1 2018-05-18 Martin Liska <mliska@suse.cz>
3 * passes.def: Remove a redundant pass.
5 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
8 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
10 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
14 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
16 (armv2, armv2a, armv3, armv3m): Delete architectures.
17 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
18 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
19 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
21 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
22 (*mulsidi3adddi): Likewise.
24 (*mulsidi3_nov6): Likewise.
25 (umulsidi3): Likewise.
26 (umulsidi3_nov6): Likewise.
27 (umaddsidi4): Likewise.
28 (*umulsidi3adddi): Likewise.
29 (smulsi3_highpart): Likewise.
30 (*smulsi3_highpart_nov6): Likewise.
31 (umulsi3_highpart): Likewise.
32 (*umulsi3_highpart_nov6): Likewise.
33 * config/arm/arm.h (arm_arch3m): Delete.
34 * config/arm/arm.c (arm_arch3m): Delete.
35 (arm_option_override_internal): Update armv3-related comment.
36 (arm_configure_build_target): Delete use of isa_bit_mode32.
37 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
38 (arm_rtx_costs_internal): Delete check of arm_arch3m.
39 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
42 * config/arm/arm-protos.h (arm_arch3m): Delete.
43 * config/arm/arm-tables.opt: Regenerate.
44 * config/arm/arm-tune.md: Likewise.
45 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
46 deleted architectures.
48 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
50 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
51 (armv5t, armv5te): New features.
52 (ARMv5, ARMv5e): Delete fgroups.
53 (ARMv5t, ARMv5te): Adjust for above changes.
55 (armv5, armv5e): Delete arches.
56 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
58 (*call_reg_arm): Likewise.
59 (*call_value_reg_armv5): Likewise.
60 (*call_value_reg_arm): Likewise.
61 (*call_symbol): Likewise.
62 (*call_value_symbol): Likewise.
63 (*sibcall_insn): Likewise.
64 (*sibcall_value_insn): Likewise.
67 (define_split and define_peephole2 dependent on arm_arch5):
69 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
71 (TARGET_ARM_QBIT): Likewise.
72 (TARGET_DSP_MULTIPLY): Likewise.
73 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
74 (arm_arch5, arm_arch5e): Delete.
75 (arm_arch5t, arm_arch5te): Declare.
76 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
77 (arm_arch5t): Declare.
78 (arm_option_reconfigure_globals): Update for the above.
79 (arm_options_perform_arch_sanity_checks): Update comment, replace
80 use of arm_arch5 with arm_arch5t.
81 (use_return_insn): Likewise.
82 (arm_emit_call_insn): Likewise.
83 (output_return_instruction): Likewise.
84 (arm_final_prescan_insn): Likewise.
85 (arm_coproc_builtin_available): Likewise.
86 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
87 arm_arch5e with arm_arch5t and arm_arch5te.
88 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
89 (arm_arch5t, arm_arch5te): Declare.
90 * config/arm/arm-tables.opt: Regenerate.
91 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
92 * config/arm/t-multilib: Likewise.
93 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
95 (*call_reg_thumb1): Likewise.
96 (*call_value_reg_thumb1_v5): Likewise.
97 (*call_value_reg_thumb1): Likewise.
98 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
100 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
102 2018-05-18 Martin Liska <mliska@suse.cz>
104 PR gcov-profile/84846
105 * doc/gcov.texi: Document -t option of gcov tool.
107 2018-05-18 Martin Liska <mliska@suse.cz>
109 PR gcov-profile/84846
110 * gcov.c (print_usage): Add new -t option.
111 (process_args): Handle the option.
112 (generate_results): Use stdout as output when requested by
115 2018-05-18 Martin Liska <mliska@suse.cz>
117 PR gcov-profile/84846
118 * coverage.c (coverage_init): Write PWD to .gcno file.
119 * doc/gcov.texi: Document how working directory is printed.
120 * gcov-dump.c (dump_gcov_file): Print PWD.
121 * gcov.c (output_intermediate_file): Likewise.
122 (read_graph_file): Read PWD string.
123 (output_lines): Print PWD.
125 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
128 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
129 for retval and return false if all args to phi are zero.
131 2018-05-18 Richard Biener <rguenther@suse.de>
133 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
135 (evrp_dom_walker::before_dom_children): Call it.
137 2018-05-18 Richard Biener <rguenther@suse.de>
139 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
140 results when processing array refs with variable index.
142 2018-05-18 Toon Moene <toon@moene.org>
144 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
145 directly after that of -floop-interchange. Indicate that both
146 options are enabled by default when specifying -O3.
148 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
150 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
151 iterator. Delete separate integer-mode vec_set<mode> expander.
152 (aarch64_simd_vec_setv2di): Delete.
153 (vec_setv2di): Delete.
154 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
155 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
156 the "w, r" alternative.
158 2018-05-18 Martin Liska <mliska@suse.cz>
160 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
161 * tree-pass.h (make_pass_lower_switch_O0): New function.
162 * tree-switch-conversion.c (node_has_low_bound): Remove.
163 (node_has_high_bound): Likewise.
164 (node_is_bounded): Likewise.
165 (class pass_lower_switch): Make it a template type and create
167 (pass_lower_switch::execute): Add template argument.
168 (make_pass_lower_switch): New function.
169 (make_pass_lower_switch_O0): New function.
170 (do_jump_if_equal): Remove.
171 (emit_case_nodes): Simplify to just handle all 3 cases and leave
172 all the hard work to tree optimization passes.
174 2018-05-18 Martin Liska <mliska@suse.cz>
176 * dbgcnt.c (limit_low): Renamed from limit.
177 (limit_high): New variable.
178 (dbg_cnt_is_enabled): Check for upper limit.
179 (dbg_cnt): Adjust dumping.
180 (dbg_cnt_set_limit_by_index): Add new argument for high
182 (dbg_cnt_set_limit_by_name): Likewise.
183 (dbg_cnt_process_single_pair): Parse new format.
184 (dbg_cnt_process_opt): Use strtok.
185 (dbg_cnt_list_all_counters): Remove 'value' and add
187 * doc/invoke.texi: Document changes.
189 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
191 * doc/sourcebuild.texi (scalar_all_fma): Document.
192 * tree.def (FMA_EXPR): Delete.
193 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
194 * internal-fn.c (ternary_direct): New macro.
195 (expand_ternary_optab_fn): Likewise.
196 (direct_ternary_optab_supported_p): Likewise.
197 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
198 * builtins.c (fold_builtin_fma): Delete.
199 (fold_builtin_3): Don't call it.
200 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
201 * expr.c (expand_expr_real_2): Likewise.
202 * fold-const.c (operand_equal_p): Likewise.
203 (fold_ternary_loc): Likewise.
204 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
205 * gimple.c (DEFTREECODE): Likewise.
206 * gimplify.c (gimplify_expr): Likewise.
207 * optabs-tree.c (optab_for_tree_code): Likewise.
208 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
209 * tree-eh.c (operation_could_trap_p): Likewise.
210 (stmt_could_throw_1_p): Likewise.
211 * tree-inline.c (estimate_operator_cost): Likewise.
212 * tree-pretty-print.c (dump_generic_node): Likewise.
213 (op_code_prio): Likewise.
214 * tree-ssa-loop-im.c (stmt_cost): Likewise.
215 * tree-ssa-operands.c (get_expr_operands): Likewise.
216 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
217 * fold-const-call.h (fold_fma): Delete.
218 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
219 CFN_FNMA and CFN_FNMS.
221 * genmatch.c (combined_fn): New enum.
222 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
223 (commutative_op): New function.
224 (commutate): Use it. Handle more than 2 operands.
225 (dt_operand::gen_gimple_expr): Use commutative_op.
226 (parser::parse_expr): Allow :c to be used with non-binary
227 operators if the commutative operand is known.
228 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
229 CFN_FMS, CFN_FNMA and CFN_FNMS.
230 (backprop::process_assign_use): Remove FMA_EXPR handling.
231 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
232 (gen_hsa_fma): New function.
233 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
234 IFN_FNMA and IFN_FNMS.
235 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
236 * gimple-fold.h (follow_all_ssa_edges): Declare.
237 * gimple-fold.c (follow_all_ssa_edges): New function.
238 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
239 gimple_build interface and use follow_all_ssa_edges to fold the result.
240 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
241 instead of checking for optabs directly.
242 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
243 rather than FMA_EXPRs.
244 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
245 call to IFN_FMA instead of an FMA_EXPR.
247 2018-05-17 Jim Wilson <jimw@sifive.com>
249 * expr.c (do_tablejump): When converting index to Pmode, if we have a
250 sign extended promoted subreg, and the range does not have the sign bit
251 set, then do a sign extend.
253 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
254 test, check for sign extended subreg and/or constant operands, and
255 do a sign extend in that case.
257 2018-05-17 Steve Ellcey <sellcey@cavium.com>
259 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
260 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
262 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
263 Change logics_shift_reg to logics_shift_imm.
264 (thunderx2t99_fp_loadpair_basic): Delete.
265 (thunderx2t99_fp_storepair_basic): Delete.
266 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
267 (thunderx2t99_asimd_polynomial): Delete.
268 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
269 and neon_fp_mul_d_scalar_q.
270 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
271 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
272 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
273 (thunderx2t99_asimd_lut): Add missing tbl types.
274 (thunderx2t99_asimd_ext): Delete.
275 (thunderx2t99_asimd_load1_1_mult): Delete.
276 (thunderx2t99_asimd_load1_2_mult): Delete.
277 (thunderx2t99_asimd_load1_ldp): New.
278 (thunderx2t99_asimd_load1): New.
279 (thunderx2t99_asimd_load2): Add missing *load2* types.
280 (thunderx2t99_asimd_load3): New.
281 (thunderx2t99_asimd_load4): New.
282 (thunderx2t99_asimd_store1_1_mult): Delete.
283 (thunderx2t99_asimd_store1_2_mult): Delete.
284 (thunderx2t99_asimd_store2_mult): Delete.
285 (thunderx2t99_asimd_store2_onelane): Delete.
286 (thunderx2t99_asimd_store_stp): New.
287 (thunderx2t99_asimd_store1): New.
288 (thunderx2t99_asimd_store2): New.
289 (thunderx2t99_asimd_store3): New.
290 (thunderx2t99_asimd_store4): New.
292 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
294 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
295 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
297 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
298 Segher Boessenkool <segher@kernel.crashing.org>
301 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
304 2018-05-17 Richard Biener <rguenther@suse.de>
306 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
307 for pruning loop and prune defs feeding only already visited PHIs.
309 2018-05-17 Richard Biener <rguenther@suse.de>
311 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
313 2018-05-17 Bin Cheng <bin.cheng@arm.com>
314 Richard Biener <rguenther@suse.de>
316 PR tree-optimization/85793
317 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
318 for VMAT_ELEMENTWISE.
320 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
322 * internal-fn.h (lookup_internal_fn): Declare
323 * internal-fn.c (lookup_internal_fn): New function.
324 * gimple.c (gimple_build_call_from_tree): Handle calls to
326 * gimple-pretty-print.c (dump_gimple_call): Print "." before
327 internal function names.
328 * tree-pretty-print.c (dump_generic_node): Likewise.
329 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
331 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
333 * gimple-fold.h (gimple_build): Make the function forms take
334 combined_fn rather than built_in_function.
335 (gimple_simplify): Likewise.
336 * gimple-match-head.c (gimple_simplify): Likewise.
337 * gimple-fold.c (gimple_build): Likewise.
338 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
339 rather than gimple_build_call_internal.
340 (get_initial_defs_for_reduction): Likewise.
341 (vect_create_epilog_for_reduction): Likewise.
342 (vectorizable_live_operation): Likewise.
344 2018-05-17 Martin Liska <mliska@suse.cz>
346 * gimple-ssa-sprintf.c (format_directive): Do not use
347 space in between 'G_' and '('.
349 2018-05-17 Jakub Jelinek <jakub@redhat.com>
352 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
353 even if the mask is not all ones.
356 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
358 (ix86_gimple_fold_builtin): Likewise.
361 * config/i386/i386.c: Include tree-vector-builder.h.
362 (ix86_vector_shift_count): New function.
363 (ix86_fold_builtin): Fold shift builtins by scalar count.
364 (ix86_gimple_fold_builtin): Likewise.
366 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
367 _mm512_setzero): New intrinsics.
369 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
370 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
372 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
373 code generation for cases where splatting a value is not useful.
374 * simplify-rtx.c (simplify_ternary_operation): Simplify
375 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
376 mode to a vec_concat.
378 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
380 * config.gcc: Support "goldmont-plus".
381 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
383 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
384 PROCESSOR_GOLDMONT_PLUS.
385 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
386 (processor_target_table): Add "goldmont-plus".
387 (PTA_GOLDMONT_PLUS): Define.
388 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
389 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
390 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
391 (fold_builtin_cpu): Add "goldmont-plus".
392 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
393 (ix86_option_override_internal): Add "goldmont-plus".
394 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
395 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
396 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
397 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
399 2018-05-17 Richard Biener <rguenther@suse.de>
401 PR tree-optimization/85757
402 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
403 remove defs that only feed that PHI from further processing.
405 2018-05-16 Jim Wilson <jimw@sifive.com>
407 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
409 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
411 2018-05-16 Mark Wielaard <mark@klomp.org>
413 * dwarf2out.c (count_index_strings): New function.
414 (output_indirect_strings): Call count_index_strings and generate
415 header for dwarf_version >= 5.
417 2018-05-16 Mark Wielaard <mark@klomp.org>
419 * dwarf2out.c (dwarf_FORM): New function.
420 (set_indirect_string): Use dwarf_FORM.
421 (reset_indirect_string): Likewise.
422 (size_of_die): Likewise.
423 (value_format): Likewise.
424 (output_die): Likewise.
425 (add_skeleton_AT_string): Likewise.
426 (output_macinfo_op): Likewise.
427 (index_string): Likewise.
428 (output_index_string_offset): Likewise.
429 (output_index_string): Likewise.
430 (count_index_strings): Likewise.
432 2018-05-16 Carl Love <cel@us.ibm.com>
434 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
435 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
437 2018-05-16 Martin Jambor <mjambor@suse.cz>
439 * ipa-prop.c (ipa_free_all_edge_args): Remove.
440 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
442 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
444 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
445 (fnma<mode>4): Likewise.
446 (fms<mode>4): Likewise.
447 (fnms<mode>4): Likewise.
448 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
449 (aarch64_fnma<mode>4): Likewise.
450 (aarch64_fms<mode>4): Likewise.
451 (aarch64_fnms<mode>4): Likewise.
452 (aarch64_fnmadd<mode>4): Likewise.
454 2018-05-16 Jason Merrill <jason@redhat.com>
456 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
458 2018-05-16 Richard Biener <rguenther@suse.de>
460 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
461 (dump_stmt_cost): Declare.
462 (add_stmt_cost): Dump cost we add.
463 (add_stmt_costs): New function.
464 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
466 (vect_analyze_stmt): Adjust prototype.
467 (vectorizable_condition): Likewise.
468 (vectorizable_live_operation): Likewise.
469 (vectorizable_reduction): Likewise.
470 (vectorizable_induction): Likewise.
471 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
472 cost vector to pass to vectorizable_ and record afterwards.
473 (vect_model_reduction_cost): Take cost vector argument and adjust.
474 (vect_model_induction_cost): Likewise.
475 (vectorizable_reduction): Likewise.
476 (vectorizable_induction): Likewise.
477 (vectorizable_live_operation): Likewise.
478 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
479 SLP_TREE_NUMBER_OF_VEC_STMTS.
480 (vect_analyze_slp_cost_1): Remove.
481 (vect_analyze_slp_cost): Likewise.
482 (vect_slp_analyze_node_operations): Take visited args and
483 a target cost vector. Avoid processing already visited stmt sets.
484 (vect_slp_analyze_operations): Use a local cost vector to gather
485 costs and register those of non-discarded instances.
486 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
487 (vect_schedule_slp_instance): Remove copying of
488 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
490 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
491 adding cost. Record cost entry location.
492 (vect_prologue_cost_for_slp_op): Function to compute cost of
493 a constant or invariant generated for SLP vect in the prologue,
494 split out from vect_analyze_slp_cost_1.
495 (vect_model_simple_cost): Make static. Adjust for SLP costing.
496 (vect_model_promotion_demotion_cost): Likewise.
497 (vect_model_store_cost): Likewise, make static.
498 (vect_model_load_cost): Likewise.
499 (vectorizable_bswap): Add cost vector arg and adjust.
500 (vectorizable_call): Likewise.
501 (vectorizable_simd_clone_call): Likewise.
502 (vectorizable_conversion): Likewise.
503 (vectorizable_assignment): Likewise.
504 (vectorizable_shift): Likewise.
505 (vectorizable_operation): Likewise.
506 (vectorizable_store): Likewise.
507 (vectorizable_load): Likewise.
508 (vectorizable_condition): Likewise.
509 (vectorizable_comparison): Likewise.
510 (can_vectorize_live_stmts): Likewise.
511 (vect_analyze_stmt): Likewise.
512 (vect_transform_stmt): Adjust calls to vectorizable_*.
513 * tree-vectorizer.c: Include gimple-pretty-print.h.
514 (dump_stmt_cost): New function.
516 2018-05-16 Richard Biener <rguenther@suse.de>
518 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
519 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
520 * tree-ssa-dse.c: Include tree-ssa-loop.h.
521 (check_name): New callback.
522 (dse_classify_store): Track cycles via a visited bitmap of PHI
523 defs and simplify handling of in-loop and across loop dead stores
524 and properly fail for loop-variant refs. Handle byte-tracking with
525 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
528 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
530 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
531 (vect_get_mask_type_for_stmt): Likewise.
532 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
534 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
535 to determine the statement's vector type and the vector type that
536 should be used for calculating nunits. Deal with cases in which
537 the type has to be deferred.
538 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
539 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
540 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
541 (vect_determine_vf_for_stmt): New functions, split out from...
542 (vect_determine_vectorization_factor): ...here.
543 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
544 (vect_get_mask_type_for_stmt): New functions, split out from
545 vect_determine_vectorization_factor.
547 2018-05-16 Richard Biener <rguenther@suse.de>
549 * tree-cfg.c (verify_gimple_assign_ternary): Properly
550 verify the [VEC_]COND_EXPR embedded comparison.
552 2018-05-15 Martin Sebor <msebor@redhat.com>
554 PR tree-optimization/85753
555 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
556 RECORD_TYPE in addition to ARRAY_TYPE.
558 2018-05-15 Martin Sebor <msebor@redhat.com>
561 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
563 2018-05-15 Richard Biener <rguenther@suse.de>
565 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
566 add by_clobber_p one. Change algorithm to collect all defs
567 representing uses we need to walk and try reducing them to
568 a single one before failing.
569 (dse_dom_walker::dse_optimize_stmt): Adjust.
571 2018-05-13 Mark Wielaard <mark@klomp.org>
573 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
574 (size_of_loc_descr): Likewise.
575 (output_loc_operands): Likewise.
576 (output_loc_operands_raw): Likewise.
577 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
578 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
579 (hash_loc_operands): Likewise.
580 (compare_loc_operands): Likewise.
582 2018-05-14 Mark Wielaard <mark@klomp.org>
584 * dwarf2out.c (count_index_addrs): New function.
585 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
587 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
589 PR tree-optimization/83648
590 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
591 return value as malloc candidate.
593 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
596 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
597 as true in call to suggest_attribute.
599 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
601 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
604 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
606 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
608 (aarch64_crypto_aesd_fused): Likewise.
610 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
612 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
613 (movsi_aarch64): Likewise.
614 (load_pairsi): Likewise.
615 (load_pairdi): Likewise.
616 (store_pairsi): Likewise.
617 (store_pairdi): Likewise.
618 (load_pairsf): Likewise.
619 (load_pairdf): Likewise.
620 (store_pairsf): Likewise.
621 (store_pairdf): Likewise.
622 (zero_extend): Likewise.
623 (trunc): Swap alternatives.
624 (fcvt_target): Add '?' to prefer w over r.
626 2018-05-14 Jakub Jelinek <jakub@redhat.com>
629 * config/i386/i386.md: Disallow non-commutative arithmetics in
630 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
631 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
632 in the peephole2 before it.
634 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
636 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
637 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
638 (ix86_handle_option): Handle -mcldemote.
639 * config.gcc: New header.
640 * config/i386/cldemoteintrin.h: New file.
641 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
642 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
644 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
645 OPTION_MASK_ISA_CLDEMOTE.
646 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
647 (ix86_valid_target_attribute_inner_p): Ditto.
648 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
649 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
650 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
651 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
652 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
654 * config/i386/i386.opt: Add -mcldemote.
655 * config/i386/x86intrin.h: New header.
656 * doc/invoke.texi: Add -mcldemote.
658 2018-05-14 Richard Biener <rguenther@suse.de>
660 * doc/match-and-simplify.texi: Adjust :s documentation.
662 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
664 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
665 intended memcpy size.
666 (REORDER_45): Likewise.
668 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
671 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
672 * vec.c (qsort_chk): Use gcc_qsort.
673 * Makefile.in (OBJS-libcommon): Add sort.o.
674 (build/sort.o): New target. Use it...
675 (BUILD_RTL): ... here, and...
676 (build/gencfn-macros): ... here, and...
677 (build/genmatch): ... here.
679 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
680 Chung-Ju Wu <jasonwucj@gmail.com>
682 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
683 * config/nds32/nds32-graywolf.md: New file.
684 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
685 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
687 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
688 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
689 * config/nds32/nds32.md (pipeline_model): Add graywolf.
690 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
691 * config/nds32/pipelines.md: Include n15 settings.
693 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
694 Chung-Ju Wu <jasonwucj@gmail.com>
696 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
697 * config/nds32/nds32-n13.md: New file.
698 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
699 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
701 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
702 * config/nds32/nds32.md (pipeline_model): Add n13.
703 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
704 * config/nds32/pipelines.md: Include n13 settings.
706 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
707 Chung-Ju Wu <jasonwucj@gmail.com>
709 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
710 * config/nds32/nds32-n10.md: New file.
711 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
712 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
714 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
715 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
716 * config/nds32/nds32.md (pipeline_model): Add n10.
717 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
718 * config/nds32/pipelines.md: Include n10 settings.
720 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
721 Kito Cheng <kito.cheng@gmail.com>
722 Chung-Ju Wu <jasonwucj@gmail.com>
724 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
725 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
726 Add enum values for DSP extension instructions.
727 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
729 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
730 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
732 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
733 * config/nds32/nds32-dspext.md: New file for DSP implementation.
734 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
735 * config/nds32/nds32-intrinsic.md: Likewise.
736 * config/nds32/nds32_intrinsic.h: Likewise.
737 * config/nds32/nds32-md-auxiliary.c: Likewise.
738 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
739 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
740 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
741 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
742 * config/nds32/nds32-protos.h: New declarations for DSP extension.
743 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
744 TYPE_DMAC in switch statement.
745 * config/nds32/nds32.c: New checking and implementation for DSP
746 extension instructions.
747 * config/nds32/nds32.h: Likewise.
748 * config/nds32/nds32.md: Likewise.
749 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
750 * config/nds32/predicates.md: Implement new predicates for DSP
753 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
755 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
756 Reformat alternatives and attributes so it is easier to identify
757 which constraints/attributes go with which instruction.
758 (mov<mode>_hardfloat32, FMOVE64): Likewise.
759 (mov<mode>_softfloat32, FMOVE64): Likewise.
760 (mov<mode>_hardfloat64, FMOVE64): Likewise.
761 (mov<mode>_softfloat64, FMOVE64): Likewise.
763 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
765 * doc/extend.texi (PowerPC Built-in Functions): Rename this
767 (Basic PowerPC Built-in Functions): The new name of the
768 subsection previously known as "PowerPC Built-in Functions".
769 (Basic PowerPC Built-in Functions Available on all Configurations):
771 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
772 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
773 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
774 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
776 2018-05-11 Martin Jambor <mjambor@suse.cz>
779 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
782 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
785 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
787 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
789 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
790 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
791 (ix86_handle_option): Handle -mwaitpkg.
792 * config.gcc: New header.
793 * config/i386/cpuid.h (bit_WAITPKG): New bit.
794 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
795 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
797 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
798 OPTION_MASK_ISA_WAITPKG.
799 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
800 (ix86_option_override_internal): Add PTA_WAITPKG.
801 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
802 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
804 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
805 __builtin_ia32_umwait and __builtin_ia32_tpause.
806 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
807 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
808 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
809 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
810 UNSPECV_TPAUSE): New.
811 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
812 * config/i386/i386.opt: Add -mwaitpkg.
813 * config/i386/waitpkgintrin.h: New file.
814 * config/i386/x86intrin.h: New header.
815 * doc/invoke.texi: Add -mwaitpkg.
817 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
820 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
822 (cortex-m0): Use armv6s-m isa.
823 (cortex-m0plus): Likewise.
824 (cortex-m1): Likewise.
825 (cortex-m0.small-multiply): Likewise.
826 (cortex-m0plus.small-multiply): Likewise.
827 (cortex-m1.small-multiply): Likewise.
829 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
830 Jakub Jelinek <jakub@redhat.com>
832 PR tree-optimization/85692
833 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
834 source permute as well.
836 2018-05-11 Martin Liska <mliska@suse.cz>
839 * doc/extend.texi: Document LLVM style format for no_sanitize
842 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
844 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
845 mode_supports_vsx_dform_quad to mode_supports_dq_form.
846 (mode_supports_vsx_dform_quad): Likewise.
847 (mode_supports_vmx_dform): Move these functions to be next to the
848 other mode_supports functions.
849 (mode_supports_dq_form): Likewise.
850 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
851 mode_supports_dq_form.
852 (reg_offset_addressing_ok_p): Likewise.
853 (offsettable_ok_by_alignment): Likewise.
854 (rs6000_legitimate_offset_address_p): Likewise.
855 (legitimate_lo_sum_address_p): Likewise.
856 (rs6000_legitimize_address): Likewise.
857 (rs6000_legitimize_reload_address): Likewise.
858 (rs6000_secondary_reload_inner): Likewise.
859 (rs6000_preferred_reload_class): Likewise.
860 (rs6000_output_move_128bit): Likewise.
862 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
864 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
865 Generate SImode target register for null target.
866 <case IX86_BUILTIN_XGETBV>: Ditto.
867 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
868 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
870 2018-05-10 Carl Love <cel@us.ibm.com>
872 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
873 dcbtt and dcbtstt if operands[2] is 0.
875 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
878 * config/i386/sse.md (usadv64qi): New expander.
880 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
882 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
883 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
884 -maltivec=be support.
885 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
886 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
887 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
888 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
889 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
890 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
891 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
892 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
893 altivec_vsumsws): Adjust.
894 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
895 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
896 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
898 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
899 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
900 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
901 (altivec_lve<VI_char>x): Delete expand.
902 (*altivec_lve<VI_char>x_internal): Rename to...
903 (altivec_lve<VI_char>x): ... this.
904 (altivec_lvxl_<mode>): Delete expand.
905 (*altivec_lvxl_<mode>_internal): Rename to ...
906 (altivec_lvxl_<mode>): ... this.
907 (altivec_stvxl_<mode>): Delete expand.
908 (*altivec_stvxl_<mode>_internal): Rename to ...
909 (altivec_stvxl_<mode>): ... this.
910 (altivec_stve<VI_char>x): Delete expand.
911 (*altivec_stve<VI_char>x_internal): Rename to ...
912 (altivec_stve<VI_char>x): ... this.
913 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
914 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
915 reduc_plus_scal_<mode>): Adjust.
916 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
918 (rs6000_cpu_cpp_builtins): Adjust.
919 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
920 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
921 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
922 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
923 -maltivec=be support.
924 (rs6000_split_vec_extract_var): Adjust.
925 (rs6000_split_v4si_init): Adjust.
926 (swap_selector_for_mode): Delete.
927 (altivec_expand_lvx_be, altivec_expand_stvx_be,
928 altivec_expand_stvex_be): Delete.
929 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
930 -maltivec=be support.
931 (rs6000_gimple_fold_builtin): Ditto.
932 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
934 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
935 (TARGET_DIRECT_MOVE_64BIT): Adjust.
936 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
937 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
938 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
939 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
940 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
941 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
942 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
943 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
944 anonymous split): Adjust.
945 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
946 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
948 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
950 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
951 when --with-gxx-include-dir is also specified.
952 * configure: Regenerate.
954 2018-05-09 Jim Wilson <jimw@sifive.com>
957 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
958 * config/riscv/t-withmultilib: New.
959 * config/riscv/withmultilib.h: New.
960 * doc/install.texi: Document RISC-V --with-multilib-list support.
962 2018-05-09 Richard Biener <rguenther@suse.de>
964 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
966 (vect_bb_vectorization_profitable_p): Adjust. Compute
967 actual scalar cost using the cost vector and the add_stmt_cost
970 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
972 PR rtl-optimization/85645
973 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
974 in the REG_CFA_REGISTER note for LR, don't leave it empty.
976 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
978 PR rtl-optimization/85645
979 * shrink-wrap.c (spread_components): Return a boolean saying if
980 anything was changed.
981 (try_shrink_wrapping_separate): Iterate spread_components until
982 nothing changes anymore.
984 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
986 PR rtl-optimization/85645
987 * regrename.c (build_def_use): Also kill the chains that include the
988 destination of a REG_CFA_REGISTER note.
990 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
992 PR rtl-optimization/85645
993 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
994 insn that has a REG_CFA_REGISTER note.
996 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
998 * cfgexpand.c (expand_clobber): New function.
999 (expand_gimple_stmt_1): Use it.
1000 * tree-vect-stmts.c (vect_clobber_variable): New function,
1002 (vectorizable_simd_clone_call): ...here.
1003 (vectorizable_store): Emit a clobber either side of an
1004 IFN_STORE_LANES sequence.
1005 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
1007 2018-05-09 Tom de Vries <tom@codesourcery.com>
1010 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
1011 (define_insn "trap_if_false"): Add exit after trap.
1013 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
1015 PR rtl-optimization/85638
1016 * bb-reorder.c: Include common/common-target.h.
1017 (create_forwarder_block): New function extracted from...
1018 (fix_up_crossing_landing_pad): ...here. Rename into...
1019 (dw2_fix_up_crossing_landing_pad): ...this.
1020 (sjlj_fix_up_crossing_landing_pad): New function.
1021 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
1022 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
1023 from both partitions and exit the loop after one iteration.
1025 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1028 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1030 (Basic PowerPC Built-in Functions): The new name of the
1031 subsection previously known as "PowerPC Built-in Functions".
1032 (Basic PowerPC Built-in Functions Available on all Configurations):
1034 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1036 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1038 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1040 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1043 2018-05-08 Jim Wilson <jimw@sifive.com>
1045 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
1046 (LD_EMUL_SUFFIX): New.
1047 (LINK_SPEC): Use it.
1049 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
1051 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1053 (Basic PowerPC Built-in Functions): The new name of the
1054 subsection previously known as "PowerPC Built-in Functions".
1055 (Basic PowerPC Built-in Functions Available on all Configurations):
1057 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
1059 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
1061 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
1063 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
1066 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1069 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
1070 after cmpelim optimization.
1072 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
1074 * config.gcc: Support "goldmont".
1075 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
1076 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1078 * config/i386/i386.c (m_GOLDMONT): Define.
1079 (processor_target_table): Add "goldmont".
1080 (PTA_GOLDMONT): Define.
1081 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
1082 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
1083 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
1084 (fold_builtin_cpu): Add "goldmont".
1085 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
1086 (ix86_option_override_internal): Add "goldmont".
1087 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
1088 (processor_type): Add PROCESSOR_GOLDMONT.
1089 * config/i386/i386.md: Add CPU "glm".
1090 * config/i386/glm.md: New file.
1091 * config/i386/x86-tune.def: Add m_GOLDMONT.
1092 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
1094 2018-05-08 Jakub Jelinek <jakub@redhat.com>
1097 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
1099 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
1100 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
1101 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
1104 * config/i386/i386.c (ix86_fold_builtin): Handle
1105 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
1108 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
1109 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
1111 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
1114 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
1115 (check_arch): Likewise.
1116 (check_fpu): Return the result rather than printing it.
1117 (end arch): Fix operator precedence.
1118 (end cpu): Likewise.
1119 (END): Print the result from check_fpu.
1121 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1122 Alan Hayward <alan.hayward@arm.com>
1123 David Sherwood <david.sherwood@arm.com>
1125 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
1126 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
1127 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
1128 (*fcmuo<mode>_and): New patterns.
1130 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
1132 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
1133 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
1134 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
1135 (cmp_op, sve_imm_con): New code attributes.
1136 (SVE_COND_INT_CMP, imm_con): Delete.
1137 (cmp_op): Remove above unspecs from int attribute.
1138 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
1140 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
1141 comparison-specific unspecs.
1142 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
1143 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
1144 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
1145 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
1146 (*vec_fcm<cmp_op><mode>): Rename to...
1147 (*fcm<cmp_op><mode>): ...this and adjust likewise.
1148 (*vec_fcmuo<mode>): Rename to...
1149 (*fcmuo<mode>): ...this and adjust likewise.
1150 (*pred_fcm<cmp_op><mode>): New pattern.
1151 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
1152 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
1154 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
1156 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
1157 (aarch64_emit_sve_predicated_cond): New function.
1158 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
1159 (aarch64_emit_unspec_cond_or): Replace with...
1160 (aarch64_emit_sve_or_conds): ...this new function. Use
1161 aarch64_emit_sve_ptrue_op for the individual comparisons and
1162 aarch64_emit_binop to OR them together.
1163 (aarch64_emit_inverted_unspec_cond): Replace with...
1164 (aarch64_emit_sve_inverted_cond): ...this new function. Use
1165 aarch64_emit_sve_ptrue_op for the comparison and
1166 aarch64_emit_unop to invert the result.
1167 (aarch64_expand_sve_vec_cmp_float): Update after the above
1168 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
1170 2018-05-07 Nathan Sidwell <nathan@acm.org>
1172 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
1173 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
1174 (Backwards Compatibility): Likewise.
1176 2018-05-07 Luis Machado <luis.machado@linaro.org>
1180 2018-05-07 Luis Machado <luis.machado@linaro.org>
1182 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1183 <prefetch_dynamic_strides>: New const bool field.
1184 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1185 prefetch_dynamic_strides.
1186 (exynosm1_prefetch_tune): Likewise.
1187 (thunderxt88_prefetch_tune): Likewise.
1188 (thunderx_prefetch_tune): Likewise.
1189 (thunderx2t99_prefetch_tune): Likewise.
1190 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1191 (aarch64_override_options_internal): Update to set
1192 PARAM_PREFETCH_DYNAMIC_STRIDES.
1193 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1194 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1195 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1196 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1197 prefetch-dynamic-strides setting.
1199 2018-05-07 Luis Machado <luis.machado@linaro.org>
1201 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1202 <minimum_stride>: New const int field.
1203 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1204 minimum_stride field.
1205 (exynosm1_prefetch_tune): Likewise.
1206 (thunderxt88_prefetch_tune): Likewise.
1207 (thunderx_prefetch_tune): Likewise.
1208 (thunderx2t99_prefetch_tune): Likewise.
1209 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1210 (aarch64_override_options_internal): Update to set
1211 PARAM_PREFETCH_MINIMUM_STRIDE.
1212 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1213 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1214 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1215 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1216 stride is constant and is below the minimum stride threshold.
1218 2018-05-07 Luis Machado <luis.machado@linaro.org>
1220 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
1223 2018-05-07 Luis Machado <luis.machado@linaro.org>
1225 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1226 <prefetch_dynamic_strides>: New const bool field.
1227 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1228 prefetch_dynamic_strides.
1229 (exynosm1_prefetch_tune): Likewise.
1230 (thunderxt88_prefetch_tune): Likewise.
1231 (thunderx_prefetch_tune): Likewise.
1232 (thunderx2t99_prefetch_tune): Likewise.
1233 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
1234 (aarch64_override_options_internal): Update to set
1235 PARAM_PREFETCH_DYNAMIC_STRIDES.
1236 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1237 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1238 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1239 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1240 prefetch-dynamic-strides setting.
1242 2018-05-07 Luis Machado <luis.machado@linaro.org>
1244 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1245 <minimum_stride>: New const int field.
1246 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1247 minimum_stride field.
1248 (exynosm1_prefetch_tune): Likewise.
1249 (thunderxt88_prefetch_tune): Likewise.
1250 (thunderx_prefetch_tune): Likewise.
1251 (thunderx2t99_prefetch_tune): Likewise.
1252 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
1253 (aarch64_override_options_internal): Update to set
1254 PARAM_PREFETCH_MINIMUM_STRIDE.
1255 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1256 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1257 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1258 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1259 stride is constant and is below the minimum stride threshold.
1261 2018-05-06 Jakub Jelinek <jakub@redhat.com>
1264 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
1265 the type is addressable. Don't force op into register if it has
1268 2018-05-05 Roland McGrath <mcgrathr@google.com>
1271 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
1272 any section for which we don't know a specific type it should have,
1273 regardless of name. Previously this was done only for the exact
1274 names ".init_array", ".fini_array", and ".preinit_array".
1275 (default_elf_asm_named_section): Add comment about
1276 relationship with default_section_type_flags and SECTION_NOTYPE.
1277 (get_section): Don't consider it a type conflict if one side has
1278 SECTION_NOTYPE and the other doesn't, as long as neither has the
1279 SECTION_BSS et al used in the default_section_type_flags logic.
1281 2018-05-05 Tom de Vries <tom@codesourcery.com>
1284 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
1285 (workaround_barsyncs): New function.
1286 (nvptx_reorg): Use workaround_barsyncs.
1287 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1288 (define_expand "nvptx_membar_cta"): New define_expand.
1289 (define_insn "*nvptx_membar_cta"): New insn.
1291 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
1293 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
1294 To improve optimization opportunities.
1295 * builtin-types.def: The new needed builtin types for the above.
1297 2018-05-04 Richard Biener <rguenther@suse.de>
1299 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
1300 * gimple-ssa-store-merging.c
1301 (imm_store_chain_info::output_merged_store): Remove redundant create,
1302 release split_store vector contents on failure.
1303 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
1304 scalar stmt vector on cache hit.
1306 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
1308 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
1310 * config.gcc (powerpc-xilinx-eabi*): Remove.
1311 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
1313 (fusion_addis_mem_combo_load): Ditto.
1314 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
1316 (rs6000_cpu_cpp_builtins): Ditto.
1317 * config/rs6000/rs6000-linux.c
1318 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
1319 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
1320 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
1322 (rs6000_setup_reg_addr_masks): Ditto.
1323 (rs6000_init_hard_regno_mode_ok): Ditto.
1324 (rs6000_option_override_internal): Ditto.
1325 (legitimate_lo_sum_address_p): Ditto.
1326 (rs6000_legitimize_address): Ditto.
1327 (rs6000_legitimize_reload_address): Ditto.
1328 (rs6000_legitimate_address_p): Ditto.
1329 (abi_v4_pass_in_fpr): Ditto.
1330 (setup_incoming_varargs): Ditto.
1331 (rs6000_gimplify_va_arg): Ditto.
1332 (rs6000_split_multireg_move): Ditto.
1333 (rs6000_savres_strategy): Ditto.
1334 (rs6000_emit_prologue_components): Ditto.
1335 (rs6000_emit_epilogue_components): Ditto.
1336 (rs6000_emit_prologue): Ditto.
1337 (rs6000_emit_epilogue): Ditto.
1338 (rs6000_elf_file_end): Ditto.
1339 (rs6000_function_value): Ditto.
1340 (rs6000_libcall_value): Ditto.
1341 * config/rs6000/rs6000.h: Ditto.
1342 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
1343 (TARGET_MINMAX): ... this. New.
1344 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
1345 * config/rs6000/rs6000.md: Remove Xilinx FP support.
1346 (*movsi_internal1_single): Delete.
1347 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
1348 mfpu=, mxilinx-fpu): Delete.
1349 * config/rs6000/singlefp.h: Delete.
1350 * config/rs6000/sysv4.h: Remove Xilinx FP support.
1351 * config/rs6000/t-rs6000: Ditto.
1352 * config/rs6000/t-xilinx: Delete.
1353 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
1354 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
1355 (VStype_simple): Delete.
1356 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
1357 * config/rs6000/xfpu.h: Delete.
1358 * config/rs6000/xfpu.md: Delete.
1359 * config/rs6000/xilinx.h: Delete.
1360 * config/rs6000/xilinx.opt: Delete.
1361 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
1362 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
1364 2018-05-04 Tom de Vries <tom@codesourcery.com>
1367 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
1370 2018-05-04 Richard Biener <rguenther@suse.de>
1373 * tree-complex.c (update_complex_assignment): We are always in SSA form.
1374 (expand_complex_div_wide): Likewise.
1375 (expand_complex_operations_1): Likewise.
1376 (expand_complex_libcall): Preserve EH info of the original stmt.
1377 (tree_lower_complex): Handle removed blocks.
1378 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
1379 on complex multiplication and division libcall builtins.
1381 2018-05-04 Richard Biener <rguenther@suse.de>
1384 * fold-const.c (negate_expr_p): Restrict negation of operand
1385 zero of a division to when we know that can happen without
1387 (fold_negate_expr_1): Likewise.
1389 2018-05-04 Jakub Jelinek <jakub@redhat.com>
1392 * real.h (real_nextafter): Declare.
1393 * real.c (real_nextafter): New function.
1394 * fold-const-call.c (fold_const_nextafter): New function.
1395 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
1396 CASE_CFN_NEXTTOWARD.
1397 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
1398 even when arg1_mode is different from arg0_mode.
1400 2018-05-03 Nathan Sidwell <nathan@acm.org>
1402 * doc/extend.texi (Deprecated Features): Remove
1404 (Backwards Compatibility): Likewise.
1405 * doc/invoke.texi (C++ Language Options): Likewise.
1406 (C++ Dialect Options): Likewise.
1408 2018-05-03 Jakub Jelinek <jakub@redhat.com>
1411 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1412 _mm512_mask_mullox_epi64): New intrinsics.
