.
10934 (vfmlalq_laneq_high_u32): Ditto.
10935 (vfmlslq_laneq_high_u32): Ditto.
10936 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
10937 (AARCH64_FL_FOR_ARCH8_4): New.
10938 (AARCH64_ISA_F16FML): New ISA flag.
10939 (TARGET_F16FML): New feature flag for fp16fml.
10940 (doc/invoke.texi): Document new fp16fml option.
10942 2018-01-10 Michael Collison <michael.collison@arm.com>
10944 * config/aarch64/aarch64-builtins.c:
10945 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
10946 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10947 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
10948 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
10949 (AARCH64_ISA_SHA3): New ISA flag.
10950 (TARGET_SHA3): New feature flag for sha3.
10951 * config/aarch64/iterators.md (sha512_op): New int attribute.
10952 (CRYPTO_SHA512): New int iterator.
10953 (UNSPEC_SHA512H): New unspec.
10954 (UNSPEC_SHA512H2): Ditto.
10955 (UNSPEC_SHA512SU0): Ditto.
10956 (UNSPEC_SHA512SU1): Ditto.
10957 * config/aarch64/aarch64-simd-builtins.def
10958 (aarch64_crypto_sha512hqv2di): New builtin.
10959 (aarch64_crypto_sha512h2qv2di): Ditto.
10960 (aarch64_crypto_sha512su0qv2di): Ditto.
10961 (aarch64_crypto_sha512su1qv2di): Ditto.
10962 (aarch64_eor3qv8hi): Ditto.
10963 (aarch64_rax1qv2di): Ditto.
10964 (aarch64_xarqv2di): Ditto.
10965 (aarch64_bcaxqv8hi): Ditto.
10966 * config/aarch64/aarch64-simd.md:
10967 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
10968 (aarch64_crypto_sha512su0qv2di): Ditto.
10969 (aarch64_crypto_sha512su1qv2di): Ditto.
10970 (aarch64_eor3qv8hi): Ditto.
10971 (aarch64_rax1qv2di): Ditto.
10972 (aarch64_xarqv2di): Ditto.
10973 (aarch64_bcaxqv8hi): Ditto.
10974 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
10975 (vsha512h2q_u64): Ditto.
10976 (vsha512su0q_u64): Ditto.
10977 (vsha512su1q_u64): Ditto.
10978 (veor3q_u16): Ditto.
10979 (vrax1q_u64): Ditto.
10980 (vxarq_u64): Ditto.
10981 (vbcaxq_u16): Ditto.
10982 * config/arm/types.md (crypto_sha512): New type attribute.
10983 (crypto_sha3): Ditto.
10984 (doc/invoke.texi): Document new sha3 option.
10986 2018-01-10 Michael Collison <michael.collison@arm.com>
10988 * config/aarch64/aarch64-builtins.c:
10989 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
10990 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10991 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
10992 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
10993 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
10994 (AARCH64_ISA_SM4): New ISA flag.
10995 (TARGET_SM4): New feature flag for sm4.
10996 * config/aarch64/aarch64-simd-builtins.def
10997 (aarch64_sm3ss1qv4si): Ditto.
10998 (aarch64_sm3tt1aq4si): Ditto.
10999 (aarch64_sm3tt1bq4si): Ditto.
11000 (aarch64_sm3tt2aq4si): Ditto.
11001 (aarch64_sm3tt2bq4si): Ditto.
11002 (aarch64_sm3partw1qv4si): Ditto.
11003 (aarch64_sm3partw2qv4si): Ditto.
11004 (aarch64_sm4eqv4si): Ditto.
11005 (aarch64_sm4ekeyqv4si): Ditto.
11006 * config/aarch64/aarch64-simd.md:
11007 (aarch64_sm3ss1qv4si): Ditto.
11008 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
11009 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
11010 (aarch64_sm4eqv4si): Ditto.
11011 (aarch64_sm4ekeyqv4si): Ditto.
11012 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
11013 (sm3part_op): Ditto.
11014 (CRYPTO_SM3TT): Ditto.
11015 (CRYPTO_SM3PART): Ditto.
11016 (UNSPEC_SM3SS1): New unspec.
11017 (UNSPEC_SM3TT1A): Ditto.
11018 (UNSPEC_SM3TT1B): Ditto.
11019 (UNSPEC_SM3TT2A): Ditto.
11020 (UNSPEC_SM3TT2B): Ditto.
11021 (UNSPEC_SM3PARTW1): Ditto.
11022 (UNSPEC_SM3PARTW2): Ditto.
11023 (UNSPEC_SM4E): Ditto.
11024 (UNSPEC_SM4EKEY): Ditto.
11025 * config/aarch64/constraints.md (Ui2): New constraint.
11026 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
11027 * config/arm/types.md (crypto_sm3): New type attribute.
11028 (crypto_sm4): Ditto.
11029 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
11030 (vsm3tt1aq_u32): Ditto.
11031 (vsm3tt1bq_u32): Ditto.
11032 (vsm3tt2aq_u32): Ditto.
11033 (vsm3tt2bq_u32): Ditto.
11034 (vsm3partw1q_u32): Ditto.
11035 (vsm3partw2q_u32): Ditto.
11036 (vsm4eq_u32): Ditto.
11037 (vsm4ekeyq_u32): Ditto.
11038 (doc/invoke.texi): Document new sm4 option.
11040 2018-01-10 Michael Collison <michael.collison@arm.com>
11042 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
11043 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
11044 (AARCH64_FL_FOR_ARCH8_4): New.
11045 (AARCH64_FL_V8_4): New flag.
11046 (doc/invoke.texi): Document new armv8.4-a option.
11048 2018-01-10 Michael Collison <michael.collison@arm.com>
11050 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11051 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
11052 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
11053 * config/aarch64/aarch64-option-extension.def: Add
11054 AARCH64_OPT_EXTENSION of 'sha2'.
11055 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
11056 (crypto): Disable sha2 and aes if crypto disabled.
11057 (crypto): Enable aes and sha2 if enabled.
11058 (simd): Disable sha2 and aes if simd disabled.
11059 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
11061 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
11062 (TARGET_SHA2): New feature flag for sha2.
11063 (TARGET_AES): New feature flag for aes.
11064 * config/aarch64/aarch64-simd.md:
11065 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
11066 conditional on TARGET_AES.
11067 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
11068 (aarch64_crypto_sha1hsi): Make pattern conditional
11070 (aarch64_crypto_sha1hv4si): Ditto.
11071 (aarch64_be_crypto_sha1hv4si): Ditto.
11072 (aarch64_crypto_sha1su1v4si): Ditto.
11073 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
11074 (aarch64_crypto_sha1su0v4si): Ditto.
11075 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
11076 (aarch64_crypto_sha256su0v4si): Ditto.
11077 (aarch64_crypto_sha256su1v4si): Ditto.
11078 (doc/invoke.texi): Document new aes and sha2 options.
11080 2018-01-10 Martin Sebor <msebor@redhat.com>
11082 PR tree-optimization/83781
11083 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
11086 2018-01-11 Martin Sebor <msebor@gmail.com>
11087 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11089 PR tree-optimization/83501
11090 PR tree-optimization/81703
11092 * tree-ssa-strlen.c (get_string_cst): Rename...
11093 (get_string_len): ...to this. Handle global constants.
11094 (handle_char_store): Adjust.
11096 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
11097 Jim Wilson <jimw@sifive.com>
11099 * config/riscv/riscv-protos.h (riscv_output_return): New.
11100 * config/riscv/riscv.c (struct machine_function): New naked_p field.
11101 (riscv_attribute_table, riscv_output_return),
11102 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
11103 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
11104 (riscv_compute_frame_info): Only compute frame->mask if not a naked
11106 (riscv_expand_prologue): Add early return for naked function.
11107 (riscv_expand_epilogue): Likewise.
11108 (riscv_function_ok_for_sibcall): Return false for naked function.
11109 (riscv_set_current_function): New.
11110 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
11111 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
11112 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
11113 * doc/extend.texi (RISC-V Function Attributes): New.
11115 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
11117 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
11118 check for 128-bit long double before checking TCmode.
11119 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
11120 128-bit long doubles before checking TFmode or TCmode.
11121 (FLOAT128_IBM_P): Likewise.
11123 2018-01-10 Martin Sebor <msebor@redhat.com>
11125 PR tree-optimization/83671
11126 * builtins.c (c_strlen): Unconditionally return zero for the empty
11128 Use -Warray-bounds for warnings.
11129 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
11130 for non-constant array indices with COMPONENT_REF, arrays of
11131 arrays, and pointers to arrays.
11132 (gimple_fold_builtin_strlen): Determine and set length range for
11133 non-constant character arrays.
11135 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
11137 PR middle-end/81897
11138 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
11141 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
11143 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
11145 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11148 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
11149 VECTOR_MEM_ALTIVEC_OR_VSX_P.
11150 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
11151 indexed_or_indirect_operand predicate.
11152 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
11153 (*vsx_le_perm_load_v8hi): Likewise.
11154 (*vsx_le_perm_load_v16qi): Likewise.
11155 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
11156 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
11157 (*vsx_le_perm_store_v8hi): Likewise.
11158 (*vsx_le_perm_store_v16qi): Likewise.
11159 (eight unnamed splitters): Likewise.
11161 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11163 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
11164 * config/rs6000/emmintrin.h: Likewise.
11165 * config/rs6000/mmintrin.h: Likewise.
11166 * config/rs6000/xmmintrin.h: Likewise.
11168 2018-01-10 David Malcolm <dmalcolm@redhat.com>
11171 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
11173 * tree.c (tree_nop_conversion): Return true for location wrapper
11175 (maybe_wrap_with_location): New function.
11176 (selftest::check_strip_nops): New function.
11177 (selftest::test_location_wrappers): New function.
11178 (selftest::tree_c_tests): Call it.