1414 2018-05-03 Tom de Vries <tom@codesourcery.com>
1417 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1418 dump files): Add offload-tree.
1420 2018-05-03 Richard Biener <rguenther@suse.de>
1422 PR tree-optimization/85615
1423 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1424 to loops not nested in BBs loop father to avoid creating multi-entry
1427 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1429 PR tree-optimization/70291
1430 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1431 arguments. Change return type to tree. Emit libcall as a new
1432 statement rather than replacing existing one when inplace_p is true.
1433 (expand_complex_multiplication_components): New function.
1434 (expand_complex_multiplication): Expand floating-point complex
1435 multiplication using the above.
1436 (expand_complex_division): Rename inner_type parameter to type.
1437 Update expand_complex_libcall call-site.
1438 (expand_complex_operations_1): Update expand_complex_multiplication
1439 and expand_complex_division call-sites.
1441 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1444 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1445 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1446 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1447 the highest significant bit of the shift count mask is clear. In
1448 check whether and[sq]i3 is needed verify that all significant bits
1449 of the shift count other than the highest are set.
1451 2018-05-02 Tom de Vries <tom@codesourcery.com>
1454 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1455 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1456 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1457 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1458 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1459 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1460 __builtin_goacc_parlevel_size.
1462 2018-05-02 Richard Biener <rguenther@suse.de>
1464 PR tree-optimization/85597
1465 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1466 do not use split vect_get_vec_defs call but call vect_get_slp_defs
1469 2018-05-02 Tom de Vries <tom@codesourcery.com>
1472 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1473 dump files): Add ltrans-tree.
1475 2018-05-02 Tom de Vries <tom@codesourcery.com>
1478 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1479 dump files): Add wpa-ipa.
1481 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
1483 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
1484 powerpc*-*-linux*paired* target.
1485 * config/rs6000/750cl.h: Delete.
1486 * config/rs6000/paired.h: Delete.
1487 * config/rs6000/paired.md: Delete.
1488 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1490 * config/rs6000/rs6000-builtin.def: Remove paired float support.
1491 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1492 comment. Remove paired float support.
1493 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1494 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1496 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1497 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1499 * config/rs6000/rs6000.c: Remove paired float support.
1500 (paired_expand_vector_init, paired_expand_vector_move,
1501 paired_emit_vector_compare, paired_emit_vector_cond_expr,
1502 (paired_expand_lv_builtin, paired_expand_stv_builtin,
1503 paired_expand_builtin, paired_expand_predicate_builtin,
1504 paired_init_builtins): Delete.
1505 * config/rs6000/rs6000.h: Remove paired float support.
1506 * config/rs6000/rs6000.md: Remove paired float support.
1507 (move_from_CR_ov_bit): Delete.
1508 * config/rs6000/rs6000.opt (mpaired): Delete.
1509 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1510 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1512 2018-05-02 Richard Biener <rguenther@suse.de>
1515 * gimplify.c (gimplify_save_expr): When in SSA form allow
1516 SAVE_EXPRs to compute to SSA vars.
1518 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1521 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1522 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1523 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1524 clobber operands[2], instead use a new pseudo. Formatting fixes.
1526 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1528 PR tree-optimization/85586
1529 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1530 exit early for statements in the same group if the accesses are
1533 2018-05-02 Tom de Vries <tom@codesourcery.com>
1536 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1539 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
1541 PR tree-optimization/85143
1542 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1544 2018-05-01 Tom de Vries <tom@codesourcery.com>
1547 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1548 not found" error message.
1550 2018-05-01 Tom de Vries <tom@codesourcery.com>
1553 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1554 * vec.c (test_ordered_remove_if): New function.
1555 (vec_c_tests): Call test_ordered_remove_if.
1556 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1557 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1558 * tree-vect-patterns.c (vect_pattern_recog_1): Use
1559 VEC_ORDERED_REMOVE_IF.
1561 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1563 PR tree-optimization/82665
1564 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1565 pointer subtraction where arguments come from a memchr call.
1567 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1569 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1570 --push-state --as-needed and --pop-state instead of --as-needed and
1571 --no-as-needed if ld supports it.
1572 * configure: Regenerated.
1575 * doc/install.texi2html: Replace _002d with - and _002a with * in
1576 generated html files using sed.
1578 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1581 * gcc-rich-location.c (blank_line_before_p): New function.
1582 (use_new_line): New function.
1583 (gcc_rich_location::add_fixit_insert_formatted): New function.
1584 * gcc-rich-location.h
1585 (gcc_rich_location::add_fixit_insert_formatted): New function.
1587 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1589 * selftest.c (assert_streq): Rename "expected" and "actual" to
1590 "val1" and "val2". Extend NULL-handling to cover both inputs
1591 symmetrically, while still requiring both to be non-NULL for a pass.
1592 * selftest.h (assert_streq): Rename "expected" and "actual" to
1594 (ASSERT_EQ): Likewise.
1595 (ASSERT_EQ_AT): Likewise.
1596 (ASSERT_KNOWN_EQ): Likewise.
1597 (ASSERT_KNOWN_EQ_AT): Likewise.
1598 (ASSERT_NE): Likewise.
1599 (ASSERT_MAYBE_NE): Likewise.
1600 (ASSERT_MAYBE_NE_AT): Likewise.
1601 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1602 the assertion to pass.
1603 (ASSERT_STREQ_AT): Likewise.
1605 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1607 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1608 interaction with -pie.
1610 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1612 * selftest.h: Fix alphabetization of per-source-file selftest
1615 2018-04-30 Jason Merrill <jason@redhat.com>
1617 PR c++/61982 - dead stores to destroyed objects.
1618 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1621 2018-04-30 Jason Merrill <jason@redhat.com>
1623 * tree.c (build_clobber): New.
1624 * tree.h: Declare it.
1625 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1627 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1629 * diagnostic-show-locus.c (layout::layout): Update for
1630 location_get_source_line returning a char_span.
1631 (struct char_span): Move to input.h.
1632 (struct correction): Update for fields in char_span becoming
1634 (struct source_line): Update for location_get_source_line
1635 returning a char_span.
1636 (layout::print_line): Likewise.
1637 * edit-context.c (edited_file::print_content): Likewise.
1638 (edited_file::print_diff_hunk): Likewise.
1639 (edited_file::print_run_of_changed_lines): Likewise.
1640 (edited_file::get_num_lines): Likewise.
1641 (edited_line::edited_line): Likewise.
1642 * final.c (asm_show_source): Likewise.
1643 * input.c (location_get_source_line): Convert return type
1644 from const char * to char_span, losing the final "line_len"
1646 (dump_location_info): Update for the above.
1647 (get_substring_ranges_for_loc): Likewise. Use a char_span
1648 when handling the literal within the line.
1649 (test_reading_source_line): Update for location_get_source_line
1650 returning a char_span.
1651 * input.h (class char_span): Move here from
1652 diagnostic-show-locus.c, converting from a struct to a class.
1653 Make data members private.
1654 (char_span::operator bool): New.
1655 (char_span::length): New.
1656 (char_span::get_buffer): New.
1657 (char_span::operator[]): New.
1658 (char_span::subspan): Make const.
1659 (char_span::xstrdup): New.
1660 (location_get_source_line): Convert return type from const char *
1661 to char_span, losing the final "line_size" param.
1663 2018-04-30 Jan Hubicka <jh@suse.cz>
1665 * lto-wrapper.c (ltrans_priorities): New static var.
1666 (cmp_priority): New.
1667 (run_gcc): Read priorities and if doing parallel build order
1668 the Makefile by them.
1670 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1672 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
1674 2018-04-30 Richard Biener <rguenther@suse.de>
1676 * tree-cfg.c (verify_address): Remove base argument, add
1677 flag whether to check TREE_ADDRESSABLE and do that.
1678 (verify_expr): Remove.
1679 (verify_types_in_gimple_reference): Add pieces from verify_expr.
1680 (verify_gimple_assign_single): Likewise.
1681 (verify_gimple_switch): Likewise.
1682 (verify_expr_location_1): Dereference tp once. Add (disabled)
1683 piece from verify_expr.
1684 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1686 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1688 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1690 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1692 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1693 (small_data_pattern): Likewise.
1694 (arc_rewrite_small_data): Likewise.
1695 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1696 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1697 (get_symbol_alignment): New function.
1698 (legitimate_small_data_address_p): Likewise.
1699 (legitimate_scaled_address): Update, call
1700 legitimate_small_data_address_p.
1701 (output_sdata): New static variable.
1702 (arc_print_operand): Update how we handle small data operands.
1703 (arc_print_operand_address): Likewise.
1704 (arc_legitimate_address_p): Update, use
1705 legitimate_small_data_address_p.
1706 (arc_rewrite_small_data_p): Remove.
1707 (arc_rewrite_small_data_1): Likewise.
1708 (arc_rewrite_small_data): Likewise.
1709 (small_data_pattern): Likewise.
1710 (compact_sda_memory_operand): Update to use
1711 legitimate_small_data_address_p and get_symbol_alignment.
1712 (prepare_move_operands): Don't rewite sdata pattern.
1713 (prepare_extend_operands): Remove.
1714 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
1716 (zero_extendqisi2): Likewise.
1717 (zero_extendhisi2): Likewise.
1718 (extendqihi2): Likewise.
1719 (extendqisi2): Likewise.
1720 (extendhisi2): Likewise.
1724 * config/arc/constraints.md (Usd): Change it to memory constraint.
1726 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1728 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
1729 as source of std instructions.
1730 * config/arc/arc.md (movsi_insn): Update pattern predicate to
1731 allow 6-bit constants as source for store instructions.
1732 (movdi_insn): Update instruction pattern to allow 6-bit constants
1733 as source for store instructions.
1735 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1737 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
1739 2018-04-30 Nathan Sidwell <nathan@acm.org>
1740 Sandra Loosemore <sandra@codesourcery.com>
1742 * dumpfile.c (dump_open): Allow '-' for stdout.
1743 * doc/invoke.texi (Developer Options): Document dump filename
1744 determination early. Document stdin/stdout selection.
1746 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
1748 Microblaze Target: PIC data text relative
1750 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
1751 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
1753 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
1754 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
1755 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
1756 New addressing mode for data-text relative position indepenedent code.
1757 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
1758 'ADDRESS_SYMBOLIC_TXT_REL'.
1759 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
1760 (microblaze_legitimate_pic_operand): Exclude function calls from
1761 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
1762 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
1764 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1765 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1766 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
1767 for 'address + offset'.
1768 (microblaze_expand_prologue): Add new function prologue call for
1770 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
1771 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
1772 table in case of TARGET_PIC_DATA_TEXT_REL.
1773 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
1774 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
1775 Add new macros 'UNSPEC_TEXT',
1776 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
1777 + exclude function calls from 'UNSPEC_PLT' in case of data text
1779 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1780 new target hook for generating address diff vector tables in case of
1782 * doc/tm.texi : Regenerate.
1783 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
1784 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
1785 of addr diff vector generation.
1786 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1787 target hook definition.
1788 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
1789 Add default function for generate_pic_addr_diff_vec -> flag_pic.
1790 * doc/invoke.texi (Add new pic option): Add new microblaze pic
1791 option for data text relative.
1793 2018-04-30 Richard Biener <rguenther@suse.de>
1795 * tree-chrec.h (evolution_function_is_constant_p): Remove
1797 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
1799 2018-04-30 Richard Biener <rguenther@suse.de>
1802 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
1804 2018-04-30 Richard Biener <rguenther@suse.de>
1806 PR tree-optimization/28364
1807 PR tree-optimization/85275
1808 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
1809 copying first exit test.
1811 2018-04-28 Mark Wielaard <mark@klomp.org>
1813 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
1815 (dwarf_AT): Handle DW_AT_addr_base.
1816 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
1818 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
1821 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
1822 (*ashl<dwi>3_doubleword_mask_1): Ditto.
1823 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
1824 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
1826 2018-04-28 Richard Biener <rguenther@suse.de>
1828 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
1829 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
1830 to reflect use. Only add interesting stmts.
1832 2018-04-27 Martin Jambor <mjambor@suse.cz>
1835 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
1836 the jump function allows for passing through aggregate values.
1838 2018-04-27 David Malcolm <dmalcolm@redhat.com>
1840 * input.h (in_system_header_at): Convert from macro to inline
1842 (from_macro_expansion_at): Likewise.
1843 (from_macro_definition_at): Likewise.
1845 2018-04-27 Jeff Law <law@redhat.com>
1847 * config.gcc: Mark tile* targets as deprecated/obsolete.
1849 2018-04-27 Richard Biener <rguenther@suse.de>
1851 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
1854 2018-04-27 Richard Biener <rguenther@suse.de>
1856 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
1858 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
1860 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
1861 with Yd constraint. Set "preferred_for_speed" attribute from
1862 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
1864 (*movdi_internal): Ditto.
1865 (movti_interunit splitters): Remove
1866 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
1867 (movdi_interunit splitters): Ditto.
1868 * config/i386/constraints.md (Ye): Remove.
1869 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
1871 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1874 * config/aarch64/constraints.md (Usg): Limit to 31.
1877 2018-04-27 Jakub Jelinek <jakub@redhat.com>
1879 PR tree-optimization/85529
1880 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
1881 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
1882 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
1883 zero extension or masking of the MSB bit.
1884 (optimize_range_tests): Add FIRST_BB argument, pass it through
1885 to optimize_range_tests_var_bound.
1886 (maybe_optimize_range_tests, reassociate_bb): Adjust
1887 optimize_range_tests callers.
1889 2018-04-26 Richard Biener <rguenther@suse.de>
1890 Jakub Jelinek <jakub@redhat.com>
1892 * cgraph.h (symbol_table): Just declare debug method here.
1893 * symtab.c (symbol_table::debug): Define.
1895 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
1897 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
1899 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
1901 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
1902 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
1903 (*movdi_internal): Substitute Yi and Yj constraint with x
1904 and Ym and Yn constraint with y constraint. Update "isa"
1905 attribute and set "preferred_for_speed" attribute from
1906 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
1907 (*movsi_internal): Ditto.
1908 (*movdf_internal): Ditto.
1909 (*movsf_internal): Ditto.
1910 (*zero_extendsidi2): Ditto.
1911 * config/i386/sse.md (vec_set<mode>_0): Ditto.
1912 (sse2_loadld): Ditto.
1913 (*vec_extract<ssevecmodelower>_0): Ditto.
1914 (*vec_extractv4si_0_zext_sse4): Ditto.
1915 (vec_concatv2di): Ditto.
1916 (*vec_dup<mode>): Ditto.
1917 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
1918 * config/i386/constraints.md (Yi): Remove.
1923 2018-04-26 Nathan Sidwell <nathan@acm.org>
1925 * dumpfile.c (dump_open): New.
1926 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
1927 (dump_finish): Detect stdio/stderr by value not name.
1929 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
1931 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
1933 2018-04-26 Tom de Vries <tom@codesourcery.com>
1936 * config/nvptx/nvptx.c (verify_neutering_jumps)
1937 (verify_neutering_labels): New function
1938 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
1940 2018-04-26 Tom de Vries <tom@codesourcery.com>
1943 * config/nvptx/nvptx.c (needs_neutering_p): New function.
1944 (nvptx_single): Use needs_neutering_p to skip over insns that do not
1947 2018-04-26 Richard Biener <rguenther@suse.de>
1948 Tom de Vries <tom@codesourcery.com>
1951 * lto-streamer-out.c (output_function): Fixup loops if required to match
1952 discovery done in the reader.
1954 2018-04-26 Richard Biener <rguenther@suse.de>
1956 PR tree-optimization/85116
1957 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
1958 have a loop exit from the single latch predecessor. Remove
1959 case of header with just condition.
1960 (ch_base::copy_headers): Exclude infinite loops from any
1962 (pass_ch::execute): Record exits.
1964 2018-04-26 Richard Biener <rguenther@suse.de>
1966 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
1967 prologue cost vector and pass it to vect_get_load_cost.
1968 (vect_get_peeling_costs_all_drs): Likewise.
1969 (vect_peeling_hash_get_lowest_cost): Likewise.
1970 (vect_enhance_data_refs_alignment): Likewise.
1972 2018-04-26 Richard Biener <rguenther@suse.de>
1975 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
1976 checking of integer<->pointer conversions.
1977 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
1978 sign-/zero-extending pointer types.
1979 (expand_omp_for_static_chunk): Likewise.
1981 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
1982 Jean Lee <xiaoyur347@gmail.com>
1984 * config/mips/mips.c (mips_asan_shadow_offset): New function.
1985 (TARGET_ASAN_SHADOW_OFFSET): Define.
1986 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
1987 true for -fsanitize=address.
1989 2018-04-25 Mark Wielaard <mark@klomp.org>
1991 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
1994 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1996 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
1997 than "alu", remove explicit "memory" and "imm_disp" attributes.
1998 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
2001 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
2002 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
2003 gen_lowpart_no_emit.
2005 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
2008 * config/i386/i386.c (ix86_expand_builtin): Change memory
2009 operand to XI, extend p0 to Pmode.
2010 * config/i386/i386.md: Change unspec volatile and operand
2011 1 mode to XI, change operand 0 mode to P.
2013 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2015 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
2016 GET_MODE_MASK before any checking.
2017 (nds32_can_use_bset_p): Likewise.
2018 (nds32_can_use_btgl_p): Likewise.
2020 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2022 * config/nds32/nds32-doubleword.md: New define_split pattern for
2023 illegal register number.
2025 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2027 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
2029 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
2031 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
2033 2018-04-25 Richard Biener <rguenther@suse.de>
2035 * lto-streamer.h (LTO_major_version): Bump to 8.
2037 2018-04-25 Jakub Jelinek <jakub@redhat.com>
2039 * BASE-VER: Set to 9.0.0.
2041 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
2043 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
2044 in __abskf2 and __powikf2.
2046 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2049 * config/aarch64/constraints.md (Usg, Usj): New constraints.
2050 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
2051 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
2052 Use the above on operand 2. Reindent.
2053 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
2055 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2058 * common/config/i386/i386-common.c (ix86_handle_option): Don't
2060 * config/i386/i386.opt (mcet): Removed.
2061 * doc/install.texi: Remove -mcet documentation.
2062 * doc/invoke.texi: Likewise.
2064 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
2067 * doc/install.texi: Remove -mcet from bootstrap-cet.
2069 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2072 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
2073 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
2077 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
2078 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
2079 containing a CONST_VECTOR.
2081 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
2083 * doc/install.texi: Update newlib dependency for nvptx.
2085 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2088 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
2089 instead of INTVAL when shifting x left.
2091 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
2093 PR tree-optimization/85478
2094 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
2095 vect_grouped_store_supported for single element vectors.
2097 2018-04-24 Richard Biener <rguenther@suse.de>
2100 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
2101 load cost increase to the case of non-constant step.
2103 2018-04-24 Jakub Jelinek <jakub@redhat.com>
2106 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
2107 destination if any_malformed_asm.
2109 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
2112 * expr.c (store_field): In the bitfield case, if the value comes from
2113 a function call and is returned in registers by means of a PARALLEL,
2114 do not change the mode of the temporary unless BLKmode and VOIDmode.
2116 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
2118 PR rtl-optimization/85423
2119 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
2120 dependencies to debug insns when the previous insn is non-debug.
2122 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2124 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
2125 enums into a single definition.
2126 (fls): Fix predicates and printing.
2129 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
2131 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
2132 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
2133 and short u6 immediate.
2134 (check_if_valid_sleep_operand): Remove.
2135 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
2137 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2139 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
2140 flag_always_save_lp condition.
2141 * config/nds32/nds32.opt (malways-save-lp): New option.
2143 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2145 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
2146 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
2147 * config/nds32/nds32.h
2148 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
2149 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
2151 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2153 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
2154 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
2156 2018-04-22 Shiva Chen <shiva0217@gmail.com>
2157 Chung-Ju Wu <jasonwucj@gmail.com>
2159 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
2161 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
2162 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
2164 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2166 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
2168 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2170 * config/nds32/nds32-protos.h (nds32_data_alignment,
2171 nds32_local_alignment): Declare.
2172 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
2173 nds32_local_alignment): New functions.
2174 (TARGET_CONSTANT_ALIGNMENT): Define.
2175 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
2177 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2179 * config/nds32/nds32.c
2180 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
2181 (TARGET_MODES_TIEABLE_P): Likewise.
2183 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
2185 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
2188 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
2189 Chung-Ju Wu <jasonwucj@gmail.com>
2191 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2192 for unaligned access.
2193 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2195 * config/nds32/nds32-intrinsic.md: Likewise.
2196 * config/nds32/nds32_intrinsic.h: Likewise.
2197 * config/nds32/nds32.h (nds32_builtins): Likewise.
2198 * config/nds32/nds32.opt (munaligned-access): New option.
2199 * config/nds32/nds32.c (nds32_asm_file_start): Display
2200 flag_unaligned_access status.
2202 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2204 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
2205 -mno-relax is present.
2206 * config/riscv/linux.h (LINK_SPEC): Ditto.
2208 2018-04-20 Martin Sebor <msebor@redhat.com>
2211 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
2213 (gimple_fold_builtin_stxcpy_chk): Same.
2214 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
2216 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
2219 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
2220 __powikf2 when long double is IEEE 128-bit.
2222 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
2224 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
2225 step to make sure stack always aligned.
2227 2018-04-20 Carl Love <cel@us.ibm.com>
2230 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
2231 size check for arg0.
2233 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
2234 Tom de Vries <tom@codesourcery.com>
2237 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
2238 Emit insns for calls too.
2239 (nvptx_find_par): Always look for worker-level predecessor insn.
2240 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
2242 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
2243 (nvptx_process_pars): Propagate frames for calls.
2245 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
2248 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
2250 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
2251 (ix86_handle_option): Don't handle OPT_mibt.
2252 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
2254 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
2256 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
2257 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
2258 (ix86_target_macros): Define __CET__ with flag_cf_protection
2259 for -fcf-protection.
2260 * config/i386/i386.c (isa2_opts): Remove -mibt.
2261 * config/i386/i386.h (TARGET_IBT): Removed.
2262 (TARGET_IBT_P): Likewise.
2263 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
2264 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
2265 * config/i386/i386.opt (mcet): Update help message.
2268 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
2269 -mcet as an alias for -mshstk.
2271 2018-04-20 Richard Biener <rguenther@suse.de>
2274 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
2275 complexity by forcing a single use of the multiply operand.
2277 2018-04-20 Martin Jambor <mjambor@suse.cz>
2280 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
2281 recursion dependency to only apply to non-clones.
2283 2018-04-20 Martin Jambor <mjambor@suse.cz>
2286 * ipa-cp.c (create_specialized_node): Check that clones of
2287 self-recursive edges exist during IPA-CP.
2289 2018-04-19 Toon Moene <toon@moene.org>
2291 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
2294 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2296 PR tree-optimization/85467
2297 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
2298 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
2299 VECTOR_CST element to type.
2301 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2304 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
2305 * config/i386/i386.md (builtin_setjmp_setup): Removed.
2306 (builtin_longjmp): Likewise.
2307 (save_stack_nonlocal): New pattern.
2308 (restore_stack_nonlocal): Likewise.
2310 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2313 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2314 Replace ASM_OUTPUT_LABEL with fprintf.
2316 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2319 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2320 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
2321 * config/i386/i386-c.c (ix86_target_macros_internal): Also
2322 define __IBT__ and __SHSTK__ for -fcf-protection.
2323 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
2325 (ix86_trampoline_init): Likewise.
2326 (x86_output_mi_thunk): Likewise.
2327 (ix86_notrack_prefixed_insn_p): Likewise.
2328 (ix86_option_override_internal): Don't disallow -fcf-protection.
2329 * config/i386/i386.md (rdssp<mode>): Also enable for
2331 (incssp<mode>): Likewise.
2332 (nop_endbr): Likewise.
2333 * config/i386/i386.opt (mcet): Change help message to built-in
2337 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
2338 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
2339 enable CET built-in functions.
2341 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
2343 * common/config/i386/i386-common.c
2344 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
2345 OPTION_MASK_ISA_MOVDIRI_UNSET,
2346 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
2347 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
2348 * config.gcc (movdirintrin.h): New header.
2349 * config/i386/cpuid.h (bit_MOVDIRI,
2350 bit_MOVDIR64B): New bits.
2351 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
2353 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
2354 (VOID, PVOID, PCVOID)): New function types.
2355 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
2356 __builtin_ia32_directstoreu_u64,
2357 __builtin_ia32_movdir64b): New builtins.
2358 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
2359 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
2361 (ix86_valid_target_attribute_inner_p): Ditto.
2362 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
2363 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
2364 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
2365 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
2366 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
2367 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
2368 (movdiri<mode>, movdir64b_<mode>): New.
2369 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
2370 * config/i386/immintrin.h: Include movdirintrin.h.
2371 * config/i386/movdirintrin.h: New file.
2372 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
2374 2018-04-19 Richard Biener <rguenther@suse.de>
2377 * cfg.c (clear_bb_flags): When loop state says we have
2378 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
2380 2018-04-19 Richard Biener <rguenther@suse.de>
2382 PR tree-optimization/84737
2383 * tree-vect-data-refs.c (vect_copy_ref_info): New function
2384 copying restrict info.
2385 (vect_setup_realignment): Use it.
2386 * tree-vectorizer.h (vect_copy_ref_info): Declare.
2387 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
2388 the first DR to all generated stores.
2389 (vectorizable_load): Likewise for loads.
2391 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2393 PR tree-optimization/85446
2394 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
2395 the integral and pointer types to have the same precision.
2397 * doc/install.texi: Document --disable-cet being the default and
2400 2018-04-18 Martin Liska <mliska@suse.cz>
2402 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
2405 2018-04-18 Martin Liska <mliska@suse.cz>
2408 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2411 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2412 arguments if they are comparable.
2414 2018-04-18 Martin Liska <mliska@suse.cz>
2417 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2420 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2423 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
2426 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2427 ENDBR after calling __morestack.
2429 2018-04-18 David Malcolm <dmalcolm@redhat.com>
2432 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2433 by using gcc_base_ver to generate a gcc_driver_version, and use
2434 it when generating GCC_DRIVER_NAME.
2435 * configure: Regenerate.
2437 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2440 * config.gcc: Obsolete powerpc*-*-*spe*.
2442 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2445 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2446 (stabstr_D): Change type of unum from unsigned int to
2447 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
2450 2018-04-17 Jim Wilson <jimw@sifive.com>
2453 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2454 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2455 Set arg_pointer_offset after using pretend_args_size.
2457 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2459 PR rtl-optimization/85431
2460 * dse.c (record_store): Ignore zero width stores.
2463 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
2464 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2465 __builtin_stack_restore rather than after it.
2466 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2467 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2468 argument instead of virtual_dynamic_stack_rtx.
2470 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
2472 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2474 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2475 Add note to error message to explain internal mapping of overloaded
2476 built-in function name to non-overloaded built-in function name.
2477 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2480 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2483 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2484 where the inputs overlap with the output.
2486 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2488 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2489 (=v, v) alternative and explicit "memory" attribute.
2490 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
2491 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2493 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2494 "sselog1" type instead of "sselog".
2495 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2496 "sselog". Remove explicit "memory" attribute.
2497 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2498 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2500 (vec_extract_hi_v32hi): Merge all alternatives into one, use
2501 "sselog1" type instead of "sselog". Remove explicit "memory"
2503 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2504 use "sselog1" type instead of "sselog". Remove explicit "memory"
2506 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2507 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2509 (vec_extract_hi_v64qi): Merge all alternatives into one, use
2510 "sselog1" type instead of "sselog". Remove explicit "memory"
2512 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2513 use "sselog1" type instead of "sselog". Remove explicit "memory"
2517 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2520 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2523 2018-04-17 Martin Jambor <mjambor@suse.cz>
2526 * ipa-cp.c (create_specialized_node): Call
2527 expand_all_artificial_thunks if necessary.
2529 2018-04-17 Martin Liska <mliska@suse.cz>
2532 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2533 in message, remote space in between '_G' and '('.
2535 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2538 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2539 avx512f_vmcmp<mode>3<round_saeonly_name>,
2540 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2541 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2542 avx512f_rndscale<mode><round_saeonly_name>,
2543 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2544 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2545 Use %<iptr>2 instead of %2 for -masm=intel.
2546 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2547 avx512f_vcvttss2usi<round_saeonly_name>,
2548 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2550 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2551 avx512f_vcvttsd2usi<round_saeonly_name>,
2552 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2553 Use %q1 instead of %1 for -masm=intel.
2554 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2555 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2556 of %3 for -masm=intel.
2557 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2559 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2561 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2563 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2564 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2566 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2568 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2569 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2570 %g1 and one with %0 and %1.
2571 (avx512er_vmrcp28<mode><round_saeonly_name>,
2572 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2574 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2575 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2576 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2577 of %0 and %{%4%} for -masm=intel.
2578 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2579 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2580 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2581 order of %0 and %{%5%}%{z%} for -masm=intel.
2583 2018-04-17 Jan Hubicka <jh@suse.cz>
2586 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2588 2018-04-17 Martin Liska <mliska@suse.cz>
2591 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2592 for target_clone error message. Make default implementation
2593 clone to be a local declaration.
2594 (separate_attrs): Add new argument and check for an empty
2596 (expand_target_clones): Handle it.
2597 (ipa_target_clone): Make redirection just for target_clones
2600 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2601 Tom de Vries <tom@codesourcery.com>
2604 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2605 tiled basic blocks without omp continue statements.
2607 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2610 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2611 vec_extract expression as having side effects to make sure it gets
2614 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2617 * config/i386/i386.c (get_builtin_code_for_version): Check
2620 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2623 * gcc/config.gcc: Support "skylake".
2624 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2626 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2627 (processor_target_table): Add "skylake".
2628 (ix86_option_override_internal): Add "skylake".
2629 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2630 PROCESSOR_CANNONLAKE.
2631 (get_builtin_code_for_version): Fix priority for
2632 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2633 PROCESSOR_SKYLAKE-AVX512.
2634 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2635 (processor_type): Add PROCESSOR_SKYLAKE.
2637 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
2638 Jason Merrill <jason@redhat.com>
2641 * convert.c (convert_to_integer_1): Use direct recursion for
2642 enumeral types and types with a precision less than the number
2643 of bits in their mode.
2645 2018-04-16 Julia Koval <julia.koval@intel.com>
2648 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2649 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2651 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
2654 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2655 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2656 and -mno-direct-move.
2658 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
2661 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
2662 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
2663 Ensure negative shifts result in {0}.
2665 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
2667 PR rtl-optimization/79916
2668 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
2669 regs (if any) to define how to gnerate SD moves when LRA is in
2672 2018-04-13 Jakub Jelinek <jakub@redhat.com>
2674 PR rtl-optimization/85393
2675 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
2676 * except.c (expand_dw2_landing_pad_for_region): Make static.
2677 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2678 a label and unconditional jump to old_bb, rather than
2679 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2682 PR rtl-optimization/85376
2683 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2684 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2685 instead of a specific value.
2687 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2688 Bin Cheng <bin.cheng@arm.com>
2690 PR tree-optimization/82965
2691 PR tree-optimization/83991
2692 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2693 by_profile_only parameter.
2694 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2695 information if the loop was predicted to iterate too many times.
2696 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2698 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2701 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
2704 2018-04-13 Martin Liska <mliska@suse.cz>
2705 Jakub Jelinek <jakub@redhat.com>
2708 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
2709 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
2710 * builtins.c (expand_builtin_memory_copy_args): Use
2711 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
2712 handle dest_addr == pc_rtx.
2714 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
2717 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
2718 asked to not generate direct moves.
2719 (fix_trunc<mode>si2_stfiwx): Similar.
2720 (fix_trunc<mode>si2_internal): Similar.
2722 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2725 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
2726 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
2727 lookup if dest in some wider mode is known to be const0_rtx and
2728 if so, record permanent equivalence for it to be ZERO_EXTEND of
2729 the narrower mode destination.
2731 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2733 * lto-streamer-out.c (output_function): Revert 259346.
2734 * omp-expand.c (expand_oacc_for): Likewise.
2736 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
2738 PR rtl-optimization/85354
2739 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
2740 * sel-sched.c (sel_global_init): ... here.
2742 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
2745 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
2746 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2747 mode for PE-COFF targets.
2748 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
2749 (i386_pe_asm_lto_end): Likewise.
2750 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
2751 (TARGET_ASM_LTO_END): Likewise.
2752 * config/i386/winnt.c (saved_debug_info_level): New static variable.
2753 (i386_pe_asm_lto_start): New function.
2754 (i386_pe_asm_lto_end): Likewise.
2756 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2757 Richard Biener <rguenther@suse.de>
2760 * lto-streamer-out.c (output_function): Fix CFG loop state before
2762 * omp-expand.c (expand_oacc_for): Handle calls to internal
2763 functions like regular functions.
2765 2018-04-12 Richard Biener <rguenther@suse.de>
2768 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
2769 for the early LTO debug to properly generate references to it
2770 during DIE emission. Do not re-use that for the skeleton for
2772 (dwarf2out_early_finish): Likewise.
2774 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2777 * config/i386/sse.md
2778 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
2779 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
2780 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
2781 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
2782 and output is a reg, avoid creating invalid lowpart subreg, but
2783 instead split into a 512-bit move. Don't split if not AVX512VL,
2784 input is xmm16+ reg and output is a mem.
2785 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
2786 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
2787 xmm16+ reg and output is a mem.
2789 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2791 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
2792 also for flag_dwarf2_cfi_asm.
2794 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2796 PR rtl-optimization/85342
2797 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
2798 a bool scalar var inside of the loop instead. Don't try to update
2799 recog_data.operand after failed apply_change_group.
2801 2018-04-12 Tom de Vries <tom@codesourcery.com>
2804 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
2805 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
2806 array with flexible array member as array without given dimension.
2807 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
2808 argument for undefined param to true.
2810 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
2813 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2814 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
2815 from PowerPC section.
2816 * config/rs6000/sysv4.opt (mcall-): Improve help text.
2817 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
2818 help text that is too long.
2819 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
2820 help text that is too long.
2821 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
2822 help text that is too long.
2824 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
2826 * config/alpha/alpha.md (stack_probe_internal): Rename
2827 from "probe_stack". Update all callers.
2829 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2831 PR rtl-optimization/84566
2832 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
2833 sched_macro_fuse_insns.
2835 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2838 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
2839 (compute_block_dependences): ... from here.
2841 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2843 PR tree-optimization/85331
2844 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
2845 from int to HOST_WIDE_INT.
2847 2018-04-11 Martin Jambor <mjambor@suse.cz>
2850 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
2851 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
2852 not the same as the source val.
2853 (cgraph_edge_brings_value_p): New parameter.
2854 (gather_edges_for_value): Pass destination value to
2855 cgraph_edge_brings_value_p.
2856 (perhaps_add_new_callers): Likewise.
2857 (get_info_about_necessary_edges): Likewise and exclude values brought
2858 only by self-recursive edges.
2859 (create_specialized_node): Redirect only clones of self-calling edges.
2860 (+self_recursive_pass_through_p): New function.
2861 (find_more_scalar_values_for_callers_subset): Use it.
2862 (find_aggregate_values_for_callers_subset): Likewise.
2863 (known_aggs_to_agg_replacement_list): Removed.
2864 (decide_whether_version_node): Re-calculate known constants for all
2865 remaining context clones.
2867 2018-04-11 Richard Biener <rguenther@suse.de>
2870 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
2871 from early DWARF output.
2872 (dwarf2out_early_finish): Output line info unconditionally into
2873 early DWARF and add reference to it.
2875 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2878 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
2879 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
2880 other than V2DFmode using iptr mode attribute.
2881 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
2883 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2885 PR rtl-optimization/84659
2886 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
2888 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2891 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
2893 (output_loc_list): Pass address of a dummy size variable even in the
2894 locview handling loop.
2895 (index_location_lists): Add comment on why skip_loc_list_entry can't
2898 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
2901 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
2904 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
2907 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2908 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
2909 and -mstring-compare-inline-limit.
2911 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2914 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
2915 for stack clash protection in a register whenever we need it to be in
2918 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2920 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
2921 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
2923 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2926 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
2928 (mlong-double-): Ditto.
2929 * config/rs6000/sysv4.opt (msdata=): Ditto.
2930 (mtls-size=): Ditto.
2932 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
2934 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2935 erroneous entries for
2936 "vector int vec_ldl (int, long int *)", and
2937 "vector unsigned int vec_ldl (int, unsigned long int *)".
2938 Add comments and entries for
2939 "vector bool char vec_ldl (int, bool char *)",
2940 "vector bool short vec_ldl (int, bool short *)",
2941 "vector bool int vec_ldl (int, bool int *)",
2942 "vector bool long long vec_ldl (int, bool long long *)",
2943 "vector pixel vec_ldl (int, pixel *)",
2944 "vector long long vec_ldl (int, long long *)",
2945 "vector unsigned long long vec_ldl (int, unsigned long long *)".
2946 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
2947 type tree bool_long_long_type_node and correct definition of
2948 bool_V2DI_type_node to make reference to this new type tree.
2949 (rs6000_mangle_type): Replace erroneous reference to
2950 bool_long_type_node with bool_long_long_type_node.
2951 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
2952 comments to emphasize sign distinctions for char and int types and
2953 replace RS6000_BTI_bool_long constant with
2954 RS6000_BTI_bool_long_long constant. Also add comment to restrict
2955 use of RS6000_BTI_pixel.
2956 (bool_long_type_node): Remove this macro definition.
2957 (bool_long_long_type_node): New macro definition
2959 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2961 PR rtl-optimization/85300
2962 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
2963 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
2964 simplify_unary_operation fails.
2966 2018-04-10 Martin Liska <mliska@suse.cz>
2968 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
2969 cgraph_edge and ipa_ref.
2971 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2975 * config/i386/sse.md
2976 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
2977 computation of the VEC_MERGE selector from mask.
2978 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
2979 Fix decoding of the VEC_MERGE selector into mask.
2981 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2983 PR tree-optimization/85286
2984 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
2986 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2988 * final.c (final_1): Set insn_last_address as well as
2989 insn_current_address.
2991 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2994 * explow.c (emit_stack_probe): Call validize_mem on memory location
2995 before passing it to gen_probe_stack. Create address operand and
2996 legitimize it for the probe_stack_address case.
2998 2018-04-09 Jan Hubicka <jh@suse.cz>
3001 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
3002 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
3003 * tree.c (free_lang_data_in_type): Fix handling of binfos;
3005 (free_lang_data): Rebuild type inheritance graph.
3007 2018-04-09 Martin Sebor <msebor@redhat.com>
3009 * invoke.texi (-finline-small-functions): Mention other optimization
3011 (-findirect-inlining, -fpartial-inlining): Same.
3012 (-finline-functions-called-once): Same.
3013 (-freorder-blocks-and-partition): Same.
3015 2018-04-09 Jan Hubicka <jh@suse.cz>
3018 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
3019 jumps; choose last target that matches the criteria (i.e.
3020 no partition changes for non-crossing jumps).
3021 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
3022 support for redirecting crossing jumps to non-crossing.
3024 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
3026 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
3027 also for naked functions.
3029 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
3031 * config/arc/arc.md (add_shift): New pattern.
3032 (add_shift2): Likewise.
3033 (sub_shift): Likewise.
3034 (sub_shift_cmp0_noout): Likewise.
3035 (compare_si_ashiftsi): Likewise.
3036 (xbfu_cmp0_noout): New combine pattern.
3037 (xbfu_cmp0"): Likewise.
3038 (movsi_set_cc_insn): Place the predicable variant first.
3039 (commutative_binary_cmp0_noout): Remove clobber.
3040 (commutative_binary_cmp0): New pattern.
3041 (noncommutative_binary_cmp0): Likewise.
3042 (noncommutative_binary_cmp0_noout): Likewise.
3043 (noncommutative_binary_comparison_result_used): Removed.
3044 (rsub_cmp0): New pattern.
3045 (rsub_cmp0_noout): Likewise.
3046 (extzvsi): Changed, keep only meaningful variants.
3047 (SQH, SEZ): New iterators.
3048 (SQH_postfix): New mode attribute.
3049 (SEZ_prefix): New code attribute.
3050 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
3051 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
3052 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
3054 (noncommutative_operator): Check the availability of barrel
3057 2018-04-09 Richard Biener <rguenther@suse.de>
3059 PR tree-optimization/85284
3060 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
3061 Only use the niter constraining form of simple_iv when the exit
3064 2018-04-09 Tom de Vries <tom@codesourcery.com>
3067 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3068 (define_expand "*memory_barrier"): New define_expand.
3069 (define_insn "memory_barrier"): New insn.
3071 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3073 PR rtl-optimization/80463
3074 PR rtl-optimization/83972
3075 PR rtl-optimization/83480
3077 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
3078 correct producer for the insn.
3079 (tidy_control_flow): Fixup seqnos in case of debug insns.
3081 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3083 PR rtl-optimization/83913
3085 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
3086 different sched-times when merging exprs.
3088 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3090 PR rtl-optimization/83962
3092 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
3093 tidy_fallthru_edge and tidy_control_flow.
3095 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
3097 PR rtl-optimization/83530
3099 * sel-sched.c (force_next_insn): New global variable.
3100 (remove_insn_for_debug): When force_next_insn is true, also leave only
3101 next insn in the ready list.
3102 (sel_sched_region): When the region wasn't scheduled, make another pass
3103 over it with force_next_insn set to 1.
3105 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3107 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
3109 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3110 for interrupt control.
3111 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3112 functions for interrupt control.
3113 * config/nds32/nds32-intrinsic.md: Likewise.
3114 * config/nds32/nds32_intrinsic.h: Likewise.
3115 * config/nds32/nds32.h (nds32_builtins): Likewise.
3117 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3119 * config/nds32/nds32.c (nds32_init_machine_status,
3120 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
3121 strict_aligned_p field.
3122 (nds32_expand_to_rtl_hook): New function.
3123 (TARGET_EXPAND_TO_RTL_HOOK): Define.
3124 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
3126 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3127 Chung-Ju Wu <jasonwucj@gmail.com>
3129 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
3130 * config/nds32/nds32-n7.md: New file.
3131 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
3132 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
3134 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
3135 * config/nds32/nds32.md (pipeline_model): Add n7.
3136 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
3137 * config/nds32/pipelines.md: Include n7 settings.
3139 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3140 Chung-Ju Wu <jasonwucj@gmail.com>
3142 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
3143 * config/nds32/nds32-e8.md: New file.
3144 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
3145 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
3147 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
3148 * config/nds32/nds32.md (pipeline_model): Add e8.
3149 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
3150 * config/nds32/pipelines.md: Include e8 settings.
3152 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3153 Chung-Ju Wu <jasonwucj@gmail.com>
3155 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
3156 * config/nds32/nds32-n8.md: New file.
3157 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
3158 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
3160 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
3161 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
3162 * config/nds32/nds32.md (pipeline_model): Add n8.
3163 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
3164 * config/nds32/pipelines.md: Include n8 settings.
3166 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
3167 Chung-Ju Wu <jasonwucj@gmail.com>
3169 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
3170 * config/nds32/nds32-n9-2r1w.md: New file.
3171 * config/nds32/nds32-n9-3r2w.md: New file.
3172 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
3173 nds32_register_ports): New or modify for cpu n9.
3174 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
3176 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
3177 * config/nds32/nds32-utils.c: New file.
3178 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
3179 TARGET_MUL_SLOW): Define.
3180 * config/nds32/nds32.md (pipeline_model): New attribute.
3181 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
3182 New options that support cpu n9.
3183 * config/nds32/pipelines.md: Include n9 settings.
3184 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
3186 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
3188 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
3189 information if necessary.
3190 (output_cond_branch_compare_zero): Likewise.
3191 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
3192 (nds32_target_alignment): Refine for alignment.
3193 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
3194 (FUNCTION_BOUNDARY): Modify.
3195 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
3197 * config/nds32/nds32.opt (malways-align, malign-functions): New.
3199 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
3201 * config/nds32/constants.md (unspec_volatile_element): Add values for
3202 TLB operation and data prefetch.
3203 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
3204 functions for TLB operation and data prefetch.
3205 * config/nds32/nds32-intrinsic.md: Likewise.
3206 * config/nds32/nds32_intrinsic.h: Likewise.
3207 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
3208 (nds32_print_operand): Likewise.
3209 * config/nds32/nds32.h (nds32_builtins): Likewise.
3211 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
3212 Andrew Pinski <pinsika@gcc.gnu.org>
3215 * match.pd: Use constant_boolean_node of correct type instead of
3216 boolean_true_node or boolean_false_node for simplifying
3217 pointer comparisons to zero.
3219 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3221 PR tree-optimization/80021
3222 * tree.c (verify_type_variant): Make error call in verify_variant_match
3223 translatable and remove final full stop.
3225 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3227 * config/nds32/constants.md (unspec_volatile_element): Add
3228 UNSPEC_VOLATILE_EH_RETURN.
3229 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
3230 nds32_output_stack_pop): Support dwarf exception handling process.
3231 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
3232 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
3233 exception handling process.
3234 (nds32_compute_stack_frame): Likewise.
3235 (nds32_return_addr_rtx): Likewise.
3236 (nds32_initial_elimination_offset): Likewise.
3237 (nds32_expand_prologue): Likewise.
3238 (nds32_expand_epilogue): Likewise.
3239 (nds32_dynamic_chain_address): New function.
3240 * config/nds32/nds32.h (machine_function): Add fields for dwarf
3242 (DYNAMIC_CHAIN_ADDRESS): Define.
3243 (EH_RETURN_DATA_REGNO): Define.
3244 (EH_RETURN_STACKADJ_RTX): Define.
3245 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
3246 patterns for dwarf exception handling.
3248 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3250 * config/nds32/nds32.h: Clean up obsolete macros.
3252 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3254 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3255 Add enum values for particular instructions.
3256 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3257 particular intrinsic functions.
3258 * config/nds32/nds32-intrinsic.md: Likewise.
3259 * config/nds32/nds32_intrinsic.h: Likewise.
3260 * config/nds32/nds32.h (nds32_builtins): Likewise.
3261 * config/nds32/nds32.md (type): Add pbsad and pbsada.
3262 (btst, ave): New patterns for particular instructions.
3264 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3266 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
3267 Add enum values for atomic load/store and memory sync.
3268 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
3270 * config/nds32/nds32-intrinsic.md: Likewise.
3271 * config/nds32/nds32_intrinsic.h: Likewise.
3272 * config/nds32/nds32.h (nds32_builtins): Likewise.
3274 2018-04-07 Jakub Jelinek <jakub@redhat.com>
3276 PR tree-optimization/85257
3277 * fold-const.c (native_encode_vector): If not all elts could fit
3278 and off is -1, return 0 rather than offset.
3279 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
3280 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
3281 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
3282 adjust buffer in native_interpret_expr call.
3284 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3286 * config/nds32/constants.md (unspec_volatile_element): Add cache
3287 control enum values.
3288 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
3289 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
3290 * config/nds32/nds32.c (nds32_cctl_names): New.
3291 (nds32_print_operand): Handle cache control register names.
3292 * config/nds32/nds32.h (nds32_builtins): New enum values.
3293 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
3295 * config/nds32/nds32.md (type): Add mmu.
3296 * config/nds32/pipelines.md (simple_insn): Add mmu.
3298 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3300 * config/nds32/nds32.md (type): Remove call.
3301 * config/nds32/pipelines.md (simple_insn): Likewise.
3303 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3305 * config/nds32/constants.md (unspec_volatile_element): Add
3306 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
3307 UNSPEC_VOLATILE_FMFCFG.
3308 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
3309 description for fmfcfg and fmfcsr.
3310 (bdesc_1arg): Add fmtcsr.
3311 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
3312 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
3313 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
3314 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
3315 unspec_fmfcfg): New patterns.
3316 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
3317 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
3318 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
3319 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
3320 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
3321 __nds32__fmfcfg): Define.
3323 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3325 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
3326 intrinsic register names.
3327 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
3328 intrinsic register enum values and macros.
3330 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3332 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
3333 for load/store addressing form.
3334 (nds32_print_operand_address): Likewise.
3336 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
3339 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
3340 based on LABEL_REF. Remove useless assertion.
3341 (pic_address_needs_scratch): Fix formatting.
3342 (sparc_legitimize_pic_address): Minor tweaks.
3343 (sparc_delegitimize_address): Adjust assertion accordingly.
3344 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
3345 into symbolic_operand.
3346 (movsi_high_pic_label_ref): Likewise.
3347 (movsi_lo_sum_pic_label_ref): Likewise.
3348 (movdi_pic_label_ref): Likewise.
3349 (movdi_high_pic_label_ref): Likewise.
3350 (movdi_lo_sum_pic_label_ref): Likewise.
3352 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
3354 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
3355 custom LIB_SPEC setup.
3357 2018-04-06 Ruslan Bukin <br@bsdpad.com>
3358 Kito Cheng <kito.cheng@gmail.com>
3360 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
3361 * config/riscv/freebsd.h: New.
3363 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3365 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
3366 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
3369 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3370 Kito Cheng <kito.cheng@gmail.com>
3372 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
3373 nds32_output_call, nds32_symbol_binds_local_p): New functions.
3374 * config/nds32/nds32-protos.h (nds32_output_call,
3375 nds32_output_return): Declare.
3376 * config/nds32/nds32.md: Refine all the call and return patterns.
3378 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3381 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
3382 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
3384 PR rtl-optimization/84872
3385 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
3386 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
3389 2018-04-06 Tamar Christina <tamar.christina@arm.com>
3391 * expr.c (copy_blkmode_to_reg): Revert 254862.
3392 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
3394 2018-04-06 Richard Biener <rguenther@suse.de>
3397 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
3398 after seeing a component reference with an adjacent field. Treat
3399 refs to arrays at struct end of external decls similar to
3400 refs to unconstrained commons.
3402 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3405 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
3406 look through SAVE_EXPRs with non-side-effects argument. Adjust
3408 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3411 2018-04-06 Richard Biener <rguenther@suse.de>
3414 * alias.c (find_base_term): New wrapper around find_base_term
3415 unwinding CSELIB_VAL_PTR changes.
3416 (find_base_term): Do not restore CSELIB_VAL_PTR during the
3419 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3421 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3423 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3424 constant definitions.
3425 ("nop"): lr 0,0 -> nopr r0
3426 ("nop_lr0", "nop_lr1"): New insn definitions.
3428 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3430 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3431 NDS32_V3PUSH_AVAILABLE_P macro.
3433 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
3434 Chung-Ju Wu <jasonwucj@gmail.com>
3436 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3437 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3438 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3439 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3440 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3441 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3442 * config/nds32/constraints.md: New constraints and checking for hard
3443 float configuration.
3444 * config/nds32/iterators.md: New mode iterator and attribute for hard
3445 float configuration.
3446 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3448 * config/nds32/nds32-fpu.md: New file.
3449 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3450 deal with hard float code generation.
3451 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3453 (abi_type, float_reg_number): New enum type.
3454 * config/nds32/nds32-predicates.c: New predicates for hard float.
3455 * config/nds32/nds32-protos.h: Declare functions for hard float.
3456 * config/nds32/nds32.c: Implementation for hard float configuration.
3457 * config/nds32/nds32.h: Definitions for hard float configuration.
3458 * config/nds32/nds32.md: Include hard float machine description and
3459 modify patterns for hard float configuration.
3460 * config/nds32/nds32.opt: New options for hard float configuration.
3461 * config/nds32/predicates.md: New predicates for hard float
3464 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3466 * common/config/nds32/nds32-common.c
3467 (nds32_option_optimization_table): Enable -mreleax-hint by default.
3469 2018-04-05 Jakub Jelinek <jakub@redhat.com>
3472 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3473 CONSTRUCTOR_ELT (ctor, ...)->value.
3475 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
3478 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3480 2018-04-05 Tom de Vries <tom@codesourcery.com>
3483 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3486 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3487 Kito Cheng <kito.cheng@gmail.com>
3489 * config/nds32/constraints.md (U33): Fine-tune checking condition.
3490 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3491 * config/nds32/nds32.h (nds32_16bit_address_type): Add
3492 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3494 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3495 Kito Cheng <kito.cheng@gmail.com>
3497 * config/nds32/constraints.md (Ufe): New memory constraint.
3498 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3499 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3500 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3502 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3503 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3505 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3507 * config/nds32/nds32.md: Use optimize_size in the condition for
3508 alu-shift instructions.
3510 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3512 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3514 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3516 * config/nds32/nds32.md (negsi2): Refine pattern.
3518 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3519 Chung-Ju Wu <jasonwucj@gmail.com>
3521 * config/nds32/iterators.md (shift_rotate): New code iterator.
3522 (shift): New code attribute.
3523 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3524 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3525 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3526 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3527 bit-wise operations.
3528 (andsi3, *andsi3): Ditto.
3529 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3530 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3531 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3532 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3533 nds32_ior_operand, nds32_xor_operand): New predicates.
3535 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3537 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3538 (addsi3, subsi3): ... this.
3540 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3542 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3544 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3546 * config/nds32/nds32.md: Adjust indention.
3548 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3550 * config/nds32/nds32.md (feature): New attribute.
3552 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3554 * config/nds32/nds32.md (subtype): New attribute.
3556 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
3559 * config/arm/arm-builtins.c (arm_expand_builtin): Change
3560 expansion to perform a bitwise AND of the argument followed by a
3561 boolean negation of the result.
3563 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3565 PR rtl-optimization/84878
3566 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3567 the basic block. Assert the use reference is not artificial and that
3568 it has an associated insn.
3570 2018-04-04 Michael Matz <matz@suse.de>
3572 * builtins.c (compute_objsize): Pass correct operand
3573 to array_at_struct_end_p.
3575 2018-04-04 Richard Biener <rguenther@suse.de>
3578 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3579 from contexts for DINFO_LEVEL_TERSE and below.
3581 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3583 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3584 resiter_operand condition.
3585 * config/nds32/nds32.md (*move<mode>): Ditto.
3587 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3588 Monk Chiang <sh.chiang04@gmail.com>
3590 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3592 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3594 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3596 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3597 Kito Cheng <kito.cheng@gmail.com>
3599 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3600 nds32_cond_code_str, output_cond_branch,
3601 output_cond_branch_compare_zero, nds32_expand_cbranch,
3602 nds32_expand_cstore, nds32_expand_movcc,
3603 nds32_output_cbranchsi4_equality_zero,
3604 nds32_output_cbranchsi4_equality_reg,
3605 nds32_output_cbranchsi4_equality_reg_or_const_int,
3606 nds32_output_cbranchsi4_greater_less_zero: New functions.
3607 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3608 nds32_expand_cstore, nds32_expand_movcc,
3609 nds32_output_cbranchsi4_equality_zero,
3610 nds32_output_cbranchsi4_equality_reg,
3611 nds32_output_cbranchsi4_equality_reg_or_const_int,
3612 nds32_output_cbranchsi4_greater_less_zero): Declare.
3613 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3614 nds32_rimm11s_operand): New predicates.
3615 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3616 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3619 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3621 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3622 * config/nds32/nds32.md: Ditto.
3623 * config/nds32/pipelines.md: Ditto.
3625 2018-04-04 Richard Biener <rguenther@suse.de>
3627 PR tree-optimization/85168
3628 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3629 propagating abnormals.
3631 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3633 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3635 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3636 Kito Cheng <kito.cheng@gmail.com>
3638 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3639 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3640 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3641 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3642 * config/nds32/nds32.md (sibcall_internal): New.
3643 (sibcall_register): Remove.
3644 (sibcall_immediate): Remove.
3645 (sibcall_value_internal): New.
3646 (sibcall_value_register): Remove.
3647 (sibcall_value_immediate): Remove.
3648 * config/nds32/predicates.md (nds32_general_register_operand): New.
3649 (nds32_call_address_operand): New.
3651 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3653 PR rtl-optimization/85167
3654 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3655 bb_defs if *split_p, instead preinitialize it to NULL.
3657 PR tree-optimization/85156
3658 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3659 evaluating the argument multiple times.
3661 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
3663 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
3665 (_mm_cvtpd_ps): Likewise.
3666 (_mm_cvttpd_epi32): Likewise.
3667 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
3668 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
3669 vector, pixel, and bool following altivec.h include.
3671 2018-04-03 Martin Sebor <msebor@redhat.com>
3673 * doc/extend.texi (Common Function Attributes): Clarify.
3674 (const attribute): Likewise.
3675 (pure attribute): Likewise.
3677 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3680 * config/i386/i386.c (ix86_expand_vector_set): Use
3681 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
3683 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
3685 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3686 instructions when changing rounding bits to preserve precision bits
3687 in the x87 control word.
3689 2018-04-03 Martin Liska <mliska@suse.cz>
3691 PR tree-optimization/82491
3692 * rtl.h (strip_offset_and_add): Replace += suboffset with
3693 poly_uint64 () + suboffset.
3695 2018-03-29 Martin Liska <mliska@suse.cz>
3696 Martin Jambor <mjambor@suse.cz>
3699 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
3700 param_type is not an integral or pointer type.
3702 2018-04-03 Richard Biener <rguenther@suse.de>
3704 * sese.h (recompute_all_dominators): Remove.
3706 2018-04-02 Martin Sebor <msebor@redhat.com>
3708 * doc/invoke.texi (-Wrestrict): Fix typos.
3710 2018-04-02 Jim Wilson <jimw@sifive.com>
3712 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
3713 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
3714 (<optab>di3, <optab>si3_extend): Likewise.
3715 (<optab>si3_mask, <optab>si3_mask_1): New.
3716 (<optab>di3_mask, <optab>di3_mask_1): New.
3717 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
3718 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
3719 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
3721 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
3723 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
3726 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
3728 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
3729 (nds32_canonicalize_comparison): New function.
3731 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3732 Kito Cheng <kito.cheng@gmail.com>
3733 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3735 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
3736 * config/nds32/constants.md (unspec_volatile_element): Add
3737 UNSPEC_VOLATILE_RELAX_GROUP.
3738 * config/nds32/nds32-relax-opt.c: New file.
3739 * config/nds32/nds32-predicates.c
3740 (nds32_symbol_load_store_p): New function.
3741 * config/nds32/nds32-protos.h
3742 (nds32_symbol_load_store_p): Declare function.
3743 (make_pass_nds32_relax_opt): Declare new rtl pass function.
3744 * config/nds32/nds32.c
3745 (nds32_register_pass): New function to register pass.
3746 (nds32_register_passes): New function to register passes.
3747 * config/nds32/nds32.md (relax_group): New pattern.
3748 * config/nds32/nds32.opt (mrelax-hint): New option.
3749 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
3751 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
3753 * config/nds32/t-nds32: Modify files dependency.
3755 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3757 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
3758 (PROFILE_HOOK): Define its implementation.
3760 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3762 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
3763 type and 32-bit size.
3765 2018-04-01 Jakub Jelinek <jakub@redhat.com>
3768 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
3769 (V_128_256): New mode iterator.
3770 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
3771 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
3772 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
3774 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
3775 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
3777 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
3780 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
3781 NaN inputs correctly.
3783 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
3786 * config/rs6000/vsx.md (??r): New mode attribute.
3787 (*vsx_mov<mode>_64bit): Use it.
3788 (*vsx_mov<mode>_32bit): Likewise.
3790 2018-03-30 Martin Sebor <msebor@redhat.com>
3792 PR tree-optimization/84818
3793 * builtins.c (check_access): Use warning_n.
3795 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3798 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
3800 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
3803 2018-03-30 Julia Koval <julia.koval@intel.com>
3806 * x86-tune.def (movx, partial_reg_dependency): Enable for
3809 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
3812 * lra-constraints.c (process_alt_operands): Move setting
3813 this_alternative_matches below.
3815 2018-03-29 Martin Liska <mliska@suse.cz>
3818 * doc/invoke.texi: Document how LTO works with debug info.
3819 Describe auto-load support of binutils. Mention 'x86-64'
3820 as valid option value of -march option.
3822 2018-03-29 Jakub Jelinek <jakub@redhat.com>
3824 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
3827 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
3828 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
3829 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
3832 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3835 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
3836 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
3837 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
3838 for RS6000_BTM_POWERPC64.
3839 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
3840 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
3841 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
3846 2018-03-28 Carl Love <cel@us.ibm.com>
3849 2017-09-27 Carl Love <cel@us.ibm.com>
3851 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
3852 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
3853 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
3856 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3858 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
3859 instead of __vector bool.
3860 (_mm_max_pu8): Likewise.
3861 (_mm_min_pi16): Likewise.
3863 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3866 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
3867 (DIVWEUO): Likewise.
3869 (DIVDEUO): Likewise.
3870 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
3871 DIVWEUO and DIVDEUO.
3872 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
3873 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
3874 (div_extend): Likewise.
3875 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
3877 (__builtin_divweuo): Likewise.
3878 (__builtin_divdeo): Likewise.
3879 (__builtin_divdeuo): Likewise.
3881 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3884 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
3885 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
3887 PR tree-optimization/82004
3888 * gimple-match-head.c (optimize_pow_to_exp): New function.
3889 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
3890 Don't fold to exp if optimize_pow_to_exp is false.
3892 2018-03-28 Martin Liska <mliska@suse.cz>
3895 * calls.c (initialize_argument_information): Fix trailing space.
3896 * common.opt: Fix typo and provide better explanation for
3897 -fsanitize-coverage option.
3898 * config/i386/i386.opt: Fix typo.
3900 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3901 Martin Liska <mliska@suse.cz>
3904 * gimplify.c (asan_poison_variable): Don't do the check for
3905 gimplify_omp_ctxp here.
3906 (gimplify_decl_expr): Do it here.
3907 (gimplify_target_expr): Likewise.
3909 2018-03-28 Martin Liska <mliska@suse.cz>
3912 * config/i386/i386.c (ix86_function_arg_advance): Do not call
3913 chkp_type_bounds_count if MPX is not enabled.
3915 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
3917 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
3919 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
3922 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
3923 function to create the function decl for complex long double
3924 multiply and divide for -mabi=ieeelongdouble.
3925 (init_float128_ieee): Call it.
3927 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
3930 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
3931 -fcf-protection=branch -mibt.
3932 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
3934 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3937 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
3939 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
3942 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
3943 extern array declarations.
3945 2018-03-27 Richard Biener <rguenther@suse.de>
3948 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
3949 explicit single_use checks.
3951 2018-03-27 Richard Biener <rguenther@suse.de>
3953 PR tree-optimization/85082
3954 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
3957 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3959 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
3960 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
3961 Turn on fasynchronous-unwind-tables and funwind-tables.
3963 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
3966 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
3967 (*bmi_blsr_<mode>_ccz): Ditto.
3969 2018-03-26 Tom de Vries <tom@codesourcery.com>
3971 PR tree-optimization/85063
3972 * omp-general.c (offloading_function_p): New function. Factor out
3974 * omp-offload.c (pass_omp_target_link::gate): ... here.
3975 * omp-general.h (offloading_function_p): Declare.
3976 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
3977 with attribute omp declare target for offloading functions.
3979 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
3981 PR tree-optimization/84005
3982 * tree-data-ref.h (get_base_for_alignment): Declare.
3983 * tree-data-ref.c (get_base_for_alignment_1): New function.
3984 (get_base_for_alignment): Likewise.
3985 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
3986 get_base_for_alignment to find a suitable base object, instead
3987 of always using drb->base_address.
3989 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3992 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
3993 known size by default.
3995 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
3998 * lra-constraints.c (process_alt_operands): Don't match BLKmode
3999 and non BLKmode operands.
4001 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4004 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
4005 Clean up attributes.
4007 2018-03-23 Richard Biener <rguenther@suse.de>
4010 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
4011 we are going to emit early debug for LTO.
4013 2018-03-23 Jakub Jelinek <jakub@redhat.com>
4016 * function.c (match_asm_constraints_1): Don't optimize if input
4017 doesn't satisfy general_operand predicate for output's mode.
4020 * alias.c (write_dependence_p): Don't require for x_canonicalized
4021 non-VOIDmode if x has VOIDmode.
4024 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
4025 just don't try to optimize it rather than assert it never happens.
4027 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4029 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
4030 macro expansions for definition of ST_INTERNAL_<mode> and
4031 LD_INTERNAL_<mode> builtins.
4032 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
4034 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
4036 (altivec_expand_st_builtin): Likewise.
4037 (altivec_expand_builtin): Remove calls to deleted functions.
4038 (rs6000_address_for_altivec): Delete this function.
4039 * config/rs6000/vector.md: Remove expands for
4040 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
4042 2018-03-22 Sudakshina Das <sudi.das@arm.com>
4045 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
4046 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
4047 re-computing once computed.
4048 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
4049 (arm_init_machine_status): Initialize
4050 machine->static_chain_stack_bytes.
4052 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4055 * doc/extend.texi: Add four new prototypes for vec_ld.
4056 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
4057 definitions for more logical presentation.
4058 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
4059 entries for V1TI variants of __builtin_altivec_ld builtin.
4060 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
4061 handling of V1TI variant of LVX icode pattern.
4062 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
4063 (rs6000_gimple_fold_builtin): Likewise.
4064 (altivec_init_builtins): Add code to define
4065 __builtin_altivec_lvx_v1ti function.
4067 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4070 * function.c (match_asm_constraints_1): Don't do the optimization
4071 if input isn't a REG, SUBREG, MEM or constant.
4073 2018-03-22 Tom de Vries <tom@codesourcery.com>
4075 PR tree-optimization/84956
4076 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
4077 bb_has_abnormal_pred.
4079 2018-03-22 Jakub Jelinek <jakub@redhat.com>
4082 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
4083 DECL_INITIAL (decl) to decl at the end.
4084 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
4087 2018-03-21 Joseph Myers <joseph@codesourcery.com>
4089 * doc/extend.texi (__builtin_tgmath): Document when complex
4090 integer types are treated as _Complex _Float64.
4092 2018-03-21 Tom de Vries <tom@codesourcery.com>
4094 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
4096 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4098 PR tree-optimization/84960
4099 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
4100 if it is ENTRY block, move them into single succ of ENTRY in that case.
4102 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
4104 PR tree-optimization/84811
4105 * poly-int.h (poly_span_traits): Remove the T3 parameter and
4106 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
4107 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
4108 (known_subrange_p): Update accordingly. Cast each value involved
4109 in the size comparison, rather than casting the result of the
4112 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4114 PR tree-optimization/84982
4115 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
4116 by flipping the least significant bit rather than all bits from
4117 bitpos to bitpos + bitsize - 1.
4119 2018-03-21 Nathan Sidwell <nathan@acm.org>
4121 * doc/extend.texi (Deprecated Features): Remove mention of
4122 long-deleted deprecations.
4124 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4127 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
4128 * configure: Regenerate.
4130 2018-03-21 Tom de Vries <tom@codesourcery.com>
4132 PR tree-optimization/83126
4133 * tree-parloops.c (num_phis): New function.
4134 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
4136 2018-03-21 Nathan Sidwell <nathan@acm.org>
4138 * doc/extend.texi (Deprecated Features): Update deprecated flags,
4139 mention anon-struct/union members and trailing attributes.
4141 2018-03-21 Bin Cheng <bin.cheng@arm.com>
4143 PR tree-optimization/84969
4144 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
4145 builtin memset partitions if they set different rhs values.
4147 2018-03-21 Jakub Jelinek <jakub@redhat.com>
4149 PR rtl-optimization/84989
4150 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
4151 VEC_DUPLICATE with scalar result mode.
4153 2018-03-21 Martin Liska <mliska@suse.cz>
4156 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
4157 not intended return statement.
4159 2018-03-21 Martin Liska <mliska@suse.cz>
4162 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
4163 (chkp_find_bound_slots_1): Limit number of iterations.
4165 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
4168 * Minor grammar fixes for x86 options.
4170 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4173 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
4174 holding REG_CFA_RESTORE notes, instead turn them into a USE.
4176 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
4179 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
4180 (altivec_lvx_<mode>_1op): Likewise.
4181 (altivec_stvx_<mode>_2op): Likewise.
4182 (altivec_stvx_<mode>_1op): Likewise.
4183 (altivec_lvx_<VM2:mode>): New define_expand.
4184 (altivec_stvx_<VM2:mode>): Likewise.
4185 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
4186 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4187 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
4188 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
4189 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
4190 (rs6000_gen_lvx): Likewise.
4191 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
4192 (altivec_expand_stv_builtin): Likewise.
4193 (altivec_expand_builtin): Likewise.
4194 * config/rs6000/vector.md: Likewise.
4196 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4199 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
4202 2018-03-20 Richard Biener <rguenther@suse.de>
4205 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
4206 sign-conversions as zero, fall back to standard scalar_stmt
4209 2018-03-20 Martin Liska <mliska@suse.cz>
4212 * predict.c (rebuild_frequencies): Handle case when we have
4213 PROFILE_ABSENT, but flag_guess_branch_prob is false.
4215 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4218 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
4219 flag_section_anchors.
4220 * varasm.c (use_blocks_for_decl_p): Remove hack for
4221 dw2_force_const_mem.
4224 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
4226 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
4227 be created, use lowpart_subreg of operands[0] rather than operands[0]
4229 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
4230 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
4231 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
4232 and n constraint instead of aarch64_shift_imm_di and Usd.
4233 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
4234 (*aarch64_<optab>_reg_minus<mode>3): ... this.
4236 2018-03-20 Sudakshina Das <sudi.das@arm.com>
4239 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
4240 to favor GPR over NEON registers.
4241 (<shift>di3_neon): Likewise.
4243 2018-03-20 Tom de Vries <tom@codesourcery.com>
4246 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
4247 (nvptx_process_pars): Emit bar.sync asap and alap.
4249 2018-03-20 Tom de Vries <tom@codesourcery.com>
4252 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
4253 seen_label if seen_label is already set.
4255 2018-03-20 Jakub Jelinek <jakub@redhat.com>
4258 * config/i386/i386.c (fold_builtin_cpu): For features above 31
4259 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
4260 Use 1U instead of 1. Formatting fixes.
4263 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
4264 instead of TREE_TYPE (s1) for the return value.
4266 2018-03-19 Jakub Jelinek <jakub@redhat.com>
4268 PR tree-optimization/84946
4269 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
4270 bitsize + bitsize in poly_uint64 rather than poly_int64.
4273 * dwarf2asm.c: Include fold-const.c.
4274 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
4275 of decl rather than decl itself.
4277 PR rtl-optimization/84643
4278 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
4280 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
4283 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
4284 calling assemble_variable.
4286 2018-03-19 Sudakshina Das <sudi.das@arm.com>
4289 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
4290 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
4292 2018-03-19 Jim Wilson <jimw@sifive.com>
4295 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
4296 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
4297 (riscv_first_stack_step): Likewise.
4298 (riscv_option_override): Use STACK_BOUNDARY instead of
4300 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
4302 (BIGGEST_ALIGNMENT): Set to 128.
4303 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
4304 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
4307 2018-03-19 Richard Biener <rguenther@suse.de>
4309 PR tree-optimization/84933
4310 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
4311 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
4313 2018-03-19 Richard Biener <rguenther@suse.de>
4315 PR tree-optimization/84859
4316 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
4317 (cond_if_else_store_replacement): Perform sinking operation on
4318 single-store BBs regardless of MAX_STORES_TO_SINK setting.
4319 Generalize what a BB with a single eligible store is.
4321 2018-03-19 Richard Biener <rguenther@suse.de>
4323 PR tree-optimization/84929
4324 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
4325 chrec_is_positive against non-chrec arg.
4327 2018-03-19 Tamar Christina <tamar.christina@arm.com>
4330 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
4332 2018-03-18 Martin Liska <mliska@suse.cz>
4334 PR rtl-optimization/84635
4335 * regrename.c (build_def_use): Use matches_mode only when
4338 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
4340 PR tree-optimization/84913
4341 * tree-vect-loop.c (vectorizable_reduction): Don't try to
4342 vectorize chains of COND_EXPRs.
4344 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4346 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
4348 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4350 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
4352 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4354 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
4356 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4357 Kito Cheng <kito.cheng@gmail.com>
4359 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
4360 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
4361 (nds32_adjust_reg_alloc_order): New function.
4362 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
4364 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
4366 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
4367 nds32_print_operand, nds32_print_operand_address): Use
4368 HOST_WIDE_INT_PRINT_DEC instead.
4370 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4372 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
4374 2018-03-17 Jakub Jelinek <jakub@redhat.com>
4377 * config/i386/i386.c (initial_ix86_tune_features,
4378 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
4380 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
4381 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
4382 rather than 1u << ix86_tune. Formatting fix.
4383 (ix86_option_override_internal): Change ix86_arch_mask from
4384 unsigned int to unsigned HOST_WIDE_INT, initialize to
4385 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
4386 (ix86_function_specific_restore): Likewise.
4388 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4391 * postreload.c (reload_combine_recognize_pattern): Perform
4392 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
4393 truncate_int_for_mode the result for the destination's mode.
4396 * hsa-gen.c (mem_type_for_type): Fix comment typo.
4397 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
4399 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4402 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
4405 * lra-assigns.c (lra_split_hard_reg_for): Don't use
4406 regno_allocno_class_array and sorted_pseudos.
4407 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
4408 insns where regno is used.
4410 2018-03-16 Martin Liska <mliska@suse.cz>
4413 * multiple_target.c (create_dispatcher_calls): Redirect
4414 reference in the symbol table.
4416 2018-03-16 Martin Liska <mliska@suse.cz>
4419 * multiple_target.c (create_dispatcher_calls): Redirect also
4422 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4426 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4427 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4428 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4430 2018-03-16 Julia Koval <julia.koval@intel.com>
4432 * doc/invoke.texi (Skylake Server): Add CLWB.
4433 Cannonlake): Remove CLWB.
4435 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4437 PR tree-optimization/84841
4438 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4440 (FLOAT_ONE_CONST_TYPE): Define.
4441 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4442 (sort_by_operand_rank): Put entries with higher constant_type last
4443 rather than first to match comments.
4445 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
4447 * config/nios2/nios2.md (movsi_internal): Fix thinko in
4450 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4453 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4454 check and use address of target always.
4456 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
4459 * config/i386/i386.c (indirect_thunk_needed): Update comments.
4460 (indirect_thunk_bnd_needed): Likewise.
4461 (indirect_thunks_used): Likewise.
4462 (indirect_thunks_bnd_used): Likewise.
4463 (indirect_return_needed): New.
4464 (indirect_return_bnd_needed): Likewise.
4465 (output_indirect_thunk_function): Add a bool argument for
4467 (output_indirect_thunk_function): Don't generate alias for
4468 function return thunk.
4469 (ix86_code_end): Call output_indirect_thunk_function to generate
4470 function return thunks.
4471 (ix86_output_function_return): Set indirect_return_bnd_needed
4472 and indirect_return_needed instead of indirect_thunk_bnd_needed
4473 and indirect_thunk_needed.
4475 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
4477 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4478 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4479 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4481 2018-03-15 David Malcolm <dmalcolm@redhat.com>
4482 Paul Hua <paul.hua.gm@gmail.com>
4485 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4487 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
4489 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4490 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4493 2018-03-15 Tamar Christina <tamar.christina@arm.com>
4496 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4497 instead of GET_MODE_SIZE when comparing Units.
4499 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4502 * varasm.c (hash_section): Return an unchangeble hash value
4503 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4504 Return !aarch64_can_use_per_function_literal_pools_p ().
4506 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4509 * optabs.c (emit_conditional_move): Pass address of cmode's copy
4510 rather than address of cmode as last argument to prepare_cmp_insn.
4512 2018-03-15 Julia Koval <julia.koval@intel.com>
4514 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4515 F_AVX512VNNI, F_AVX512BITALG): New.
4517 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4520 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
4521 insn for floating-point loads and stores.
4523 2018-03-14 Carl Love <cel@us.ibm.com>
4525 * config/rs6000/rs6000-c.c: Add macro definitions for
4526 ALTIVEC_BUILTIN_VEC_PERMXOR.
4527 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4528 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4529 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4530 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4532 * config/doc/extend.texi: Add prototypes for vec_permxor.
4534 2018-03-14 David Malcolm <dmalcolm@redhat.com>
4537 * diagnostic-show-locus.c (class layout_point): Convert m_line
4538 from int to linenum_type.
4539 (line_span::comparator): Use linenum "compare" function when
4540 comparing line numbers.
4541 (test_line_span): New function.
4542 (layout_range::contains_point): Convert param "row" from int to
4544 (layout_range::intersects_line_p): Likewise.
4545 (layout::will_show_line_p): Likewise.
4546 (layout::print_source_line): Likewise.
4547 (layout::should_print_annotation_line_p): Likewise.
4548 (layout::print_annotation_line): Likewise.
4549 (layout::print_leading_fixits): Likewise.
4550 (layout::annotation_line_showed_range_p): Likewise.
4551 (struct line_corrections): Likewise for field m_row.
4552 (line_corrections::line_corrections): Likewise for param "row".
4553 (layout::print_trailing_fixits): Likewise.
4554 (layout::get_state_at_point): Likewise.
4555 (layout::get_x_bound_for_row): Likewise.
4556 (layout::print_line): Likewise.
4557 (diagnostic_show_locus): Likewise for locals "last_line" and
4559 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4560 * input.c (selftest::test_linenum_comparisons): New function.
4561 (selftest::input_c_tests): Call it.
4562 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4563 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4564 * selftest.h (ASSERT_GT): New macro.
4565 (ASSERT_GT_AT): New macro.
4566 (ASSERT_LT): New macro.
4567 (ASSERT_LT_AT): New macro.
4569 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4571 PR rtl-optimization/84780
4572 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4574 2018-03-14 Martin Liska <mliska@suse.cz>
4576 * tree.c (record_node_allocation_statistics): Use
4577 get_stats_node_kind.
4578 (get_stats_node_kind): New function extracted from
4579 record_node_allocation_statistics.
4580 (free_node): Use get_stats_node_kind.
4582 2018-03-14 Richard Biener <rguenther@suse.de>
4584 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4585 that the value-set of ANTIC_IN doesn't grow.
4588 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4590 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4591 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4593 2018-03-14 Julia Koval <julia.koval@intel.com>
4595 * config.gcc (icelake-client, icelake-server): New.
4597 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4598 (initial_ix86_arch_features): Ditto.
4599 (PTA_SKYLAKE): Add SGX.
4600 (PTA_ICELAKE): Remove.
4601 (PTA_ICELAKE_CLIENT): New.
4602 (PTA_ICELAKE_SERVER): New.
4603 (ix86_option_override_internal): Split up icelake on icelake client and
4605 (get_builtin_code_for_version): Ditto.
4606 (fold_builtin_cpu): Ditto.
4607 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4608 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4609 * config/i386/i386.h (processor_type): Ditto.
4610 * doc/invoke.texi: Ditto.
4612 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4615 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4616 INTEGER_CST offset, add it together with bitpos / 8 and
4617 sign extend based on POINTER_SIZE.
4621 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4624 * config/i386/constraints.md (Yc): New register constraint.
4625 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4626 Use Yc constraint for alternative 2 of operand 0. Remove
4627 preferred_for_speed attribute.
4629 2018-03-14 Richard Biener <rguenther@suse.de>
4631 PR tree-optimization/84830
4632 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4633 with the old one to avoid oscillations.
4635 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
4638 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4640 (assign_by_spills): Return a flag of reload assignment failure.
4641 Do not process the reload assignment failures. Do not spill other
4642 reload pseudos if they has the same reg class. Update n if
4644 (lra_assign): Add a return arg. Set up from the result of
4645 assign_by_spills call.
4646 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4647 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4648 usage_insns if it is not NULL.
4649 (spill_hard_reg_in_range): New function.
4650 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4651 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4652 function prototypes.
4653 (lra_assign): Change prototype.
4654 * lra.c (lra): Add code to deal with fails by splitting hard reg
4657 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
4659 * config/riscv/riscv.opt (mrelax): New option.
4660 * config/riscv/riscv.c (riscv_file_start): Emit ".option
4661 "norelax" when riscv_mrelax is disabled.
4662 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
4664 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4667 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
4668 reassociation for int modes.
4670 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4672 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
4673 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
4675 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
4676 * config/aarch64/aarch64-sve.md
4677 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4678 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4679 (*extend<mode><Vwide>2): Rename to...
4680 (aarch64_sve_extend<mode><Vwide>2): ...this.
4681 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4682 renaming the old pattern to...
4683 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
4685 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4686 define_expand, renaming the old pattern to...
4687 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4688 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4689 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4690 account when deciding which SVE instruction the optab should use.
4691 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4693 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4695 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4696 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4697 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4698 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4699 (tlsdesc_small_<mode>): Turn a define_expand and use
4700 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
4701 (tlsdesc_small_advsimd_<mode>): ...this.
4702 (tlsdesc_small_sve_<mode>): New pattern.
4704 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4706 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
4707 (UNSPEC_UMUL_HIGHPART): New constants.
4708 (MUL_HIGHPART): New int iteraor.
4709 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
4710 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
4712 (*<su>mul<mode>3_highpart): New define_insn.
4714 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4717 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4720 2018-03-13 Martin Liska <mliska@suse.cz>
4723 * (sem_item_optimizer::sem_item_optimizer): Initialize new
4725 (sem_item_optimizer::~sem_item_optimizer): Release it.
4726 (sem_item_optimizer::merge_classes): Register variable aliases.
4727 (sem_item_optimizer::fixup_pt_set): New function.
4728 (sem_item_optimizer::fixup_points_to_sets): Likewise.
4729 * ipa-icf.h: Declare new variables and functions.
4731 2018-03-13 Jakub Jelinek <jakub@redhat.com>
4734 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
4735 integer_pow2p@2 and test integer_pow2p in condition.
4736 (A < 0 ? C : 0): Similarly for @1.
4739 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
4740 characters starting at p contain '\0' character, don't look beyond
4744 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
4745 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
4748 * reg-stack.c (change_stack): Change update_end var from int to
4749 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
4750 also call set_block_for_insn on the newly added insns and rescan.
4753 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
4754 on the last operand.
4757 * tree.c (stabilize_reference_1): Return save_expr (e) for
4758 STATEMENT_LIST even if it doesn't have side-effects.
4760 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
4762 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
4764 2018-03-12 Renlin Li <renlin.li@arm.com>
4766 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
4767 aarch64_output_scalar_simd_mov_immediate.
4769 2018-03-12 Martin Sebor <msebor@redhat.com>
4771 PR tree-optimization/83456
4772 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
4773 for perfectly overlapping calls to memcpy.
4774 (gimple_fold_builtin_memory_chk): Same.
4775 (gimple_fold_builtin_strcpy): Handle no-warning.
4776 (gimple_fold_builtin_stxcpy_chk): Same.
4777 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
4779 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4781 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
4782 parameter. Use it for SFmode.
4783 (rs6000_function_arg_advance_1): Adjust.
4784 (rs6000_function_arg): Adjust.
4785 (rs6000_gimplify_va_arg): Pass false for that new parameter.
4787 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4789 PR rtl-optimization/84169
4790 PR rtl-optimization/84780
4791 * combine.c (can_combine_p): Check for a 2-insn combination whether
4792 the destination register is used between the two insns, too.
4794 2018-03-12 Richard Biener <rguenther@suse.de>
4796 PR tree-optimization/84803
4797 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
4798 for refs DR analysis didn't process.
4800 2018-03-12 Richard Biener <rguenther@suse.de>
4802 PR tree-optimization/84777
4803 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
4804 force-vectorize loops ignore whether we are optimizing for size.
4806 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
4808 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
4809 (TARGET_MD_ASM_ADJUST): Define.