11179 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
11180 (maybe_wrap_with_location): New decl.
11181 (EXPR_LOCATION_WRAPPER_P): New macro.
11182 (location_wrapper_p): New inline function.
11183 (tree_strip_any_location_wrapper): New inline function.
11185 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
11188 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
11189 stack_realign_offset for the largest alignment of stack slot
11191 (ix86_find_max_used_stack_alignment): New function.
11192 (ix86_finalize_stack_frame_flags): Use it. Set
11193 max_used_stack_alignment if we don't realign stack.
11194 * config/i386/i386.h (machine_function): Add
11195 max_used_stack_alignment.
11197 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
11199 * config/arm/arm.opt (-mbranch-cost): New option.
11200 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
11203 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
11206 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
11207 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
11209 2018-01-10 Richard Biener <rguenther@suse.de>
11212 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
11213 early out so it also covers the case where we have a non-NULL
11216 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11218 PR tree-optimization/83753
11219 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
11220 for non-strided grouped accesses if the number of elements is 1.
11222 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11225 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
11226 * i386.h (TARGET_USE_GATHER): Define.
11227 * x86-tune.def (X86_TUNE_USE_GATHER): New.
11229 2018-01-10 Martin Liska <mliska@suse.cz>
11232 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
11233 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
11235 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
11236 CLEANUP_NO_PARTITIONING is not set.
11238 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11240 * doc/rtl.texi: Remove documentation of (const ...) wrappers
11241 for vectors, as a partial revert of r254296.
11242 * rtl.h (const_vec_p): Delete.
11243 (const_vec_duplicate_p): Don't test for vector CONSTs.
11244 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
11245 * expmed.c (make_tree): Likewise.
11248 * common.md (E, F): Use CONSTANT_P instead of checking for
11250 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
11251 checking for CONST_VECTOR.
11253 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11255 PR middle-end/83575
11256 * predict.c (force_edge_cold): Handle in more sane way edges
11257 with no prediction.
11259 2018-01-09 Carl Love <cel@us.ibm.com>
11261 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
11263 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
11264 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
11265 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
11266 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
11267 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
11268 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
11269 * config/rs6000/rs6000-protos.h: Add extern defition for
11270 rs6000_generate_float2_double_code.
11271 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
11273 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
11274 (float2_v2df): Add define_expand.
11276 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
11279 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
11280 op_mode in the force_to_mode call.
11282 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11284 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
11285 instead of checking each element individually.
11286 (aarch64_evpc_uzp): Likewise.
11287 (aarch64_evpc_zip): Likewise.
11288 (aarch64_evpc_ext): Likewise.
11289 (aarch64_evpc_rev): Likewise.
11290 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
11291 instead of checking each element individually. Return true without
11293 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
11294 whether all selected elements come from the same input, instead of
11295 checking each element individually. Remove calls to gen_rtx_REG,
11296 start_sequence and end_sequence and instead assert that no rtl is
11299 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11301 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
11302 order of HIGH and CONST checks.
11304 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11306 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
11307 if the destination isn't an SSA_NAME.
11309 2018-01-09 Richard Biener <rguenther@suse.de>
11311 PR tree-optimization/83668
11312 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
11314 (canonicalize_loop_form): ... here, renamed from ...
11315 (canonicalize_loop_closed_ssa_form): ... this and amended to
11316 swap successor edges for loop exit blocks to make us use
11317 the RPO order we need for initial schedule generation.
11319 2018-01-09 Joseph Myers <joseph@codesourcery.com>
11321 PR tree-optimization/64811
11322 * match.pd: When optimizing comparisons with Inf, avoid
11323 introducing or losing exceptions from comparisons with NaN.
11325 2018-01-09 Martin Liska <mliska@suse.cz>
11328 * asan.c (shadow_mem_size): Add gcc_assert.
11330 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
11332 Don't save registers in main().
11335 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
11336 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
11337 * config/avr/avr.c (avr_set_current_function): Don't error if
11338 naked, OS_task or OS_main are specified at the same time.
11339 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
11341 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
11343 * common/config/avr/avr-common.c (avr_option_optimization_table):
11344 Switch on -mmain-is-OS_task for optimizing compilations.
11346 2018-01-09 Richard Biener <rguenther@suse.de>
11348 PR tree-optimization/83572
11349 * graphite.c: Include cfganal.h.
11350 (graphite_transform_loops): Connect infinite loops to exit
11351 and remove fake edges at the end.
11353 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11355 * ipa-inline.c (edge_badness): Revert accidental checkin.
11357 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11360 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
11361 symbols; not inline clones.
11363 2018-01-09 Jakub Jelinek <jakub@redhat.com>
11366 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
11367 hard registers. Formatting fixes.
11369 PR preprocessor/83722
11370 * gcc.c (try_generate_repro): Pass
11371 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
11372 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
11375 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
11376 Kito Cheng <kito.cheng@gmail.com>
11378 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
11379 (riscv_leaf_function_p): Delete.
11380 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
11382 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11384 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
11386 (do_ifelse): New function.
11387 (do_isel): New function.
11388 (do_sub3): New function.
11389 (do_add3): New function.
11390 (do_load_mask_compare): New function.
11391 (do_overlap_load_compare): New function.
11392 (expand_compare_loop): New function.
11393 (expand_block_compare): Call expand_compare_loop() when appropriate.
11394 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
11395 option description.
11396 (-mblock-compare-inline-loop-limit): New option.
11398 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11401 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
11402 Reverse order of second and third operands in first alternative.
11403 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
11404 of first and second elements in UNSPEC_VPERMR vector.
11405 (altivec_expand_vec_perm_le): Likewise.
11407 2018-01-08 Jeff Law <law@redhat.com>
11409 PR rtl-optimizatin/81308
11410 * tree-switch-conversion.c (cfg_altered): New file scoped static.
11411 (process_switch): If group_case_labels makes a change, then set
11413 (pass_convert_switch::execute): If a switch is converted, then
11414 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
11416 PR rtl-optimization/81308
11417 * recog.c (split_all_insns): Conditionally cleanup the CFG after
11420 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
11422 PR target/83663 - Revert r255946
11423 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11424 generation for cases where splatting a value is not useful.
11425 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11426 across a vec_duplicate and a paradoxical subreg forming a vector
11427 mode to a vec_concat.
11429 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11431 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11432 -march=armv8.3-a variants.
11433 * config/arm/t-multilib: Likewise.
11434 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11436 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11438 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11440 (cceq_ior_compare_complement): Give it a name so I can use it, and
11441 change boolean_or_operator predicate to boolean_operator so it can
11442 be used to generate a crand.
11443 (eqne): New code iterator.
11444 (bd/bd_neg): New code_attrs.
11445 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11446 a single define_insn.
11447 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11448 decrement (bdnzt/bdnzf/bdzt/bdzf).
11449 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11450 with the new names of the branch decrement patterns, and added the
11451 names of the branch decrement conditional patterns.
11453 2018-01-08 Richard Biener <rguenther@suse.de>
11455 PR tree-optimization/83563
11456 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11459 2018-01-08 Richard Biener <rguenther@suse.de>
11461 PR middle-end/83713
11462 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11464 2018-01-08 Richard Biener <rguenther@suse.de>
11466 PR tree-optimization/83685
11467 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11468 references to abnormals.
11470 2018-01-08 Richard Biener <rguenther@suse.de>
11473 * dwarf2out.c (output_indirect_strings): Handle empty
11474 skeleton_debug_str_hash.
11475 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11477 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11479 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11480 (emit_store_direct): Likewise.
11481 (arc_trampoline_adjust_address): Likewise.
11482 (arc_asm_trampoline_template): New function.
11483 (arc_initialize_trampoline): Use asm_trampoline_template.
11484 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11485 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11486 * config/arc/arc.md (flush_icache): Delete pattern.
11488 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11490 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11491 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11494 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11497 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11498 by not USED_FOR_TARGET.
11499 (make_pass_resolve_sw_modes): Likewise.
11501 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11503 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11506 2018-01-08 Richard Biener <rguenther@suse.de>
11508 PR middle-end/83580
11509 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11511 2018-01-08 Richard Biener <rguenther@suse.de>
11513 PR middle-end/83517
11514 * match.pd ((t * 2) / 2) -> t): Add missing :c.
11516 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
11518 PR middle-end/81897
11519 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11520 basic blocks with a small number of successors.
11521 (convert_control_dep_chain_into_preds): Improve handling of
11523 (dump_predicates): Split apart into...
11524 (dump_pred_chain): ...here...
11525 (dump_pred_info): ...and here.
11526 (can_one_predicate_be_invalidated_p): Add debugging printfs.
11527 (can_chain_union_be_invalidated_p): Improve check for invalidation
11529 (uninit_uses_cannot_happen): Avoid unnecessary if
11530 convert_control_dep_chain_into_preds yielded nothing.
11532 2018-01-06 Martin Sebor <msebor@redhat.com>
11534 PR tree-optimization/83640
11535 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11536 subtracting negative offset from size.
11537 (builtin_access::overlap): Adjust offset bounds of the access to fall
11538 within the size of the object if possible.
11540 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
11542 PR rtl-optimization/83699
11543 * expmed.c (extract_bit_field_1): Restrict the vector usage of
11544 extract_bit_field_as_subreg to cases in which the extracted
11545 value is also a vector.
11547 * lra-constraints.c (process_alt_operands): Test for the equivalence
11548 substitutions when detecting a possible reload cycle.
11550 2018-01-06 Jakub Jelinek <jakub@redhat.com>
11553 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11554 by default if flag_selective_schedling{,2}. Formatting fixes.
11556 PR rtl-optimization/83682
11557 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11558 if it has non-VECTOR_MODE element mode.
11559 (vec_duplicate_p): Likewise.