4811 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
4812 Kito Cheng <kito.cheng@gmail.com>
4813 Chung-Ju Wu <jasonwucj@gmail.com>
4815 * config/nds32/nds32.c (nds32_compute_stack_frame,
4816 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
4817 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
4818 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
4819 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
4820 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
4821 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
4822 * config/nds32/nds32.md (prologue, epilogue): Use macro
4823 NDS32_V3PUSH_AVAILABLE_P to do checking.
4825 2018-03-11 Jakub Jelinek <jakub@redhat.com>
4828 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
4829 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
4830 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
4831 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
4832 addition of most attributes on !orig_type_die or the attribute not
4833 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
4835 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4836 Chung-Ju Wu <jasonwucj@gmail.com>
4838 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
4840 * config/nds32/nds32.opt (mvh): New option.
4842 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4843 Chung-Ju Wu <jasonwucj@gmail.com>
4845 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
4847 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
4848 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
4851 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4852 Chung-Ju Wu <jasonwucj@gmail.com>
4854 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
4856 * config/nds32/nds32-multiple.md (strlensi): New pattern.
4857 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
4859 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
4860 Kito Cheng <kito.cheng@gmail.com>
4861 Chung-Ju Wu <jasonwucj@gmail.com>
4863 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
4864 UNSPEC_FFMISM and UNSPEC_FLMISM.
4865 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
4866 for ffb, ffmism and flmism.
4867 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
4868 (unspec_ffmism): Ditto.
4869 (unspec_flmism): Ditto.
4870 (nds32_expand_builtin_impl): Check if string extension is available.
4871 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
4872 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
4874 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
4877 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4880 * lra-assigns.c (assign_by_spills): Return a flag of reload
4881 assignment failure. Do not process the reload assignment
4882 failures. Do not spill other reload pseudos if they has the same
4884 (lra_assign): Add a return arg. Set up from the result of
4885 assign_by_spills call.
4886 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4887 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4888 usage_insns if it is not NULL.
4889 (spill_hard_reg_in_range): New function.
4890 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4891 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4892 function prototypes.
4893 (lra_assign): Change prototype.
4894 * lra.c (lra): Add code to deal with fails by splitting hard reg
4897 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
4900 * config/i386/i386.opt: Replace Enforcment with Enforcement.
4902 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
4905 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
4906 (dw_val_node): Add val_symbolic_view.
4907 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
4908 (symview_upper_bound): New.
4909 (new_line_info_table): Initialize symviews_since_reset.
4910 (dwarf2out_source_line): Count symviews_since_reset and set
4911 symview_upper_bound.
4912 (dw_val_equal_p): Handle symview.
4913 (add_AT_symview): New.
4914 (print_dw_val): Handle symview.
4915 (attr_checksum, attr_checksum_ordered): Likewise.
4916 (same_dw_val_p, size_of_die): Likewise.
4917 (value_format, output_die): Likewise.
4918 (add_high_low_attributes): Use add_AT_symview for entry_view.
4919 (dwarf2out_finish): Reset symview_upper_bound, clear
4922 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
4925 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
4926 Add strict argument and use it.
4927 (rs6000_split_multireg_move): Update for new strict argument.
4928 (mem_operand_gpr): Disallow all non-offsettable addresses.
4929 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
4931 2018-03-09 Jakub Jelinek <jakub@redhat.com>
4934 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
4935 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
4936 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
4939 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
4940 decl, use remap_type if we want to use the type.
4942 2018-03-09 Martin Sebor <msebor@redhat.com>
4944 PR tree-optimization/84526
4945 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4947 (builtin_access::generic_overlap): Be prepared to handle non-array
4950 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
4952 PR rtl-optimization/84682
4953 * lra-constraints.c (process_address_1): Check is_address flag
4954 for address constraints.
4955 (process_alt_operands): Likewise.
4956 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
4957 preprocess_constraints.
4958 * recog.h (preprocess_constraints): Add oploc parameter.
4960 * recog.c (preprocess_constraints): Test address_operand for
4961 CT_ADDRESS constraints.
4963 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4966 * lra-assigns.c (assign_by_spills): Return a flag of reload
4967 assignment failure. Do not process the reload assignment
4968 failures. Do not spill other reload pseudos if they has the same
4970 (lra_assign): Add a return arg. Set up from the result of
4971 assign_by_spills call.
4972 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4973 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4974 usage_insns if it is not NULL.
4975 (spill_hard_reg_in_range): New function.
4976 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4977 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4978 function prototypes.
4979 (lra_assign): Change prototype.
4980 * lra.c (lra): Add code to deal with fails by splitting hard reg
4983 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4986 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
4987 Accept complain bool parameter. Only emit errors if it is true.
4988 (arm_parse_cpu_option_name): Likewise.
4989 (arm_target_thumb_only): Adjust callers of the above.
4990 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
4991 prototype to take a default true bool parameter.
4992 (arm_parse_arch_option_name): Likewise.
4994 2018-03-09 David Malcolm <dmalcolm@redhat.com>
4995 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
4999 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
5000 * configure: Regenerate.
5001 * configure.ac ("linker --version-script option"): New.
5002 ("linker soname option"): New.
5004 2018-03-09 Richard Biener <rguenther@suse.de>
5006 PR tree-optimization/84775
5007 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
5008 immediate uses of predicate stmts and mark them modified.
5011 PR tree-optimization/84178
5012 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5014 (version_loop_for_if_conversion): Delay update_ssa call.
5015 (tree_if_conversion): Delay update_ssa until after predicate
5018 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
5021 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
5022 when the function accesses prior frames.
5024 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5027 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
5028 gen_llsym, otherwise call maybe_gen_llsym.
5031 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
5032 has ',' character inside of it.
5034 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5037 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
5038 as clobbering CC_REGNUM.
5040 2018-03-08 Richard Biener <rguenther@suse.de>
5043 * tree-scalar-evolution.c: Include tree-into-ssa.h.
5044 (follow_copies_to_constant): Do not follow SSA names registered
5047 2018-03-08 Richard Biener <rguenther@suse.de>
5049 PR tree-optimization/84178
5050 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
5052 (version_loop_for_if_conversion): Delay update_ssa call.
5053 (tree_if_conversion): Delay update_ssa until after predicate
5056 2018-03-08 David Malcolm <dmalcolm@redhat.com>
5058 PR tree-optimization/84178
5059 * tree-if-conv.c (release_bb_predicate): Remove the
5060 the assertion that the stmts have NULL use_ops.
5061 Discard the statements, asserting that they haven't
5062 yet been added to a BB.
5064 2018-03-08 Richard Biener <rguenther@suse.de>
5066 PR tree-optimization/84746
5067 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
5068 (phi_translate): Pass in destination ANTIC_OUT set.
5069 (phi_translate_1): Likewise. For a simplified result lookup
5070 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
5071 (phi_translate_set): Adjust.
5072 (do_pre_regular_insertion): Likewise.
5073 (do_pre_partial_partial_insertion): Likewise.
5075 2018-03-08 Martin Liska <mliska@suse.cz>
5077 PR gcov-profile/84735
5078 * doc/gcov.texi: Document usage of profile files.
5079 * gcov-io.h: Document changes in the format.
5081 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
5085 * dwarf2out.c (struct dw_line_info_table): Update comments for
5087 (FORCE_RESET_NEXT_VIEW): New.
5088 (FORCE_RESETTING_VIEW_P): New.
5089 (RESETTING_VIEW_P): Check for -1 too.
5090 (ZERO_VIEW_P): Likewise.
5091 (new_line_info_table): Force-reset next view.
5092 (dwarf2out_begin_function): Likewise.
5093 (dwarf2out_source_line): Simplify zero_view_p initialization.
5094 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
5095 view directly. Omit view when omitting .loc at line 0.
5097 2018-03-08 Jakub Jelinek <jakub@redhat.com>
5099 PR tree-optimization/84740
5100 * tree-switch-conversion.c (process_switch): Call build_constructors
5101 only if info.phi_count is non-zero.
5103 PR tree-optimization/84739
5104 * tree-tailcall.c (find_tail_calls): Check call arguments against
5105 DECL_ARGUMENTS (current_function_decl) rather than
5106 DECL_ARGUMENTS (func) when checking for tail recursion.
5108 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5110 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
5111 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
5112 Volker Reichelt's entry and add entries for people that perform
5113 GCC fuzzy testing and report numerous bugs.
5115 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
5118 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
5119 readonly data in sdata, if that is disabled.
5120 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
5121 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
5122 -mreadonly-in-sdata option.
5124 2018-03-07 Martin Sebor <msebor@redhat.com>
5126 PR tree-optimization/84468
5127 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
5128 basic block when looking for nul assignment.
5130 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
5133 * except.h (output_function_exception_table): Adjust prototype.
5134 * except.c (output_function_exception_table): Remove FNNAME parameter
5135 and add SECTION parameter. Ouput one part of the table at a time.
5136 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
5137 the first part of the exception table and emit unwind directives.
5138 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
5139 (i386_pe_seh_cold_init): Likewise.
5140 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
5141 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
5142 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
5143 (ix86_output_call_insn): Emit a nop in one more case for SEH.
5144 * config/i386/winnt.c: Include except.h.
5145 (struct seh_frame_state): Add reg_offset, after_prologue and
5146 in_cold_section fields.
5147 (i386_pe_seh_end_prologue): Set seh->after_prologue.
5148 (i386_pe_seh_cold_init): New function.
5149 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
5150 to seh->in_cold_section.
5151 (seh_emit_push): Record the offset of the push.
5152 (seh_emit_save): Record the offet of the save.
5153 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
5154 Test seh->after_prologue to disregard the epilogue.
5155 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
5156 (i386_pe_end_cold_function): New function.
5158 2018-03-07 Jakub Jelinek <jakub@redhat.com>
5161 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
5162 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
5165 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
5167 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
5168 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
5171 * multiple_target.c: Include tree-inline.h and intl.h.
5172 (expand_target_clones): Diagnose and fail if node->definition and
5173 !tree_versionable_function_p (node->decl).
5175 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
5177 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
5179 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
5180 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
5181 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
5183 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5186 * combine.c (try_combine): Use reg_or_subregno instead of handling
5187 just paradoxical SUBREGs and REGs.
5189 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
5191 * config/arc/arc.c (arc_finalize_pic): Remove function.
5192 (arc_must_save_register): We use single base PIC register, remove
5193 checks to save/restore the PIC register.
5194 (arc_expand_prologue): Likewise.
5195 * config/arc/arc-protos.h (arc_set_default_type_attributes):
5197 (arc_verify_short): Likewise.
5198 (arc_attr_type): Likewise.
5199 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
5200 (walk_stores): Likewise.
5201 (arc_address_cost): Make it static.
5202 (arc_verify_short): Likewise.
5203 (branch_dest): Likewise.
5204 (arc_attr_type): Likewise.
5205 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
5206 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
5207 (arc_final_prescan_insn): Remove inserting the nops due to
5208 hardware hazards. It is done in reorg step.
5209 (insn_length_variant_t): Remove.
5210 (insn_length_parameters_t): Likewise.
5211 (arc_insn_length_parameters): Likewise.
5212 (arc_get_insn_variants): Likewise.
5213 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
5215 2018-03-06 Jakub Jelinek <jakub@redhat.com>
5218 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
5221 PR tree-optimization/84687
5222 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
5224 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
5226 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5228 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
5229 Rename to ppc_speculation_barrier.
5230 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
5231 __builtin_ppc_speculation_barrier.
5233 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5236 * combine.c (combine_simplify_rtx): Don't try to simplify if
5237 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
5240 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
5242 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
5243 to 32 bytes when compiling for POWER9.
5245 2018-03-05 Jakub Jelinek <jakub@redhat.com>
5248 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
5249 regparm >= 3 with no arg reg available also for calls with
5250 flag_force_indirect_call. Pass decl to ix86_function_regparm.
5253 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
5255 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
5257 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
5260 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
5262 2018-03-05 Richard Biener <rguenther@suse.de>
5264 PR tree-optimization/84486
5265 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
5266 When inserting a __builtin_assume_aligned call set the LHS
5267 SSA name alignment info accordingly.
5269 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
5271 PR tree-optimization/84114
5272 * config/aarch64/aarch64.c (aarch64_reassociation_width)
5273 Avoid reassociation of FLOAT_MODE addition.
5275 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
5277 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
5278 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
5279 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
5280 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
5281 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
5282 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
5283 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
5285 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
5286 __builtin_ia32_wbinvd): New builtins.
5287 (SPECIAL_ARGS2): New.
5288 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
5289 (SPECIAL_ARGS2): New.
5290 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
5291 (ix86_valid_target_attribute_inner_p): Ditto.
5292 (ix86_init_mmx_sse_builtins): Add special_args2.
5293 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
5294 TARGET_WBNOINVD_P): New.
5295 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
5296 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
5297 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
5298 * config/i386/immintrin.h (_wbinvd): New intrinsic.
5299 * config/i386/pconfigintrin.h: New file.
5300 * config/i386/wbnoinvdintrin.h: Ditto.
5301 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
5302 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
5304 2018-03-05 Richard Biener <rguenther@suse.de>
5306 PR tree-optimization/84670
5307 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5309 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5310 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5311 (compute_antic_aux): Only assert the number of values in ANTIC_IN
5312 doesn't grow if all successors (recursively) were visited at least
5315 2018-03-05 Richard Biener <rguenther@suse.de>
5317 PR tree-optimization/84650
5318 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
5319 if executed in the loop pipeline.
5321 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
5323 * doc/configfiles.texi (Configuration Files): Move info about
5324 conditionalizing $target-protos.h to...
5325 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
5326 differs from $target-protos.h.
5328 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
5329 Chung-Ju Wu <jasonwucj@gmail.com>
5331 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
5332 * config/nds32/nds32-multiple.md (setmemsi): Define.
5333 * config/nds32/nds32-memory-manipulation.c
5334 (nds32_gen_dup_4_byte_to_word_value): New.
5335 (emit_setmem_word_loop): New.
5336 (emit_setmem_byte_loop): New.
5337 (nds32_expand_setmem_loop): New.
5338 (nds32_expand_setmem_loop_v3m): New.
5339 (nds32_expand_setmem_unroll): New.
5340 (nds32_expand_setmem): New.
5342 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5343 Chung-Ju Wu <jasonwucj@gmail.com>
5345 * config/nds32/nds32-memory-manipulation.c
5346 (nds32_emit_load_store): New.
5347 (nds32_emit_post_inc_load_store): New.
5348 (nds32_emit_mem_move): New.
5349 (nds32_emit_mem_move_block): New.
5350 (nds32_expand_movmemsi_loop_unknown_size): New.
5351 (nds32_expand_movmemsi_loop_known_size): New.
5352 (nds32_expand_movmemsi_loop): New.
5353 (nds32_expand_movmemsi_unroll): New.
5354 (nds32_expand_movmemqi): Rename ...
5355 (nds32_expand_movmemsi): ... to this.
5356 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
5357 (movmemsi): ... to this.
5358 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
5359 (nds32_expand_movmemsi): ... to this.
5361 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5362 Monk Chiang <sh.chiang04@gmail.com>
5363 Chung-Ju Wu <jasonwucj@gmail.com>
5365 * config/nds32/nds32-protos.h
5366 (nds32_expand_load_multiple): New arguments.
5367 (nds32_expand_store_multiple): Ditto.
5368 (nds32_valid_multiple_load_store): Rename ...
5369 (nds32_valid_multiple_load_store_p): ... to this.
5370 * config/nds32/nds32-memory-manipulation.c
5371 (nds32_expand_load_multiple): Refine implementation.
5372 (nds32_expand_store_multiple): Ditto.
5373 * config/nds32/nds32-multiple.md
5374 (load_multiple): Update nds32_expand_load_multiple interface.
5375 (store_multiple): Update nds32_expand_store_multiple interface.
5376 * config/nds32/nds32-predicates.c
5377 (nds32_valid_multiple_load_store): Rename ...
5378 (nds32_valid_multiple_load_store_p): ... to this and refine
5380 * config/nds32/predicates.md
5381 (nds32_load_multiple_and_update_address_operation): New predicate.
5382 (nds32_store_multiple_and_update_address_operation): New predicate.
5384 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5385 Chung-Ju Wu <jasonwucj@gmail.com>
5387 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
5388 (combo): New attribute.
5389 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
5391 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
5393 * config/nds32/nds32.opt: Change -mcmodel= default value.
5395 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
5396 Monk Chiang <sh.chiang04@gmail.com>
5397 Chung-Ju Wu <jasonwucj@gmail.com>
5399 * config/nds32/constants.md (unspec_element): New enum.
5400 * config/nds32/constraints.md (Umw): New constraint.
5401 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
5402 * config/nds32/nds32-intrinsic.md: Likewise.
5403 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
5404 (nds32_valid_smw_lwm_base_p): New.
5405 (nds32_output_smw_single_word): New.
5406 (nds32_output_lmw_single_word): New.
5407 (nds32_expand_unaligned_load): New.
5408 (nds32_expand_unaligned_store): New.
5409 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5410 (nds32_output_smw_single_word): Declare.
5411 (nds32_output_lmw_single_word): Declare.
5412 (nds32_expand_unaligned_load): Declare.
5413 (nds32_expand_unaligned_store): Declare.
5414 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5415 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5416 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5417 NDS32_BUILTIN_UASTORE_DW.
5418 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5421 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
5422 Kito Cheng <kito.cheng@gmail.com>
5423 Chung-Ju Wu <jasonwucj@gmail.com>
5425 * config/nds32/nds32-intrinsic.c
5426 (nds32_expand_builtin_null_ftype_reg): Delete.
5427 (nds32_expand_builtin_reg_ftype_imm): Ditto.
5428 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5429 (nds32_read_argument): New.
5430 (nds32_legitimize_target): Ditto.
5431 (nds32_legitimize_argument): Ditto.
5432 (nds32_check_constant_argument): Ditto.
5433 (nds32_expand_unop_builtin): Ditto.
5434 (nds32_expand_unopimm_builtin): Ditto.
5435 (nds32_expand_binop_builtin): Ditto.
5436 (nds32_builtin_decl_impl): Ditto.
5437 (builtin_description): Ditto.
5438 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5439 (nds32_init_builtins_impl): Ditto.
5440 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5441 (nds32_builtin_decl): New.
5442 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5443 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5445 2018-03-02 Jeff Law <law@redhat.com>
5447 * reorg.c (stop_search_p): Handle DEBUG_INSN.
5448 (redundant_insn, fill_simple_delay_slots): Likewise.
5449 (fill_slots_from_thread): Likewise.
5450 * resource.c (mark_referenced_resources): Likewise.
5451 (mark_set_resources, find_dead_or_set_registers): Likewise.
5453 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5455 * substring-locations.h (format_warning_va): Formatting fix for
5457 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5459 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5460 * substring-locations.c: Include intl.h.
5461 (format_warning_va): Turned into small wrapper around
5462 format_warning_n_va, renamed to ...
5463 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5464 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5466 (format_warning_at_substring_n): New function.
5467 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5468 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
5469 format_warning_at_substring with just a shorter name instead of
5470 const function pointer.
5471 (fmtwarn_n): New function.
5472 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5473 appropriate, get rid of all the fmtstr temporaries, move conditionals
5474 with G_() wrapped string literals directly into fmtwarn arguments,
5475 cast dir.len to (int), formatting fixes.
5477 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
5479 * doc/invoke.texi: Remove "Cilk Plus" references.
5481 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5482 Richard Biener <rguenther@suse.de>
5485 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5486 for error or warning attributes if CALL_FROM_THUNK_P is set.
5489 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5492 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5493 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5496 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5497 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5499 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5500 instead of -1U in last predictors element's probability member.
5502 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5505 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5506 arguments if they are comparable.
5508 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5510 PR tree-optimization/84634
5511 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5512 masks and masked_loop_p with a single loop_masks, making sure it's
5513 null for bb vectorization.
5515 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5517 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5518 (vect_analyze_data_ref_access): Use loop->safe_len rather than
5519 loop->force_vectorize to check whether there is no alias.
5521 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5524 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5526 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5528 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5529 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5530 instead of a loop around prev_real_insn.
5531 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5535 * config/i386/i386.c (ix86_print_operand): Use conditional
5536 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5539 2018-03-02 Richard Biener <rguenther@suse.de>
5541 PR tree-optimization/84427
5542 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5543 (bitmap_set_subtract_values): Rewrite to handle multiple
5546 (prune_clobbered_mems): Likewise.
5547 (phi_translate): Take edge instead of pred/phiblock.
5548 (phi_translate_1): Likewise.
5549 (phi_translate_set): Likewise. Insert all translated
5550 exprs for a value into the set, keeping possibly multiple
5551 expressions per value.
5552 (compute_antic_aux): Adjust for phi_translate changes.
5553 When intersecting union the expressions and prune those
5554 not in the final value set, keeping possibly multiple
5555 expressions per value. Do not use value-insertion
5556 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5557 all expressions. Add verification that the value-sets
5558 only shrink during iteration.
5559 (compute_partial_antic_aux): Adjust for the phi_translate changes.
5560 (do_pre_regular_insertion): Likewise.
5561 (do_pre_partial_partial_insertion): Likewise.
5563 2018-03-02 Richard Biener <rguenther@suse.de>
5566 * config/darwin.c (saved_debug_info_level): New static global.
5567 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5568 (darwin_asm_lto_end): Restore debug info generation settings.
5570 2018-03-01 Martin Liska <mliska@suse.cz>
5573 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5576 2018-03-01 Richard Biener <rguenther@suse.de>
5579 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5580 type annotation with LTO when debug was disabled at compile-time.
5582 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5584 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5586 (mips_final_postscan_insn): Likewise.
5588 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5590 PR rtl-optimization/84528
5591 * alias.c (init_alias_target): Add commentary.
5592 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5593 a unique base value if the frame pointer is not eliminated
5594 to the stack pointer.
5596 2018-03-01 Tom de Vries <tom@codesourcery.com>
5598 PR rtl-optimization/83327
5599 * lra-int.h (hard_regs_spilled_into): Declare.
5600 * lra.c (hard_regs_spilled_into): Define.
5601 (init_reg_info): Init hard_regs_spilled_into.
5602 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5603 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5604 (process_bb_lives): Handle hard_regs_spilled_into.
5605 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5606 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5608 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5610 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5611 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5612 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5613 * config/rs6000/aix72.h: New file.
5615 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5617 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5618 instead of warning_at with conditional singular and plural messages
5622 * stor-layout.c (update_alignment_for_field): For
5623 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5624 && !DECL_PACKED (field), do the alignment update, just use
5625 only desired_align instead of MAX (type_align, desired_align)
5627 (place_field): Don't do known_align < desired_align handling
5628 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5629 is non-NULL, instead do it after rli->prev_field handling and
5630 only if not within a bitfield word. For DECL_PACKED (field)
5631 use type_align of BITS_PER_UNIT.
5633 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5635 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5636 superfluous parentheses and trailing spaces.
5638 2018-02-28 Richard Biener <rguenther@suse.de>
5640 PR tree-optimization/84584
5641 * graphite-scop-detection.c (scop_detection::add_scop): Discard
5642 SCoPs with fake exit edge.
5644 2018-02-28 Martin Liska <mliska@suse.cz>
5647 * timevar.c (timer::print): Fix format to properly print 100%
5650 2018-02-28 Richard Biener <rguenther@suse.de>
5653 * genmatch.c (capture_info::walk_match): Do not mark
5654 captured expressions without operands as expr_p given
5655 they act more like predicates and should be subject to
5656 "lost tail" side-effect preserving.
5658 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
5660 PR rtl-optimization/81611
5661 * auto-inc-dec.c (attempt_change): Move dead note from
5662 mem_insn if it's the next use of regno
5663 (find_address): Take address use of reg holding
5664 non-incremented value. Add parm to limit search to the named
5666 (merge_in_block): Attempt to use a mem insn that is the next
5667 use of the original regno.
5669 2018-02-27 Martin Sebor <msebor@redhat.com>
5672 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
5673 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
5675 2018-02-27 Martin Sebor <msebor@redhat.com>
5677 PR translation/84207
5678 * diagnostic-core.h (warning_n, error_n, inform_n): Change
5679 n argument to unsigned HOST_WIDE_INT.
5680 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5681 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
5682 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5683 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5685 2018-02-27 Richard Biener <rguenther@suse.de>
5687 PR tree-optimization/84512
5688 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5689 Do not use the estimate returned from record_stmt_cost for
5690 the scalar iteration cost but sum properly using add_stmt_cost.
5692 2018-02-27 Richard Biener <rguenther@suse.de>
5694 PR tree-optimization/84466
5695 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5696 Adjust last change to less strictly validate use operands.
5698 2018-02-27 Martin Liska <mliska@suse.cz>
5700 PR gcov-profile/84548
5701 * gcov.c (process_file): Allow partial overlap and consider it
5702 also as group functions.
5703 (output_lines): Properly calculate range of lines for a group.
5705 2018-02-27 Martin Liska <mliska@suse.cz>
5707 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
5708 'ggc' suffixes. Change first column width.
5709 (timer::print): Fix formatting of the column.
5711 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
5713 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
5714 preserve inline entry blocks for the sake of debug inline
5715 entry point markers alone.
5716 (remove_unused_locals): Suggest in comments a better place to
5717 force the preservation of inline entry blocks that are
5718 otherwise unused, but do not preserve them.
5720 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5722 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
5724 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5727 * config/i386/constraints.md (Bs): Replace
5728 ix86_indirect_branch_register with
5729 TARGET_INDIRECT_BRANCH_REGISTER.
5731 * config/i386/i386.md (indirect_jump): Likewise.
5732 (tablejump): Likewise.
5733 (*sibcall_memory): Likewise.
5734 (*sibcall_value_memory): Likewise.
5735 Peepholes of indirect call and jump via memory: Likewise.
5736 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
5737 (*sibcall_value_GOT_32): Likewise.
5738 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5739 (GOT_memory_operand): Likewise.
5740 (call_insn_operand): Likewise.
5741 (sibcall_insn_operand): Likewise.
5742 (GOT32_symbol_operand): Likewise.
5743 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
5745 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5747 PR rtl-optimization/83496
5748 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
5749 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
5750 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
5751 redundant insn, if any.
5752 (relax_delay_slots): Likewise.
5753 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
5755 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
5757 PR tree-optimization/83965
5758 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
5759 that grouped statements are part of a reduction chain. Return
5760 true if the statement is not marked as a reduction itself but
5762 (vect_recog_dot_prod_pattern): Don't check whether the statement
5763 is part of a group here.
5764 (vect_recog_sad_pattern): Likewise.
5765 (vect_recog_widen_sum_pattern): Likewise.
5767 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5770 * final.c (rest_of_clean_state): Also look for calls inside sequences.
5772 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5775 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
5777 (ix86_output_indirect_function_return): New prototype.
5778 (ix86_split_simple_return_pop_internal): Likewise.
5779 * config/i386/i386.c (indirect_return_via_cx): New.
5780 (indirect_return_via_cx_bnd): Likewise.
5781 (indirect_thunk_name): Handle return va CX_REG.
5782 (output_indirect_thunk_function): Create alias for
5783 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
5784 (ix86_output_indirect_jmp): Remove the bool argument.
5785 (ix86_output_indirect_function_return): New function.
5786 (ix86_split_simple_return_pop_internal): Likewise.
5787 * config/i386/i386.md (*indirect_jump): Don't pass false
5788 to ix86_output_indirect_jmp.
5789 (*tablejump_1): Likewise.
5790 (simple_return_pop_internal): Change it to define_insn_and_split.
5791 Call ix86_split_simple_return_pop_internal to split it for
5793 (simple_return_indirect_internal): Call
5794 ix86_output_indirect_function_return instead of
5795 ix86_output_indirect_jmp.
5797 2018-02-26 Jakub Jelinek <jakub@redhat.com>
5800 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
5801 memset and value initialization afterwards.
5803 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
5805 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
5807 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5810 * common/config/aarch64/aarch64-common.c
5811 (aarch_option_optimization_table[]): Switch
5812 off fomit-frame-pointer
5814 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5815 Chung-Ju Wu <jasonwucj@gmail.com>
5817 * config/nds32/nds32-multiple.md (load_multiple): Disallow
5819 (store_multiple): Ditto.
5821 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5823 * config.gcc: Add --with-cpu support for nds32 target.
5824 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
5825 * config/nds32/nds32.opt: Add -mcpu= option.
5827 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
5829 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
5830 isel=yes): Warn for these deprecated options.
5832 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
5834 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
5835 ISA_2_5_MASKS_EMBEDDED.
5837 2018-02-23 Jakub Jelinek <jakub@redhat.com>
5839 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
5840 p->max as pointers rather than using iterative_hash_expr.
5842 2018-02-23 Carl Love <cel@us.ibm.com>
5844 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
5845 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
5847 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
5848 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
5849 P8V_BUILTIN_VEC_VUNSIGNED2.
5851 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
5854 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
5855 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
5857 * lra-constraints.c (curr_insn_transform): Set up
5858 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
5860 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
5861 * lra-eliminations.c (spill_pseudos): Ditto.
5862 (process_insn_for_elimination): Ditto.
5863 * lra-lives.c (reg_early_clobber_p): Use the new macros.
5864 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
5865 LRA_NON_CLOBBERED_ALT.
5867 2018-02-22 Martin Sebor <msebor@redhat.com>
5869 PR tree-optimization/84480
5870 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
5871 to maybe_diag_stxncpy_trunc. Call it.
5872 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
5873 from gimple_fold_builtin_strcpy. Print inlining stack.
5874 (handle_builtin_stxncpy): Print inlining stack.
5875 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
5877 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
5880 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
5881 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
5882 and -fcheck-pointer-bounds are used together.
5883 (indirect_thunk_prefix): New enum.
5884 (indirect_thunk_need_prefix): New function.
5885 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
5886 "_nt" instead of "_bnd" for NOTRACK prefix.
5887 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
5888 (output_indirect_thunk_function): Likewise.
5890 (ix86_code_end): Update output_indirect_thunk_function calls.
5891 (ix86_output_indirect_branch_via_reg): Replace
5892 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
5893 (ix86_output_indirect_branch_via_push): Likewise.
5894 (ix86_output_function_return): Likewise.
5895 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
5896 incompatible with -fcf-protection=branch and
5897 -fcheck-pointer-bounds.
5899 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5902 * config/aarch64/aarch64.c (aarch64_print_address_internal):
5903 Change gcc_assert call to output_operand_lossage.
5905 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5907 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
5909 2018-02-22 DJ Delorie <dj@redhat.com>
5910 Sebastian Perta <sebastian.perta@renesas.com>
5911 Oleg Endo <olegendo@gcc.gnu.org>
5913 * config/rx/rx.c (rx_rtx_costs): New function.
5914 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
5916 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
5918 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
5920 2018-02-22 Martin Liska <mliska@suse.cz>
5923 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
5924 Add "native" as a possible value.
5926 2018-02-22 Martin Liska <mliska@suse.cz>
5929 * config/i386/i386.c (ix86_option_override_internal):
5930 Add "native" as a possible value for -march and -mtune.
5932 2018-02-22 Jakub Jelinek <jakub@redhat.com>
5935 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
5936 to all type variants.
5938 PR tree-optimization/84503
5939 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
5940 width as info->bitpos + info->bitsize - start.
5941 (merged_store_group::merge_overlapping): Simplify width computation.
5942 (check_no_overlap): New function.
5943 (imm_store_chain_info::try_coalesce_bswap): Compute expected
5944 start + width and last_order of the group, fail if check_no_overlap
5946 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
5947 to group if check_no_overlap fails.
5949 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5951 * config/rs6000/altivec.md: Delete contraint arguments to
5952 define_expand, define_split, and define_peephole2, and in
5953 define_insn_and_split if always unused.
5954 * config/rs6000/darwin.md: Ditto.
5955 * config/rs6000/dfp.md: Ditto.
5956 * config/rs6000/rs6000.md: Ditto.
5957 * config/rs6000/sync.md: Ditto.
5958 * config/rs6000/vector.md: Ditto.
5959 * config/rs6000/vsx.md: Ditto.
5961 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5963 * config/rs6000/altivec.md: Write output control strings as braced
5964 blocks instead of double-quoted strings.
5965 * config/rs6000/darwin.md: Ditto.
5966 * config/rs6000/rs6000.md: Ditto.
5967 * config/rs6000/vector.md: Ditto.
5968 * config/rs6000/vsx.md: Ditto.
5970 2018-02-21 Jason Merrill <jason@redhat.com>
5972 PR c++/84314 - ICE with templates and fastcall attribute.
5973 * attribs.c (build_type_attribute_qual_variant): Remove assert.
5975 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5977 * ipa-cp.c (determine_versionability): Fix comment typos.
5979 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5982 * ipa-cp.c (determine_versionability): Do not version functions caling
5985 2018-02-21 Martin Liska <mliska@suse.cz>
5988 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
5989 Add "native" as a possible value.
5990 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
5991 the macro when native cpu detection is available.
5993 2018-02-21 Martin Liska <mliska@suse.cz>
5996 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
5997 Add "native" as a possible value.
5998 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
5999 when native cpu detection is available.
6001 2018-02-21 Jakub Jelinek <jakub@redhat.com>
6002 Martin Sebor <msebor@redhat.com>
6004 PR tree-optimization/84478
6005 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
6007 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
6008 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
6009 support which is conservatively correct, for 2 only stay conservative
6010 for maxlen. Formatting and comment capitalization fixes. Add STRICT
6011 argument to the 2 argument get_range_strlen, adjust 6 arg
6012 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
6014 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
6015 (gimple_fold_builtin_strlen): Pass true as last argument to
6018 2018-02-20 Martin Sebor <msebor@redhat.com>
6021 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
6022 (builtin_memref::set_base_and_offset): Same. Handle inner references.
6023 (builtin_memref::builtin_memref): Factor out parts into
6024 set_base_and_offset and call it.
6026 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6029 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
6030 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
6031 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
6032 search at the associated MODE_INT.
6034 2018-02-20 Jeff Law <law@redhat.com>
6037 PR tree-optimization/81592
6039 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
6040 for range data rather than using global data.
6041 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
6042 range data rather than using global data.
6043 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
6044 pass it to children as needed.
6045 (struct directive::fmtresult): Similarly.
6046 (struct directive::set_width): Similarly.
6047 (struct directive::set_precision): Similarly.
6048 (format_integer, format_directive, parse_directive): Similarly.
6049 (format_none): Accept unnamed vr_values parameter.
6050 (format_percent, format_floating, format_character): Similarly.
6051 (format_string, format_plain): Similarly.
6052 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
6053 the EVRP range analyzer for range data rather than using global data.
6054 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
6055 gimple-ssa-evrp-analyze.h
6056 (class sprintf_dom_walker): Add after_dom_children member function.
6057 Add evrp_range_analyzer member.
6058 (sprintf_dom_walker::before_dom_children): Call into the EVRP
6059 range analyzer as needed.
6060 (sprintf_dom_walker::after_dom_children): New member function.
6061 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
6063 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6064 (evrp_range_analyzer::pop_to_marker): Likewise.
6066 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
6068 PR tree-optimization/84419
6069 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
6070 with the required type if its current type is compatible but
6073 2018-02-20 Jakub Jelinek <jakub@redhat.com>
6076 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
6077 after vectorization.
6079 2018-02-20 Martin Liska <mliska@suse.cz>
6082 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
6083 possible values if we don't have a hint.
6085 2018-02-20 Martin Liska <mliska@suse.cz>
6089 * final.c (shorten_branches): Build align_tab array with one
6091 * opts.c (finish_options): Add alignment option limit check.
6092 (MAX_CODE_ALIGN): Likewise.
6093 (MAX_CODE_ALIGN_VALUE): Likewise.
6094 * doc/invoke.texi: Document maximum allowed option value for
6095 all -falign-* options.
6097 2018-02-19 Jakub Jelinek <jakub@redhat.com>
6100 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
6101 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
6102 * var-tracking.c (emit_note_insn_var_location): Remove all references
6103 to NOTE_INSN_CALL_ARG_LOCATION.
6104 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
6105 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
6106 Use copy_rtx_if_shared.
6107 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
6108 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
6109 (dwarf2out_var_location): Remove handling of
6110 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
6112 * final.c (final_scan_insn): Remove all references to
6113 NOTE_INSN_CALL_ARG_LOCATION.
6114 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
6115 before dumping final insns.
6116 * except.c (emit_note_eh_region_end): Remove all references to
6117 NOTE_INSN_CALL_ARG_LOCATION.
6118 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
6119 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
6120 * config/arc/arc.c (hwloop_optimize): Likewise.
6121 * config/arm/arm.c (create_fix_barrier): Likewise.
6122 * config/s390/s390.c (s390_chunkify_start): Likewise.
6123 * config/sh/sh.c (find_barrier): Likewise.
6124 * config/i386/i386.c (rest_of_insert_endbranch,
6125 ix86_seh_fixup_eh_fallthru): Likewise.
6126 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
6127 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
6128 * config/frv/frv.c (frv_function_prologue): Likewise.
6129 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
6131 (note_outside_basic_block_p): Remove all references to
6132 NOTE_INSN_CALL_ARG_LOCATION.
6133 * gengtype.c (adjust_field_rtx_def): Likewise.
6134 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
6136 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
6137 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
6140 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
6143 PR tree-optimization/84452
6144 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
6145 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
6148 2018-02-19 Martin Liska <mliska@suse.cz>
6151 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
6153 2018-02-19 Martin Liska <mliska@suse.cz>
6154 Richard Sandiford <richard.sandiford@linaro.org>
6156 PR tree-optimization/82491
6157 * gimple-fold.c (get_base_constructor): Make earlier bail out
6160 2018-02-19 Carl Love <cel@us.ibm.com>
6162 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
6163 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
6165 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
6166 P8V_BUILTIN_VEC_NEG.
6168 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
6170 * config/rl78/rl78.md (movdf): New define expand.
6172 2018-02-19 Martin Liska <mliska@suse.cz>
6175 * doc/invoke.texi: Fix typo.
6176 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
6178 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
6180 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
6181 handle rs6000_single_float and rs6000_double_float specially for
6184 2018-02-16 Jeff Law <law@redhat.com>
6186 * config/rx/rx.c (add_pop_cfi_notes): New function.;
6189 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6192 * ipa-inline.c (inline_small_functions): Fix a typo.
6194 2018-02-16 Nathan Sidwell <nathan@acm.org>
6196 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
6198 2018-02-16 Carl Love <cel@us.ibm.com>
6200 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
6201 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
6202 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
6203 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
6204 expansion to P8V_BUILTIN_VEC_FLOAT2.
6206 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
6208 PR rtl-optimization/70023
6209 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
6210 src_regno into account.
6212 2018-02-16 Carl Love <cel@us.ibm.com>
6214 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
6215 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
6216 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
6217 * config/rs6000/rs6000.c: Remove case statements for
6218 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
6219 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
6220 and P9V_BUILTIN_VEC_VINSERT4B.
6221 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
6222 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
6223 * config/rs6000/vsx.md:
6224 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
6227 2018-02-16 Carl Love <cel@us.ibm.com>
6229 * config/rs6000/altivec.h: Add builtin names vec_extract4b
6231 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
6233 * config/rs6000/rs6000-c.c: Add the definitions for
6234 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
6235 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
6236 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
6237 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
6238 definition for insert4b and define insn *insert3b_internal.
6239 * doc/extend.texi: Add documentation for vec_extract4b.
6241 2018-02-16 Nathan Sidwell <nathan@acm.org>
6243 * doc/extend.texi (Backwards Compatibility): Mention friend
6244 injection. Note for-scope is deprecated.
6245 * doc/invoke.texi (-ffriend-injection): Deprecate.
6247 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
6249 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
6250 that moved to I2, also allow destinations that are a paradoxical
6251 subreg (instead of a normal reg).
6253 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
6256 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
6259 2018-02-16 Richard Biener <rguenther@suse.de>
6261 PR tree-optimization/84037
6262 PR tree-optimization/84016
6264 * config/i386/i386.c (ix86_builtin_vectorization_cost):
6265 Adjust vec_construct for the fact we need additional higher latency
6266 128bit inserts for AVX256 and AVX512 vector builds.
6267 (ix86_add_stmt_cost): Scale vector construction cost for
6270 2018-02-16 Richard Biener <rguenther@suse.de>
6272 PR tree-optimization/84417
6273 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
6274 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
6275 (non_rewritable_lvalue_p): Likewise, use poly-ints.
6277 2018-02-16 Martin Liska <mliska@suse.cz>
6280 * internal-fn.def (ASAN_CHECK): Set proper flags.
6281 (ASAN_MARK): Likewise.
6283 2018-02-16 Julia Koval <julia.koval@intel.com>
6285 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
6286 from PTA_CANNONLAKE.
6288 2018-02-16 Jakub Jelinek <jakub@redhat.com>
6291 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
6292 Use ++iter rather than iter++ for std::list iterators.
6293 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
6294 defer deleting them until all nodes in the forest are processed. Do
6295 free even leaf nodes. Change to_process into auto_vec.
6298 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
6299 * vec.h (vec_default_construct): Use memset instead of placement new
6300 if BROKEN_VALUE_INITIALIZATION is defined.
6301 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
6302 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
6305 PR rtl-optimization/83723
6306 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
6307 * lra.c (lra_substitute_pseudo): Likewise. If true, use
6308 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
6310 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
6312 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
6314 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
6316 PR rtl-optimization/81443
6317 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
6318 from inner REGs to paradoxical SUBREGs.
6320 2018-02-16 Richard Biener <rguenther@suse.de>
6322 PR tree-optimization/84399
6323 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6324 For operands we can analyze at their definition make sure we can
6325 analyze them at each use as well.
6327 2018-02-16 Richard Biener <rguenther@suse.de>
6329 PR tree-optimization/84190
6330 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
6331 volatile accesses if the decl isn't volatile.
6333 2018-02-15 Jason Merrill <jason@redhat.com>
6335 PR c++/84314 - ICE with templates and fastcall attribute.
6336 * attribs.c (build_type_attribute_qual_variant): Don't clobber
6337 TYPE_CANONICAL on an existing type.
6339 2018-02-15 Jakub Jelinek <jakub@redhat.com>
6341 PR tree-optimization/84383
6342 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
6343 dstoff nor call operand_equal_p if dstbase is NULL.
6345 PR tree-optimization/84334
6346 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
6347 also a CONSTANT_CLASS_P, punt.
6349 2018-02-14 Jim Wilson <jimw@sifive.com>
6351 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
6352 first SMALL_OPERAND check. New local min_second_step. Move assert
6353 to where locals are set. Add TARGET_RVC support.
6354 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
6356 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
6358 * doc/invoke.texi: Correct -Wformat-overflow code sample.
6360 2018-02-14 Martin Sebor <msebor@redhat.com>
6362 PR tree-optimization/83698
6363 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
6364 arrays constrain the offset range to their bounds.
6365 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
6366 (builtin_access::overlap): Avoid setting the size of overlap if it's
6368 (maybe_diag_overlap): Also consider arrays when deciding what values
6369 of offsets to include in diagnostics.
6371 2018-02-14 Martin Sebor <msebor@redhat.com>
6374 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
6375 that correspond to the kind of a declaration.
6377 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
6380 * config/pa/pa.md: Load address of PIC label using the linkage table
6381 if the label is nonlocal.
6383 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
6385 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
6386 warning message if user requests -maltivec=be.
6387 * doc/invoke.texi: Document deprecation of -maltivec=be.
6389 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
6392 * config/rs6000/rs6000-c.c: Update definitions for
6393 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
6394 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
6396 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6399 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
6400 add _get_ssp intrinsics. Remove argument from
6401 __builtin_ia32_rdssp[d|q].
6402 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
6403 * config/i386/i386-builtin.def: Remove argument from
6404 __builtin_ia32_rdssp[d|q].
6405 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
6406 ix86_expand_special_args_builtin for _rdssp[d|q].
6407 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
6408 Clear register before usage.
6409 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6410 Add documentation for new _get_ssp and _inc_ssp intrinsics.
6412 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
6414 PR tree-optimization/84357
6415 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6416 operand 1 of an ARRAY_REF too.
6418 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
6421 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6422 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6424 (set_of_reg): New struct.
6425 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6426 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6427 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6429 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6430 Split into bitclr, bitset, bitinvert patterns if appropriate.
6431 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6432 use rx_fuse_in_memory_bitop.
6433 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6434 to named insn, correct maximum insn length.
6436 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6439 * machmode.def: Define a complex mode for PARTIAL_INT.
6440 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6442 * doc/rtl.texi: Document CSPImode.
6443 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6445 (msp430_hard_regno_nregs_with_padding): Likewise.
6447 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
6450 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6452 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
6454 PR rtl-optimization/84169
6455 * combine.c (try_combine): New variable split_i2i3. Set it to true if
6456 we generated a parallel as new i3 and we split that to new i2 and i3
6457 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
6458 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
6459 those to i2, not i1. Partially rewrite this scan code.
6461 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6464 * stor-layout.c (place_field): For variable length fields, adjust
6465 offset_align afterwards not just based on the field's alignment,
6466 but also on the size.
6469 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6470 of exps and logs in the use_exp2 case.
6472 2018-02-13 Jeff Law <law@redhat.com>
6474 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6477 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6480 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
6484 * common.opt (gas-loc-support, gas-locview-support): New.
6485 (ginline-points, ginternal-reset-location-views): New.
6486 * doc/invoke.texi: Document them. Use @itemx where intended.
6487 (gvariable-location-views): Adjust.
6488 * target.def (reset_location_view): New.
6489 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6490 (TARGET_RESET_LOCATION_VIEW): New.
6491 * doc/tm.texi: Rebuilt.
6492 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6493 (dwarf2out_default_as_locview_support): New.
6494 (output_asm_line_debug_info): Use option variables.
6495 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6496 (output_loc_list): Likewise.
6497 (add_high_low_attributes): Check option variables.
6498 Don't output entry view attribute in strict mode.
6499 (gen_inlined_subroutine_die): Check option variables.
6500 (dwarf2out_inline_entry): Likewise.
6501 (init_sections_and_labels): Likewise.
6502 (dwarf2out_early_finish): Likewise.
6503 (maybe_reset_location_view): New, from...
6504 (dwarf2out_var_location): ... here. Call it.
6505 * debug.h (dwarf2out_default_as_loc_support): Declare.
6506 (dwarf2out_default_as_locview_support): Declare.
6507 * hooks.c (hook_int_rtx_insn_0): New.
6508 * hooks.h (hook_int_rtx_insn_0): Declare.
6509 * toplev.c (process_options): Take -gas-loc-support and
6510 -gas-locview-support from dwarf2out. Enable
6511 -gvariable-location-views by default only with locview
6512 assembler support. Enable -ginternal-reset-location-views by
6513 default only if the target defines the corresponding hook.
6514 Enable -ginline-points by default if location views are
6515 enabled; force it disabled if statement frontiers are
6517 * tree-inline.c (expand_call_inline): Check option variables.
6518 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6520 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
6522 PR tree-optimization/84321
6523 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6524 handling. Also check whether the anti-range contains any values
6525 that satisfy the mask; switch to a VR_RANGE if not.
6527 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
6530 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6532 2018-02-13 Martin Jambor <mjambor@suse.cz>
6535 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6536 of call statements, also set location of a load to a temporary.
6538 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
6540 * config/rl78/rl78.c (add_vector_labels): New function.
6541 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6542 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6543 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6544 which checks that no arguments are passed.
6545 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6546 * doc/extend.texi: Documentation for the new attribute.
6548 2018-02-13 Andreas Schwab <schwab@suse.de>
6550 * config/riscv/linux.h (CPP_SPEC): Define.
6552 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6555 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6556 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6557 OPTION_MASK_ISA_AES as first argument to def_builtin_const
6558 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6559 instead of OPTION_MASK_ISA_PCLMUL as first argument to
6560 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6561 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6562 temporarily for AES and PCLMUL builtins.
6564 PR tree-optimization/84339
6565 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6566 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6570 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6571 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6572 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6574 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6576 * omp-simd-clone.h: New file.
6577 * omp-simd-clone.c: Include omp-simd-clone.h.
6578 (expand_simd_clones): No longer static.
6579 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6580 cgraph.h and omp-simd-clone.h.
6581 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6582 (vect_recog_widen_shift_pattern): Formatting fix.
6583 (vect_pattern_recog_1): Don't check optab for calls.
6586 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6587 operands[2] into a REG before using gen_lowpart on it.
6589 2018-02-12 Jeff Law <law@redhat.com>
6592 * config/sh/sh.c (find_barrier): Consider a sibling call
6595 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6596 successfully back substituting a reg.
6598 2018-02-12 Richard Biener <rguenther@suse.de>
6600 PR tree-optimization/84037
6601 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6602 parameter, move visited init to caller.
6603 (vect_slp_analyze_operations): Separate cost from validity
6604 check, initialize visited once for all instances.
6605 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6607 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6609 (vect_model_promotion_demotion_cost): Likewise.
6610 (vectorizable_bswap): Guard cost modeling with !slp_node
6611 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6613 (vectorizable_call): Likewise.
6614 (vectorizable_conversion): Likewise.
6615 (vectorizable_assignment): Likewise.
6616 (vectorizable_shift): Likewise.
6617 (vectorizable_operation): Likewise.
6618 (vectorizable_store): Likewise.
6619 (vectorizable_load): Likewise.
6620 (vectorizable_condition): Likewise.
6621 (vectorizable_comparison): Likewise.
6623 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6626 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6627 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6628 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6630 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6633 * params.def (inline-min-speedup): Increase from 8 to 15.
6634 (max-inline-insns-auto): Decrease from 40 to 30.
6635 * ipa-split.c (consider_split): Add some buffer for function to
6636 be considered inlining candidate.
6637 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6640 2018-02-12 Richard Biener <rguenther@suse.de>
6642 PR tree-optimization/84037
6643 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6644 matched stmts if we cannot swap the non-matched ones.
6646 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6648 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6649 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6650 _mm_maskz_scalef_round_ss): New intrinsics.
6651 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6652 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6653 __builtin_ia32_scalefss_round): Remove.
6654 (__builtin_ia32_scalefsd_mask_round,
6655 __builtin_ia32_scalefss_mask_round): New intrinsics.
6656 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6657 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6658 ((match_operand:VF_128 2 "<round_nimm_predicate>"
6659 "<round_constraint>")): Changed to ...
6660 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
6661 "<round_scalar_constraint>")): ... this.
6662 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
6663 %0, %1, %2<round_op3>}"): Changed to ...
6664 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
6665 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
6666 %2<round_scalar_mask_op3>}"): ... this.
6667 * config/i386/subst.md (round_scalar_nimm_predicate): New.
6669 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6671 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
6672 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
6673 (_mm_maskz_sqrt_round_ss): New intrinsics.
6674 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
6675 (__builtin_ia32_sqrtsd_mask_round)
6676 (__builtin_ia32_sqrtss_mask_round): New builtins.
6677 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6678 (__builtin_ia32_sqrtss_round): Remove.
6679 (__builtin_ia32_sqrtsd_mask_round)
6680 (__builtin_ia32_sqrtss_mask_round): New builtins.
6681 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6682 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6683 ((match_operand:VF_128 1 "vector_operand"
6684 "xBm,<round_constraint>")): Changed to ...
6685 ((match_operand:VF_128 1 "vector_operand"
6686 "xBm,<round_scalar_constraint>")): ... this.
6687 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6688 %0, %2, %<iptr>1<round_op3>}): Changed to ...
6689 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6690 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6691 %<iptr>1<round_scalar_mask_op3>}): ... this.
6692 ((set_attr "prefix" "<round_prefix>")): Changed to ...
6693 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6695 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
6698 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6699 Cast vec_cmpeq result to correct type.
6700 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
6701 Cast vec_cmpgt result to correct type.
6703 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
6705 * final.c (final_scan_insn_1): Renamed from...
6706 (final_scan_insn): ... this. New wrapper, to recover
6707 seen from the outermost call in recursive ones.
6708 * config/sparc/sparc.c (output_return): Drop seen from call.
6709 (output_sibcall): Likewise.
6710 * config/visium/visium.c (output_branch): Likewise.
6712 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
6714 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
6717 2018-02-10 Alan Modra <amodra@gmail.com>
6720 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
6721 Specify LR as an input.
6723 2018-02-10 Jakub Jelinek <jakub@redhat.com>
6726 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
6727 remove_member_access_dummy_vars): New functions.
6728 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
6729 lower_omp_1, execute_lower_omp): Use them.
6731 PR rtl-optimization/84308
6732 * shrink-wrap.c (spread_components): Release todo vector.
6734 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
6736 PR rtl-optimization/57193
6737 * ira-color.c (struct allocno_color_data): Add member
6738 conflict_allocno_hard_prefs.
6739 (update_conflict_allocno_hard_prefs): New.
6740 (bucket_allocno_compare_func): Add a preference based on
6741 conflict_allocno_hard_prefs.
6742 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
6743 (color_allocnos): Remove a dead code. Initiate
6744 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
6746 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6749 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
6750 constraint from =wa to wa. Avoid a subreg on the output operand,
6751 instead use a pseudo and subreg it in a move.
6752 (p9_xxbrd_<mode>): Changed to ...
6753 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
6754 (p9_xxbrd_v2df): New expander.
6755 (p9_xxbrw_<mode>): Changed to ...
6756 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
6757 (p9_xxbrw_v4sf): New expander.
6759 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6761 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
6763 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
6766 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
6767 multiply in 32-bit mode.
6768 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
6769 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
6772 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6774 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
6775 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
6776 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
6777 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
6779 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
6782 * dwarf2out.c (is_trivial_indirect_ref): New function.
6783 (dwarf2out_late_global_decl): Do not generate a location
6784 attribute for variables that have a non-trivial DECL_VALUE_EXPR
6785 and that are not defined in the current unit.
6787 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6789 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
6790 instead of a libcall for UNORDERED.
6792 2018-02-09 Tamar Christina <tamar.christina@arm.com>
6795 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
6796 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
6798 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6801 * config/s390/s390.c (s390_set_current_function): Invoke
6802 s390_indirect_branch_settings also if fndecl didn't change.
6804 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6806 * config/rs6000/rs6000.md (blockage): Set length to zero.
6808 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6810 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
6812 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6815 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
6816 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
6820 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
6821 PARALLEL incoming that failed vt_get_decl_and_offset check.
6824 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
6825 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
6827 (get_variable_section): For decls in named .bss* sections pass true as
6828 second argument to bss_initializer_p.
6830 2018-02-09 Marek Polacek <polacek@redhat.com>
6831 Jakub Jelinek <jakub@redhat.com>
6834 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
6835 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
6836 Sync some changes from cxx_fold_indirect_ref.
6838 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6840 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
6842 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
6843 (BLOCK_INLINE_ENTRY_LABEL): New.
6844 (dwarf2out_var_location): Disregard inline entry markers.
6845 (inline_entry_data): New struct.
6846 (inline_entry_data_hasher): New hashtable type.
6847 (inline_entry_data_hasher::hash): New.
6848 (inline_entry_data_hasher::equal): New.
6849 (inline_entry_data_table): New variable.
6850 (add_high_low_attributes): Add DW_AT_entry_pc and
6851 DW_AT_GNU_entry_view attributes if a pending entry is found
6852 in inline_entry_data_table. Add old entry_pc attribute only
6853 if debug nonbinding markers are disabled.
6854 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
6855 markers are enabled.
6856 (block_within_block_p, dwarf2out_inline_entry): New.
6857 (dwarf2out_finish): Check that no entries remained in
6858 inline_entry_data_table.
6859 * final.c (reemit_insn_block_notes): Handle inline entry notes.
6860 (final_scan_insn, notice_source_line): Likewise.
6861 (rest_of_clean_state): Skip inline entry markers.
6862 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
6864 * gimple.c (gimple_build_debug_inline_entry): New.
6865 * gimple.h (enum gimple_debug_subcode): Add
6866 GIMPLE_DEBUG_INLINE_ENTRY.
6867 (gimple_build_debug_inline_entry): Declare.
6868 (gimple_debug_inline_entry_p): New.
6869 (gimple_debug_nonbind_marker_p): Adjust.
6870 * insn-notes.def (INLINE_ENTRY): New.
6871 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
6872 inline entry marker notes.
6873 (print_insn): Likewise.
6874 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
6875 (INSN_DEBUG_MARKER_KIND): Likewise.
6876 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
6877 * tree-inline.c (expand_call_inline): Build and insert
6878 debug_inline_entry stmt.
6879 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
6880 inline entry blocks early, if nonbind markers are enabled.
6881 (dump_scope_block): Dump fragment info.
6882 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
6883 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
6884 (gimple_build_debug_inline_entry): New.
6885 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
6886 Enable/disable inline entry points too.
6887 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
6888 (DEBUG_INSN): Describe inline entry markers.
6890 * common.opt (gvariable-location-views): New.
6891 (gvariable-location-views=incompat5): New.
6892 * config.in: Rebuilt.
6893 * configure: Rebuilt.
6894 * configure.ac: Test assembler for view support.
6895 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
6896 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
6897 * dwarf2out.c (var_loc_view): New typedef.
6898 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
6899 (dwarf2out_locviews_in_attribute): New.
6900 (dwarf2out_locviews_in_loclist): New.
6901 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
6902 (enum dw_line_info_opcode): Add LI_adv_address.
6903 (struct dw_line_info_table): Add view.
6904 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
6905 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
6906 (zero_view_p): New variable.
6907 (ZERO_VIEW_P): New macro.
6908 (output_asm_line_debug_info): New.
6909 (struct var_loc_node): Add view.
6910 (add_AT_view_list, AT_loc_list): New.
6911 (add_var_loc_to_decl): Add view param. Test it against last.
6912 (new_loc_list): Add view params. Record them.
6913 (AT_loc_list_ptr): Handle loc and view lists.
6914 (view_list_to_loc_list_val_node): New.
6915 (print_dw_val): Handle dw_val_class_view_list.
6916 (size_of_die): Likewise.
6917 (value_format): Likewise.
6918 (loc_list_has_views): New.
6919 (gen_llsym): Set vl_symbol too.
6920 (maybe_gen_llsym, skip_loc_list_entry): New.
6921 (dwarf2out_maybe_output_loclist_view_pair): New.
6922 (output_loc_list): Output view list or entries too.
6923 (output_view_list_offset): New.
6924 (output_die): Handle dw_val_class_view_list.
6925 (output_dwarf_version): New.
6926 (output_compilation_unit_header): Use it.
6927 (output_skeleton_debug_sections): Likewise.
6928 (output_rnglists, output_line_info): Likewise.
6929 (output_pubnames, output_aranges): Update version comments.
6930 (output_one_line_info_table): Output view numbers in asm comments.
6931 (dw_loc_list): Determine current endview, pass it to new_loc_list.
6932 Call maybe_gen_llsym.
6933 (loc_list_from_tree_1): Adjust.
6934 (add_AT_location_description): Create view list attribute if
6935 needed, check it's absent otherwise.
6936 (convert_cfa_to_fb_loc_list): Adjust.
6937 (maybe_emit_file): Call output_asm_line_debug_info for test.
6938 (dwarf2out_var_location): Reset views as needed. Precompute
6939 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
6940 attribute. Set view.
6941 (new_line_info_table): Reset next view.
6942 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
6943 (dwarf2out_source_line): Likewise. Output view resets and labels to
6944 the assembler, or select appropriate line info opcodes.
6945 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
6946 (optimize_string_length): Catch it. Adjust.
6947 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
6948 dw_val_class_view_list, and remove it if no longer needed.
6949 (hash_loc_list): Hash view numbers.
6950 (loc_list_hasher::equal): Compare them.
6951 (optimize_location_lists): Check whether a view list symbol is
6952 needed, and whether the locview attribute is present, and
6953 whether they match. Remove the locview attribute if no longer
6955 (index_location_lists): Call skip_loc_list_entry for test.
6956 (dwarf2out_finish): Call output_asm_line_debug_info for test.
6957 Use output_dwarf_version.
6958 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
6959 (struct dw_val_node): Add val_view_list.
6960 * final.c (SEEN_NEXT_VIEW): New.
6961 (set_next_view_needed): New.
6962 (clear_next_view_needed): New.
6963 (maybe_output_next_view): New.
6964 (final_start_function): Rename to...
6965 (final_start_function_1): ... this. Take pointer to FIRST,
6966 add SEEN parameter. Emit param bindings in the initial view.
6967 (final_start_function): Reintroduce SEEN-less interface.
6968 (final): Rename to...
6969 (final_1): ... this. Take SEEN parameter. Output final pending
6970 next view at the end.
6971 (final): Reintroduce seen-less interface.
6972 (final_scan_insn): Output pending next view before switching
6973 sections or ending a block. Mark the next view as needed when
6974 outputting variable locations. Notify debug backend of section
6975 changes, and of location view changes.
6976 (rest_of_handle_final): Adjust.
6977 * toplev.c (process_options): Autodetect value for debug variable
6978 location views option. Warn on incompat5 without -gdwarf-5.
6979 * doc/invoke.texi (gvariable-location-views): New.
6980 (gvariable-location-views=incompat5): New.
6981 (gno-variable-location-views): New.
6983 2018-02-08 David Malcolm <dmalcolm@redhat.com>
6985 PR tree-optimization/84136
6986 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
6987 that the result of find_edge is non-NULL.
6989 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6992 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
6993 storing integer register in SImode. Fix cost of 256 and 512
6994 byte aligned SSE register store.
6996 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6998 * config/i386/i386.c (ix86_multiplication_cost): Fix
6999 multiplication cost for TARGET_AVX512DQ.
7001 2018-02-08 Marek Polacek <polacek@redhat.com>
7003 PR tree-optimization/84238
7004 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
7007 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7009 PR tree-optimization/84265
7010 * tree-vect-stmts.c (vectorizable_store): Don't treat
7011 VMAT_CONTIGUOUS accesses as grouped.
7012 (vectorizable_load): Likewise.
7014 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7016 PR tree-optimization/81635
7017 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
7018 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
7019 (test_round_for_mask): New functions.
7020 (wide_int_cc_tests): Call test_round_for_mask.
7021 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
7022 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
7023 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
7024 range returned by get_range_info.
7026 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7029 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
7030 * symtab.c: Include builtins.h
7031 (symtab_node::output_to_lto_symbol_table_p): Move here
7032 from lto-streamer-out.c:output_symbol_p.
7033 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
7034 (output_symbol_p): Move all logic to symtab.c
7035 (produce_symtab): Update.
7037 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7039 * config/s390/s390-opts.h (enum indirect_branch): Define.
7040 * config/s390/s390-protos.h (s390_return_addr_from_memory)
7041 (s390_indirect_branch_via_thunk)
7042 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
7043 (enum s390_indirect_branch_type): Define.
7044 * config/s390/s390.c (struct s390_frame_layout, struct
7045 machine_function): Remove.
7046 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
7047 (indirect_branch_table_label_no, indirect_branch_table_name):
7049 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
7050 (enum s390_indirect_branch_option): Define.
7051 (s390_return_addr_from_memory): New function.
7052 (s390_handle_string_attribute): New function.
7053 (s390_attribute_table): Add new attribute handler.
7054 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
7055 (s390_indirect_branch_via_thunk): New function.
7056 (s390_indirect_branch_via_inline_thunk): New function.
7057 (s390_function_ok_for_sibcall): When jumping via thunk disallow
7058 sibling call optimization for non z10 compiles.
7059 (s390_emit_call): Force indirect branch target to be a single
7060 register. Add r1 clobber for non-z10 compiles.
7061 (s390_emit_epilogue): Emit return jump via return_use expander.
7062 (s390_reorg): Handle JUMP_INSNs as execute targets.
7063 (s390_option_override_internal): Perform validity checks for the
7064 new command line options.
7065 (s390_indirect_branch_attrvalue): New function.
7066 (s390_indirect_branch_settings): New function.
7067 (s390_set_current_function): Invoke s390_indirect_branch_settings.
7068 (s390_output_indirect_thunk_function): New function.
7069 (s390_code_end): Implement target hook.
7070 (s390_case_values_threshold): Implement target hook.
7071 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
7073 * config/s390/s390.h (struct s390_frame_layout)
7074 (struct machine_function): Move here from s390.c.
7075 (TARGET_INDIRECT_BRANCH_NOBP_RET)
7076 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
7077 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
7078 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
7079 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
7080 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
7081 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
7082 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
7083 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
7084 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
7085 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
7086 (mnemonic attribute): Add values which aren't recognized
7088 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
7089 pattern for branch conversion. Fix mnemonic attribute.
7090 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
7091 indirect branch via thunk if requested.
7092 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
7093 ("*indirect_jump"): Disable for branch conversion using out of
7095 ("indirect_jump_via_thunk<mode>_z10")
7096 ("indirect_jump_via_thunk<mode>")
7097 ("indirect_jump_via_inlinethunk<mode>_z10")
7098 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
7099 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
7100 ("casesi_jump_via_inlinethunk<mode>_z10")
7101 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
7102 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
7103 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
7104 ("*indirect2_jump"): Disable for branch conversion.
7105 ("casesi_jump"): Turn into expander and expand patterns for branch
7107 ("return_use"): New expander.
7108 ("*return"): Emit return via thunk and rename it to ...
7109 ("*return<mode>"): ... this one.
7110 * config/s390/s390.opt: Add new options and and enum for the
7113 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
7115 * lra-constraints.c (match_reload): Unconditionally use
7116 gen_lowpart_SUBREG, rather than selecting between that
7117 and equivalent gen_rtx_SUBREG code.
7119 2018-02-08 Richard Biener <rguenther@suse.de>
7121 PR tree-optimization/84233
7122 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
7123 changed flag instead of boguously re-using phi_inserted.
7125 2018-02-08 Martin Jambor <mjambor@suse.cz>
7127 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
7128 static local variables.
7130 2018-02-08 Richard Biener <rguenther@suse.de>
7132 PR tree-optimization/84278
7133 * tree-vect-stmts.c (vectorizable_store): When looking for
7134 smaller vector types to perform grouped strided loads/stores
7135 make sure the mode is supported by the target.
7136 (vectorizable_load): Likewise.
7138 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7140 * config/aarch64/aarch64.c (aarch64_components_for_bb):
7141 Increase LDP/STP opportunities by adding adjacent callee-saves.
7143 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
7145 PR rtl-optimization/84068
7146 PR rtl-optimization/83459
7147 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
7149 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
7151 PR tree-optimization/84224
7152 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
7153 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
7156 2018-02-07 Iain Sandoe <iain@codesourcery.com>
7159 * config/rs6000/altivec.md (*restore_world): Remove LR use.
7160 * config/rs6000/predicates.md (restore_world_operation): Adjust op
7161 count, remove one USE.
7163 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
7165 * doc/install.texi (Configuration): Document the
7166 --with-long-double-format={ibm,ieee} PowerPC configuration
7170 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
7171 Convert from define_expand to be define_insn_and_split. Rework
7172 float/double/_Float128 conversions to QI/HI/SImode to work with
7173 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
7174 conversions to QI/HImode types did a store and then a load to
7175 truncate the value. For conversions to VSX registers, don't split
7176 the insn, instead emit the code directly. Use the code iterator
7177 any_fix to combine signed and unsigned conversions.
7178 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
7179 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
7180 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7181 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
7182 (fix_<mode>di2_hw): Likewise.
7183 (fixuns_<mode>di2_hw): Likewise.
7184 (fix_<mode>si2_hw): Likewise.
7185 (fixuns_<mode>si2_hw): Likewise.
7186 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
7187 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
7188 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
7189 fix<uns>_trunc<SFDF:mode>si2_p8.
7190 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
7192 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
7193 (fix<uns>_<mode>_mem): Likewise.
7194 (fctiw<u>z_<mode>_mem): Likewise.
7195 (fix<uns>_<mode>_mem): Likewise.
7196 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
7197 the register allocator from doing a direct move to the GPRs to do
7198 a store, and instead use the ISA 3.0 store byte/half-word from
7199 vector register instruction. For IEEE 128-bit floating point,
7200 also optimize stores of 32-bit ints.
7201 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
7203 2018-02-07 Alan Hayward <alan.hayward@arm.com>
7205 * genextract.c (push_pathstr_operand): New function to support
7207 (walk_rtx): Call push_pathstr_operand.
7208 (print_path): Support [a-zA-Z].
7210 2018-02-07 Richard Biener <rguenther@suse.de>
7212 PR tree-optimization/84037
7213 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
7214 (cse_and_gimplify_to_preheader): Declare.
7215 (vect_get_place_in_interleaving_chain): Likewise.
7216 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7218 (_loop_vec_info::~_loop_vec_info): Delete it.
7219 (cse_and_gimplify_to_preheader): New function.
7220 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
7221 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
7222 (vectorizable_load): Likewise. For grouped stores always base
7223 the IV on the first element.
7224 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
7225 condition before gimplifying.
7227 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7229 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
7230 *DIV_EXPR and *MOD_EXPR.
7232 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
7235 * config/i386/i386.c (ix86_option_override_internal): Mask out
7236 the CF_SET bit when checking -fcf-protection.
7238 2018-02-07 Tom de Vries <tom@codesourcery.com>
7241 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
7244 2018-02-07 Richard Biener <rguenther@suse.de>
7246 PR tree-optimization/84204
7247 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
7250 PR tree-optimization/84205
7251 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
7252 special-case isl_ast_op_zdiv_r.
7254 PR tree-optimization/84223
7255 * graphite-scop-detection.c (gather_bbs::before_dom_children):
7256 Only add conditions from within the region.
7257 (gather_bbs::after_dom_children): Adjust.
7259 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
7262 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
7263 * config/avr/avr.md: Only post-reload split REG-REG moves if
7264 either register is GENERAL_REG_P.
7266 2018-02-07 Jakub Jelinek <jakub@redhat.com>
7268 PR tree-optimization/84235
7269 * tree-ssa-scopedtables.c
7270 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
7271 if the subtraction is performed in floating point type where NaNs are
7272 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
7273 build 1. Formatting fix.
7275 2018-02-06 Jakub Jelinek <jakub@redhat.com>
7278 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
7279 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
7280 and skip it regardless of bb boundaries. Use CALL_P macro,
7281 don't test INSN_P (insn) together with CALL_P or JUMP_P check
7282 unnecessarily, formatting fix.
7284 2018-02-06 Michael Collison <michael.collison@arm.com>
7286 * config/arm/thumb2.md:
7287 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
7288 (*thumb_mov_notscc): Ditto.
7290 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
7293 * config/rs6000/rs6000.md (su code attribute): Use "u" for
7294 unsigned_fix, not "s".
7296 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7298 * configure.ac (gcc_fn_eh_frame_ro): New function.
7299 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
7300 correct .eh_frame permissions.
7301 * configure: Regenerate.
7303 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
7305 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
7308 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7310 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7311 Display warning message for -mno-speculate-indirect-jumps.
7313 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
7315 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
7317 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
7319 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
7321 PR tree-optimization/84225
7322 * tree-eh.c (find_trapping_overflow): Only call
7323 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
7325 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7328 * config/i386/i386.c: Reimplement the check of possible options
7329 -mibt/-mshstk conbination. Change error messages.
7330 * doc/invoke.texi: Fix a typo: remove extra '='.
7332 2018-02-06 Marek Polacek <polacek@redhat.com>
7334 PR tree-optimization/84228
7335 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
7337 2018-02-06 Tamar Christina <tamar.christina@arm.com>
7340 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
7341 emitted arch directives.
7342 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
7343 __ARM_FEATURE_COPROC before changing architectures.
7345 2018-02-06 Richard Biener <rguenther@suse.de>
7347 * config/i386/i386.c (print_reg): Fix typo.
7348 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
7350 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
7352 * configure: Regenerate.
7354 2018-02-05 Martin Sebor <msebor@redhat.com>
7356 PR tree-optimization/83369
7357 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
7360 2018-02-05 Martin Liska <mliska@suse.cz>
7362 * doc/invoke.texi: Cherry-pick upstream r323995.
7364 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
7366 * ira.c (ira_init_register_move_cost): Adjust comment.
7368 2018-02-05 Martin Liska <mliska@suse.cz>
7370 PR gcov-profile/84137
7371 * doc/gcov.texi: Fix typo in documentation.
7373 2018-02-05 Martin Liska <mliska@suse.cz>
7375 PR gcov-profile/83879
7376 * doc/gcov.texi: Document necessity of --dynamic-list-data when
7377 using dlopen functionality.
7379 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
7381 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
7382 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
7383 _mm_maskz_range_ss, _mm_mask_range_round_ss,
7384 _mm_maskz_range_round_ss): New intrinsics.
7385 (__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/i386-builtin.def (__builtin_ia32_rangesd128_round,
7390 __builtin_ia32_rangess128_round): Remove.
7391 (__builtin_ia32_rangesd128_mask_round,
7392 __builtin_ia32_rangess128_mask_round): New builtins.
7393 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
7394 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
7395 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
7396 "<round_saeonly_constraint>")): Changed to ...
7397 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
7398 "<round_saeonly_scalar_constraint>")): ... this.
7399 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
7400 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
7401 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
7402 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
7403 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
7405 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
7407 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
7409 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7410 Remove all values except native, 8540 and 8548.
7412 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
7414 * config/i386/i386.c (ix86_output_function_return): Pass
7415 INVALID_REGNUM, instead of -1, as invalid register number to
7416 indirect_thunk_name and output_indirect_thunk.
7418 2018-02-02 Julia Koval <julia.koval@intel.com>
7420 * config.gcc: Add -march=icelake.
7421 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7422 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7423 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7424 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7425 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7426 (processor_target_table): Add icelake.
7427 (ix86_option_override_internal): Handle new PTAs.
7428 (get_builtin_code_for_version): Handle icelake.
7429 (M_INTEL_COREI7_ICELAKE): New.
7430 (fold_builtin_cpu): Handle icelake.
7431 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7432 * doc/invoke.texi: Add -march=icelake.
7434 2018-02-02 Julia Koval <julia.koval@intel.com>
7436 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7437 to wide_int_bitmask.
7438 * wide-int-bitmask.h: New.
7440 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7443 * config/i386/i386.md: Replace Pmode with word_mode in
7444 builtin_setjmp_setup and builtin_longjmp to support x32.
7446 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
7450 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7452 (rs6000_supported_cpu_names): New static variable.
7453 (linux_cpu_translation_table): Likewise.
7454 (elf_platform) <cpu>: Define new static variable and use it.
7455 Translate kernel AT_PLATFORM name to canonical name if needed.
7456 Error if platform name is unknown.
7458 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
7461 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7463 2018-02-01 Jeff Law <law@redhat.com>
7466 * config/i386/i386.c (release_scratch_register_on_entry): Add new
7467 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
7468 the scratch if RELEASE_VIA_POP is false.
7469 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7470 If we have to save a temporary register, decrement SIZE appropriately.
7471 Pass new arguments to release_scratch_register_on_entry.
7472 (ix86_adjust_stack_and_probe): Likewise.
7473 (ix86_emit_probe_stack_range): Pass new arguments to
7474 release_scratch_register_on_entry.
7476 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
7478 PR rtl-optimization/84157
7479 * combine.c (change_zero_ext): Use REG_P predicate in
7480 front of HARD_REGISTER_P predicate.
7482 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
7484 * config/avr/avr.c (avr_option_override): Move disabling of
7485 -fdelete-null-pointer-checks to...
7486 * common/config/avr/avr-common.c (avr_option_optimization_table):
7489 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7491 PR tree-optimization/81635
7492 * tree-data-ref.c (split_constant_offset_1): For types that
7493 wrap on overflow, try to use range info to prove that wrapping
7496 2018-02-01 Renlin Li <renlin.li@arm.com>
7499 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7501 (aarch64_register_move_cost): Likewise.
7502 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7504 (REG_CLASS_NAMES): Likewise.
7505 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7506 TAILCALL_ADDR_REGS. Remove IP registers.
7507 * config/aarch64/aarch64.md (Ucs): Update register constraint.
7509 2018-02-01 Richard Biener <rguenther@suse.de>
7511 * domwalk.h (dom_walker::dom_walker): Add additional constructor
7512 for specifying RPO order and allow NULL for that.
7513 * domwalk.c (dom_walker::dom_walker): Likewise.
7514 (dom_walker::walk): Handle NULL RPO order.
7515 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7517 (rewrite_update_dom_walker): Likewise.
7518 (mark_def_dom_walker): Likewise.
7520 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7522 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7523 (aarch64_maybe_expand_sve_subreg_move): Declare.
7524 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7525 * config/aarch64/predicates.md (aarch64_any_register_operand): New
7527 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7528 that are semantically a reverse operation.
7529 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7530 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7531 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7533 (aarch64_can_change_mode_class): For big-endian, forbid changes
7534 between two SVE modes if they have different element sizes.
7536 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7538 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7539 the TImode handling for big-endian targets.
7541 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7543 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7544 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
7546 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7547 Remove BSWAP handing for big-endian targets and use the form of
7548 LD1RQ appropariate for the mode.
7550 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7552 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7553 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7556 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7559 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7560 check for operands that need to go through aarch64_sve_reload_be.
7562 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7564 PR tree-optimization/81661
7565 PR tree-optimization/84117
7566 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7567 * tree-eh.c: Include gimplify.h.
7568 (find_trapping_overflow, replace_trapping_overflow,
7569 rewrite_to_non_trapping_overflow): New functions.
7570 * tree-vect-loop.c: Include tree-eh.h.
7571 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7572 * tree-data-ref.c: Include tree-eh.h.
7573 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7575 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7577 PR rtl-optimization/84123
7578 * combine.c (change_zero_ext): Check if hard register satisfies
7579 can_change_dest_mode before calling gen_lowpart_SUBREG.
7581 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7584 * ira.c (ira_init_register_move_cost): Remove assert.
7586 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7588 PR rtl-optimization/84071
7589 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7590 * doc/tm.texi: Regenerate.
7592 2018-01-31 Richard Biener <rguenther@suse.de>
7594 PR tree-optimization/84132
7595 * tree-data-ref.c (analyze_miv_subscript): Properly
7596 check whether evolution_function_is_affine_multivariate_p
7597 before calling gcd_of_steps_may_divide_p.
7599 2018-01-31 Julia Koval <julia.koval@intel.com>
7602 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7603 * config/i386/i386.md (rdpid_rex64) New.
7604 (rdpid): Make 32bit only.
7606 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7609 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7610 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7612 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7615 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7617 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7619 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7621 PR rtl-optimization/84071
7622 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7623 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7625 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7627 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7628 (arc_attribute_table): Add 'aux' attribute.
7629 (arc_in_small_data_p): Consider aux like variables.
7630 (arc_is_aux_reg_p): New function.
7631 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7632 (arc_get_aux_arg): New function.
7633 (prepare_move_operands): Handle aux-register access.
7634 (arc_handle_aux_attribute): New function.
7635 * doc/extend.texi (ARC Variable attributes): Add subsection.
7637 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7639 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7640 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7641 (arc_attribute_table): Add 'uncached' attribute.
7642 (arc_print_operand): Print '.di' flag for uncached memory
7644 (arc_in_small_data_p): Do not consider for small data the uncached
7646 (arc_is_uncached_mem_p): New function.
7647 * config/arc/predicates.md (compact_store_memory_operand): Check
7648 for uncached memory accesses.
7649 (nonvol_nonimm_operand): Likewise.
7650 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7652 2018-01-31 Jakub Jelinek <jakub@redhat.com>
7655 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7656 falign-loops=): Add Optimization flag.
7658 2018-01-30 Jeff Law <law@redhat.com>
7661 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
7662 INT_REGISTERS_SAVED. Check it prior to calling
7663 get_scratch_register_on_entry.
7664 (ix86_adjust_stack_and_probe): Similarly.
7665 (ix86_emit_probe_stack_range): Similarly.
7666 (ix86_expand_prologue): Corresponding changes.
7668 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7671 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
7672 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
7674 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
7677 * lra-constraints.c (curr_insn_transform): Process AND in the
7680 2018-01-30 Jakub Jelinek <jakub@redhat.com>
7682 PR rtl-optimization/83986
7683 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7684 dependence against last_pending_memory_flush in addition to
7687 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
7689 PR tree-optimization/81611
7690 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7693 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7696 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7699 2018-01-30 Richard Biener <rguenther@suse.de>
7700 Jakub Jelinek <jakub@redhat.com>
7702 PR tree-optimization/84111
7703 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
7704 inner loops added during recursion, as they don't have up-to-date
7707 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7710 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
7711 (can_inline_edge_by_limits_p): ... here.
7712 (can_early_inline_edge_p, check_callers,
7713 update_caller_keys, update_callee_keys, recursive_inlining,
7714 add_new_edges_to_heap, speculation_useful_p,
7715 inline_small_functions,
7716 inline_small_functions, flatten_function,
7717 inline_to_all_callers_1): Update.
7719 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7721 * profile-count.c (profile_count::combine_with_ipa_count): Handle
7724 2018-01-30 Richard Biener <rguenther@suse.de>
7726 PR tree-optimization/83008
7727 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
7728 invariant and constant vector uses in stmts when they need
7731 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7734 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
7735 * configure: Regenerate.
7737 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7739 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
7741 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
7742 Use gen_rtx_REG rather than gen_lowpart.
7744 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7746 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
7747 rather than 0 when creating partial subregs.
7749 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7751 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
7754 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
7757 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
7758 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
7759 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
7760 flags. This restores the settings used before the 2017-07-24.
7761 Turning off pre increment/decrement/modify allows IVOPTS to
7762 optimize DF/SF loops where the index is an int.
7764 2018-01-29 Richard Biener <rguenther@suse.de>
7765 Kelvin Nilsen <kelvin@gcc.gnu.org>
7768 * tree-vect-stmts.c (vectorizable_call): Don't call
7769 targetm.vectorize_builtin_md_vectorized_function if callee is
7772 2018-01-22 Carl Love <cel@us.ibm.com>
7774 * doc/extend.tex: Fix typo in second arg in
7775 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
7777 2018-01-29 Richard Biener <rguenther@suse.de>
7779 PR tree-optimization/84086
7780 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
7781 (flush_ssaname_freelist): When SSA names were released reset
7782 the SCEV hash table.
7784 2018-01-29 Richard Biener <rguenther@suse.de>
7786 PR tree-optimization/84057
7787 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
7788 removed paths when removing edges.
7790 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
7792 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
7793 -mfunction-return=@var{choice}.
7795 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
7798 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
7800 (layout::print_source_line): Likewise.
7801 (test_get_line_width_without_trailing_whitespace): Add test cases.
7803 2018-01-27 Jakub Jelinek <jakub@redhat.com>
7806 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
7809 2018-01-26 Jim Wilson <jimw@sifive.com>
7811 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
7813 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
7816 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7818 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
7819 and CMP + SUB-immediate -> SUBS.
7821 2018-01-26 Martin Sebor <msebor@redhat.com>
7823 PR tree-optimization/83896
7824 * tree-ssa-strlen.c (get_string_len): Rename...
7825 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
7826 Avoid assuming length is constant.
7827 (handle_char_store): Use HOST_WIDE_INT for string length.
7829 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
7832 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
7833 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
7835 2018-01-26 Richard Biener <rguenther@suse.de>
7837 PR rtl-optimization/84003
7838 * dse.c (record_store): Only record redundant stores when
7839 the earlier store aliases at least all accesses the later one does.
7841 2018-01-26 Jakub Jelinek <jakub@redhat.com>
7843 PR rtl-optimization/83985
7844 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
7845 REG_CFA_RESTORE insns.
7846 (delete_unmarked_insns): Don't ignore separate shrink wrapping
7847 REG_CFA_RESTORE insns here.
7850 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
7851 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
7853 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7855 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
7856 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
7857 (arc_init): Likewise.
7858 (arc_override_options): Likewise.
7859 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
7861 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
7863 * config/arc/arc.h (TARGET_DBNZ): Define.
7864 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
7865 properly set the tune attribute.
7866 (dbnz): Use TARGET_DBNZ guard.
7867 * config/arc/arc.opt (mtune): Add core3 option.
7869 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7871 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
7872 recognize new pic like addresses.
7873 (arc_delegitimize_address): Clean up.