11561 PR middle-end/83694
11562 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11563 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11565 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11568 * config/i386/i386-builtin.def
11569 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11570 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11571 Require also OPTION_MASK_ISA_AVX512F in addition to
11572 OPTION_MASK_ISA_GFNI.
11573 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11574 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11575 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11576 to OPTION_MASK_ISA_GFNI.
11577 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11578 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11579 OPTION_MASK_ISA_AVX512BW.
11580 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11581 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11582 addition to OPTION_MASK_ISA_GFNI.
11583 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11584 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11585 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11586 to OPTION_MASK_ISA_GFNI.
11587 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11588 a requirement for all ISAs rather than any of them with a few
11590 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11592 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11593 bitmasks to be enabled with 3 exceptions, instead of requiring any
11594 enabled ISA with lots of exceptions.
11595 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11596 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11597 Change avx512bw in isa attribute to avx512f.
11598 * config/i386/sgxintrin.h: Add license boilerplate.
11599 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11600 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11601 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11602 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11604 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11605 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11606 temporarily sse2 rather than sse if not enabled already.
11609 * config/i386/sse.md (VI248_VLBW): Rename to ...
11610 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11611 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11612 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11613 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11614 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11615 mode iterator instead of VI248_VLBW.
11617 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11619 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11620 (record_modified): Skip clobbers; add debug output.
11621 (param_change_prob): Use sreal frequencies.
11623 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11625 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11626 punt for user-aligned variables.
11628 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11630 * tree-chrec.c (chrec_contains_symbols): Return true for
11633 2018-01-05 Sudakshina Das <sudi.das@arm.com>
11636 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11637 of (x|y) == x for BICS pattern.
11639 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11641 PR tree-optimization/83605
11642 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11643 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11646 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
11648 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11649 * config/epiphany/rtems.h: New file.
11651 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11652 Uros Bizjak <ubizjak@gmail.com>
11655 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
11656 QIreg_operand instead of register_operand predicate.
11657 * config/i386/i386.c (ix86_rop_should_change_byte_p,
11658 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
11659 comments instead of -fmitigate[-_]rop.
11661 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11664 * cgraphunit.c (symbol_table::compile): Switch to text_section
11665 before calling assembly_start debug hook.
11666 * run-rtl-passes.c (run_rtl_passes): Likewise.
11669 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11671 * tree-vrp.c (extract_range_from_binary_expr_1): Check
11672 range_int_cst_p rather than !symbolic_range_p before calling
11673 extract_range_from_multiplicative_op_1.
11675 2018-01-04 Jeff Law <law@redhat.com>
11677 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11678 redundant test in assertion.
11680 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11682 * doc/rtl.texi: Document machine_mode wrapper classes.
11684 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11686 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11687 using tree_to_uhwi.
11689 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11691 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11692 the VEC_PERM_EXPR fold to fail.
11694 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11697 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
11698 to switched_sections.
11700 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11703 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
11704 test for d.testing.
11706 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
11709 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
11710 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
11712 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11715 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
11716 is BLKmode and bitpos not zero or mode change is needed.
11718 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11721 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
11724 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
11727 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
11728 instead of MULT rtx. Update all corresponding splitters.
11729 (*saddl_se): Ditto.
11730 (*ssub<modesuffix>): Ditto.
11731 (*ssubl_se): Ditto.
11732 (*cmp_sadd_di): Update split patterns.
11733 (*cmp_sadd_si): Ditto.
11734 (*cmp_sadd_sidi): Ditto.
11735 (*cmp_ssub_di): Ditto.
11736 (*cmp_ssub_si): Ditto.
11737 (*cmp_ssub_sidi): Ditto.
11738 * config/alpha/predicates.md (const23_operand): New predicate.
11739 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
11740 Look for ASHIFT, not MULT inner operand.
11741 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
11743 2018-01-04 Martin Liska <mliska@suse.cz>
11745 PR gcov-profile/83669
11746 * gcov.c (output_intermediate_file): Add version to intermediate
11748 * doc/gcov.texi: Document new field 'version' in intermediate
11749 file format. Fix location of '-k' option of gcov command.
11751 2018-01-04 Martin Liska <mliska@suse.cz>
11754 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
11756 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11758 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
11760 2018-01-03 Martin Sebor <msebor@redhat.com>
11762 PR tree-optimization/83655
11763 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
11764 checking calls with invalid arguments.
11766 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11768 * tree-vect-stmts.c (vect_get_store_rhs): New function.
11769 (vectorizable_mask_load_store): Delete.
11770 (vectorizable_call): Return false for masked loads and stores.
11771 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
11772 instead of gimple_assign_rhs1.
11773 (vectorizable_load): Handle IFN_MASK_LOAD.
11774 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
11776 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11778 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
11780 (vectorizable_mask_load_store): ...here.
11781 (vectorizable_load): ...and here.
11783 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11785 * tree-vect-stmts.c (vect_build_all_ones_mask)
11786 (vect_build_zero_merge_argument): New functions, split out from...
11787 (vectorizable_load): ...here.
11789 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11791 * tree-vect-stmts.c (vect_check_store_rhs): New function,
11793 (vectorizable_mask_load_store): ...here.
11794 (vectorizable_store): ...and here.
11796 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11798 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
11800 (vectorizable_mask_load_store): ...here.
11802 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11804 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
11805 (vect_model_store_cost): Take a vec_load_store_type instead of a
11807 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
11808 (vect_model_store_cost): Take a vec_load_store_type instead of a
11810 (vectorizable_mask_load_store): Update accordingly.
11811 (vectorizable_store): Likewise.
11812 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
11814 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11816 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
11817 IFN_MASK_LOAD calls here rather than...
11818 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
11820 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11821 Alan Hayward <alan.hayward@arm.com>
11822 David Sherwood <david.sherwood@arm.com>
11824 * expmed.c (extract_bit_field_1): For vector extracts,
11825 fall back to extract_bit_field_as_subreg if vec_extract
11828 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11829 Alan Hayward <alan.hayward@arm.com>
11830 David Sherwood <david.sherwood@arm.com>
11832 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
11833 they are variable or constant sized.
11834 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
11835 slots for constant-sized data.
11837 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11838 Alan Hayward <alan.hayward@arm.com>
11839 David Sherwood <david.sherwood@arm.com>
11841 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
11842 handling COND_EXPRs with boolean comparisons, try to find a better
11843 basis for the mask type than the boolean itself.
11845 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11847 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
11848 is calculated and how it can be overridden.
11849 * genmodes.c (max_bitsize_mode_any_mode): New variable.
11850 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
11852 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
11855 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11856 Alan Hayward <alan.hayward@arm.com>
11857 David Sherwood <david.sherwood@arm.com>
11859 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
11860 Remove the mode argument.
11861 (aarch64_simd_valid_immediate): Remove the mode and inverse
11863 * config/aarch64/iterators.md (bitsize): New iterator.
11864 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
11865 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
11866 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
11867 aarch64_simd_valid_immediate.
11868 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
11869 (aarch64_reg_or_bic_imm): Likewise.
11870 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
11871 with an insn_type enum and msl with a modifier_type enum.
11872 Replace element_width with a scalar_mode. Change the shift
11873 to unsigned int. Add constructors for scalar_float_mode and
11874 scalar_int_mode elements.
11875 (aarch64_vect_float_const_representable_p): Delete.
11876 (aarch64_can_const_movi_rtx_p)
11877 (aarch64_simd_scalar_immediate_valid_for_move)
11878 (aarch64_simd_make_constant): Update call to
11879 aarch64_simd_valid_immediate.
11880 (aarch64_advsimd_valid_immediate_hs): New function.
11881 (aarch64_advsimd_valid_immediate): Likewise.
11882 (aarch64_simd_valid_immediate): Remove mode and inverse
11883 arguments. Rewrite to use the above. Use const_vec_duplicate_p
11884 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
11885 and aarch64_float_const_representable_p on the result.
11886 (aarch64_output_simd_mov_immediate): Remove mode argument.
11887 Update call to aarch64_simd_valid_immediate and use of
11888 simd_immediate_info.
11889 (aarch64_output_scalar_simd_mov_immediate): Update call
11892 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11893 Alan Hayward <alan.hayward@arm.com>
11894 David Sherwood <david.sherwood@arm.com>
11896 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
11897 (mode_nunits): Likewise CONST_MODE_NUNITS.
11898 * machmode.def (ADJUST_NUNITS): Document.
11899 * genmodes.c (mode_data::need_nunits_adj): New field.
11900 (blank_mode): Update accordingly.
11901 (adj_nunits): New variable.
11902 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
11904 (emit_mode_size_inline): Set need_bytesize_adj for all modes
11905 listed in adj_nunits.
11906 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
11907 listed in adj_nunits. Don't emit case statements for such modes.
11908 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
11909 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
11910 nothing if adj_nunits is nonnull.
11911 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
11912 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
11913 (emit_mode_fbit): Update use of print_maybe_const_decl.
11914 (emit_move_size): Likewise. Treat the array as non-const
11916 (emit_mode_adjustments): Handle adj_nunits.
11918 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11920 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
11921 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
11922 (VECTOR_MODES): Use it.
11923 (make_vector_modes): Take the prefix as an argument.
11925 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11926 Alan Hayward <alan.hayward@arm.com>
11927 David Sherwood <david.sherwood@arm.com>
11929 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
11930 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
11931 for MODE_VECTOR_BOOL.
11932 * machmode.def (VECTOR_BOOL_MODE): Document.
11933 * genmodes.c (VECTOR_BOOL_MODE): New macro.
11934 (make_vector_bool_mode): New function.
11935 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
11937 * lto-streamer-in.c (lto_input_mode_table): Likewise.
11938 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
11940 * stor-layout.c (int_mode_for_mode): Likewise.
11941 * tree.c (build_vector_type_for_mode): Likewise.
11942 * varasm.c (output_constant_pool_2): Likewise.
11943 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
11944 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
11945 for MODE_VECTOR_BOOL.