7875 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7877 * config/arc/arc-arches.def: Option mrf16 valid for all
7879 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
7880 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
7881 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
7882 * config/arc/arc-tables.opt: Regenerate.
7883 * config/arc/arc.c (arc_conditional_register_usage): Handle
7884 reduced register file case.
7885 (arc_file_start): Set must have build attributes.
7886 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
7888 * config/arc/arc.opt (mrf16): Add new option.
7889 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
7890 * config/arc/genmultilib.awk: Handle new mrf16 option.
7891 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
7892 * config/arc/t-multilib: Regenerate.
7893 * doc/invoke.texi (ARC Options): Document mrf16 option.
7895 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7897 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
7898 * config/arc/arc.c (arc_handle_secure_attribute): New function.
7899 (arc_attribute_table): Add 'secure_call' attribute.
7900 (arc_print_operand): Print secure call operand.
7901 (arc_function_ok_for_sibcall): Don't optimize tail calls when
7903 (arc_is_secure_call_p): New function. * config/arc/arc.md
7904 (call_i): Add support for sjli instruction.
7905 (call_value_i): Likewise.
7906 * config/arc/constraints.md (Csc): New constraint.
7908 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7909 John Eric Martin <John.Martin@emmicro-us.com>
7911 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
7912 * config/arc/arc.c (_arc_jli_section): New struct.
7913 (arc_jli_section): New type.
7914 (rc_jli_sections): New static variable.
7915 (arc_handle_jli_attribute): New function.
7916 (arc_attribute_table): Add jli_always and jli_fixed attribute.
7917 (arc_file_end): New function.
7918 (TARGET_ASM_FILE_END): Define.
7919 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
7920 (arc_add_jli_section): New function.
7921 (jli_call_scan): Likewise.
7922 (arc_reorg): Call jli_call_scan.
7923 (arc_output_addsi): Remove 'S' from printing asm operand.
7924 (arc_is_jli_call_p): New function.
7925 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
7927 (movhi_insn): Likewise.
7928 (movsi_insn): Likewise.
7929 (movsi_set_cc_insn): Likewise.
7930 (loadqi_update): Likewise.
7931 (load_zeroextendqisi_update): Likewise.
7932 (load_signextendqisi_update): Likewise.
7933 (loadhi_update): Likewise.
7934 (load_zeroextendhisi_update): Likewise.
7935 (load_signextendhisi_update): Likewise.
7936 (loadsi_update): Likewise.
7937 (loadsf_update): Likewise.
7938 (movsicc_insn): Likewise.
7939 (bset_insn): Likewise.
7940 (bxor_insn): Likewise.
7941 (bclr_insn): Likewise.
7942 (bmsk_insn): Likewise.
7943 (bicsi3_insn): Likewise.
7944 (cmpsi_cc_c_insn): Likewise.
7945 (movsi_ne): Likewise.
7946 (movsi_cond_exec): Likewise.
7947 (clrsbsi2): Likewise.
7958 (call_i): Remove 'S' asm letter, add jli instruction.
7959 (call_value_i): Likewise.
7960 * config/arc/arc.op (mjli-always): New option.
7961 * config/arc/constraints.md (Cji): New constraint.
7962 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
7964 (subsf3_fpx): Likewise.
7965 (mulsf3_fpx): Likewise.
7966 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
7968 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
7970 * doc/invoke.texi (ARC): Document mjli-always option.
7972 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
7974 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
7975 avoid addition with 0 and use incw and decw where possible.
7977 2018-01-26 Richard Biener <rguenther@suse.de>
7979 PR tree-optimization/81082
7980 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
7981 association if it requires casting to unsigned.
7982 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
7983 from fold_plusminus_mult_expr to catch important cases late when
7984 range info is available.
7986 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7988 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
7989 * configure.ac (hidden_linkonce): New test.
7990 * configure: Regenerate.
7991 * config.in: Regenerate.
7993 2018-01-26 Julia Koval <julia.koval@intel.com>
7995 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
7996 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
7997 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
7998 _mm_mask_bitshuffle_epi64_mask): Fix type.
7999 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
8000 USI_FTYPE_V4DI_V4DI_USI): Remove.
8001 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
8002 __builtin_ia32_vpshufbitqmb256_mask,
8003 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
8004 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
8005 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
8007 2018-01-26 Alan Modra <amodra@gmail.com>
8010 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
8011 UNSPEC_VBPERMQ. Sort other unspecs.
8013 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
8015 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
8017 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
8020 * predict.c (drop_profile): Do not push/pop cfun; update also
8022 (handle_missing_profiles): Fix logic looking for zero profiles.
8024 2018-01-25 Jakub Jelinek <jakub@redhat.com>
8027 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
8028 on functions with #pragma omp declare simd or functions with simd
8030 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
8031 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
8032 Remove trailing \n from warning_at calls.
8034 2018-01-25 Tom de Vries <tom@codesourcery.com>
8037 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8038 for neutered workers.
8040 2018-01-24 Joseph Myers <joseph@codesourcery.com>
8043 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
8044 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
8046 2018-01-24 Jeff Law <law@redhat.com>
8049 * i386.c (get_probe_interval): Move to earlier point.
8050 (ix86_compute_frame_layout): If -fstack-clash-protection and
8051 the frame is larger than the probe interval, then use pushes
8052 to save registers rather than reg->mem moves.
8053 (ix86_expand_prologue): Remove conditional for int_registers_saved
8056 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
8059 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
8060 min/max for never referenced object.
8062 2018-01-24 Jakub Jelinek <jakub@redhat.com>
8065 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
8067 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
8068 attributes from DECL_ATTRIBUTES (decl) without affecting
8069 DECL_ATTRIBUTES (current_function_decl).
8070 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
8071 functions with non-NULL DECL_ABSTRACT_ORIGIN.
8073 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
8075 PR tree-optimization/83979
8076 * fold-const.c (fold_comparison): Use constant_boolean_node
8077 instead of boolean_{true,false}_node.
8079 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
8081 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
8084 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8086 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8087 Simplify the clause that sets the length attribute.
8088 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8089 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
8090 clause that sets the length attribute.
8091 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8093 2018-01-24 Tom de Vries <tom@codesourcery.com>
8096 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
8097 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
8098 Add strict parameter.
8099 (prevent_branch_around_nothing): Insert dummy insn between branch to
8100 label and label with no ptx insn inbetween.
8101 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
8103 2018-01-24 Tom de Vries <tom@codesourcery.com>
8106 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
8107 for neutered threads in warp.
8108 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
8110 2018-01-24 Richard Biener <rguenther@suse.de>
8112 PR tree-optimization/83176
8113 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
8116 2018-01-24 Richard Biener <rguenther@suse.de>
8118 PR tree-optimization/82819
8119 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
8120 code generating pluses that are no-ops in the target precision.
8122 2018-01-24 Richard Biener <rguenther@suse.de>
8125 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
8127 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8129 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
8130 to merge probabilities.
8131 * predict.c (probably_never_executed): Also mark as cold functions
8132 with global 0 profile and guessed local profile.
8133 * profile-count.c (profile_probability::combine_with_count): New
8135 * profile-count.h (profile_probability::operator*,
8136 profile_probability::operator*=, profile_probability::operator/,
8137 profile_probability::operator/=): Reduce precision to adjusted
8138 and set value to guessed on contradictory divisions.
8139 (profile_probability::combine_with_freq): Remove.
8140 (profile_probability::combine_wiht_count): Declare.
8141 (profile_count::force_nonzero):: Set to adjusted.
8142 (profile_count::probability_in):: Set quality to adjusted.
8143 * tree-ssa-tail-merge.c (replace_block_by): Use
8146 2018-01-23 Andrew Waterman <andrew@sifive.com>
8147 Jim Wilson <jimw@sifive.com>
8149 * config/riscv/riscv.c (riscv_stack_boundary): New.
8150 (riscv_option_override): Set riscv_stack_boundary. Handle
8151 riscv_preferred_stack_boundary_arg.
8152 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
8153 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
8154 (STACK_BOUNDARY): Set to riscv_stack_boundary.
8155 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
8156 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
8157 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
8159 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
8162 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
8163 of struct ix86_frame.
8164 (ix86_expand_epilogue): Likewise. Add a local variable for
8165 the reg_save_offset field in struct ix86_frame.
8167 2018-01-23 Bin Cheng <bin.cheng@arm.com>
8169 PR tree-optimization/82604
8170 * tree-loop-distribution.c (enum partition_kind): New enum item
8171 PKIND_PARTIAL_MEMSET.
8172 (partition_builtin_p): Support above new enum item.
8173 (generate_code_for_partition): Ditto.
8174 (compute_access_range): Differentiate cases that equality can be
8175 proven at all loops, the innermost loops or no loops.
8176 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
8177 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
8178 (finalize_partitions, distribute_loop): Don't fuse partition of
8179 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
8180 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
8183 2018-01-23 Martin Liska <mliska@suse.cz>
8185 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
8186 order to ignore the predictor.
8187 (PRED_POLYMORPHIC_CALL): Likewise.
8188 (PRED_RECURSIVE_CALL): Likewise.
8190 2018-01-23 Martin Liska <mliska@suse.cz>
8192 * tree-profile.c (tree_profiling): Print function header to
8193 aware reader which function we are working on.
8194 * value-prof.c (gimple_find_values_to_profile): Do not print
8195 not interesting value histograms.
8197 2018-01-23 Martin Liska <mliska@suse.cz>
8199 * profile-count.h (enum profile_quality): Add
8200 profile_uninitialized as the first value. Do not number values
8201 as they are zero based.
8202 (profile_count::verify): Update sanity check.
8203 (profile_probability::verify): Likewise.
8205 2018-01-23 Nathan Sidwell <nathan@acm.org>
8207 * doc/invoke.texi (ffor-scope): Deprecate.
8209 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8211 PR tree-optimization/83510
8212 * domwalk.c (set_all_edges_as_executable): New function.
8213 (dom_walker::dom_walker): Convert bool param
8214 "skip_unreachable_blocks" to enum reachability. Move setup of
8215 edge flags to set_all_edges_as_executable and only do it when
8216 reachability is REACHABLE_BLOCKS.
8217 * domwalk.h (enum dom_walker::reachability): New enum.
8218 (dom_walker::dom_walker): Convert bool param
8219 "skip_unreachable_blocks" to enum reachability.
8220 (set_all_edges_as_executable): New decl.
8221 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
8222 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
8224 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
8225 but converting true to REACHABLE_BLOCKS.
8226 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
8228 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
8229 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
8230 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
8232 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
8233 if check_all_array_refs will be called.
8235 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8237 * tree.c (selftest::test_location_wrappers): Add more test
8240 2018-01-23 David Malcolm <dmalcolm@redhat.com>
8242 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
8243 (selftest::test_bit_in_range): Likewise.
8245 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8248 * doc/sourcebuild.texi (vect_float): Say that the selector
8249 only describes the situation when -funsafe-math-optimizations is on.
8250 (vect_float_strict): Document.
8252 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
8254 PR tree-optimization/83965
8255 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
8256 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
8257 instead of checking only for a reduction.
8258 (vect_recog_widen_sum_pattern): Likewise.
8260 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8262 * predict.c (probably_never_executed): Only use precise profile info.
8263 (compute_function_frequency): Skip after inlining hack since we now
8264 have quality checking.
8266 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
8268 * profile-count.h (profile_probability::very_unlikely,
8269 profile_probability::unlikely, profile_probability::even): Set
8270 precision to guessed.
8272 2018-01-23 Richard Biener <rguenther@suse.de>
8274 PR tree-optimization/83963
8275 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
8276 Properly terminate dominator walk when crossing the exit edge not
8277 when visiting its source block.
8279 2018-01-23 Jakub Jelinek <jakub@redhat.com>
8282 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
8283 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
8285 2018-01-22 Jakub Jelinek <jakub@redhat.com>
8287 PR tree-optimization/83957
8288 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
8289 semicolon after for body surrounded by braces.
8291 PR tree-optimization/83081
8292 * profile-count.h (profile_probability::split): New method.
8293 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
8294 Use profile_probability::split.
8295 (do_compare_rtx_and_jump): Fix adjustment of probabilities
8296 when splitting a single conditional jump into 2.
8298 2018-01-22 David Malcolm <dmalcolm@redhat.com>
8300 PR tree-optimization/69452
8301 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
8304 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8306 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
8307 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
8308 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
8310 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8312 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
8313 * config/rl78/rl78.md (movdi): New define_expand.
8314 * config/rl78/rl78.c (rl78_split_movdi): New function.
8316 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
8319 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
8321 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
8322 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
8323 128-bit to produce an UNSPEC move to get the double word with the
8324 signbit and then a shift directly to do signbit.
8325 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
8326 implementation with a new version that just does either a direct
8327 move or a regular move. Move memory interface to separate insns.
8328 Move insns so they are next to the expander.
8329 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
8330 with signbit move. Split big and little endian case.
8331 (signbit<mode>2_dm_mem_le): Likewise.
8332 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
8333 (signbit<mode>2_dm2): Likewise.
8335 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8337 * config/rl78/rl78.md (anddi3): New define_expand.
8339 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8341 * config/rl78/rl78.md (umindi3): New define_expand.
8343 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8345 * config/rl78/rl78.md (smindi3): New define_expand.
8347 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8349 * config/rl78/rl78.md (smaxdi3): New define_expand.
8351 2018-01-22 Carl Love <cel@us.ibm.com>
8353 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
8354 LVX_V1TI): Add macro expansion.
8355 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
8356 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
8357 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
8358 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
8359 Change check to determine if the instruction is a byte reversing
8360 entry. Fix typo in comment.
8361 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
8362 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
8363 Add def_builtin calls for new builtins.
8364 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
8365 Add define_insn expansion.
8367 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8369 * config/rl78/rl78.md (umaxdi3): New define_expand.
8371 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8373 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
8374 for non-QImode registers.
8376 2018-01-22 Richard Biener <rguenther@suse.de>
8378 PR tree-optimization/83963
8379 * graphite-scop-detection.c (scop_detection::get_sese): Delay
8380 including the loop exit block.
8381 (scop_detection::merge_sese): Likewise.
8382 (scop_detection::add_scop): Do it here instead.
8384 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8386 * doc/sourcebuild.texi (arm_softfloat): Document.
8388 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
8391 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
8392 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
8393 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
8395 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8396 David Edelsohn <dje.gcc@gmail.com>
8399 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8400 Change "crset eq" to "crset 2".
8401 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8402 (*call_indirect_aix<mode>_nospec): Likewise.
8403 (*call_value_indirect_aix<mode>_nospec): Likewise.
8404 (*call_indirect_elfv2<mode>_nospec): Likewise.
8405 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
8406 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
8407 change assembly output from . to $.
8408 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8409 (indirect_jump<mode>_nospec): Change assembly output from . to $.
8410 (*tablejump<mode>_internal1_nospec): Likewise.
8412 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
8415 * config/sh/sh_optimize_sett_clrt.cc:
8416 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8418 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
8420 PR tree-optimization/83940
8421 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8422 offset_dt to vect_constant_def rather than vect_unknown_def_type.
8423 (vect_check_load_store_mask): Add a mask_dt_out parameter and
8424 use it to pass back the definition type.
8425 (vect_check_store_rhs): Likewise rhs_dt_out.
8426 (vect_build_gather_load_calls): Add a mask_dt argument and use
8427 it instead of a call to vect_is_simple_use.
8428 (vectorizable_store): Update calls to vect_check_load_store_mask
8429 and vect_check_store_rhs. Use the dt returned by the latter instead
8430 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
8431 instead of calls to vect_is_simple_use. Pass the scalar rather
8432 than the vector operand to vect_is_simple_use when handling
8433 second and subsequent copies of an rhs value.
8434 (vectorizable_load): Update calls to vect_check_load_store_mask
8435 and vect_build_gather_load_calls. Use the cached mask_dt and
8436 gs_info.offset_dt instead of calls to vect_is_simple_use.
8438 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8441 * tree-emutls.c: Include gimplify.h.
8442 (lower_emutls_2): New function.
8443 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8444 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8445 it before further processing.
8448 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8449 UINTVAL (trueop1) instead of INTVAL (op1).
8451 2018-01-19 Jakub Jelinek <jakub@redhat.com>
8455 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8456 INCOMING_FRAME_SP_OFFSET if not defined.
8457 (scan_trace): Add ENTRY argument. If true and
8458 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8459 emit a note to adjust the CFA offset.
8460 (create_cfi_notes): Adjust scan_trace callers.
8461 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8462 INCOMING_FRAME_SP_OFFSET in the CIE.
8463 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8464 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8466 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8467 * doc/tm.texi: Regenerated.
8469 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8471 PR rtl-optimization/83147
8472 * lra-constraints.c (remove_inheritance_pseudos): Use
8473 lra_substitute_pseudo_within_insn.
8475 2018-01-19 Tom de Vries <tom@codesourcery.com>
8476 Cesar Philippidis <cesar@codesourcery.com>
8479 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8481 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
8484 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8485 spaces for function labels.
8487 2018-01-19 Martin Liska <mliska@suse.cz>
8489 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8490 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8491 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8492 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8493 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8494 (PRED_CONST_RETURN): Change from 69 to 65.
8495 (PRED_NULL_RETURN): Change from 91 to 71.
8496 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8497 (PRED_LOOP_GUARD): Change from 66 to 73.
8499 2018-01-19 Martin Liska <mliska@suse.cz>
8501 * predict.c (predict_insn_def): Add new assert.
8502 (struct branch_predictor): Change type to signed integer.
8503 (test_prediction_value_range): Amend test to cover
8505 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8506 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8507 (PRED_LOOP_ITERATIONS_MAX): Likewise.
8508 (PRED_LOOP_IV_COMPARE): Likewise.
8509 * predict.h (PROB_UNINITIALIZED): Define new constant.
8511 2018-01-19 Martin Liska <mliska@suse.cz>
8513 * predict.c (dump_prediction): Add new format for
8514 analyze_brprob.py script which is enabled with -details
8516 * profile-count.h (precise_p): New function.
8518 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8520 PR tree-optimization/83922
8521 * tree-vect-loop.c (vect_verify_full_masking): Return false if
8522 there are no statements that need masking.
8523 (vect_active_double_reduction_p): New function.
8524 (vect_analyze_loop_operations): Use it when handling phis that
8525 are not in the loop header.
8527 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8529 PR tree-optimization/83914
8530 * tree-vect-loop.c (vectorizable_induction): Don't convert
8531 init_expr or apply the peeling adjustment for inductions
8532 that are nested within the vectorized loop.
8534 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8536 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8539 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8543 * function.h (gimplify_parameters): Add gimple_seq * argument.
8544 * function.c: Include gimple.h and options.h.
8545 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8546 for the added local temporaries if needed.
8547 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8548 if there are any parameter cleanups, wrap whole body into a
8549 try/finally with the cleanups.
8551 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8554 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8555 Use GET_MODE_CLASS for scalar floating point.
8557 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8561 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8562 Fix call of call_cgraph_insertion_hooks.
8564 2018-01-18 Martin Sebor <msebor@redhat.com>
8566 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8568 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8571 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8574 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8577 * common.opt: (-ffile-prefix-map): New option.
8578 * opts.c (common_handle_option): Defer it.
8579 * opts-global.c (handle_common_deferred_options): Handle it.
8580 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8581 * file-prefix-map.h: New file.
8582 (remap_debug_filename, add_debug_prefix_map): ...here.
8583 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8584 * final.c (debug_prefix_map, add_debug_prefix_map
8585 remap_debug_filename): Move to...
8586 * file-prefix-map.c: New file.
8587 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8588 generalize, get rid of alloca(), use strrchr() instead of strchr().
8589 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8590 Implement in terms of add_prefix_map().
8591 (remap_macro_filename, remap_debug_filename): Implement in term of
8593 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8594 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8595 * dbxout.c: Include file-prefix-map.h.
8596 * varasm.c: Likewise.
8597 * vmsdbgout.c: Likewise.
8598 * xcoffout.c: Likewise.
8599 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8600 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8601 * doc/invoke.texi (-ffile-prefix-map): Document.
8602 (-fdebug-prefix-map): Update description.
8604 2018-01-18 Martin Liska <mliska@suse.cz>
8606 * config/i386/i386.c (indirect_thunk_name): Document that also
8608 (output_indirect_thunk): Document why both instructions
8609 (pause and lfence) are generated.
8611 2018-01-18 Richard Biener <rguenther@suse.de>
8613 PR tree-optimization/83887
8614 * graphite-scop-detection.c
8615 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8616 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8617 (scop_detection::merge_sese): Re-implement with a flood-fill
8618 algorithm that properly finds a SESE region if it exists.
8620 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8623 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8624 pointer_diff optimizations use view_convert instead of convert.
8626 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8628 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8629 Generate different code for -mno-speculate-indirect-jumps.
8630 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8631 (*call_indirect_aix<mode>): Disable for
8632 -mno-speculate-indirect-jumps.
8633 (*call_indirect_aix<mode>_nospec): New define_insn.
8634 (*call_value_indirect_aix<mode>): Disable for
8635 -mno-speculate-indirect-jumps.
8636 (*call_value_indirect_aix<mode>_nospec): New define_insn.
8637 (*sibcall_nonlocal_sysv<mode>): Generate different code for
8638 -mno-speculate-indirect-jumps.
8639 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8641 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8643 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8644 long double type, set the flags for noting the default long double
8645 type, even if we don't pass or return a long double type.
8647 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
8650 * ipa-inline.c (flatten_function): Do not overwrite final inlining
8653 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
8655 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8656 support for merge[hl].
8657 (fold_mergehl_helper): New helper function.
8658 (tree-vector-builder.h): New #include for tree_vector_builder usage.
8659 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
8660 (altivec_vmrglw_direct): Add xxmrglw insn.
8662 2018-01-17 Andrew Waterman <andrew@sifive.com>
8664 * config/riscv/riscv.c (riscv_conditional_register_usage): If
8665 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
8667 2018-01-17 David Malcolm <dmalcolm@redhat.com>
8670 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
8671 call the lto_location_cache before reading the
8672 DECL_SOURCE_LOCATION of the types.
8674 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
8675 Richard Sandiford <richard.sandiford@linaro.org>
8677 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8678 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8679 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8680 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8681 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8683 * config/aarch64/constraints.md (aarch64_movti_operand):
8685 * config/aarch64/predicates.md (Uti): Add new constraint.
8687 2018-01-17 Carl Love <cel@us.ibm.com>
8689 * config/rs6000/vsx.md (define_expand xl_len_r,
8690 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8691 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8693 (define_expand, define_insn): Move the shift left from the
8694 define_insn to the define_expand for lxvl and stxvl instructions.
8695 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8696 and XL_LEN_R definitions to PURE.
8698 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
8700 * config/i386/i386.c (indirect_thunk_name): Declare regno
8701 as unsigned int. Compare regno with INVALID_REGNUM.
8702 (output_indirect_thunk): Ditto.
8703 (output_indirect_thunk_function): Ditto.
8704 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
8705 in the call to output_indirect_thunk_function.
8707 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
8710 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
8711 rather than the size of inner_type to determine the stack slot size
8712 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
8714 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
8717 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
8720 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
8722 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
8723 endian Linux systems to optionally enable multilibs for selecting
8724 the long double type if the user configured an explicit type.
8725 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
8726 have no long double multilibs if not defined.
8727 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
8728 warn if the user used -mabi={ieee,ibm}longdouble and we built
8729 multilibs for long double.
8730 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
8731 appropriate multilib option.
8732 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
8734 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
8735 for building long double multilibs.
8736 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
8738 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
8740 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
8743 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
8745 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
8748 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
8751 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
8754 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8756 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
8757 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
8759 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8761 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
8762 different rtl trees depending on TARGET_64BIT.
8763 (rs6000_gen_lvx): Likewise.
8765 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8767 * config/visium/visium.md (nop): Tweak comment.
8768 (hazard_nop): Likewise.
8770 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8772 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
8773 -mspeculate-indirect-jumps.
8774 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
8775 for -mno-speculate-indirect-jumps.
8776 (*call_indirect_elfv2<mode>_nospec): New define_insn.
8777 (*call_value_indirect_elfv2<mode>): Disable for
8778 -mno-speculate-indirect-jumps.
8779 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
8780 (indirect_jump): Emit different RTL for
8781 -mno-speculate-indirect-jumps.
8782 (*indirect_jump<mode>): Disable for
8783 -mno-speculate-indirect-jumps.
8784 (*indirect_jump<mode>_nospec): New define_insn.
8785 (tablejump): Emit different RTL for
8786 -mno-speculate-indirect-jumps.
8787 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
8788 (tablejumpsi_nospec): New define_expand.
8789 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
8790 (tablejumpdi_nospec): New define_expand.
8791 (*tablejump<mode>_internal1): Disable for
8792 -mno-speculate-indirect-jumps.
8793 (*tablejump<mode>_internal1_nospec): New define_insn.
8794 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
8797 2018-01-16 Artyom Skrobov tyomitch@gmail.com
8799 * caller-save.c (insert_save): Drop unnecessary parameter. All
8802 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8803 Richard Biener <rguenth@suse.de>
8806 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
8807 return early, inline manually is_gimple_sizepos. Make sure if we
8808 call gimplify_expr we don't end up with a gimple constant.
8809 * tree.c (variably_modified_type_p): Don't return true for
8810 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
8811 * gimplify.h (is_gimple_sizepos): Remove.
8813 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8815 PR tree-optimization/83857
8816 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
8817 vectorizable_live_operation for pure SLP statements.
8818 (vectorizable_live_operation): Handle PHIs.
8820 2018-01-16 Richard Biener <rguenther@suse.de>
8822 PR tree-optimization/83867
8823 * tree-vect-stmts.c (vect_transform_stmt): Precompute
8824 nested_in_vect_loop_p since the scalar stmt may get invalidated.
8826 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8829 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
8830 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
8831 If off is not INTEGER_CST, issue a may not be aligned warning
8832 rather than isn't aligned. Use isn%'t rather than isn't.
8833 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
8835 <case MULT_EXPR>: Improve the case when bottom and one of the
8836 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
8837 operand, in that case check if the other operand is multiple of
8838 bottom divided by the INTEGER_CST operand.
8840 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8843 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
8844 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
8845 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
8846 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
8847 * config/pa/pa.c (pa_function_arg_advance): Likewise.
8848 (pa_function_arg, pa_arg_partial_bytes): Likewise.
8849 (pa_function_arg_size): New function.
8851 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8853 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
8854 in a separate statement.
8856 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8858 PR tree-optimization/83847
8859 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
8860 group gathers and scatters.
8862 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8864 PR rtl-optimization/86620
8865 * params.def (max-sched-ready-insns): Bump minimum value to 1.
8867 PR rtl-optimization/83213
8868 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
8869 to last if both are JUMP_INSNs.
8871 PR tree-optimization/83843
8872 * gimple-ssa-store-merging.c
8873 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
8874 store_immediate_info for bswap/nop orig_stores.
8876 2018-01-15 Andrew Waterman <andrew@sifive.com>
8878 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
8880 <UDIV>: Increase cost if !TARGET_DIV.
8882 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
8884 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
8885 (define_attr "cr_logical_3op"): New.
8886 (cceq_ior_compare): Adjust.
8887 (cceq_ior_compare_complement): Adjust.
8888 (*cceq_rev_compare): Adjust.
8889 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
8890 (is_cracked_insn): Adjust.
8891 (insn_must_be_first_in_group): Adjust.
8892 * config/rs6000/40x.md: Adjust.
8893 * config/rs6000/440.md: Adjust.
8894 * config/rs6000/476.md: Adjust.
8895 * config/rs6000/601.md: Adjust.
8896 * config/rs6000/603.md: Adjust.
8897 * config/rs6000/6xx.md: Adjust.
8898 * config/rs6000/7450.md: Adjust.
8899 * config/rs6000/7xx.md: Adjust.
8900 * config/rs6000/8540.md: Adjust.
8901 * config/rs6000/cell.md: Adjust.
8902 * config/rs6000/e300c2c3.md: Adjust.
8903 * config/rs6000/e500mc.md: Adjust.
8904 * config/rs6000/e500mc64.md: Adjust.
8905 * config/rs6000/e5500.md: Adjust.
8906 * config/rs6000/e6500.md: Adjust.
8907 * config/rs6000/mpc.md: Adjust.
8908 * config/rs6000/power4.md: Adjust.
8909 * config/rs6000/power5.md: Adjust.
8910 * config/rs6000/power6.md: Adjust.
8911 * config/rs6000/power7.md: Adjust.
8912 * config/rs6000/power8.md: Adjust.
8913 * config/rs6000/power9.md: Adjust.
8914 * config/rs6000/rs64.md: Adjust.
8915 * config/rs6000/titan.md: Adjust.
8917 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8919 * config/i386/predicates.md (indirect_branch_operand): Rewrite
8920 ix86_indirect_branch_register logic.
8922 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8924 * config/i386/constraints.md (Bs): Update
8925 ix86_indirect_branch_register check. Don't check
8926 ix86_indirect_branch_register with GOT_memory_operand.
8928 * config/i386/predicates.md (GOT_memory_operand): Don't check
8929 ix86_indirect_branch_register here.
8930 (GOT32_symbol_operand): Likewise.
8932 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8934 * config/i386/predicates.md (constant_call_address_operand):
8935 Rewrite ix86_indirect_branch_register logic.
8936 (sibcall_insn_operand): Likewise.
8938 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8940 * config/i386/constraints.md (Bs): Replace
8941 ix86_indirect_branch_thunk_register with
8942 ix86_indirect_branch_register.
8944 * config/i386/i386.md (indirect_jump): Likewise.
8945 (tablejump): Likewise.
8946 (*sibcall_memory): Likewise.
8947 (*sibcall_value_memory): Likewise.
8948 Peepholes of indirect call and jump via memory: Likewise.
8949 * config/i386/i386.opt: Likewise.
8950 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8951 (GOT_memory_operand): Likewise.
8952 (call_insn_operand): Likewise.
8953 (sibcall_insn_operand): Likewise.
8954 (GOT32_symbol_operand): Likewise.
8956 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8959 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
8960 type rather than type addr's type points to.
8961 (expand_omp_atomic_mutex): Likewise.
8962 (expand_omp_atomic): Likewise.
8964 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8967 * config/i386/i386.c (output_indirect_thunk_function): Use
8968 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
8969 for __x86_return_thunk.
8971 2018-01-15 Richard Biener <rguenther@suse.de>
8974 * expmed.c (extract_bit_field_1): Fix typo.
8976 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8979 * config/arm/iterators.md (VF): New mode iterator.
8980 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
8981 Remove integer-related logic from pattern.
8982 (neon_vabd<mode>_3): Likewise.
8984 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8987 * common.opt (fstrict-overflow): No longer an alias.
8988 (fwrapv-pointer): New option.
8989 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
8990 also for pointer types based on flag_wrapv_pointer.
8991 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
8992 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
8993 opts->x_flag_wrapv got set.
8994 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
8995 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
8996 POINTER_TYPE_OVERFLOW_UNDEFINED.
8997 * match.pd: Likewise in address comparison pattern.
8998 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
9000 2018-01-15 Richard Biener <rguenther@suse.de>
9003 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
9004 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
9005 Reset type names to their identifier if their TYPE_DECL doesn't
9006 have linkage (and thus is used for ODR and devirt).
9007 (save_debug_info_for_decl): Remove.
9008 (save_debug_info_for_type): Likewise.
9009 (add_tree_to_fld_list): Adjust.
9010 * tree-pretty-print.c (dump_generic_node): Make dumping of
9011 type names more robust.
9013 2018-01-15 Richard Biener <rguenther@suse.de>
9015 * BASE-VER: Bump to 8.0.1.
9017 2018-01-14 Martin Sebor <msebor@redhat.com>
9020 * builtins.c (check_access): Avoid warning when the no-warning bit
9023 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
9025 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
9026 * ira-color (allocno_hard_regs_compare): Likewise.
9028 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
9031 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
9032 Use .pushsection/.popsection.
9034 2018-01-14 Martin Sebor <msebor@redhat.com>
9037 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
9039 2018-01-14 Jakub Jelinek <jakub@redhat.com>
9041 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
9042 entry from extra_headers.
9043 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
9044 extra_headers, make the list bitwise identical to the i?86-*-* one.
9046 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9048 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
9049 -mcmodel=large with -mindirect-branch=thunk,
9050 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
9051 -mfunction-return=thunk-extern.
9052 * doc/invoke.texi: Document -mcmodel=large is incompatible with
9053 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
9054 -mfunction-return=thunk and -mfunction-return=thunk-extern.
9056 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9058 * config/i386/i386.c (print_reg): Print the name of the full
9059 integer register without '%'.
9060 (ix86_print_operand): Handle 'V'.
9061 * doc/extend.texi: Document 'V' modifier.
9063 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9065 * config/i386/constraints.md (Bs): Disallow memory operand for
9066 -mindirect-branch-register.
9068 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9069 (GOT_memory_operand): Likewise.
9070 (call_insn_operand): Likewise.
9071 (sibcall_insn_operand): Likewise.
9072 (GOT32_symbol_operand): Likewise.
9073 * config/i386/i386.md (indirect_jump): Call convert_memory_address
9074 for -mindirect-branch-register.
9075 (tablejump): Likewise.
9076 (*sibcall_memory): Likewise.
9077 (*sibcall_value_memory): Likewise.
9078 Disallow peepholes of indirect call and jump via memory for
9079 -mindirect-branch-register.
9080 (*call_pop): Replace m with Bw.
9081 (*call_value_pop): Likewise.
9082 (*sibcall_pop_memory): Replace m with Bs.
9083 * config/i386/i386.opt (mindirect-branch-register): New option.
9084 * doc/invoke.texi: Document -mindirect-branch-register option.
9086 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9088 * config/i386/i386-protos.h (ix86_output_function_return): New.
9089 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
9090 set function_return_type.
9091 (indirect_thunk_name): Add ret_p to indicate thunk for function
9093 (output_indirect_thunk_function): Pass false to
9094 indirect_thunk_name.
9095 (ix86_output_indirect_branch_via_reg): Likewise.
9096 (ix86_output_indirect_branch_via_push): Likewise.
9097 (output_indirect_thunk_function): Create alias for function
9098 return thunk if regno < 0.
9099 (ix86_output_function_return): New function.
9100 (ix86_handle_fndecl_attribute): Handle function_return.
9101 (ix86_attribute_table): Add function_return.
9102 * config/i386/i386.h (machine_function): Add
9103 function_return_type.
9104 * config/i386/i386.md (simple_return_internal): Use
9105 ix86_output_function_return.
9106 (simple_return_internal_long): Likewise.
9107 * config/i386/i386.opt (mfunction-return=): New option.
9108 (indirect_branch): Mention -mfunction-return=.
9109 * doc/extend.texi: Document function_return function attribute.
9110 * doc/invoke.texi: Document -mfunction-return= option.
9112 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
9114 * config/i386/i386-opts.h (indirect_branch): New.
9115 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
9116 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
9117 with local indirect jump when converting indirect call and jump.
9118 (ix86_set_indirect_branch_type): New.
9119 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
9120 (indirectlabelno): New.
9121 (indirect_thunk_needed): Likewise.
9122 (indirect_thunk_bnd_needed): Likewise.
9123 (indirect_thunks_used): Likewise.
9124 (indirect_thunks_bnd_used): Likewise.
9125 (INDIRECT_LABEL): Likewise.
9126 (indirect_thunk_name): Likewise.
9127 (output_indirect_thunk): Likewise.
9128 (output_indirect_thunk_function): Likewise.
9129 (ix86_output_indirect_branch_via_reg): Likewise.
9130 (ix86_output_indirect_branch_via_push): Likewise.
9131 (ix86_output_indirect_branch): Likewise.
9132 (ix86_output_indirect_jmp): Likewise.
9133 (ix86_code_end): Call output_indirect_thunk_function if needed.
9134 (ix86_output_call_insn): Call ix86_output_indirect_branch if
9136 (ix86_handle_fndecl_attribute): Handle indirect_branch.
9137 (ix86_attribute_table): Add indirect_branch.
9138 * config/i386/i386.h (machine_function): Add indirect_branch_type
9139 and has_local_indirect_jump.
9140 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
9142 (tablejump): Likewise.
9143 (*indirect_jump): Use ix86_output_indirect_jmp.
9144 (*tablejump_1): Likewise.
9145 (simple_return_indirect_internal): Likewise.
9146 * config/i386/i386.opt (mindirect-branch=): New option.
9147 (indirect_branch): New.
9150 (thunk-inline): Likewise.
9151 (thunk-extern): Likewise.
9152 * doc/extend.texi: Document indirect_branch function attribute.
9153 * doc/invoke.texi: Document -mindirect-branch= option.
9155 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
9158 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
9160 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
9162 * ipa-inline.c (want_inline_small_function_p): Return false if
9163 inlining has already failed with CIF_FINAL_ERROR.
9164 (update_caller_keys): Call want_inline_small_function_p before
9166 (update_callee_keys): Likewise.
9168 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9170 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
9172 (rs6000_quadword_masked_address_p): Likewise.
9173 (quad_aligned_load_p): Likewise.
9174 (quad_aligned_store_p): Likewise.
9175 (const_load_sequence_p): Add comment to describe the outer-most loop.
9176 (mimic_memory_attributes_and_flags): New function.
9177 (rs6000_gen_stvx): Likewise.
9178 (replace_swapped_aligned_store): Likewise.
9179 (rs6000_gen_lvx): Likewise.
9180 (replace_swapped_aligned_load): Likewise.
9181 (replace_swapped_load_constant): Capitalize argument name in
9182 comment describing this function.
9183 (rs6000_analyze_swaps): Add a third pass to search for vector loads
9184 and stores that access quad-word aligned addresses and replace
9185 with stvx or lvx instructions when appropriate.
9186 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
9187 New function prototype.
9188 (rs6000_quadword_masked_address_p): Likewise.
9189 (rs6000_gen_lvx): Likewise.
9190 (rs6000_gen_stvx): Likewise.
9191 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
9192 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
9193 when memory address is aligned.
9194 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
9195 this split to select lvx instruction when memory address is aligned.
9196 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
9197 instruction when memory address is aligned.
9198 (*vsx_le_perm_load_v16qi): Likewise.
9199 (four unnamed splitters): Modify to select the stvx instruction
9200 when memory is aligned.
9202 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
9204 * predict.c (determine_unlikely_bbs): Handle correctly BBs
9205 which appears in the queue multiple times.
9207 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9208 Alan Hayward <alan.hayward@arm.com>
9209 David Sherwood <david.sherwood@arm.com>
9211 * tree-vectorizer.h (vec_lower_bound): New structure.
9212 (_loop_vec_info): Add check_nonzero and lower_bounds.
9213 (LOOP_VINFO_CHECK_NONZERO): New macro.
9214 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
9215 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
9216 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
9217 fields. Make seg_len the distance travelled, not including the
9219 (dr_direction_indicator): Declare.
9220 (dr_zero_step_indicator): Likewise.
9221 (dr_known_forward_stride_p): Likewise.
9222 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
9224 (runtime_alias_check_p): Allow runtime alias checks with
9226 (operator ==): Compare access_size and align.
9227 (prune_runtime_alias_test_list): Rework for new distinction between
9228 the access_size and seg_len.
9229 (create_intersect_range_checks_index): Likewise. Cope with polynomial
9231 (get_segment_min_max): New function.
9232 (create_intersect_range_checks): Use it.
9233 (dr_step_indicator): New function.
9234 (dr_direction_indicator): Likewise.
9235 (dr_zero_step_indicator): Likewise.
9236 (dr_known_forward_stride_p): Likewise.
9237 * tree-loop-distribution.c (data_ref_segment_size): Return
9238 DR_STEP * (niters - 1).
9239 (compute_alias_check_pairs): Update call to the dr_with_seg_len
9241 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
9242 (vect_preserves_scalar_order_p): New function, split out from...
9243 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
9244 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
9245 (vect_vfa_access_size): New function.
9246 (vect_vfa_align): Likewise.
9247 (vect_compile_time_alias): Take access_size_a and access_b arguments.
9248 (dump_lower_bound): New function.
9249 (vect_check_lower_bound): Likewise.
9250 (vect_small_gap_p): Likewise.
9251 (vectorizable_with_step_bound_p): Likewise.
9252 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
9253 depencies if the vectorization factor is 1. Convert the checks
9254 for nonzero steps into checks on the bounds of DR_STEP. Try using
9255 a bunds check for variable steps if the minimum required step is
9256 relatively small. Update calls to the dr_with_seg_len
9257 constructor and to vect_compile_time_alias.
9258 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
9260 (vect_loop_versioning): Call it.
9261 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
9263 (vect_estimate_min_profitable_iters): Account for any bounds checks.
9265 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9266 Alan Hayward <alan.hayward@arm.com>
9267 David Sherwood <david.sherwood@arm.com>
9269 * doc/sourcebuild.texi (vect_scatter_store): Document.
9270 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
9272 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
9274 * genopinit.c (main): Add supports_vec_scatter_store and
9275 supports_vec_scatter_store_cached to target_optabs.
9276 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
9277 IFN_MASK_SCATTER_STORE.
9278 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
9280 * internal-fn.h (internal_store_fn_p): Declare.
9281 (internal_fn_stored_value_index): Likewise.
9282 * internal-fn.c (scatter_store_direct): New macro.
9283 (expand_scatter_store_optab_fn): New function.
9284 (direct_scatter_store_optab_supported_p): New macro.
9285 (internal_store_fn_p): New function.
9286 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
9287 IFN_MASK_SCATTER_STORE.
9288 (internal_fn_mask_index): Likewise.
9289 (internal_fn_stored_value_index): New function.
9290 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
9292 * optabs-query.h (supports_vec_scatter_store_p): Declare.
9293 * optabs-query.c (supports_vec_scatter_store_p): New function.
9294 * tree-vectorizer.h (vect_get_store_rhs): Declare.
9295 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
9296 true for scatter stores.
9297 (vect_gather_scatter_fn_p): Handle scatter stores too.
9298 (vect_check_gather_scatter): Consider using scatter stores if
9299 supports_vec_scatter_store_p.
9300 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
9302 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9303 internal_fn_stored_value_index.
9304 (check_load_store_masking): Handle scatter stores too.
9305 (vect_get_store_rhs): Make public.
9306 (vectorizable_call): Use internal_store_fn_p.
9307 (vectorizable_store): Handle scatter store internal functions.
9308 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
9309 when deciding whether the end of the group has been reached.
9310 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
9311 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
9312 (mask_scatter_store<mode>): New insns.
9314 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9315 Alan Hayward <alan.hayward@arm.com>
9316 David Sherwood <david.sherwood@arm.com>
9318 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
9319 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
9320 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
9322 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
9323 Use vect_truncate_gather_scatter_offset if we can't treat the
9324 operation as a normal gather load or scatter store.
9325 (get_group_load_store_type): Take the gather_scatter_info
9326 as argument. Try using a gather load or scatter store for
9327 single-element groups.
9328 (get_load_store_type): Update calls to get_group_load_store_type
9329 and vect_use_strided_gather_scatters_p.
9331 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9332 Alan Hayward <alan.hayward@arm.com>
9333 David Sherwood <david.sherwood@arm.com>
9335 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
9336 optional tree argument.
9337 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
9339 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
9340 but continue to use the current value as a fallback.
9341 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
9342 to compare the updates.
9343 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
9344 (get_load_store_type): Use it when handling a strided access.
9345 (vect_get_strided_load_store_ops): New function.
9346 (vect_get_data_ptr_increment): Likewise.
9347 (vectorizable_load): Handle strided gather loads. Always pass
9348 a step to vect_create_data_ref_ptr and bump_vector_ptr.
9350 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9351 Alan Hayward <alan.hayward@arm.com>
9352 David Sherwood <david.sherwood@arm.com>
9354 * doc/md.texi (gather_load@var{m}): Document.
9355 (mask_gather_load@var{m}): Likewise.
9356 * genopinit.c (main): Add supports_vec_gather_load and
9357 supports_vec_gather_load_cached to target_optabs.
9358 * optabs-tree.c (init_tree_optimization_optabs): Use
9359 ggc_cleared_alloc to allocate target_optabs.
9360 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
9361 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
9363 * internal-fn.h (internal_load_fn_p): Declare.
9364 (internal_gather_scatter_fn_p): Likewise.
9365 (internal_fn_mask_index): Likewise.
9366 (internal_gather_scatter_fn_supported_p): Likewise.
9367 * internal-fn.c (gather_load_direct): New macro.
9368 (expand_gather_load_optab_fn): New function.
9369 (direct_gather_load_optab_supported_p): New macro.
9370 (direct_internal_fn_optab): New function.
9371 (internal_load_fn_p): Likewise.
9372 (internal_gather_scatter_fn_p): Likewise.
9373 (internal_fn_mask_index): Likewise.
9374 (internal_gather_scatter_fn_supported_p): Likewise.
9375 * optabs-query.c (supports_at_least_one_mode_p): New function.
9376 (supports_vec_gather_load_p): Likewise.
9377 * optabs-query.h (supports_vec_gather_load_p): Declare.
9378 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
9379 and memory_type field.
9380 (NUM_PATTERNS): Bump to 15.
9381 * tree-vect-data-refs.c: Include internal-fn.h.
9382 (vect_gather_scatter_fn_p): New function.
9383 (vect_describe_gather_scatter_call): Likewise.
9384 (vect_check_gather_scatter): Try using internal functions for
9385 gather loads. Recognize existing calls to a gather load function.
9386 (vect_analyze_data_refs): Consider using gather loads if
9387 supports_vec_gather_load_p.
9388 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
9389 (vect_get_gather_scatter_offset_type): Likewise.
9390 (vect_convert_mask_for_vectype): Likewise.
9391 (vect_add_conversion_to_patterm): Likewise.
9392 (vect_try_gather_scatter_pattern): Likewise.
9393 (vect_recog_gather_scatter_pattern): New pattern recognizer.
9394 (vect_vect_recog_func_ptrs): Add it.
9395 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9396 internal_fn_mask_index and internal_gather_scatter_fn_p.
9397 (check_load_store_masking): Take the gather_scatter_info as an
9398 argument and handle gather loads.
9399 (vect_get_gather_scatter_ops): New function.
9400 (vectorizable_call): Check internal_load_fn_p.
9401 (vectorizable_load): Likewise. Handle gather load internal
9403 (vectorizable_store): Update call to check_load_store_masking.
9404 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
9405 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
9406 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
9407 (aarch64_gather_scale_operand_d): New predicates.
9408 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
9409 (mask_gather_load<mode>): New insns.
9411 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9412 Alan Hayward <alan.hayward@arm.com>
9413 David Sherwood <david.sherwood@arm.com>
9415 * optabs.def (fold_left_plus_optab): New optab.
9416 * doc/md.texi (fold_left_plus_@var{m}): Document.
9417 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9418 * internal-fn.c (fold_left_direct): Define.
9419 (expand_fold_left_optab_fn): Likewise.
9420 (direct_fold_left_optab_supported_p): Likewise.
9421 * fold-const-call.c (fold_const_fold_left): New function.
9422 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9423 * tree-parloops.c (valid_reduction_p): New function.
9424 (gather_scalar_reductions): Use it.
9425 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9426 (vect_finish_replace_stmt): Declare.
9427 * tree-vect-loop.c (fold_left_reduction_fn): New function.
9428 (needs_fold_left_reduction_p): New function, split out from...
9429 (vect_is_simple_reduction): ...here. Accept reductions that
9430 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9431 (vect_force_simple_reduction): Also store the reduction type in
9432 the assignment's STMT_VINFO_REDUC_TYPE.
9433 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9434 (merge_with_identity): New function.
9435 (vect_expand_fold_left): Likewise.
9436 (vectorize_fold_left_reduction): Likewise.
9437 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9438 scalar phi in place for it. Check for target support and reject
9439 cases that would reassociate the operation. Defer the transform
9440 phase to vectorize_fold_left_reduction.
9441 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9442 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9443 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9445 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9447 * tree-if-conv.c (predicate_mem_writes): Remove redundant
9448 call to ifc_temp_var.
9450 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9451 Alan Hayward <alan.hayward@arm.com>
9452 David Sherwood <david.sherwood@arm.com>
9454 * target.def (legitimize_address_displacement): Take the original
9455 offset as a poly_int.
9456 * targhooks.h (default_legitimize_address_displacement): Update
9458 * targhooks.c (default_legitimize_address_displacement): Likewise.
9459 * doc/tm.texi: Regenerate.
9460 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9461 as an argument, moving assert of ad->disp == ad->disp_term to...
9462 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
9463 Try calling targetm.legitimize_address_displacement before expanding
9464 the address rather than afterwards, and adjust for the new interface.
9465 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9466 Match the new hook interface. Handle SVE addresses.
9467 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9470 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9472 * Makefile.in (OBJS): Add early-remat.o.
9473 * target.def (select_early_remat_modes): New hook.
9474 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9475 * doc/tm.texi: Regenerate.
9476 * targhooks.h (default_select_early_remat_modes): Declare.
9477 * targhooks.c (default_select_early_remat_modes): New function.
9478 * timevar.def (TV_EARLY_REMAT): New timevar.
9479 * passes.def (pass_early_remat): New pass.
9480 * tree-pass.h (make_pass_early_remat): Declare.
9481 * early-remat.c: New file.
9482 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9484 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
9486 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9487 Alan Hayward <alan.hayward@arm.com>
9488 David Sherwood <david.sherwood@arm.com>
9490 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9491 vfm1 with a bound_epilog parameter.
9492 (vect_do_peeling): Update calls accordingly, and move the prologue
9493 call earlier in the function. Treat the base bound_epilog as 0 for
9494 fully-masked loops and retain vf - 1 for other loops. Add 1 to
9495 this base when peeling for gaps.
9496 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9497 with fully-masked loops.
9498 (vect_estimate_min_profitable_iters): Handle the single peeled
9499 iteration in that case.
9501 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9502 Alan Hayward <alan.hayward@arm.com>
9503 David Sherwood <david.sherwood@arm.com>
9505 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9506 single-element interleaving even if the size is not a power of 2.
9507 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9508 accesses for single-element interleaving if the group size is
9511 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9512 Alan Hayward <alan.hayward@arm.com>
9513 David Sherwood <david.sherwood@arm.com>
9515 * doc/md.texi (fold_extract_last_@var{m}): Document.
9516 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9517 * optabs.def (fold_extract_last_optab): New optab.
9518 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9519 * internal-fn.c (fold_extract_direct): New macro.
9520 (expand_fold_extract_optab_fn): Likewise.
9521 (direct_fold_extract_optab_supported_p): Likewise.
9522 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9523 * tree-vect-loop.c (vect_model_reduction_cost): Handle
9524 EXTRACT_LAST_REDUCTION.
9525 (get_initial_def_for_reduction): Do not create an initial vector
9526 for EXTRACT_LAST_REDUCTION reductions.
9527 (vectorizable_reduction): Leave the scalar phi in place for
9528 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
9529 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
9530 epilogue code for EXTRACT_LAST_REDUCTION and defer the
9531 transform phase to vectorizable_condition.
9532 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9534 (vect_finish_stmt_generation): ...here.
9535 (vect_finish_replace_stmt): New function.
9536 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9537 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9539 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9541 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9542 Alan Hayward <alan.hayward@arm.com>
9543 David Sherwood <david.sherwood@arm.com>
9545 * doc/md.texi (extract_last_@var{m}): Document.
9546 * optabs.def (extract_last_optab): New optab.
9547 * internal-fn.def (EXTRACT_LAST): New internal function.
9548 * internal-fn.c (cond_unary_direct): New macro.
9549 (expand_cond_unary_optab_fn): Likewise.
9550 (direct_cond_unary_optab_supported_p): Likewise.
9551 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9552 loops using EXTRACT_LAST.
9553 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9554 (extract_last_<mode>): ...this optab.
9555 (vec_extract<mode><Vel>): Update accordingly.
9557 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9558 Alan Hayward <alan.hayward@arm.com>
9559 David Sherwood <david.sherwood@arm.com>
9561 * target.def (empty_mask_is_expensive): New hook.
9562 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9563 * doc/tm.texi: Regenerate.
9564 * targhooks.h (default_empty_mask_is_expensive): Declare.
9565 * targhooks.c (default_empty_mask_is_expensive): New function.
9566 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9567 if the target says that empty masks are expensive.
9568 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9570 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9572 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9573 Alan Hayward <alan.hayward@arm.com>
9574 David Sherwood <david.sherwood@arm.com>
9576 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9577 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9578 (vect_use_loop_mask_for_alignment_p): New function.
9579 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9580 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9581 niters_skip argument. Make sure that the first niters_skip elements
9582 of the first iteration are inactive.
9583 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9584 Update call to vect_set_loop_masks_directly.
9585 (get_misalign_in_elems): New function, split out from...
9586 (vect_gen_prolog_loop_niters): ...here.
9587 (vect_update_init_of_dr): Take a code argument that specifies whether
9588 the adjustment should be added or subtracted.
9589 (vect_update_init_of_drs): Likewise.
9590 (vect_prepare_for_masked_peels): New function.
9591 (vect_do_peeling): Skip prologue peeling if we're using a mask
9592 instead. Update call to vect_update_inits_of_drs.
9593 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9595 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9596 alignment. Do not include the number of peeled iterations in
9597 the minimum threshold in that case.
9598 (vectorizable_induction): Adjust the start value down by
9599 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9600 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9601 Take the number of skipped iterations into account when calculating
9603 * tree-vect-stmts.c (vect_gen_while_not): New function.
9605 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9606 Alan Hayward <alan.hayward@arm.com>
9607 David Sherwood <david.sherwood@arm.com>
9609 * doc/sourcebuild.texi (vect_fully_masked): Document.
9610 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9612 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9614 (vect_analyze_loop_2): ...here. Don't check the vectorization
9615 factor against the number of loop iterations if the loop is
9618 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9619 Alan Hayward <alan.hayward@arm.com>
9620 David Sherwood <david.sherwood@arm.com>
9622 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9623 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9624 (dump_groups): Update accordingly.
9625 (iv_use::mem_type): New member variable.
9626 (address_p): New function.
9627 (record_use): Add a mem_type argument and initialize the new
9629 (record_group_use): Add a mem_type argument. Use address_p.
9630 Remove obsolete null checks of base_object. Update call to record_use.
9631 (find_interesting_uses_op): Update call to record_group_use.
9632 (find_interesting_uses_cond): Likewise.
9633 (find_interesting_uses_address): Likewise.
9634 (get_mem_type_for_internal_fn): New function.
9635 (find_address_like_use): Likewise.
9636 (find_interesting_uses_stmt): Try find_address_like_use before
9637 calling find_interesting_uses_op.
9638 (addr_offset_valid_p): Use the iv mem_type field as the type
9639 of the addressed memory.
9640 (add_autoinc_candidates): Likewise.
9641 (get_address_cost): Likewise.
9642 (split_small_address_groups_p): Use address_p.
9643 (split_address_groups): Likewise.
9644 (add_iv_candidate_for_use): Likewise.
9645 (autoinc_possible_for_pair): Likewise.
9646 (rewrite_groups): Likewise.
9647 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9648 (determine_group_iv_cost): Update after split of USE_ADDRESS.
9649 (get_alias_ptr_type_for_ptr_address): New function.
9650 (rewrite_use_address): Rewrite address uses in calls that were
9651 identified by find_address_like_use.
9653 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9654 Alan Hayward <alan.hayward@arm.com>
9655 David Sherwood <david.sherwood@arm.com>
9657 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9659 * gimple-expr.h (is_gimple_addressable: Likewise.
9660 * gimple-expr.c (is_gimple_address): Likewise.
9661 * internal-fn.c (expand_call_mem_ref): New function.
9662 (expand_mask_load_optab_fn): Use it.
9663 (expand_mask_store_optab_fn): Likewise.
9665 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9666 Alan Hayward <alan.hayward@arm.com>
9667 David Sherwood <david.sherwood@arm.com>
9669 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
9670 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
9671 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
9672 (cond_umax@var{mode}): Document.
9673 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
9674 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
9675 (cond_umin_optab, cond_umax_optab): New optabs.
9676 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
9677 (COND_IOR, COND_XOR): New internal functions.
9678 * internal-fn.h (get_conditional_internal_fn): Declare.
9679 * internal-fn.c (cond_binary_direct): New macro.
9680 (expand_cond_binary_optab_fn): Likewise.
9681 (direct_cond_binary_optab_supported_p): Likewise.
9682 (get_conditional_internal_fn): New function.
9683 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9684 Cope with reduction statements that are vectorized as calls rather
9686 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9687 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9688 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9689 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9690 (UNSPEC_COND_EOR): New unspecs.
9691 (optab): Add mappings for them.
9692 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9693 (sve_int_op, sve_fp_op): New int attributes.
9695 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9696 Alan Hayward <alan.hayward@arm.com>
9697 David Sherwood <david.sherwood@arm.com>
9699 * optabs.def (while_ult_optab): New optab.
9700 * doc/md.texi (while_ult@var{m}@var{n}): Document.
9701 * internal-fn.def (WHILE_ULT): New internal function.
9702 * internal-fn.h (direct_internal_fn_supported_p): New override
9703 that takes two types as argument.
9704 * internal-fn.c (while_direct): New macro.
9705 (expand_while_optab_fn): New function.
9706 (convert_optab_supported_p): Likewise.
9707 (direct_while_optab_supported_p): New macro.
9708 * wide-int.h (wi::udiv_ceil): New function.
9709 * tree-vectorizer.h (rgroup_masks): New structure.
9710 (vec_loop_masks): New typedef.
9711 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
9713 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
9714 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
9715 (vect_max_vf): New function.
9716 (slpeel_make_loop_iterate_ntimes): Delete.
9717 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
9718 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
9719 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
9720 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
9721 internal-fn.h, stor-layout.h and optabs-query.h.
9722 (vect_set_loop_mask): New function.
9723 (add_preheader_seq): Likewise.
9724 (add_header_seq): Likewise.
9725 (interleave_supported_p): Likewise.
9726 (vect_maybe_permute_loop_masks): Likewise.
9727 (vect_set_loop_masks_directly): Likewise.
9728 (vect_set_loop_condition_masked): Likewise.
9729 (vect_set_loop_condition_unmasked): New function, split out from
9730 slpeel_make_loop_iterate_ntimes.
9731 (slpeel_make_loop_iterate_ntimes): Rename to..
9732 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
9733 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
9734 (vect_do_peeling): Update call accordingly.
9735 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
9737 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9738 mask_compare_type, can_fully_mask_p and fully_masked_p.
9739 (release_vec_loop_masks): New function.
9740 (_loop_vec_info): Use it to free the loop masks.
9741 (can_produce_all_loop_masks_p): New function.
9742 (vect_get_max_nscalars_per_iter): Likewise.
9743 (vect_verify_full_masking): Likewise.
9744 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
9745 retries, and free the mask rgroups before retrying. Check loop-wide
9746 reasons for disallowing fully-masked loops. Make the final decision
9747 about whether use a fully-masked loop or not.
9748 (vect_estimate_min_profitable_iters): Do not assume that peeling
9749 for the number of iterations will be needed for fully-masked loops.
9750 (vectorizable_reduction): Disable fully-masked loops.
9751 (vectorizable_live_operation): Likewise.
9752 (vect_halve_mask_nunits): New function.
9753 (vect_double_mask_nunits): Likewise.
9754 (vect_record_loop_mask): Likewise.
9755 (vect_get_loop_mask): Likewise.
9756 (vect_transform_loop): Handle the case in which the final loop
9757 iteration might handle a partial vector. Call vect_set_loop_condition
9758 instead of slpeel_make_loop_iterate_ntimes.
9759 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
9760 (check_load_store_masking): New function.
9761 (prepare_load_store_mask): Likewise.
9762 (vectorizable_store): Handle fully-masked loops.
9763 (vectorizable_load): Likewise.
9764 (supportable_widening_operation): Use vect_halve_mask_nunits for
9766 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
9767 (vect_gen_while): New function.
9768 * config/aarch64/aarch64.md (umax<mode>3): New expander.
9769 (aarch64_uqdec<mode>): New insn.
9771 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9772 Alan Hayward <alan.hayward@arm.com>
9773 David Sherwood <david.sherwood@arm.com>
9775 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
9776 (reduc_xor_scal_optab): New optabs.
9777 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
9778 (reduc_xor_scal_@var{m}): Document.
9779 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
9780 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
9782 * fold-const-call.c (fold_const_call): Handle them.
9783 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
9784 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
9785 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
9786 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
9787 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
9788 (UNSPEC_XORV): New unspecs.
9789 (optab): Add entries for them.
9790 (BITWISEV): New int iterator.
9791 (bit_reduc_op): New int attributes.
9793 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9794 Alan Hayward <alan.hayward@arm.com>
9795 David Sherwood <david.sherwood@arm.com>
9797 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
9798 * internal-fn.def (VEC_SHL_INSERT): New internal function.
9799 * optabs.def (vec_shl_insert_optab): New optab.
9800 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
9801 (duplicate_and_interleave): Likewise.
9802 * tree-vect-loop.c: Include internal-fn.h.
9803 (neutral_op_for_slp_reduction): New function, split out from
9804 get_initial_defs_for_reduction.
9805 (get_initial_def_for_reduction): Handle option 2 for variable-length
9806 vectors by loading the neutral value into a vector and then shifting
9807 the initial value into element 0.
9808 (get_initial_defs_for_reduction): Replace the code argument with
9809 the neutral value calculated by neutral_op_for_slp_reduction.
9810 Use gimple_build_vector for constant-length vectors.
9811 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
9812 but the first group_size elements have a neutral value.
9813 Use duplicate_and_interleave otherwise.
9814 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
9815 Update call to get_initial_defs_for_reduction. Handle SLP
9816 reductions for variable-length vectors by creating one vector
9817 result for each scalar result, with the elements associated
9818 with other scalar results stubbed out with the neutral value.
9819 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
9820 Require IFN_VEC_SHL_INSERT for double reductions on
9821 variable-length vectors, or SLP reductions that have
9822 a neutral value. Require can_duplicate_and_interleave_p
9823 support for variable-length unchained SLP reductions if there
9824 is no neutral value, such as for MIN/MAX reductions. Also require
9825 the number of vector elements to be a multiple of the number of
9826 SLP statements when doing variable-length unchained SLP reductions.
9827 Update call to vect_create_epilog_for_reduction.
9828 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
9829 and remove initial values.
9830 (duplicate_and_interleave): Make public.
9831 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
9832 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
9834 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9835 Alan Hayward <alan.hayward@arm.com>
9836 David Sherwood <david.sherwood@arm.com>
9838 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
9839 (can_duplicate_and_interleave_p): New function.
9840 (vect_get_and_check_slp_defs): Take the vector of statements
9841 rather than just the current one. Remove excess parentheses.
9842 Restriction rejectinon of vect_constant_def and vect_external_def
9843 for variable-length vectors to boolean types, or types for which
9844 can_duplicate_and_interleave_p is false.
9845 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
9846 (duplicate_and_interleave): New function.
9847 (vect_get_constant_vectors): Use gimple_build_vector for
9848 constant-length vectors and suitable variable-length constant
9849 vectors. Use duplicate_and_interleave for other variable-length
9850 vectors. Don't defer the update when inserting new statements.
9852 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9853 Alan Hayward <alan.hayward@arm.com>
9854 David Sherwood <david.sherwood@arm.com>
9856 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
9857 min_profitable_iters doesn't go negative.
9859 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9860 Alan Hayward <alan.hayward@arm.com>
9861 David Sherwood <david.sherwood@arm.com>
9863 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
9864 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
9865 * optabs.def (vec_mask_load_lanes_optab): New optab.
9866 (vec_mask_store_lanes_optab): Likewise.
9867 * internal-fn.def (MASK_LOAD_LANES): New internal function.
9868 (MASK_STORE_LANES): Likewise.
9869 * internal-fn.c (mask_load_lanes_direct): New macro.
9870 (mask_store_lanes_direct): Likewise.
9871 (expand_mask_load_optab_fn): Handle masked operations.
9872 (expand_mask_load_lanes_optab_fn): New macro.
9873 (expand_mask_store_optab_fn): Handle masked operations.
9874 (expand_mask_store_lanes_optab_fn): New macro.
9875 (direct_mask_load_lanes_optab_supported_p): Likewise.
9876 (direct_mask_store_lanes_optab_supported_p): Likewise.
9877 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
9879 (vect_load_lanes_supported): Likewise.
9880 * tree-vect-data-refs.c (strip_conversion): New function.
9881 (can_group_stmts_p): Likewise.
9882 (vect_analyze_data_ref_accesses): Use it instead of checking
9883 for a pair of assignments.
9884 (vect_store_lanes_supported): Take a masked_p parameter.
9885 (vect_load_lanes_supported): Likewise.
9886 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
9887 vect_store_lanes_supported and vect_load_lanes_supported.
9888 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9889 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
9890 parameter. Don't allow gaps for masked accesses.
9891 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
9892 and vect_load_lanes_supported.
9893 (get_load_store_type): Take a masked_p parameter and update
9894 call to get_group_load_store_type.
9895 (vectorizable_store): Update call to get_load_store_type.
9896 Handle IFN_MASK_STORE_LANES.
9897 (vectorizable_load): Update call to get_load_store_type.
9898 Handle IFN_MASK_LOAD_LANES.
9900 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9901 Alan Hayward <alan.hayward@arm.com>
9902 David Sherwood <david.sherwood@arm.com>
9904 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
9906 * config/aarch64/aarch64-protos.h
9907 (aarch64_sve_struct_memory_operand_p): Declare.
9908 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
9909 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
9910 (VPRED, vpred): Handle SVE structure modes.
9911 * config/aarch64/constraints.md (Utx): New constraint.
9912 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
9913 (aarch64_sve_struct_nonimmediate_operand): New predicates.
9914 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
9915 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
9916 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
9917 structure modes. Split into pieces after RA.
9918 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
9919 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
9921 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
9922 SVE structure modes.
9923 (aarch64_classify_address): Likewise.
9924 (sizetochar): Move earlier in file.
9925 (aarch64_print_operand): Handle SVE register lists.
9926 (aarch64_array_mode): New function.
9927 (aarch64_sve_struct_memory_operand_p): Likewise.
9928 (TARGET_ARRAY_MODE): Redefine.
9930 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9931 Alan Hayward <alan.hayward@arm.com>
9932 David Sherwood <david.sherwood@arm.com>
9934 * target.def (array_mode): New target hook.
9935 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
9936 * doc/tm.texi: Regenerate.
9937 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
9938 * hooks.c (hook_optmode_mode_uhwi_none): New function.
9939 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
9941 * stor-layout.c (mode_for_array): Likewise. Support polynomial
9944 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9945 Alan Hayward <alan.hayward@arm.com>
9946 David Sherwood <david.sherwood@arm.com>
9948 * fold-const.c (fold_binary_loc): Check the argument types
9949 rather than the result type when testing for a vector operation.
9951 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9953 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
9954 * doc/tm.texi: Regenerate.
9956 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9957 Alan Hayward <alan.hayward@arm.com>
9958 David Sherwood <david.sherwood@arm.com>
9960 * doc/invoke.texi (-msve-vector-bits=): Document new option.
9961 (sve): Document new AArch64 extension.
9962 * doc/md.texi (w): Extend the description of the AArch64
9963 constraint to include SVE vectors.
9964 (Upl, Upa): Document new AArch64 predicate constraints.
9965 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
9967 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
9968 (msve-vector-bits=): New option.
9969 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
9970 SVE when these are disabled.
9971 (sve): New extension.
9972 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
9973 modes. Adjust their number of units based on aarch64_sve_vg.
9974 (MAX_BITSIZE_MODE_ANY_MODE): Define.
9975 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
9976 aarch64_addr_query_type.
9977 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
9978 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
9979 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
9980 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
9981 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
9982 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
9983 (aarch64_simd_imm_zero_p): Delete.
9984 (aarch64_check_zero_based_sve_index_immediate): Declare.
9985 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9986 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9987 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9988 (aarch64_sve_float_mul_immediate_p): Likewise.
9989 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9991 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
9992 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
9993 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
9994 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
9995 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
9996 (aarch64_regmode_natural_size): Likewise.
9997 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
9998 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
10000 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
10001 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
10002 for VG and the SVE predicate registers.
10003 (V_ALIASES): Add a "z"-prefixed alias.
10004 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
10005 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
10006 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
10007 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
10008 (REG_CLASS_NAMES): Add entries for them.
10009 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
10010 and the predicate registers.
10011 (aarch64_sve_vg): Declare.
10012 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
10013 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
10014 (REGMODE_NATURAL_SIZE): Define.
10015 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
10017 * config/aarch64/aarch64.c: Include cfgrtl.h.
10018 (simd_immediate_info): Add a constructor for series vectors,
10019 and an associated step field.
10020 (aarch64_sve_vg): New variable.
10021 (aarch64_dbx_register_number): Handle VG and the predicate registers.
10022 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
10023 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
10024 (VEC_ANY_DATA, VEC_STRUCT): New constants.
10025 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
10026 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
10027 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
10028 (aarch64_get_mask_mode): New functions.
10029 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
10030 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
10031 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
10032 predicate modes and predicate registers. Explicitly restrict
10033 GPRs to modes of 16 bytes or smaller. Only allow FP registers
10034 to store a vector mode if it is recognized by
10035 aarch64_classify_vector_mode.
10036 (aarch64_regmode_natural_size): New function.
10037 (aarch64_hard_regno_caller_save_mode): Return the original mode
10039 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
10040 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
10041 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
10042 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
10044 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
10045 does not overlap dest if the function is frame-related. Handle
10047 (aarch64_split_add_offset): New function.
10048 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
10049 them aarch64_add_offset.
10050 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
10051 and update call to aarch64_sub_sp.
10052 (aarch64_add_cfa_expression): New function.
10053 (aarch64_expand_prologue): Pass extra temporary registers to the
10054 functions above. Handle the case in which we need to emit new
10055 DW_CFA_expressions for registers that were originally saved
10056 relative to the stack pointer, but now have to be expressed
10057 relative to the frame pointer.
10058 (aarch64_output_mi_thunk): Pass extra temporary registers to the
10060 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
10061 IP0 and IP1 values for SVE frames.
10062 (aarch64_expand_vec_series): New function.
10063 (aarch64_expand_sve_widened_duplicate): Likewise.
10064 (aarch64_expand_sve_const_vector): Likewise.
10065 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
10066 Handle SVE constants. Use emit_move_insn to move a force_const_mem
10067 into the register, rather than emitting a SET directly.
10068 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
10069 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
10070 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
10071 (offset_9bit_signed_scaled_p): New functions.
10072 (aarch64_replicate_bitmask_imm): New function.
10073 (aarch64_bitmask_imm): Use it.
10074 (aarch64_cannot_force_const_mem): Reject expressions involving
10075 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
10076 (aarch64_classify_index): Handle SVE indices, by requiring
10077 a plain register index with a scale that matches the element size.
10078 (aarch64_classify_address): Handle SVE addresses. Assert that
10079 the mode of the address is VOIDmode or an integer mode.
10080 Update call to aarch64_classify_symbol.
10081 (aarch64_classify_symbolic_expression): Update call to
10082 aarch64_classify_symbol.
10083 (aarch64_const_vec_all_in_range_p): New function.
10084 (aarch64_print_vector_float_operand): Likewise.
10085 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
10086 "vN" for FP registers with SVE modes. Handle (const ...) vectors
10087 and the FP immediates 1.0 and 0.5.
10088 (aarch64_print_address_internal): Handle SVE addresses.
10089 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
10090 (aarch64_regno_regclass): Handle predicate registers.
10091 (aarch64_secondary_reload): Handle big-endian reloads of SVE
10093 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
10094 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
10095 (aarch64_convert_sve_vector_bits): New function.
10096 (aarch64_override_options): Use it to handle -msve-vector-bits=.
10097 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
10098 rather than an rtx.
10099 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
10100 Handle SVE vector and predicate modes. Accept VL-based constants
10101 that need only one temporary register, and VL offsets that require
10102 no temporary registers.
10103 (aarch64_conditional_register_usage): Mark the predicate registers
10104 as fixed if SVE isn't available.
10105 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
10106 Return true for SVE vector and predicate modes.
10107 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
10108 rather than an unsigned int. Handle SVE modes.
10109 (aarch64_preferred_simd_mode): Update call accordingly. Handle
10111 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
10113 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
10114 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
10115 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
10116 (aarch64_sve_float_mul_immediate_p): New functions.
10117 (aarch64_sve_valid_immediate): New function.
10118 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
10119 Explicitly reject structure modes. Check for INDEX constants.
10120 Handle PTRUE and PFALSE constants.
10121 (aarch64_check_zero_based_sve_index_immediate): New function.
10122 (aarch64_simd_imm_zero_p): Delete.
10123 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
10124 vector modes. Accept constants in the range of CNT[BHWD].
10125 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
10126 ask for an Advanced SIMD mode.
10127 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
10128 (aarch64_simd_vector_alignment): Handle SVE predicates.
10129 (aarch64_vectorize_preferred_vector_alignment): New function.
10130 (aarch64_simd_vector_alignment_reachable): Use it instead of
10132 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
10133 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
10135 (MAX_VECT_LEN): Delete.
10136 (expand_vec_perm_d): Add a vec_flags field.
10137 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
10138 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
10139 (aarch64_evpc_ext): Don't apply a big-endian lane correction
10141 (aarch64_evpc_rev): Rename to...
10142 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
10143 (aarch64_evpc_rev_global): New function.
10144 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
10145 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
10147 (aarch64_evpc_sve_tbl): New function.
10148 (aarch64_expand_vec_perm_const_1): Update after rename of
10149 aarch64_evpc_rev. Handle SVE permutes too, trying
10150 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
10151 than aarch64_evpc_tbl.
10152 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
10153 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
10154 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
10155 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
10156 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
10157 (aarch64_expand_sve_vcond): New functions.
10158 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
10159 of aarch64_vector_mode_p.
10160 (aarch64_dwarf_poly_indeterminate_value): New function.
10161 (aarch64_compute_pressure_classes): Likewise.
10162 (aarch64_can_change_mode_class): Likewise.
10163 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
10164 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
10165 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
10166 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
10167 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
10168 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
10169 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
10170 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
10172 (Dn, Dl, Dr): Accept const as well as const_vector.
10173 (Dz): Likewise. Compare against CONST0_RTX.
10174 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
10175 of "vector" where appropriate.
10176 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
10177 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
10178 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
10179 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
10180 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
10181 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
10182 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
10183 (v_int_equiv): Extend to SVE modes.
10184 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
10186 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
10187 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
10188 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
10189 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
10190 (SVE_COND_FP_CMP): New int iterators.
10191 (perm_hilo): Handle the new unpack unspecs.
10192 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
10194 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
10195 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
10196 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
10197 (aarch64_equality_operator, aarch64_constant_vector_operand)
10198 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
10199 (aarch64_sve_nonimmediate_operand): Likewise.
10200 (aarch64_sve_general_operand): Likewise.
10201 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
10202 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
10203 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
10204 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
10205 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
10206 (aarch64_sve_float_arith_immediate): Likewise.
10207 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
10208 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
10209 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
10210 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
10211 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
10212 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
10213 (aarch64_sve_float_arith_operand): Likewise.
10214 (aarch64_sve_float_arith_with_sub_operand): Likewise.
10215 (aarch64_sve_float_mul_operand): Likewise.
10216 (aarch64_sve_vec_perm_operand): Likewise.
10217 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
10218 (aarch64_mov_operand): Accept const_poly_int and const_vector.
10219 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
10220 as well as const_vector.
10221 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
10222 in file. Use CONST0_RTX and CONSTM1_RTX.
10223 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
10224 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
10225 Use aarch64_simd_imm_zero.
10226 * config/aarch64/aarch64-sve.md: New file.
10227 * config/aarch64/aarch64.md: Include it.
10228 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
10229 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
10230 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
10231 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
10232 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
10233 (sve): New attribute.
10234 (enabled): Disable instructions with the sve attribute unless
10236 (movqi, movhi): Pass CONST_POLY_INT operaneds through
10237 aarch64_expand_mov_immediate.
10238 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
10239 CNT[BHSD] immediates.
10240 (movti): Split CONST_POLY_INT moves into two halves.
10241 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
10242 Split additions that need a temporary here if the destination
10243 is the stack pointer.
10244 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
10245 (*add<mode>3_poly_1): New instruction.
10246 (set_clobber_cc): New expander.
10248 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10250 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
10251 parameter and use it instead of GET_MODE_SIZE (innermode). Use
10252 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
10253 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
10254 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
10255 Change innermode from fixed_mode_size to machine_mode.
10256 (simplify_subreg): Update call accordingly. Handle a constant-sized
10257 subreg of a variable-length CONST_VECTOR.
10259 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10260 Alan Hayward <alan.hayward@arm.com>
10261 David Sherwood <david.sherwood@arm.com>
10263 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
10264 (add_offset_to_base): New function, split out from...
10265 (create_mem_ref): ...here. When handling a scale other than 1,
10266 check first whether the address is valid without the offset.
10267 Add it into the base if so, leaving the index and scale as-is.
10269 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10272 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
10273 fold_for_warn before checking if arg2 is INTEGER_CST.
10275 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
10277 * config/rs6000/predicates.md (load_multiple_operation): Delete.
10278 (store_multiple_operation): Delete.
10279 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
10280 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
10281 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
10282 guarded by TARGET_STRING.
10283 (rs6000_output_load_multiple): Delete.
10284 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
10285 OPTION_MASK_STRING / TARGET_STRING handling.
10286 (print_operand) <'N', 'O'>: Add comment that these are unused now.
10287 (const rs6000_opt_masks) <"string">: Change mask to 0.
10288 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
10289 (MASK_STRING): Delete.
10290 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
10292 (load_multiple): Delete.
10299 (store_multiple): Delete.
10306 (movmemsi_8reg): Delete.
10307 (corresponding unnamed define_insn): Delete.
10308 (movmemsi_6reg): Delete.
10309 (corresponding unnamed define_insn): Delete.
10310 (movmemsi_4reg): Delete.
10311 (corresponding unnamed define_insn): Delete.
10312 (movmemsi_2reg): Delete.
10313 (corresponding unnamed define_insn): Delete.
10314 (movmemsi_1reg): Delete.
10315 (corresponding unnamed define_insn): Delete.
10316 * config/rs6000/rs6000.opt (mno-string): New.
10317 (mstring): Replace by deprecation warning stub.
10318 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
10320 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10322 * regrename.c (regrename_do_replace): If replacing the same
10323 reg multiple times, try to reuse last created gen_raw_REG.
10326 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
10327 main to workaround a bug in GDB.
10329 2018-01-12 Tom de Vries <tom@codesourcery.com>
10332 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
10334 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
10336 PR rtl-optimization/80481
10337 * ira-color.c (get_cap_member): New function.
10338 (allocnos_conflict_by_live_ranges_p): Use it.
10339 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
10340 (setup_slot_coalesced_allocno_live_ranges): Ditto.
10342 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
10345 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
10346 (*saddl_se_1): Ditto.
10347 (*ssubsi_1): Ditto.
10348 (*ssubl_se_1): Ditto.
10350 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10352 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
10353 rather than wi::to_widest for DR_INITs.
10354 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
10355 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
10356 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
10358 (vect_analyze_group_access_1): Note that here.
10360 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10362 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
10363 polynomial type sizes.
10365 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10367 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
10368 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
10369 (gimple_add_tmp_var): Likewise.
10371 2018-01-12 Martin Liska <mliska@suse.cz>
10373 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
10374 (gimple_alloc_sizes): Likewise.
10375 (dump_gimple_statistics): Use PRIu64 in printf format.
10376 * gimple.h: Change uint64_t to int.
10378 2018-01-12 Martin Liska <mliska@suse.cz>
10380 * tree-core.h: Use uint64_t instead of int.
10381 * tree.c (tree_node_counts): Likewise.
10382 (tree_node_sizes): Likewise.
10383 (dump_tree_statistics): Use PRIu64 in printf format.
10385 2018-01-12 Martin Liska <mliska@suse.cz>
10387 * Makefile.in: As qsort_chk is implemented in vec.c, add
10388 vec.o to linkage of gencfn-macros.
10389 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
10390 passing the info to record_node_allocation_statistics.
10391 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
10393 * ggc-common.c (struct ggc_usage): Add operator== and use
10394 it in operator< and compare function.
10395 * mem-stats.h (struct mem_usage): Likewise.
10396 * vec.c (struct vec_usage): Remove operator< and compare
10397 function. Can be simply inherited.
10399 2018-01-12 Martin Jambor <mjambor@suse.cz>
10402 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
10403 * tree-ssa-math-opts.c: Include domwalk.h.
10404 (convert_mult_to_fma_1): New function.
10405 (fma_transformation_info): New type.
10406 (fma_deferring_state): Likewise.
10407 (cancel_fma_deferring): New function.
10408 (result_of_phi): Likewise.
10409 (last_fma_candidate_feeds_initial_phi): Likewise.
10410 (convert_mult_to_fma): Added deferring logic, split actual
10411 transformation to convert_mult_to_fma_1.
10412 (math_opts_dom_walker): New type.
10413 (math_opts_dom_walker::after_dom_children): New method, body moved
10414 here from pass_optimize_widening_mul::execute, added deferring logic
10416 (pass_optimize_widening_mul::execute): Moved most of code to
10417 math_opts_dom_walker::after_dom_children.
10418 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10419 * config/i386/i386.c (ix86_option_override_internal): Added
10420 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10422 2018-01-12 Richard Biener <rguenther@suse.de>
10425 * dwarf2out.c (gen_variable_die): Do not reset old_die for
10426 inline instance vars.
10428 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
10431 * config/rx/rx.c (rx_is_restricted_memory_address):
10432 Handle SUBREG case.
10434 2018-01-12 Richard Biener <rguenther@suse.de>
10436 PR tree-optimization/80846
10437 * target.def (split_reduction): New target hook.
10438 * targhooks.c (default_split_reduction): New function.
10439 * targhooks.h (default_split_reduction): Declare.
10440 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10441 target requests first reduce vectors by combining low and high
10443 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10444 (get_vectype_for_scalar_type_and_size): Export.
10445 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10446 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10447 * doc/tm.texi: Regenerate.
10448 * config/i386/i386.c (ix86_split_reduction): Implement
10449 TARGET_VECTORIZE_SPLIT_REDUCTION.
10451 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10454 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10455 in PIC mode except for TARGET_VXWORKS_RTP.
10456 * config/sparc/sparc.c: Include cfgrtl.h.
10457 (TARGET_INIT_PIC_REG): Define.
10458 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10459 (sparc_pic_register_p): New predicate.
10460 (sparc_legitimate_address_p): Use it.
10461 (sparc_legitimize_pic_address): Likewise.
10462 (sparc_delegitimize_address): Likewise.
10463 (sparc_mode_dependent_address_p): Likewise.
10464 (gen_load_pcrel_sym): Remove 4th parameter.
10465 (load_got_register): Adjust call to above. Remove obsolete stuff.
10466 (sparc_expand_prologue): Do not call load_got_register here.
10467 (sparc_flat_expand_prologue): Likewise.
10468 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10469 (sparc_use_pseudo_pic_reg): New function.
10470 (sparc_init_pic_reg): Likewise.
10471 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10472 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10474 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
10476 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10477 Add item for branch_cost.
10479 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10481 PR rtl-optimization/83565
10482 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10483 not extend the result to a larger mode for rotate operations.
10484 (num_sign_bit_copies1): Likewise.
10486 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10489 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10491 Use values-Xc.o for -pedantic.
10492 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10494 2018-01-12 Martin Liska <mliska@suse.cz>
10497 * ipa-devirt.c (final_warning_record::grow_type_warnings):
10499 (possible_polymorphic_call_targets): Use it.
10500 (ipa_devirt): Likewise.