11946 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
11947 of mode class checks.
11948 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
11949 instead of a list of mode class checks.
11950 (expand_vector_scalar_condition): Likewise.
11951 (type_for_widest_vector_mode): Handle BImode as an inner mode.
11953 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11954 Alan Hayward <alan.hayward@arm.com>
11955 David Sherwood <david.sherwood@arm.com>
11957 * machmode.h (mode_size): Change from unsigned short to
11959 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
11960 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11961 or if measurement_type is not polynomial.
11962 (fixed_size_mode::includes_p): Check for constant-sized modes.
11963 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
11964 return a poly_uint16 rather than an unsigned short.
11965 (emit_mode_size): Change the type of mode_size from unsigned short
11966 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
11967 (emit_mode_adjustments): Cope with polynomial vector sizes.
11968 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11970 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11972 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
11973 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
11974 * caller-save.c (setup_save_areas): Likewise.
11975 (replace_reg_with_saved_mem): Likewise.
11976 * calls.c (emit_library_call_value_1): Likewise.
11977 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
11978 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
11979 (gen_lowpart_for_combine): Likewise.
11980 * convert.c (convert_to_integer_1): Likewise.
11981 * cse.c (equiv_constant, cse_insn): Likewise.
11982 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
11983 (cselib_subst_to_values): Likewise.
11984 * dce.c (word_dce_process_block): Likewise.
11985 * df-problems.c (df_word_lr_mark_ref): Likewise.
11986 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
11987 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
11988 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
11989 (rtl_for_decl_location): Likewise.
11990 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
11991 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
11992 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
11993 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
11994 (expand_expr_real_1): Likewise.
11995 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
11996 (pad_below): Likewise.
11997 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11998 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
11999 * ira.c (get_subreg_tracking_sizes): Likewise.
12000 * ira-build.c (ira_create_allocno_objects): Likewise.
12001 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
12002 (ira_sort_regnos_for_alter_reg): Likewise.
12003 * ira-costs.c (record_operand_costs): Likewise.
12004 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
12005 (resolve_simple_move): Likewise.
12006 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
12007 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
12008 (lra_constraints): Likewise.
12009 (CONST_POOL_OK_P): Reject variable-sized modes.
12010 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
12011 (add_pseudo_to_slot, lra_spill): Likewise.
12012 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12013 * optabs-query.c (get_best_extraction_insn): Likewise.
12014 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12015 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
12016 (expand_mult_highpart, valid_multiword_target_p): Likewise.
12017 * recog.c (offsettable_address_addr_space_p): Likewise.
12018 * regcprop.c (maybe_mode_change): Likewise.
12019 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
12020 * regrename.c (build_def_use): Likewise.
12021 * regstat.c (dump_reg_info): Likewise.
12022 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
12023 (find_reloads, find_reloads_subreg_address): Likewise.
12024 * reload1.c (eliminate_regs_1): Likewise.
12025 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
12026 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
12027 (simplify_binary_operation_1, simplify_subreg): Likewise.
12028 * targhooks.c (default_function_arg_padding): Likewise.
12029 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
12030 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
12031 (verify_gimple_assign_ternary): Likewise.
12032 * tree-inline.c (estimate_move_cost): Likewise.
12033 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12034 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
12035 (get_address_cost_ainc): Likewise.
12036 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
12037 (vect_supportable_dr_alignment): Likewise.
12038 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12039 (vectorizable_reduction): Likewise.
12040 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
12041 (vectorizable_operation, vectorizable_load): Likewise.
12042 * tree.c (build_same_sized_truth_vector_type): Likewise.
12043 * valtrack.c (cleanup_auto_inc_dec): Likewise.
12044 * var-tracking.c (emit_note_insn_var_location): Likewise.
12045 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
12046 (ADDR_VEC_ALIGN): Likewise.
12048 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12049 Alan Hayward <alan.hayward@arm.com>
12050 David Sherwood <david.sherwood@arm.com>
12052 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
12054 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12055 or if measurement_type is polynomial.
12056 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
12057 * combine.c (make_extraction): Likewise.
12058 * dse.c (find_shift_sequence): Likewise.
12059 * dwarf2out.c (mem_loc_descriptor): Likewise.
12060 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
12061 (extract_bit_field, extract_low_bits): Likewise.
12062 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
12063 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
12064 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
12065 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
12066 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12067 * reload.c (find_reloads): Likewise.
12068 * reload1.c (alter_reg): Likewise.
12069 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
12070 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
12071 * tree-if-conv.c (predicate_mem_writes): Likewise.
12072 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
12073 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
12074 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
12075 * valtrack.c (dead_debug_insert_temp): Likewise.
12076 * varasm.c (mergeable_constant_section): Likewise.
12077 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
12079 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12080 Alan Hayward <alan.hayward@arm.com>
12081 David Sherwood <david.sherwood@arm.com>
12083 * expr.c (expand_assignment): Cope with polynomial mode sizes
12084 when assigning to a CONCAT.
12086 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12087 Alan Hayward <alan.hayward@arm.com>
12088 David Sherwood <david.sherwood@arm.com>
12090 * machmode.h (mode_precision): Change from unsigned short to
12092 (mode_to_precision): Return a poly_uint16 rather than an unsigned
12094 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
12095 or if measurement_type is not polynomial.
12096 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
12097 in which the mode is already known to be a scalar_int_mode.
12098 * genmodes.c (emit_mode_precision): Change the type of mode_precision
12099 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
12101 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12102 for GET_MODE_PRECISION.
12103 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12104 for GET_MODE_PRECISION.
12105 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
12107 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
12108 (expand_field_assignment, make_extraction): Likewise.
12109 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
12110 (get_last_value): Likewise.
12111 * convert.c (convert_to_integer_1): Likewise.
12112 * cse.c (cse_insn): Likewise.
12113 * expr.c (expand_expr_real_1): Likewise.
12114 * lra-constraints.c (simplify_operand_subreg): Likewise.
12115 * optabs-query.c (can_atomic_load_p): Likewise.
12116 * optabs.c (expand_atomic_load): Likewise.
12117 (expand_atomic_store): Likewise.
12118 * ree.c (combine_reaching_defs): Likewise.
12119 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
12120 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
12121 * tree.h (type_has_mode_precision_p): Likewise.
12122 * ubsan.c (instrument_si_overflow): Likewise.
12124 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12125 Alan Hayward <alan.hayward@arm.com>
12126 David Sherwood <david.sherwood@arm.com>
12128 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
12129 polynomial numbers of units.
12130 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
12131 (valid_vector_subparts_p): New function.
12132 (build_vector_type): Remove temporary shim and take the number
12133 of units as a poly_uint64 rather than an int.
12134 (build_opaque_vector_type): Take the number of units as a
12135 poly_uint64 rather than an int.
12136 * tree.c (build_vector_from_ctor): Handle polynomial
12137 TYPE_VECTOR_SUBPARTS.
12138 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
12139 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
12140 (build_vector_from_val): If the number of units is variable,
12141 use build_vec_duplicate_cst for constant operands and
12142 VEC_DUPLICATE_EXPR otherwise.
12143 (make_vector_type): Remove temporary is_constant ().
12144 (build_vector_type, build_opaque_vector_type): Take the number of
12145 units as a poly_uint64 rather than an int.
12146 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
12148 * cfgexpand.c (expand_debug_expr): Likewise.
12149 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
12150 (store_constructor, expand_expr_real_1): Likewise.
12151 (const_scalar_mask_from_tree): Likewise.
12152 * fold-const-call.c (fold_const_reduction): Likewise.
12153 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
12154 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
12155 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
12156 (fold_relational_const): Likewise.
12157 (native_interpret_vector): Likewise. Change the size from an
12158 int to an unsigned int.
12159 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
12160 TYPE_VECTOR_SUBPARTS.
12161 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
12162 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
12163 duplicating a non-constant operand into a variable-length vector.
12164 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
12165 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
12166 * ipa-icf.c (sem_variable::equals): Likewise.
12167 * match.pd: Likewise.
12168 * omp-simd-clone.c (simd_clone_subparts): Likewise.
12169 * print-tree.c (print_node): Likewise.
12170 * stor-layout.c (layout_type): Likewise.
12171 * targhooks.c (default_builtin_vectorization_cost): Likewise.
12172 * tree-cfg.c (verify_gimple_comparison): Likewise.
12173 (verify_gimple_assign_binary): Likewise.
12174 (verify_gimple_assign_ternary): Likewise.
12175 (verify_gimple_assign_single): Likewise.
12176 * tree-pretty-print.c (dump_generic_node): Likewise.
12177 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12178 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
12179 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
12180 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
12181 (vect_shift_permute_load_chain): Likewise.
12182 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
12183 (expand_vector_condition, optimize_vector_constructor): Likewise.
12184 (lower_vec_perm, get_compute_type): Likewise.
12185 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12186 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
12187 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
12188 (vect_recog_mask_conversion_pattern): Likewise.
12189 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
12190 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
12191 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12192 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
12193 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
12194 (vectorizable_shift, vectorizable_operation, vectorizable_store)
12195 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
12196 (supportable_widening_operation): Likewise.
12197 (supportable_narrowing_operation): Likewise.
12198 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
12200 * varasm.c (output_constant): Likewise.
12202 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12203 Alan Hayward <alan.hayward@arm.com>
12204 David Sherwood <david.sherwood@arm.com>
12206 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
12207 so that both the length == 3 and length != 3 cases set up their
12208 own permute vectors. Add comments explaining why we know the
12209 number of elements is constant.
12210 (vect_permute_load_chain): Likewise.
12212 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12213 Alan Hayward <alan.hayward@arm.com>
12214 David Sherwood <david.sherwood@arm.com>
12216 * machmode.h (mode_nunits): Change from unsigned char to
12218 (ONLY_FIXED_SIZE_MODES): New macro.
12219 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
12220 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
12221 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
12223 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
12224 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
12225 or if measurement_type is not polynomial.
12226 * genmodes.c (ZERO_COEFFS): New macro.
12227 (emit_mode_nunits_inline): Make mode_nunits_inline return a
12229 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
12230 Use ZERO_COEFFS when emitting initializers.
12231 * data-streamer.h (bp_pack_poly_value): New function.
12232 (bp_unpack_poly_value): Likewise.
12233 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12234 for GET_MODE_NUNITS.
12235 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12236 for GET_MODE_NUNITS.
12237 * tree.c (make_vector_type): Remove temporary shim and make
12238 the real function take the number of units as a poly_uint64
12239 rather than an int.
12240 (build_vector_type_for_mode): Handle polynomial nunits.
12241 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
12242 * emit-rtl.c (const_vec_series_p_1): Likewise.
12243 (gen_rtx_CONST_VECTOR): Likewise.
12244 * fold-const.c (test_vec_duplicate_folding): Likewise.
12245 * genrecog.c (validate_pattern): Likewise.
12246 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
12247 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12248 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
12249 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
12250 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
12251 * rtlanal.c (subreg_get_info): Likewise.
12252 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12253 (vect_grouped_load_supported): Likewise.
12254 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
12255 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
12256 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12257 (simplify_const_unary_operation, simplify_binary_operation_1)
12258 (simplify_const_binary_operation, simplify_ternary_operation)
12259 (test_vector_ops_duplicate, test_vector_ops): Likewise.
12260 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
12261 instead of CONST_VECTOR_NUNITS.
12262 * varasm.c (output_constant_pool_2): Likewise.
12263 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
12264 explicit-encoded elements in the XVEC for variable-length vectors.
12266 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12268 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
12270 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12271 Alan Hayward <alan.hayward@arm.com>
12272 David Sherwood <david.sherwood@arm.com>
12274 * coretypes.h (fixed_size_mode): Declare.
12275 (fixed_size_mode_pod): New typedef.
12276 * builtins.h (target_builtins::x_apply_args_mode)
12277 (target_builtins::x_apply_result_mode): Change type to
12278 fixed_size_mode_pod.
12279 * builtins.c (apply_args_size, apply_result_size, result_vector)
12280 (expand_builtin_apply_args_1, expand_builtin_apply)
12281 (expand_builtin_return): Update accordingly.
12283 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12285 * cse.c (hash_rtx_cb): Hash only the encoded elements.
12286 * cselib.c (cselib_hash_rtx): Likewise.
12287 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
12288 CONST_VECTOR encoding.
12290 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12291 Jeff Law <law@redhat.com>
12294 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
12295 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
12296 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
12297 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
12300 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
12301 explicitly probe *sp in a noreturn function if there were any callee
12302 register saves or frame pointer is needed.
12304 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12307 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
12308 BLKmode for ternary, binary or unary expressions.
12311 * var-tracking.c (delete_vta_debug_insn): New inline function.
12312 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
12313 insns from get_insns () to NULL instead of each bb separately.
12314 Use delete_vta_debug_insn. No longer static.
12315 (vt_debug_insns_local, variable_tracking_main_1): Adjust
12316 delete_vta_debug_insns callers.
12317 * rtl.h (delete_vta_debug_insns): Declare.
12318 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
12319 instead of variable_tracking_main.
12321 2018-01-03 Martin Sebor <msebor@redhat.com>
12323 PR tree-optimization/83603
12324 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
12325 arguments past the endof the argument list in functions declared
12326 without a prototype.
12327 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
12328 Avoid checking when arguments are null.
12330 2018-01-03 Martin Sebor <msebor@redhat.com>
12333 * doc/extend.texi (attribute const): Fix a typo.
12334 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
12335 issuing -Wsuggest-attribute for void functions.
12337 2018-01-03 Martin Sebor <msebor@redhat.com>
12339 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
12340 offset_int::from instead of wide_int::to_shwi.
12341 (maybe_diag_overlap): Remove assertion.
12342 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
12343 * gimple-ssa-sprintf.c (format_directive): Same.
12344 (parse_directive): Same.
12345 (sprintf_dom_walker::compute_format_length): Same.
12346 (try_substitute_return_value): Same.
12348 2018-01-03 Jeff Law <law@redhat.com>
12350 PR middle-end/83654
12351 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
12352 non-constant residual for zero at runtime and avoid probing in
12353 that case. Reorganize code for trailing problem to mirror handling
12356 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12358 PR tree-optimization/83501
12359 * tree-ssa-strlen.c (get_string_cst): New.
12360 (handle_char_store): Call get_string_cst.
12362 2018-01-03 Martin Liska <mliska@suse.cz>
12364 PR tree-optimization/83593
12365 * tree-ssa-strlen.c: Include tree-cfg.h.
12366 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
12367 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
12368 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
12370 (strlen_dom_walker::before_dom_children): Call
12371 gimple_purge_dead_eh_edges. Dump tranformation with details
12373 (strlen_dom_walker::before_dom_children): Update call by adding
12374 new argument cleanup_eh.
12375 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
12377 2018-01-03 Martin Liska <mliska@suse.cz>
12380 * cif-code.def (VARIADIC_THUNK): New enum value.
12381 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
12384 2018-01-03 Jan Beulich <jbeulich@suse.com>
12386 * sse.md (mov<mode>_internal): Tighten condition for when to use
12387 vmovdqu<ssescalarsize> for TI and OI modes.
12389 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12391 Update copyright years.
12393 2018-01-03 Martin Liska <mliska@suse.cz>
12396 * ipa-visibility.c (function_and_variable_visibility): Skip
12397 functions with noipa attribure.
12399 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12401 * gcc.c (process_command): Update copyright notice dates.
12402 * gcov-dump.c (print_version): Ditto.
12403 * gcov.c (print_version): Ditto.
12404 * gcov-tool.c (print_version): Ditto.
12405 * gengtype.c (create_file): Ditto.
12406 * doc/cpp.texi: Bump @copying's copyright year.
12407 * doc/cppinternals.texi: Ditto.
12408 * doc/gcc.texi: Ditto.
12409 * doc/gccint.texi: Ditto.
12410 * doc/gcov.texi: Ditto.
12411 * doc/install.texi: Ditto.
12412 * doc/invoke.texi: Ditto.
12414 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12416 * vector-builder.h (vector_builder::m_full_nelts): Change from
12417 unsigned int to poly_uint64.
12418 (vector_builder::full_nelts): Update prototype accordingly.
12419 (vector_builder::new_vector): Likewise.
12420 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12421 (vector_builder::operator ==): Likewise.
12422 (vector_builder::finalize): Likewise.
12423 * int-vector-builder.h (int_vector_builder::int_vector_builder):
12424 Take the number of elements as a poly_uint64 rather than an
12426 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12427 from unsigned int to poly_uint64.
12428 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12429 (vec_perm_indices::new_vector): Likewise.
12430 (vec_perm_indices::length): Likewise.
12431 (vec_perm_indices::nelts_per_input): Likewise.
12432 (vec_perm_indices::input_nelts): Likewise.
12433 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12434 number of elements per input as a poly_uint64 rather than an
12435 unsigned int. Use the original encoding for variable-length
12436 vectors, rather than clamping each individual element.
12437 For the second and subsequent elements in each pattern,
12438 clamp the step and base before clamping their sum.
12439 (vec_perm_indices::series_p): Handle polynomial element counts.
12440 (vec_perm_indices::all_in_range_p): Likewise.
12441 (vec_perm_indices_to_tree): Likewise.
12442 (vec_perm_indices_to_rtx): Likewise.
12443 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12444 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12445 (tree_vector_builder::new_binary_operation): Handle polynomial
12446 element counts. Return false if we need to know the number
12447 of elements at compile time.
12448 * fold-const.c (fold_vec_perm): Punt if the number of elements
12449 isn't known at compile time.
12451 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12453 * vec-perm-indices.h (vec_perm_builder): Change element type
12454 from HOST_WIDE_INT to poly_int64.
12455 (vec_perm_indices::element_type): Update accordingly.
12456 (vec_perm_indices::clamp): Handle polynomial element_types.
12457 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12458 (vec_perm_indices::all_in_range_p): Likewise.
12459 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12461 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12462 polynomial vec_perm_indices element types.
12463 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12464 * fold-const.c (fold_vec_perm): Likewise.
12465 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12466 * tree-vect-generic.c (lower_vec_perm): Likewise.
12467 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12468 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12469 element type to HOST_WIDE_INT.
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 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12476 rather than an int. Use plus_constant.
12477 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12478 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12480 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12481 Alan Hayward <alan.hayward@arm.com>
12482 David Sherwood <david.sherwood@arm.com>
12484 * calls.c (emit_call_1, expand_call): Change struct_value_size from
12485 a HOST_WIDE_INT to a poly_int64.
12487 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12488 Alan Hayward <alan.hayward@arm.com>
12489 David Sherwood <david.sherwood@arm.com>
12491 * calls.c (load_register_parameters): Cope with polynomial
12492 mode sizes. Require a constant size for BLKmode parameters
12493 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
12494 forces a parameter to be padded at the lsb end in order to
12495 fill a complete number of words, require the parameter size
12496 to be ordered wrt UNITS_PER_WORD.
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 * reload1.c (spill_stack_slot_width): Change element type
12503 from unsigned int to poly_uint64_pod.
12504 (alter_reg): Treat mode sizes as polynomial.
12506 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12507 Alan Hayward <alan.hayward@arm.com>
12508 David Sherwood <david.sherwood@arm.com>
12510 * reload.c (complex_word_subreg_p): New function.
12511 (reload_inner_reg_of_subreg, push_reload): Use it.
12513 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12514 Alan Hayward <alan.hayward@arm.com>
12515 David Sherwood <david.sherwood@arm.com>
12517 * lra-constraints.c (process_alt_operands): Reject matched
12518 operands whose sizes aren't ordered.
12519 (match_reload): Refer to this check here.
12521 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12522 Alan Hayward <alan.hayward@arm.com>
12523 David Sherwood <david.sherwood@arm.com>
12525 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12526 that the mode size is in the set {1, 2, 4, 8, 16}.
12528 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12529 Alan Hayward <alan.hayward@arm.com>
12530 David Sherwood <david.sherwood@arm.com>
12532 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12533 Use plus_constant instead of gen_rtx_PLUS.
12535 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12536 Alan Hayward <alan.hayward@arm.com>
12537 David Sherwood <david.sherwood@arm.com>
12539 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12540 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12541 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12542 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12543 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12544 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12545 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12546 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12547 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12548 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12550 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12551 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12552 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12553 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12554 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12555 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12556 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12557 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12558 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12559 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12561 * expr.c (emit_move_resolve_push): Treat the input and result
12562 of PUSH_ROUNDING as a poly_int64.
12563 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12564 (emit_push_insn): Likewise.
12565 * lra-eliminations.c (mark_not_eliminable): Likewise.
12566 * recog.c (push_operand): Likewise.
12567 * reload1.c (elimination_effects): Likewise.
12568 * rtlanal.c (nonzero_bits1): Likewise.
12569 * calls.c (store_one_arg): Likewise. Require the padding to be
12570 known at compile time.
12572 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12573 Alan Hayward <alan.hayward@arm.com>
12574 David Sherwood <david.sherwood@arm.com>
12576 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12577 Use plus_constant instead of gen_rtx_PLUS.
12579 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12580 Alan Hayward <alan.hayward@arm.com>
12581 David Sherwood <david.sherwood@arm.com>
12583 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12584 rather than an int.
12586 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12587 Alan Hayward <alan.hayward@arm.com>
12588 David Sherwood <david.sherwood@arm.com>
12590 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12591 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12592 via stack temporaries. Treat the mode size as polynomial too.
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 * expr.c (expand_expr_real_2): When handling conversions involving
12599 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12600 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12601 as a poly_uint64 too.
12603 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12604 Alan Hayward <alan.hayward@arm.com>
12605 David Sherwood <david.sherwood@arm.com>
12607 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
12609 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12610 Alan Hayward <alan.hayward@arm.com>
12611 David Sherwood <david.sherwood@arm.com>
12613 * combine.c (can_change_dest_mode): Handle polynomial
12614 REGMODE_NATURAL_SIZE.
12615 * expmed.c (store_bit_field_1): Likewise.
12616 * expr.c (store_constructor): Likewise.
12617 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12618 and polynomial REGMODE_NATURAL_SIZE.
12619 (gen_lowpart_common): Likewise.
12620 * reginfo.c (record_subregs_of_mode): Likewise.
12621 * rtlanal.c (read_modify_subreg_p): Likewise.
12623 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12624 Alan Hayward <alan.hayward@arm.com>
12625 David Sherwood <david.sherwood@arm.com>
12627 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12628 numbers of elements.
12630 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12631 Alan Hayward <alan.hayward@arm.com>
12632 David Sherwood <david.sherwood@arm.com>
12634 * match.pd: Cope with polynomial numbers of vector elements.
12636 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12637 Alan Hayward <alan.hayward@arm.com>
12638 David Sherwood <david.sherwood@arm.com>
12640 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12641 in a POINTER_PLUS_EXPR.
12643 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12644 Alan Hayward <alan.hayward@arm.com>
12645 David Sherwood <david.sherwood@arm.com>
12647 * omp-simd-clone.c (simd_clone_subparts): New function.
12648 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12649 (ipa_simd_modify_function_body): Likewise.
12651 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12652 Alan Hayward <alan.hayward@arm.com>
12653 David Sherwood <david.sherwood@arm.com>
12655 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
12656 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
12657 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
12658 (expand_vector_condition, vector_element): Likewise.
12659 (subparts_gt): New function.
12660 (get_compute_type): Use subparts_gt.
12661 (count_type_subparts): Delete.
12662 (expand_vector_operations_1): Use subparts_gt instead of
12663 count_type_subparts.
12665 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12666 Alan Hayward <alan.hayward@arm.com>
12667 David Sherwood <david.sherwood@arm.com>
12669 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
12670 (vect_compile_time_alias): ...this new function. Do the calculation
12671 on poly_ints rather than trees.
12672 (vect_prune_runtime_alias_test_list): Update call accordingly.
12674 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12675 Alan Hayward <alan.hayward@arm.com>
12676 David Sherwood <david.sherwood@arm.com>
12678 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
12680 (vect_schedule_slp_instance): Likewise.
12682 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12683 Alan Hayward <alan.hayward@arm.com>
12684 David Sherwood <david.sherwood@arm.com>
12686 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12687 constant and extern definitions for variable-length vectors.
12688 (vect_get_constant_vectors): Note that the number of units
12689 is known to be constant.
12691 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12692 Alan Hayward <alan.hayward@arm.com>
12693 David Sherwood <david.sherwood@arm.com>
12695 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
12696 of units as polynomial. Choose between WIDE and NARROW based
12699 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12700 Alan Hayward <alan.hayward@arm.com>
12701 David Sherwood <david.sherwood@arm.com>
12703 * tree-vect-stmts.c (simd_clone_subparts): New function.
12704 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
12706 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12707 Alan Hayward <alan.hayward@arm.com>
12708 David Sherwood <david.sherwood@arm.com>
12710 * tree-vect-stmts.c (vectorizable_call): Treat the number of
12711 vectors as polynomial. Use build_index_vector for
12712 IFN_GOMP_SIMD_LANE.
12714 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12715 Alan Hayward <alan.hayward@arm.com>
12716 David Sherwood <david.sherwood@arm.com>
12718 * tree-vect-stmts.c (get_load_store_type): Treat the number of
12719 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
12720 for variable-length vectors.
12721 (vectorizable_mask_load_store): Treat the number of units as
12722 polynomial, asserting that it is constant if the condition has
12723 already been enforced.
12724 (vectorizable_store, vectorizable_load): Likewise.
12726 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12727 Alan Hayward <alan.hayward@arm.com>
12728 David Sherwood <david.sherwood@arm.com>
12730 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
12731 of units as polynomial. Punt if we can't tell at compile time
12732 which vector contains the final result.
12734 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12735 Alan Hayward <alan.hayward@arm.com>
12736 David Sherwood <david.sherwood@arm.com>
12738 * tree-vect-loop.c (vectorizable_induction): Treat the number
12739 of units as polynomial. Punt on SLP inductions. Use an integer
12740 VEC_SERIES_EXPR for variable-length integer reductions. Use a
12741 cast of such a series for variable-length floating-point
12744 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12745 Alan Hayward <alan.hayward@arm.com>
12746 David Sherwood <david.sherwood@arm.com>
12748 * tree.h (build_index_vector): Declare.
12749 * tree.c (build_index_vector): New function.
12750 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
12751 of units as polynomial, forcibly converting it to a constant if
12752 vectorizable_reduction has already enforced the condition.
12753 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
12754 to create a {1,2,3,...} vector.
12755 (vectorizable_reduction): Treat the number of units as polynomial.
12756 Choose vectype_in based on the largest scalar element size rather
12757 than the smallest number of units. Enforce the restrictions
12760 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12761 Alan Hayward <alan.hayward@arm.com>
12762 David Sherwood <david.sherwood@arm.com>
12764 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
12765 number of units as polynomial.
12767 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12768 Alan Hayward <alan.hayward@arm.com>
12769 David Sherwood <david.sherwood@arm.com>
12771 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
12772 * target.def (autovectorize_vector_sizes): Return the vector sizes
12773 by pointer, using vector_sizes rather than a bitmask.
12774 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
12775 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
12776 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
12778 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
12779 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
12780 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
12781 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
12782 * omp-general.c (omp_max_vf): Likewise.
12783 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12784 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
12785 * tree-vect-loop.c (vect_analyze_loop): Likewise.
12786 * tree-vect-slp.c (vect_slp_bb): Likewise.
12787 * doc/tm.texi: Regenerate.
12788 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
12790 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
12791 the vector size as a poly_uint64 rather than an unsigned int.
12792 (current_vector_size): Change from an unsigned int to a poly_uint64.
12793 (get_vectype_for_scalar_type): Update accordingly.
12794 * tree.h (build_truth_vector_type): Take the size and number of
12795 units as a poly_uint64 rather than an unsigned int.
12796 (build_vector_type): Add a temporary overload that takes
12797 the number of units as a poly_uint64 rather than an unsigned int.
12798 * tree.c (make_vector_type): Likewise.
12799 (build_truth_vector_type): Take the number of units as a poly_uint64
12800 rather than an unsigned int.
12802 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12803 Alan Hayward <alan.hayward@arm.com>
12804 David Sherwood <david.sherwood@arm.com>
12806 * target.def (get_mask_mode): Take the number of units and length
12807 as poly_uint64s rather than unsigned ints.
12808 * targhooks.h (default_get_mask_mode): Update accordingly.
12809 * targhooks.c (default_get_mask_mode): Likewise.
12810 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
12811 * doc/tm.texi: Regenerate.
12813 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12814 Alan Hayward <alan.hayward@arm.com>
12815 David Sherwood <david.sherwood@arm.com>
12817 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
12818 * omp-general.c (omp_max_vf): Likewise.
12819 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
12820 (expand_omp_simd): Handle polynomial safelen.
12821 * omp-low.c (omplow_simd_context): Add a default constructor.
12822 (omplow_simd_context::max_vf): Change from int to poly_uint64.
12823 (lower_rec_simd_input_clauses): Update accordingly.
12824 (lower_rec_input_clauses): Likewise.
12826 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12827 Alan Hayward <alan.hayward@arm.com>
12828 David Sherwood <david.sherwood@arm.com>
12830 * tree-vectorizer.h (vect_nunits_for_cost): New function.
12831 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
12832 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
12833 (vect_analyze_slp_cost): Likewise.
12834 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
12835 (vect_model_load_cost): Likewise.
12837 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12838 Alan Hayward <alan.hayward@arm.com>
12839 David Sherwood <david.sherwood@arm.com>
12841 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
12842 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
12843 from an unsigned int * to a poly_uint64_pod *.
12844 (calculate_unrolling_factor): New function.
12845 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
12847 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12848 Alan Hayward <alan.hayward@arm.com>
12849 David Sherwood <david.sherwood@arm.com>
12851 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
12852 from an unsigned int to a poly_uint64.
12853 (_loop_vec_info::slp_unrolling_factor): Likewise.
12854 (_loop_vec_info::vectorization_factor): Change from an int
12856 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
12857 (vect_get_num_vectors): New function.
12858 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
12859 (vect_get_num_copies): Use vect_get_num_vectors.
12860 (vect_analyze_data_ref_dependences): Change max_vf from an int *
12861 to an unsigned int *.
12862 (vect_analyze_data_refs): Change min_vf from an int * to a
12864 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12865 than an unsigned HOST_WIDE_INT.
12866 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
12867 (vect_analyze_data_ref_dependence): Change max_vf from an int *
12868 to an unsigned int *.
12869 (vect_analyze_data_ref_dependences): Likewise.
12870 (vect_compute_data_ref_alignment): Handle polynomial vf.
12871 (vect_enhance_data_refs_alignment): Likewise.
12872 (vect_prune_runtime_alias_test_list): Likewise.
12873 (vect_shift_permute_load_chain): Likewise.
12874 (vect_supportable_dr_alignment): Likewise.
12875 (dependence_distance_ge_vf): Take the vectorization factor as a
12876 poly_uint64 rather than an unsigned HOST_WIDE_INT.
12877 (vect_analyze_data_refs): Change min_vf from an int * to a
12879 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
12880 vfm1 as a poly_uint64 rather than an int. Make the same change
12881 for the returned bound_scalar.
12882 (vect_gen_vector_loop_niters): Handle polynomial vf.
12883 (vect_do_peeling): Likewise. Update call to
12884 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
12885 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
12887 * tree-vect-loop.c (vect_determine_vectorization_factor)
12888 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
12889 (vect_get_known_peeling_cost): Likewise.
12890 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
12891 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
12892 (vect_transform_loop): Likewise. Use the lowest possible VF when
12893 updating the upper bounds of the loop.
12894 (vect_min_worthwhile_factor): Make static. Return an unsigned int
12895 rather than an int.
12896 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
12897 polynomial unroll factors.
12898 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
12899 (vect_make_slp_decision): Likewise.
12900 (vect_supported_load_permutation_p): Likewise, and polynomial
12902 (vect_analyze_slp_cost): Handle polynomial vf.
12903 (vect_slp_analyze_node_operations): Likewise.
12904 (vect_slp_analyze_bb_1): Likewise.
12905 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12906 than an unsigned HOST_WIDE_INT.
12907 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
12908 (vectorizable_load): Handle polynomial vf.
12909 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
12911 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
12913 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12914 Alan Hayward <alan.hayward@arm.com>
12915 David Sherwood <david.sherwood@arm.com>
12917 * match.pd: Handle bit operations involving three constants
12918 and try to fold one pair.
12920 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12922 * tree-vect-loop-manip.c: Include gimple-fold.h.
12923 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
12924 niters_maybe_zero parameters. Handle other cases besides a step of 1.
12925 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
12926 Add a path that uses a step of VF instead of 1, but disable it
12928 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
12929 and niters_no_overflow parameters. Update calls to
12930 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
12931 Create a new SSA name if the latter choses to use a ste other
12932 than zero, and return it via niters_vector_mult_vf_var.
12933 * tree-vect-loop.c (vect_transform_loop): Update calls to
12934 vect_do_peeling, vect_gen_vector_loop_niters and
12935 slpeel_make_loop_iterate_ntimes.
12936 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
12937 (vect_gen_vector_loop_niters): Update declarations after above changes.
12939 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
12941 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
12942 128-bit round to integer instructions.
12943 (ceil<mode>2): Likewise.
12944 (btrunc<mode>2): Likewise.
12945 (round<mode>2): Likewise.
12947 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12949 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
12950 unaligned VSX load/store on P8/P9.
12951 (expand_block_clear): Allow the use of unaligned VSX
12952 load/store on P8/P9.
12954 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12956 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
12958 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
12959 swap associated with both a load and a store.
12961 2018-01-02 Andrew Waterman <andrew@sifive.com>
12963 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
12964 * config/riscv/riscv.md (clear_cache): Use it.
12966 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
12968 * web.c: Remove out-of-date comment.
12970 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12972 * expr.c (fixup_args_size_notes): Check that any existing
12973 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
12974 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
12975 (emit_single_push_insn): ...here.
12977 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12979 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
12980 (const_vector_encoded_nelts): New function.
12981 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
12982 (const_vector_int_elt, const_vector_elt): Declare.
12983 * emit-rtl.c (const_vector_int_elt_1): New function.
12984 (const_vector_elt): Likewise.
12985 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
12986 of CONST_VECTOR_ELT.
12988 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12990 * expr.c: Include rtx-vector-builder.h.
12991 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
12992 directly on the tree encoding.
12993 (const_vector_from_tree): Likewise.
12994 * optabs.c: Include rtx-vector-builder.h.
12995 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
12996 sequence of "u" values.
12997 * vec-perm-indices.c: Include rtx-vector-builder.h.
12998 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
12999 directly on the vec_perm_indices encoding.
13001 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13003 * doc/rtl.texi (const_vector): Describe new encoding scheme.
13004 * Makefile.in (OBJS): Add rtx-vector-builder.o.
13005 * rtx-vector-builder.h: New file.
13006 * rtx-vector-builder.c: Likewise.
13007 * rtl.h (rtx_def::u2): Add a const_vector field.
13008 (CONST_VECTOR_NPATTERNS): New macro.
13009 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
13010 (CONST_VECTOR_DUPLICATE_P): Likewise.
13011 (CONST_VECTOR_STEPPED_P): Likewise.
13012 (CONST_VECTOR_ENCODED_ELT): Likewise.
13013 (const_vec_duplicate_p): Check for a duplicated vector encoding.
13014 (unwrap_const_vec_duplicate): Likewise.
13015 (const_vec_series_p): Check for a non-duplicated vector encoding.
13016 Say that the function only returns true for integer vectors.
13017 * emit-rtl.c: Include rtx-vector-builder.h.
13018 (gen_const_vec_duplicate_1): Delete.
13019 (gen_const_vector): Call gen_const_vec_duplicate instead of
13020 gen_const_vec_duplicate_1.
13021 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
13022 (gen_const_vec_duplicate): Use rtx_vector_builder.
13023 (gen_const_vec_series): Likewise.
13024 (gen_rtx_CONST_VECTOR): Likewise.
13025 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
13026 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13027 Build a new vector rather than modifying a CONST_VECTOR in-place.
13028 (handle_special_swappables): Update call accordingly.
13029 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
13030 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13031 Build a new vector rather than modifying a CONST_VECTOR in-place.
13032 (handle_special_swappables): Update call accordingly.
13034 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13036 * simplify-rtx.c (simplify_const_binary_operation): Use
13037 CONST_VECTOR_ELT instead of XVECEXP.
13039 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13041 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
13042 the selector elements to be different from the data elements
13043 if the selector is a VECTOR_CST.
13044 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
13045 ssizetype for the selector.
13047 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13049 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
13050 before testing each element individually.
13051 * tree-vect-generic.c (lower_vec_perm): Likewise.
13053 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13055 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
13056 * selftest-run-tests.c (selftest::run_tests): Call it.
13057 * vector-builder.h (vector_builder::operator ==): New function.
13058 (vector_builder::operator !=): Likewise.
13059 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
13060 (vec_perm_indices::all_from_input_p): New function.
13061 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13062 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
13063 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
13064 instead of reading the VECTOR_CST directly. Detect whether both
13065 vector inputs are the same before constructing the vec_perm_indices,
13066 and update the number of inputs argument accordingly. Use the
13067 utility functions added above. Only construct sel2 if we need to.
13069 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13071 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
13072 the broadcast of the low byte.
13073 (expand_mult_highpart): Use an explicit encoding for the permutes.
13074 * optabs-query.c (can_mult_highpart_p): Likewise.
13075 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
13076 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13077 (vectorizable_bswap): Likewise.
13078 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
13079 explicit encoding for the power-of-2 permutes.
13080 (vect_permute_store_chain): Likewise.
13081 (vect_grouped_load_supported): Likewise.
13082 (vect_permute_load_chain): Likewise.
13084 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13086 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
13087 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
13088 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
13089 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13090 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
13091 (vect_gen_perm_mask_any): Likewise.
13093 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13095 * int-vector-builder.h: New file.
13096 * vec-perm-indices.h: Include int-vector-builder.h.
13097 (vec_perm_indices): Redefine as an int_vector_builder.
13098 (auto_vec_perm_indices): Delete.
13099 (vec_perm_builder): Redefine as a stand-alone class.
13100 (vec_perm_indices::vec_perm_indices): New function.
13101 (vec_perm_indices::clamp): Likewise.
13102 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
13103 (vec_perm_indices::new_vector): New function.
13104 (vec_perm_indices::new_expanded_vector): Update for new
13105 vec_perm_indices class.
13106 (vec_perm_indices::rotate_inputs): New function.
13107 (vec_perm_indices::all_in_range_p): Operate directly on the
13108 encoded form, without computing elided elements.
13109 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
13110 encoding. Update for new vec_perm_indices class.
13111 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
13112 the given vec_perm_builder.
13113 (expand_vec_perm_var): Update vec_perm_builder constructor.
13114 (expand_mult_highpart): Use vec_perm_builder instead of
13115 auto_vec_perm_indices.
13116 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
13117 vec_perm_indices instead of auto_vec_perm_indices. Use a single
13118 or double series encoding as appropriate.
13119 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
13120 vec_perm_indices instead of auto_vec_perm_indices.
13121 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13122 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13123 (vect_permute_store_chain): Likewise.
13124 (vect_grouped_load_supported): Likewise.
13125 (vect_permute_load_chain): Likewise.
13126 (vect_shift_permute_load_chain): Likewise.
13127 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13128 (vect_transform_slp_perm_load): Likewise.
13129 (vect_schedule_slp_instance): Likewise.
13130 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13131 (vectorizable_mask_load_store): Likewise.
13132 (vectorizable_bswap): Likewise.
13133 (vectorizable_store): Likewise.
13134 (vectorizable_load): Likewise.
13135 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
13136 vec_perm_indices instead of auto_vec_perm_indices. Use
13137 tree_to_vec_perm_builder to read the vector from a tree.
13138 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
13139 vec_perm_builder instead of a vec_perm_indices.
13140 (have_whole_vector_shift): Use vec_perm_builder and
13141 vec_perm_indices instead of auto_vec_perm_indices. Leave the
13142 truncation to calc_vec_perm_mask_for_shift.
13143 (vect_create_epilog_for_reduction): Likewise.
13144 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
13145 from auto_vec_perm_indices to vec_perm_indices.
13146 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13147 instead of changing individual elements.
13148 (aarch64_vectorize_vec_perm_const): Use new_vector to install
13149 the vector in d.perm.
13150 * config/arm/arm.c (expand_vec_perm_d::perm): Change
13151 from auto_vec_perm_indices to vec_perm_indices.
13152 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13153 instead of changing individual elements.
13154 (arm_vectorize_vec_perm_const): Use new_vector to install
13155 the vector in d.perm.
13156 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
13157 Update vec_perm_builder constructor.
13158 (rs6000_expand_interleave): Likewise.
13159 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
13160 (rs6000_expand_interleave): Likewise.
13162 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13164 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
13165 to qimode could truncate the indices.
13166 * optabs.c (expand_vec_perm_var): Likewise.
13168 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13170 * Makefile.in (OBJS): Add vec-perm-indices.o.
13171 * vec-perm-indices.h: New file.
13172 * vec-perm-indices.c: Likewise.
13173 * target.h (vec_perm_indices): Replace with a forward class
13175 (auto_vec_perm_indices): Move to vec-perm-indices.h.
13176 * optabs.h: Include vec-perm-indices.h.
13177 (expand_vec_perm): Delete.
13178 (selector_fits_mode_p, expand_vec_perm_var): Declare.
13179 (expand_vec_perm_const): Declare.
13180 * target.def (vec_perm_const_ok): Replace with...
13181 (vec_perm_const): ...this new hook.
13182 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
13183 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
13184 * doc/tm.texi: Regenerate.
13185 * optabs.def (vec_perm_const): Delete.
13186 * doc/md.texi (vec_perm_const): Likewise.
13187 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
13188 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
13189 expand_vec_perm for constant permutation vectors. Assert that
13190 the mode of variable permutation vectors is the integer equivalent
13191 of the mode that is being permuted.
13192 * optabs-query.h (selector_fits_mode_p): Declare.
13193 * optabs-query.c: Include vec-perm-indices.h.
13194 (selector_fits_mode_p): New function.
13195 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
13196 is defined, instead of checking whether the vec_perm_const_optab
13197 exists. Use targetm.vectorize.vec_perm_const instead of
13198 targetm.vectorize.vec_perm_const_ok. Check whether the indices
13199 fit in the vector mode before using a variable permute.
13200 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
13201 vec_perm_indices instead of an rtx.
13202 (expand_vec_perm): Replace with...
13203 (expand_vec_perm_const): ...this new function. Take the selector
13204 as a vec_perm_indices rather than an rtx. Also take the mode of
13205 the selector. Update call to shift_amt_for_vec_perm_mask.
13206 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
13207 Use vec_perm_indices::new_expanded_vector to expand the original
13208 selector into bytes. Check whether the indices fit in the vector
13209 mode before using a variable permute.
13210 (expand_vec_perm_var): Make global.
13211 (expand_mult_highpart): Use expand_vec_perm_const.
13212 * fold-const.c: Includes vec-perm-indices.h.
13213 * tree-ssa-forwprop.c: Likewise.
13214 * tree-vect-data-refs.c: Likewise.
13215 * tree-vect-generic.c: Likewise.
13216 * tree-vect-loop.c: Likewise.
13217 * tree-vect-slp.c: Likewise.
13218 * tree-vect-stmts.c: Likewise.
13219 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
13221 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
13222 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
13223 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
13224 (aarch64_vectorize_vec_perm_const): ...this new function.
13225 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13226 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13227 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
13228 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
13229 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13230 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13231 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
13233 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
13234 check for NEON modes.
13235 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
13236 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
13237 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
13238 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
13240 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
13241 the old VEC_PERM_CONST conditions.
13242 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
13243 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
13244 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
13245 (ia64_vectorize_vec_perm_const_ok): Merge into...
13246 (ia64_vectorize_vec_perm_const): ...this new function.
13247 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
13248 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
13249 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
13250 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
13251 * config/mips/mips.c (mips_expand_vec_perm_const)
13252 (mips_vectorize_vec_perm_const_ok): Merge into...
13253 (mips_vectorize_vec_perm_const): ...this new function.
13254 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
13255 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
13256 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
13257 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
13258 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
13259 (rs6000_expand_vec_perm_const): Delete.
13260 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
13262 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13263 (altivec_expand_vec_perm_const_le): Take each operand individually.
13264 Operate on constant selectors rather than rtxes.
13265 (altivec_expand_vec_perm_const): Likewise. Update call to
13266 altivec_expand_vec_perm_const_le.
13267 (rs6000_expand_vec_perm_const): Delete.
13268 (rs6000_vectorize_vec_perm_const_ok): Delete.
13269 (rs6000_vectorize_vec_perm_const): New function.
13270 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13271 an element count and rtx array.
13272 (rs6000_expand_extract_even): Update call accordingly.
13273 (rs6000_expand_interleave): Likewise.
13274 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
13275 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
13276 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
13277 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
13278 (rs6000_expand_vec_perm_const): Delete.
13279 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13280 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13281 (altivec_expand_vec_perm_const_le): Take each operand individually.
13282 Operate on constant selectors rather than rtxes.
13283 (altivec_expand_vec_perm_const): Likewise. Update call to
13284 altivec_expand_vec_perm_const_le.
13285 (rs6000_expand_vec_perm_const): Delete.
13286 (rs6000_vectorize_vec_perm_const_ok): Delete.
13287 (rs6000_vectorize_vec_perm_const): New function. Remove stray
13288 reference to the SPE evmerge intructions.
13289 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13290 an element count and rtx array.
13291 (rs6000_expand_extract_even): Update call accordingly.
13292 (rs6000_expand_interleave): Likewise.
13293 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
13294 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
13296 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13298 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13300 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
13301 vector mode and that that mode matches the mode of the data
13303 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
13304 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
13305 directly using expand_vec_perm_1 when forcing selectors into
13307 (expand_vec_perm_var): New function, split out from expand_vec_perm.
13309 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13311 * optabs-query.h (can_vec_perm_p): Delete.
13312 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
13313 * optabs-query.c (can_vec_perm_p): Split into...
13314 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
13315 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
13316 particular selector is valid.
13317 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13318 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13319 (vect_grouped_load_supported): Likewise.
13320 (vect_shift_permute_load_chain): Likewise.
13321 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13322 (vect_transform_slp_perm_load): Likewise.
13323 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13324 (vectorizable_bswap): Likewise.
13325 (vect_gen_perm_mask_checked): Likewise.
13326 * fold-const.c (fold_ternary_loc): Likewise. Don't take
13327 implementations of variable permutation vectors into account
13328 when deciding which selector to use.
13329 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
13330 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
13331 with a false third argument.
13332 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
13333 to test whether the constant selector is valid and can_vec_perm_var_p
13334 to test whether a variable selector is valid.
13336 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13338 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
13339 * optabs-query.c (can_vec_perm_p): Likewise.
13340 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
13341 instead of vec_perm_indices.
13342 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
13343 (vect_gen_perm_mask_checked): Likewise,
13344 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
13345 (vect_gen_perm_mask_checked): Likewise,
13347 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13349 * optabs-query.h (qimode_for_vec_perm): Declare.
13350 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
13351 (qimode_for_vec_perm): ...this new function.
13352 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
13354 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13356 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
13357 does not have a conditional at the top.
13359 2018-01-02 Richard Biener <rguenther@suse.de>
13361 * ipa-inline.c (big_speedup_p): Fix expression.
13363 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13366 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
13369 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13373 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
13374 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
13375 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
13376 cond_taken_branch_cost 3->4.
13378 2018-01-01 Jakub Jelinek <jakub@redhat.com>
13380 PR tree-optimization/83581
13381 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
13382 TODO_cleanup_cfg if any changes have been made.
13384 PR middle-end/83608
13385 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
13386 convert_modes if target mode has the right side, but different mode
13389 PR middle-end/83609
13390 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
13391 last argument when extracting from CONCAT. If either from_real or
13392 from_imag is NULL, use expansion through memory. If result is not
13393 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
13394 the parts directly to inner mode, if even that fails, use expansion
13397 PR middle-end/83623
13398 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
13399 check for bswap in mode rather than HImode and use that in expand_unop
13402 Copyright (C) 2018 Free Software Foundation, Inc.
13404 Copying and distribution of this file, with or without modification,
13405 are permitted in any medium without royalty provided the copyright
13406 notice and this notice are preserved.