10502 2018-01-12 Martin Liska <mliska@suse.cz>
10504 * profile-count.h (enum profile_quality): Use 0 as invalid
10505 enum value of profile_quality.
10507 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
10509 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10510 -mext-string options.
10512 2018-01-12 Richard Biener <rguenther@suse.de>
10514 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10515 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10516 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10518 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10520 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10522 * configure.ac (--with-long-double-format): Add support for the
10523 configuration option to change the default long double format on
10525 * config.gcc (powerpc*-linux*-*): Likewise.
10526 * configure: Regenerate.
10527 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10528 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10529 used without modification.
10531 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10533 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10534 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10535 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10536 MISC_BUILTIN_SPEC_BARRIER.
10537 (rs6000_init_builtins): Likewise.
10538 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10540 (speculation_barrier): New define_insn.
10541 * doc/extend.texi: Document __builtin_speculation_barrier.
10543 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10546 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10547 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10548 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10550 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10551 integral modes instead of "ss" and "sd".
10552 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10553 vectors with 32-bit and 64-bit elements.
10554 (vecdupssescalarmodesuffix): New mode attribute.
10555 (vec_dup<mode>): Use it.
10557 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
10560 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10561 frame if argument is passed on stack.
10563 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10566 * ree.c (combine_reaching_defs): Optimize also
10567 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10568 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10570 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10572 PR middle-end/83189
10573 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10575 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10577 PR middle-end/83718
10578 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10579 after they are computed.
10581 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10583 PR tree-optimization/83695
10584 * gimple-loop-linterchange.cc
10585 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10586 reset cached scev information after interchange.
10587 (pass_linterchange::execute): Remove call to scev_reset_htab.
10589 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10591 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10592 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10593 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10594 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10595 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10596 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10597 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10598 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10599 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10600 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10601 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10602 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10603 (V_lane_reg): Likewise.
10604 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10606 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10607 (vfmal_lane_low<mode>_intrinsic,
10608 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10609 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10610 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10611 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10612 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10613 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10615 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10617 * config/arm/arm-cpus.in (fp16fml): New feature.
10618 (ALL_SIMD): Add fp16fml.
10619 (armv8.2-a): Add fp16fml as an option.
10620 (armv8.3-a): Likewise.
10621 (armv8.4-a): Add fp16fml as part of fp16.
10622 * config/arm/arm.h (TARGET_FP16FML): Define.
10623 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10625 * config/arm/arm-modes.def (V2HF): Define.
10626 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10627 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10628 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10629 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10630 vfmsl_low, vfmsl_high): New set of builtins.
10631 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10632 (vfml_op): New code attribute.
10633 (VFMLHALVES): New int iterator.
10634 (VFML, VFMLSEL): New mode attributes.
10635 (V_reg): Define mapping for V2HF.
10636 (V_hi, V_lo): New mode attributes.
10637 (VF_constraint): Likewise.
10638 (vfml_half, vfml_half_selector): New int attributes.
10639 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10641 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10642 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10644 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10645 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10646 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10647 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
10649 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10650 Document new effective target and option set.
10652 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10654 * config/arm/arm-cpus.in (armv8_4): New feature.
10655 (ARMv8_4a): New fgroup.
10656 (armv8.4-a): New arch.
10657 * config/arm/arm-tables.opt: Regenerate.
10658 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10659 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
10660 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
10661 Add matching rules for -march=armv8.4-a and extensions.
10662 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
10664 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
10667 * config/rx/rx.md (BW): New mode attribute.
10668 (sync_lock_test_and_setsi): Add mode suffix to insn output.
10670 2018-01-11 Richard Biener <rguenther@suse.de>
10672 PR tree-optimization/83435
10673 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
10674 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
10675 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
10677 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10678 Alan Hayward <alan.hayward@arm.com>
10679 David Sherwood <david.sherwood@arm.com>
10681 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
10683 (aarch64_classify_address): Initialize it. Track polynomial offsets.
10684 (aarch64_print_address_internal): Use it to check for a zero offset.
10686 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10687 Alan Hayward <alan.hayward@arm.com>
10688 David Sherwood <david.sherwood@arm.com>
10690 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10691 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10692 Return a poly_int64 rather than a HOST_WIDE_INT.
10693 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10694 rather than a HOST_WIDE_INT.
10695 * config/aarch64/aarch64.h (aarch64_frame): Protect with
10696 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
10697 hard_fp_offset, frame_size, initial_adjust, callee_offset and
10698 final_offset from HOST_WIDE_INT to poly_int64.
10699 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
10700 to_constant when getting the number of units in an Advanced SIMD
10702 (aarch64_builtin_vectorized_function): Check for a constant number
10704 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
10706 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
10707 attribute instead of GET_MODE_NUNITS.
10708 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
10709 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
10710 GET_MODE_SIZE for fixed-size registers.
10711 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
10712 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
10713 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
10714 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
10715 (aarch64_print_operand, aarch64_print_address_internal)
10716 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
10717 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
10718 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
10719 Handle polynomial GET_MODE_SIZE.
10720 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
10721 wider than SImode without modification.
10722 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
10723 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
10724 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
10725 passing and returning SVE modes.
10726 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
10727 rather than GEN_INT.
10728 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
10729 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
10730 (aarch64_allocate_and_probe_stack_space): Likewise.
10731 (aarch64_layout_frame): Cope with polynomial offsets.
10732 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
10733 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
10734 polynomial offsets.
10735 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
10736 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
10737 poly_int64 rather than a HOST_WIDE_INT.
10738 (aarch64_get_separate_components, aarch64_process_components)
10739 (aarch64_expand_prologue, aarch64_expand_epilogue)
10740 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
10741 (aarch64_anchor_offset): New function, split out from...
10742 (aarch64_legitimize_address): ...here.
10743 (aarch64_builtin_vectorization_cost): Handle polynomial
10744 TYPE_VECTOR_SUBPARTS.
10745 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
10747 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
10748 number of elements from the PARALLEL rather than the mode.
10749 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
10750 rather than GET_MODE_BITSIZE.
10751 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
10752 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
10753 (aarch64_expand_vec_perm_const_1): Handle polynomial
10754 d->perm.length () and d->perm elements.
10755 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
10756 Apply to_constant to d->perm elements.
10757 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
10758 polynomial CONST_VECTOR_NUNITS.
10759 (aarch64_move_pointer): Take amount as a poly_int64 rather
10761 (aarch64_progress_pointer): Avoid temporary variable.
10762 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
10763 the mode attribute instead of GET_MODE.
10765 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10766 Alan Hayward <alan.hayward@arm.com>
10767 David Sherwood <david.sherwood@arm.com>
10769 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
10770 x exists before using it.
10771 (aarch64_add_constant_internal): Rename to...
10772 (aarch64_add_offset_1): ...this. Replace regnum with separate
10773 src and dest rtxes. Handle the case in which they're different,
10774 including when the offset is zero. Replace scratchreg with an rtx.
10775 Use 2 additions if there is no spare register into which we can
10776 move a 16-bit constant.
10777 (aarch64_add_constant): Delete.
10778 (aarch64_add_offset): Replace reg with separate src and dest
10779 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
10780 Use aarch64_add_offset_1.
10781 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
10782 an rtx rather than an int. Take the delta as a poly_int64
10783 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
10784 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
10785 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
10786 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
10787 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
10788 and aarch64_add_sp.
10789 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
10790 aarch64_add_constant.
10792 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10794 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
10795 Use scalar_float_mode.
10797 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10799 * config/aarch64/aarch64-simd.md
10800 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
10801 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
10802 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
10803 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
10804 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
10805 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
10806 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
10807 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
10808 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
10809 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
10811 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10814 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
10815 targ_options->x_arm_arch_string is non NULL.
10817 2018-01-11 Tamar Christina <tamar.christina@arm.com>
10819 * config/aarch64/aarch64.h
10820 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
10822 2018-01-11 Sudakshina Das <sudi.das@arm.com>
10825 * expmed.c (emit_store_flag_force): Swap if const op0
10826 and change VOIDmode to mode of op0.
10828 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10830 PR rtl-optimization/83761
10831 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
10832 than bytes to mode_for_size.
10834 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10836 PR middle-end/83189
10837 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
10838 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
10841 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10843 PR middle-end/83575
10844 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
10845 when in layout mode.
10846 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
10847 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
10850 2018-01-10 Michael Collison <michael.collison@arm.com>
10852 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
10853 * config/aarch64/aarch64-option-extension.def: Add
10854 AARCH64_OPT_EXTENSION of 'fp16fml'.
10855 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10856 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
10857 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
10858 * config/aarch64/constraints.md (Ui7): New constraint.
10859 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
10860 (VFMLA_SEL_W): Ditto.
10863 (VFMLA16_LOW): New int iterator.
10864 (VFMLA16_HIGH): Ditto.
10865 (UNSPEC_FMLAL): New unspec.
10866 (UNSPEC_FMLSL): Ditto.
10867 (UNSPEC_FMLAL2): Ditto.
10868 (UNSPEC_FMLSL2): Ditto.
10869 (f16mac): New code attribute.
10870 * config/aarch64/aarch64-simd-builtins.def
10871 (aarch64_fmlal_lowv2sf): Ditto.
10872 (aarch64_fmlsl_lowv2sf): Ditto.
10873 (aarch64_fmlalq_lowv4sf): Ditto.
10874 (aarch64_fmlslq_lowv4sf): Ditto.
10875 (aarch64_fmlal_highv2sf): Ditto.
10876 (aarch64_fmlsl_highv2sf): Ditto.
10877 (aarch64_fmlalq_highv4sf): Ditto.
10878 (aarch64_fmlslq_highv4sf): Ditto.
10879 (aarch64_fmlal_lane_lowv2sf): Ditto.
10880 (aarch64_fmlsl_lane_lowv2sf): Ditto.
10881 (aarch64_fmlal_laneq_lowv2sf): Ditto.
10882 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
10883 (aarch64_fmlalq_lane_lowv4sf): Ditto.
10884 (aarch64_fmlsl_lane_lowv4sf): Ditto.
10885 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
10886 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
10887 (aarch64_fmlal_lane_highv2sf): Ditto.
10888 (aarch64_fmlsl_lane_highv2sf): Ditto.
10889 (aarch64_fmlal_laneq_highv2sf): Ditto.
10890 (aarch64_fmlsl_laneq_highv2sf): Ditto.
10891 (aarch64_fmlalq_lane_highv4sf): Ditto.
10892 (aarch64_fmlsl_lane_highv4sf): Ditto.
10893 (aarch64_fmlalq_laneq_highv4sf): Ditto.
10894 (aarch64_fmlsl_laneq_highv4sf): Ditto.
10895 * config/aarch64/aarch64-simd.md:
10896 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
10897 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10898 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
10899 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10900 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
10901 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
10902 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
10903 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
10904 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
10905 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
10906 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
10907 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
10908 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
10909 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
10910 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
10911 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
10912 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
10913 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
10914 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
10915 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
10916 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
10917 (vfmlsl_low_u32): Ditto.
10918 (vfmlalq_low_u32): Ditto.
10919 (vfmlslq_low_u32): Ditto.
10920 (vfmlal_high_u32): Ditto.
10921 (vfmlsl_high_u32): Ditto.
10922 (vfmlalq_high_u32): Ditto.
10923 (vfmlslq_high_u32): Ditto.
10924 (vfmlal_lane_low_u32): Ditto.
10925 (vfmlsl_lane_low_u32): Ditto.
10926 (vfmlal_laneq_low_u32): Ditto.
10927 (vfmlsl_laneq_low_u32): Ditto.
10928 (vfmlalq_lane_low_u32): Ditto.
10929 (vfmlslq_lane_low_u32): Ditto.
10930 (vfmlalq_laneq_low_u32): Ditto.
10931 (vfmlslq_laneq_low_u32): Ditto.
10932 (vfmlal_lane_high_u32): Ditto.
10933 (vfmlsl_lane_high_u32): Ditto.
10934 (vfmlal_laneq_high_u32): Ditto.
10935 (vfmlsl_laneq_high_u32): Ditto.
10936 (vfmlalq_lane_high_u32): Ditto.
10937 (vfmlslq_lane_high_u32): Ditto.
10938 (vfmlalq_laneq_high_u32): Ditto.
10939 (vfmlslq_laneq_high_u32): Ditto.
10940 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
10941 (AARCH64_FL_FOR_ARCH8_4): New.
10942 (AARCH64_ISA_F16FML): New ISA flag.
10943 (TARGET_F16FML): New feature flag for fp16fml.
10944 (doc/invoke.texi): Document new fp16fml option.
10946 2018-01-10 Michael Collison <michael.collison@arm.com>
10948 * config/aarch64/aarch64-builtins.c:
10949 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
10950 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10951 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
10952 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
10953 (AARCH64_ISA_SHA3): New ISA flag.
10954 (TARGET_SHA3): New feature flag for sha3.
10955 * config/aarch64/iterators.md (sha512_op): New int attribute.
10956 (CRYPTO_SHA512): New int iterator.
10957 (UNSPEC_SHA512H): New unspec.
10958 (UNSPEC_SHA512H2): Ditto.
10959 (UNSPEC_SHA512SU0): Ditto.
10960 (UNSPEC_SHA512SU1): Ditto.
10961 * config/aarch64/aarch64-simd-builtins.def
10962 (aarch64_crypto_sha512hqv2di): New builtin.
10963 (aarch64_crypto_sha512h2qv2di): Ditto.
10964 (aarch64_crypto_sha512su0qv2di): Ditto.
10965 (aarch64_crypto_sha512su1qv2di): Ditto.
10966 (aarch64_eor3qv8hi): Ditto.
10967 (aarch64_rax1qv2di): Ditto.
10968 (aarch64_xarqv2di): Ditto.
10969 (aarch64_bcaxqv8hi): Ditto.
10970 * config/aarch64/aarch64-simd.md:
10971 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
10972 (aarch64_crypto_sha512su0qv2di): Ditto.
10973 (aarch64_crypto_sha512su1qv2di): Ditto.
10974 (aarch64_eor3qv8hi): Ditto.
10975 (aarch64_rax1qv2di): Ditto.
10976 (aarch64_xarqv2di): Ditto.
10977 (aarch64_bcaxqv8hi): Ditto.
10978 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
10979 (vsha512h2q_u64): Ditto.
10980 (vsha512su0q_u64): Ditto.
10981 (vsha512su1q_u64): Ditto.
10982 (veor3q_u16): Ditto.
10983 (vrax1q_u64): Ditto.
10984 (vxarq_u64): Ditto.
10985 (vbcaxq_u16): Ditto.
10986 * config/arm/types.md (crypto_sha512): New type attribute.
10987 (crypto_sha3): Ditto.
10988 (doc/invoke.texi): Document new sha3 option.
10990 2018-01-10 Michael Collison <michael.collison@arm.com>
10992 * config/aarch64/aarch64-builtins.c:
10993 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
10994 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10995 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
10996 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
10997 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
10998 (AARCH64_ISA_SM4): New ISA flag.
10999 (TARGET_SM4): New feature flag for sm4.
11000 * config/aarch64/aarch64-simd-builtins.def
11001 (aarch64_sm3ss1qv4si): Ditto.
11002 (aarch64_sm3tt1aq4si): Ditto.
11003 (aarch64_sm3tt1bq4si): Ditto.
11004 (aarch64_sm3tt2aq4si): Ditto.
11005 (aarch64_sm3tt2bq4si): Ditto.
11006 (aarch64_sm3partw1qv4si): Ditto.
11007 (aarch64_sm3partw2qv4si): Ditto.
11008 (aarch64_sm4eqv4si): Ditto.
11009 (aarch64_sm4ekeyqv4si): Ditto.
11010 * config/aarch64/aarch64-simd.md:
11011 (aarch64_sm3ss1qv4si): Ditto.
11012 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
11013 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
11014 (aarch64_sm4eqv4si): Ditto.
11015 (aarch64_sm4ekeyqv4si): Ditto.
11016 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
11017 (sm3part_op): Ditto.
11018 (CRYPTO_SM3TT): Ditto.
11019 (CRYPTO_SM3PART): Ditto.
11020 (UNSPEC_SM3SS1): New unspec.
11021 (UNSPEC_SM3TT1A): Ditto.
11022 (UNSPEC_SM3TT1B): Ditto.
11023 (UNSPEC_SM3TT2A): Ditto.
11024 (UNSPEC_SM3TT2B): Ditto.
11025 (UNSPEC_SM3PARTW1): Ditto.
11026 (UNSPEC_SM3PARTW2): Ditto.
11027 (UNSPEC_SM4E): Ditto.
11028 (UNSPEC_SM4EKEY): Ditto.
11029 * config/aarch64/constraints.md (Ui2): New constraint.
11030 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
11031 * config/arm/types.md (crypto_sm3): New type attribute.
11032 (crypto_sm4): Ditto.
11033 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
11034 (vsm3tt1aq_u32): Ditto.
11035 (vsm3tt1bq_u32): Ditto.
11036 (vsm3tt2aq_u32): Ditto.
11037 (vsm3tt2bq_u32): Ditto.
11038 (vsm3partw1q_u32): Ditto.
11039 (vsm3partw2q_u32): Ditto.
11040 (vsm4eq_u32): Ditto.
11041 (vsm4ekeyq_u32): Ditto.
11042 (doc/invoke.texi): Document new sm4 option.
11044 2018-01-10 Michael Collison <michael.collison@arm.com>
11046 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
11047 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
11048 (AARCH64_FL_FOR_ARCH8_4): New.
11049 (AARCH64_FL_V8_4): New flag.
11050 (doc/invoke.texi): Document new armv8.4-a option.
11052 2018-01-10 Michael Collison <michael.collison@arm.com>
11054 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
11055 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
11056 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
11057 * config/aarch64/aarch64-option-extension.def: Add
11058 AARCH64_OPT_EXTENSION of 'sha2'.
11059 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
11060 (crypto): Disable sha2 and aes if crypto disabled.
11061 (crypto): Enable aes and sha2 if enabled.
11062 (simd): Disable sha2 and aes if simd disabled.
11063 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
11065 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
11066 (TARGET_SHA2): New feature flag for sha2.
11067 (TARGET_AES): New feature flag for aes.
11068 * config/aarch64/aarch64-simd.md:
11069 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
11070 conditional on TARGET_AES.
11071 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
11072 (aarch64_crypto_sha1hsi): Make pattern conditional
11074 (aarch64_crypto_sha1hv4si): Ditto.
11075 (aarch64_be_crypto_sha1hv4si): Ditto.
11076 (aarch64_crypto_sha1su1v4si): Ditto.
11077 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
11078 (aarch64_crypto_sha1su0v4si): Ditto.
11079 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
11080 (aarch64_crypto_sha256su0v4si): Ditto.
11081 (aarch64_crypto_sha256su1v4si): Ditto.
11082 (doc/invoke.texi): Document new aes and sha2 options.
11084 2018-01-10 Martin Sebor <msebor@redhat.com>
11086 PR tree-optimization/83781
11087 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
11090 2018-01-11 Martin Sebor <msebor@gmail.com>
11091 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11093 PR tree-optimization/83501
11094 PR tree-optimization/81703
11096 * tree-ssa-strlen.c (get_string_cst): Rename...
11097 (get_string_len): ...to this. Handle global constants.
11098 (handle_char_store): Adjust.
11100 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
11101 Jim Wilson <jimw@sifive.com>
11103 * config/riscv/riscv-protos.h (riscv_output_return): New.
11104 * config/riscv/riscv.c (struct machine_function): New naked_p field.
11105 (riscv_attribute_table, riscv_output_return),
11106 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
11107 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
11108 (riscv_compute_frame_info): Only compute frame->mask if not a naked
11110 (riscv_expand_prologue): Add early return for naked function.
11111 (riscv_expand_epilogue): Likewise.
11112 (riscv_function_ok_for_sibcall): Return false for naked function.
11113 (riscv_set_current_function): New.
11114 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
11115 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
11116 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
11117 * doc/extend.texi (RISC-V Function Attributes): New.
11119 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
11121 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
11122 check for 128-bit long double before checking TCmode.
11123 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
11124 128-bit long doubles before checking TFmode or TCmode.
11125 (FLOAT128_IBM_P): Likewise.
11127 2018-01-10 Martin Sebor <msebor@redhat.com>
11129 PR tree-optimization/83671
11130 * builtins.c (c_strlen): Unconditionally return zero for the empty
11132 Use -Warray-bounds for warnings.
11133 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
11134 for non-constant array indices with COMPONENT_REF, arrays of
11135 arrays, and pointers to arrays.
11136 (gimple_fold_builtin_strlen): Determine and set length range for
11137 non-constant character arrays.
11139 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
11141 PR middle-end/81897
11142 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
11145 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
11147 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
11149 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11152 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
11153 VECTOR_MEM_ALTIVEC_OR_VSX_P.
11154 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
11155 indexed_or_indirect_operand predicate.
11156 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
11157 (*vsx_le_perm_load_v8hi): Likewise.
11158 (*vsx_le_perm_load_v16qi): Likewise.
11159 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
11160 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
11161 (*vsx_le_perm_store_v8hi): Likewise.
11162 (*vsx_le_perm_store_v16qi): Likewise.
11163 (eight unnamed splitters): Likewise.
11165 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
11167 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
11168 * config/rs6000/emmintrin.h: Likewise.
11169 * config/rs6000/mmintrin.h: Likewise.
11170 * config/rs6000/xmmintrin.h: Likewise.
11172 2018-01-10 David Malcolm <dmalcolm@redhat.com>
11175 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
11177 * tree.c (tree_nop_conversion): Return true for location wrapper
11179 (maybe_wrap_with_location): New function.
11180 (selftest::check_strip_nops): New function.
11181 (selftest::test_location_wrappers): New function.
11182 (selftest::tree_c_tests): Call it.
11183 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
11184 (maybe_wrap_with_location): New decl.
11185 (EXPR_LOCATION_WRAPPER_P): New macro.
11186 (location_wrapper_p): New inline function.
11187 (tree_strip_any_location_wrapper): New inline function.
11189 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
11192 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
11193 stack_realign_offset for the largest alignment of stack slot
11195 (ix86_find_max_used_stack_alignment): New function.
11196 (ix86_finalize_stack_frame_flags): Use it. Set
11197 max_used_stack_alignment if we don't realign stack.
11198 * config/i386/i386.h (machine_function): Add
11199 max_used_stack_alignment.
11201 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
11203 * config/arm/arm.opt (-mbranch-cost): New option.
11204 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
11207 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
11210 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
11211 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
11213 2018-01-10 Richard Biener <rguenther@suse.de>
11216 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
11217 early out so it also covers the case where we have a non-NULL
11220 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11222 PR tree-optimization/83753
11223 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
11224 for non-strided grouped accesses if the number of elements is 1.
11226 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
11229 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
11230 * i386.h (TARGET_USE_GATHER): Define.
11231 * x86-tune.def (X86_TUNE_USE_GATHER): New.
11233 2018-01-10 Martin Liska <mliska@suse.cz>
11236 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
11237 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
11239 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
11240 CLEANUP_NO_PARTITIONING is not set.
11242 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
11244 * doc/rtl.texi: Remove documentation of (const ...) wrappers
11245 for vectors, as a partial revert of r254296.
11246 * rtl.h (const_vec_p): Delete.
11247 (const_vec_duplicate_p): Don't test for vector CONSTs.
11248 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
11249 * expmed.c (make_tree): Likewise.
11252 * common.md (E, F): Use CONSTANT_P instead of checking for
11254 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
11255 checking for CONST_VECTOR.
11257 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11259 PR middle-end/83575
11260 * predict.c (force_edge_cold): Handle in more sane way edges
11261 with no prediction.
11263 2018-01-09 Carl Love <cel@us.ibm.com>
11265 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
11267 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
11268 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
11269 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
11270 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
11271 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
11272 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
11273 * config/rs6000/rs6000-protos.h: Add extern defition for
11274 rs6000_generate_float2_double_code.
11275 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
11277 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
11278 (float2_v2df): Add define_expand.
11280 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
11283 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
11284 op_mode in the force_to_mode call.
11286 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11288 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
11289 instead of checking each element individually.
11290 (aarch64_evpc_uzp): Likewise.
11291 (aarch64_evpc_zip): Likewise.
11292 (aarch64_evpc_ext): Likewise.
11293 (aarch64_evpc_rev): Likewise.
11294 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
11295 instead of checking each element individually. Return true without
11297 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
11298 whether all selected elements come from the same input, instead of
11299 checking each element individually. Remove calls to gen_rtx_REG,
11300 start_sequence and end_sequence and instead assert that no rtl is
11303 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11305 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
11306 order of HIGH and CONST checks.
11308 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11310 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
11311 if the destination isn't an SSA_NAME.
11313 2018-01-09 Richard Biener <rguenther@suse.de>
11315 PR tree-optimization/83668
11316 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
11318 (canonicalize_loop_form): ... here, renamed from ...
11319 (canonicalize_loop_closed_ssa_form): ... this and amended to
11320 swap successor edges for loop exit blocks to make us use
11321 the RPO order we need for initial schedule generation.
11323 2018-01-09 Joseph Myers <joseph@codesourcery.com>
11325 PR tree-optimization/64811
11326 * match.pd: When optimizing comparisons with Inf, avoid
11327 introducing or losing exceptions from comparisons with NaN.
11329 2018-01-09 Martin Liska <mliska@suse.cz>
11332 * asan.c (shadow_mem_size): Add gcc_assert.
11334 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
11336 Don't save registers in main().
11339 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
11340 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
11341 * config/avr/avr.c (avr_set_current_function): Don't error if
11342 naked, OS_task or OS_main are specified at the same time.
11343 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
11345 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
11347 * common/config/avr/avr-common.c (avr_option_optimization_table):
11348 Switch on -mmain-is-OS_task for optimizing compilations.
11350 2018-01-09 Richard Biener <rguenther@suse.de>
11352 PR tree-optimization/83572
11353 * graphite.c: Include cfganal.h.
11354 (graphite_transform_loops): Connect infinite loops to exit
11355 and remove fake edges at the end.
11357 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11359 * ipa-inline.c (edge_badness): Revert accidental checkin.
11361 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11364 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
11365 symbols; not inline clones.
11367 2018-01-09 Jakub Jelinek <jakub@redhat.com>
11370 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
11371 hard registers. Formatting fixes.
11373 PR preprocessor/83722
11374 * gcc.c (try_generate_repro): Pass
11375 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
11376 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
11379 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
11380 Kito Cheng <kito.cheng@gmail.com>
11382 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
11383 (riscv_leaf_function_p): Delete.
11384 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
11386 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11388 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
11390 (do_ifelse): New function.
11391 (do_isel): New function.
11392 (do_sub3): New function.
11393 (do_add3): New function.
11394 (do_load_mask_compare): New function.
11395 (do_overlap_load_compare): New function.
11396 (expand_compare_loop): New function.
11397 (expand_block_compare): Call expand_compare_loop() when appropriate.
11398 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
11399 option description.
11400 (-mblock-compare-inline-loop-limit): New option.
11402 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11405 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
11406 Reverse order of second and third operands in first alternative.
11407 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
11408 of first and second elements in UNSPEC_VPERMR vector.
11409 (altivec_expand_vec_perm_le): Likewise.
11411 2018-01-08 Jeff Law <law@redhat.com>
11413 PR rtl-optimizatin/81308
11414 * tree-switch-conversion.c (cfg_altered): New file scoped static.
11415 (process_switch): If group_case_labels makes a change, then set
11417 (pass_convert_switch::execute): If a switch is converted, then
11418 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
11420 PR rtl-optimization/81308
11421 * recog.c (split_all_insns): Conditionally cleanup the CFG after
11424 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
11426 PR target/83663 - Revert r255946
11427 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11428 generation for cases where splatting a value is not useful.
11429 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11430 across a vec_duplicate and a paradoxical subreg forming a vector
11431 mode to a vec_concat.
11433 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11435 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11436 -march=armv8.3-a variants.
11437 * config/arm/t-multilib: Likewise.
11438 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11440 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11442 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11444 (cceq_ior_compare_complement): Give it a name so I can use it, and
11445 change boolean_or_operator predicate to boolean_operator so it can
11446 be used to generate a crand.
11447 (eqne): New code iterator.
11448 (bd/bd_neg): New code_attrs.
11449 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11450 a single define_insn.
11451 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11452 decrement (bdnzt/bdnzf/bdzt/bdzf).
11453 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11454 with the new names of the branch decrement patterns, and added the
11455 names of the branch decrement conditional patterns.
11457 2018-01-08 Richard Biener <rguenther@suse.de>
11459 PR tree-optimization/83563
11460 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11463 2018-01-08 Richard Biener <rguenther@suse.de>
11465 PR middle-end/83713
11466 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11468 2018-01-08 Richard Biener <rguenther@suse.de>
11470 PR tree-optimization/83685
11471 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11472 references to abnormals.
11474 2018-01-08 Richard Biener <rguenther@suse.de>
11477 * dwarf2out.c (output_indirect_strings): Handle empty
11478 skeleton_debug_str_hash.
11479 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11481 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11483 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11484 (emit_store_direct): Likewise.
11485 (arc_trampoline_adjust_address): Likewise.
11486 (arc_asm_trampoline_template): New function.
11487 (arc_initialize_trampoline): Use asm_trampoline_template.
11488 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11489 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11490 * config/arc/arc.md (flush_icache): Delete pattern.
11492 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11494 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11495 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11498 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11501 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11502 by not USED_FOR_TARGET.
11503 (make_pass_resolve_sw_modes): Likewise.
11505 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11507 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11510 2018-01-08 Richard Biener <rguenther@suse.de>
11512 PR middle-end/83580
11513 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11515 2018-01-08 Richard Biener <rguenther@suse.de>
11517 PR middle-end/83517
11518 * match.pd ((t * 2) / 2) -> t): Add missing :c.
11520 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
11522 PR middle-end/81897
11523 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11524 basic blocks with a small number of successors.
11525 (convert_control_dep_chain_into_preds): Improve handling of
11527 (dump_predicates): Split apart into...
11528 (dump_pred_chain): ...here...
11529 (dump_pred_info): ...and here.
11530 (can_one_predicate_be_invalidated_p): Add debugging printfs.
11531 (can_chain_union_be_invalidated_p): Improve check for invalidation
11533 (uninit_uses_cannot_happen): Avoid unnecessary if
11534 convert_control_dep_chain_into_preds yielded nothing.
11536 2018-01-06 Martin Sebor <msebor@redhat.com>
11538 PR tree-optimization/83640
11539 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11540 subtracting negative offset from size.
11541 (builtin_access::overlap): Adjust offset bounds of the access to fall
11542 within the size of the object if possible.
11544 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
11546 PR rtl-optimization/83699
11547 * expmed.c (extract_bit_field_1): Restrict the vector usage of
11548 extract_bit_field_as_subreg to cases in which the extracted
11549 value is also a vector.
11551 * lra-constraints.c (process_alt_operands): Test for the equivalence
11552 substitutions when detecting a possible reload cycle.
11554 2018-01-06 Jakub Jelinek <jakub@redhat.com>
11557 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11558 by default if flag_selective_schedling{,2}. Formatting fixes.
11560 PR rtl-optimization/83682
11561 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11562 if it has non-VECTOR_MODE element mode.
11563 (vec_duplicate_p): Likewise.
11565 PR middle-end/83694
11566 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11567 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11569 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11572 * config/i386/i386-builtin.def
11573 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11574 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11575 Require also OPTION_MASK_ISA_AVX512F in addition to
11576 OPTION_MASK_ISA_GFNI.
11577 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11578 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11579 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11580 to OPTION_MASK_ISA_GFNI.
11581 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11582 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11583 OPTION_MASK_ISA_AVX512BW.
11584 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11585 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11586 addition to OPTION_MASK_ISA_GFNI.
11587 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11588 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11589 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11590 to OPTION_MASK_ISA_GFNI.
11591 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11592 a requirement for all ISAs rather than any of them with a few
11594 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11596 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11597 bitmasks to be enabled with 3 exceptions, instead of requiring any
11598 enabled ISA with lots of exceptions.
11599 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11600 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11601 Change avx512bw in isa attribute to avx512f.
11602 * config/i386/sgxintrin.h: Add license boilerplate.
11603 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11604 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11605 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11606 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11608 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11609 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11610 temporarily sse2 rather than sse if not enabled already.
11613 * config/i386/sse.md (VI248_VLBW): Rename to ...
11614 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11615 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11616 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11617 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11618 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11619 mode iterator instead of VI248_VLBW.
11621 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11623 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11624 (record_modified): Skip clobbers; add debug output.
11625 (param_change_prob): Use sreal frequencies.
11627 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11629 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11630 punt for user-aligned variables.
11632 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11634 * tree-chrec.c (chrec_contains_symbols): Return true for
11637 2018-01-05 Sudakshina Das <sudi.das@arm.com>
11640 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11641 of (x|y) == x for BICS pattern.
11643 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11645 PR tree-optimization/83605
11646 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11647 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11650 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
11652 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11653 * config/epiphany/rtems.h: New file.
11655 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11656 Uros Bizjak <ubizjak@gmail.com>
11659 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
11660 QIreg_operand instead of register_operand predicate.
11661 * config/i386/i386.c (ix86_rop_should_change_byte_p,
11662 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
11663 comments instead of -fmitigate[-_]rop.
11665 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11668 * cgraphunit.c (symbol_table::compile): Switch to text_section
11669 before calling assembly_start debug hook.
11670 * run-rtl-passes.c (run_rtl_passes): Likewise.
11673 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11675 * tree-vrp.c (extract_range_from_binary_expr_1): Check
11676 range_int_cst_p rather than !symbolic_range_p before calling
11677 extract_range_from_multiplicative_op_1.
11679 2018-01-04 Jeff Law <law@redhat.com>
11681 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11682 redundant test in assertion.
11684 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11686 * doc/rtl.texi: Document machine_mode wrapper classes.
11688 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11690 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11691 using tree_to_uhwi.
11693 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11695 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11696 the VEC_PERM_EXPR fold to fail.
11698 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11701 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
11702 to switched_sections.
11704 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11707 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
11708 test for d.testing.
11710 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
11713 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
11714 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
11716 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11719 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
11720 is BLKmode and bitpos not zero or mode change is needed.
11722 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11725 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
11728 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
11731 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
11732 instead of MULT rtx. Update all corresponding splitters.
11733 (*saddl_se): Ditto.
11734 (*ssub<modesuffix>): Ditto.
11735 (*ssubl_se): Ditto.
11736 (*cmp_sadd_di): Update split patterns.
11737 (*cmp_sadd_si): Ditto.
11738 (*cmp_sadd_sidi): Ditto.
11739 (*cmp_ssub_di): Ditto.
11740 (*cmp_ssub_si): Ditto.
11741 (*cmp_ssub_sidi): Ditto.
11742 * config/alpha/predicates.md (const23_operand): New predicate.
11743 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
11744 Look for ASHIFT, not MULT inner operand.
11745 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
11747 2018-01-04 Martin Liska <mliska@suse.cz>
11749 PR gcov-profile/83669
11750 * gcov.c (output_intermediate_file): Add version to intermediate
11752 * doc/gcov.texi: Document new field 'version' in intermediate
11753 file format. Fix location of '-k' option of gcov command.
11755 2018-01-04 Martin Liska <mliska@suse.cz>
11758 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
11760 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11762 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
11764 2018-01-03 Martin Sebor <msebor@redhat.com>
11766 PR tree-optimization/83655
11767 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
11768 checking calls with invalid arguments.
11770 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11772 * tree-vect-stmts.c (vect_get_store_rhs): New function.
11773 (vectorizable_mask_load_store): Delete.
11774 (vectorizable_call): Return false for masked loads and stores.
11775 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
11776 instead of gimple_assign_rhs1.
11777 (vectorizable_load): Handle IFN_MASK_LOAD.
11778 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
11780 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11782 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
11784 (vectorizable_mask_load_store): ...here.
11785 (vectorizable_load): ...and here.
11787 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11789 * tree-vect-stmts.c (vect_build_all_ones_mask)
11790 (vect_build_zero_merge_argument): New functions, split out from...
11791 (vectorizable_load): ...here.
11793 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11795 * tree-vect-stmts.c (vect_check_store_rhs): New function,
11797 (vectorizable_mask_load_store): ...here.
11798 (vectorizable_store): ...and here.
11800 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11802 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
11804 (vectorizable_mask_load_store): ...here.
11806 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11808 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
11809 (vect_model_store_cost): Take a vec_load_store_type instead of a
11811 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
11812 (vect_model_store_cost): Take a vec_load_store_type instead of a
11814 (vectorizable_mask_load_store): Update accordingly.
11815 (vectorizable_store): Likewise.
11816 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
11818 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11820 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
11821 IFN_MASK_LOAD calls here rather than...
11822 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
11824 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11825 Alan Hayward <alan.hayward@arm.com>
11826 David Sherwood <david.sherwood@arm.com>
11828 * expmed.c (extract_bit_field_1): For vector extracts,
11829 fall back to extract_bit_field_as_subreg if vec_extract
11832 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11833 Alan Hayward <alan.hayward@arm.com>
11834 David Sherwood <david.sherwood@arm.com>
11836 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
11837 they are variable or constant sized.
11838 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
11839 slots for constant-sized data.
11841 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11842 Alan Hayward <alan.hayward@arm.com>
11843 David Sherwood <david.sherwood@arm.com>
11845 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
11846 handling COND_EXPRs with boolean comparisons, try to find a better
11847 basis for the mask type than the boolean itself.
11849 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11851 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
11852 is calculated and how it can be overridden.
11853 * genmodes.c (max_bitsize_mode_any_mode): New variable.
11854 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
11856 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
11859 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11860 Alan Hayward <alan.hayward@arm.com>
11861 David Sherwood <david.sherwood@arm.com>
11863 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
11864 Remove the mode argument.
11865 (aarch64_simd_valid_immediate): Remove the mode and inverse
11867 * config/aarch64/iterators.md (bitsize): New iterator.
11868 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
11869 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
11870 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
11871 aarch64_simd_valid_immediate.
11872 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
11873 (aarch64_reg_or_bic_imm): Likewise.
11874 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
11875 with an insn_type enum and msl with a modifier_type enum.
11876 Replace element_width with a scalar_mode. Change the shift
11877 to unsigned int. Add constructors for scalar_float_mode and
11878 scalar_int_mode elements.
11879 (aarch64_vect_float_const_representable_p): Delete.
11880 (aarch64_can_const_movi_rtx_p)
11881 (aarch64_simd_scalar_immediate_valid_for_move)
11882 (aarch64_simd_make_constant): Update call to
11883 aarch64_simd_valid_immediate.
11884 (aarch64_advsimd_valid_immediate_hs): New function.
11885 (aarch64_advsimd_valid_immediate): Likewise.
11886 (aarch64_simd_valid_immediate): Remove mode and inverse
11887 arguments. Rewrite to use the above. Use const_vec_duplicate_p
11888 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
11889 and aarch64_float_const_representable_p on the result.
11890 (aarch64_output_simd_mov_immediate): Remove mode argument.
11891 Update call to aarch64_simd_valid_immediate and use of
11892 simd_immediate_info.
11893 (aarch64_output_scalar_simd_mov_immediate): Update call
11896 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11897 Alan Hayward <alan.hayward@arm.com>
11898 David Sherwood <david.sherwood@arm.com>
11900 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
11901 (mode_nunits): Likewise CONST_MODE_NUNITS.
11902 * machmode.def (ADJUST_NUNITS): Document.
11903 * genmodes.c (mode_data::need_nunits_adj): New field.
11904 (blank_mode): Update accordingly.
11905 (adj_nunits): New variable.
11906 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
11908 (emit_mode_size_inline): Set need_bytesize_adj for all modes
11909 listed in adj_nunits.
11910 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
11911 listed in adj_nunits. Don't emit case statements for such modes.
11912 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
11913 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
11914 nothing if adj_nunits is nonnull.
11915 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
11916 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
11917 (emit_mode_fbit): Update use of print_maybe_const_decl.
11918 (emit_move_size): Likewise. Treat the array as non-const
11920 (emit_mode_adjustments): Handle adj_nunits.
11922 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11924 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
11925 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
11926 (VECTOR_MODES): Use it.
11927 (make_vector_modes): Take the prefix as an argument.
11929 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11930 Alan Hayward <alan.hayward@arm.com>
11931 David Sherwood <david.sherwood@arm.com>
11933 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
11934 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
11935 for MODE_VECTOR_BOOL.
11936 * machmode.def (VECTOR_BOOL_MODE): Document.
11937 * genmodes.c (VECTOR_BOOL_MODE): New macro.
11938 (make_vector_bool_mode): New function.
11939 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
11941 * lto-streamer-in.c (lto_input_mode_table): Likewise.
11942 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
11944 * stor-layout.c (int_mode_for_mode): Likewise.
11945 * tree.c (build_vector_type_for_mode): Likewise.
11946 * varasm.c (output_constant_pool_2): Likewise.
11947 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
11948 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
11949 for MODE_VECTOR_BOOL.
11950 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
11951 of mode class checks.
11952 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
11953 instead of a list of mode class checks.
11954 (expand_vector_scalar_condition): Likewise.
11955 (type_for_widest_vector_mode): Handle BImode as an inner mode.
11957 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11958 Alan Hayward <alan.hayward@arm.com>
11959 David Sherwood <david.sherwood@arm.com>
11961 * machmode.h (mode_size): Change from unsigned short to
11963 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
11964 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11965 or if measurement_type is not polynomial.
11966 (fixed_size_mode::includes_p): Check for constant-sized modes.
11967 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
11968 return a poly_uint16 rather than an unsigned short.
11969 (emit_mode_size): Change the type of mode_size from unsigned short
11970 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
11971 (emit_mode_adjustments): Cope with polynomial vector sizes.
11972 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11974 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11976 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
11977 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
11978 * caller-save.c (setup_save_areas): Likewise.
11979 (replace_reg_with_saved_mem): Likewise.
11980 * calls.c (emit_library_call_value_1): Likewise.
11981 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
11982 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
11983 (gen_lowpart_for_combine): Likewise.
11984 * convert.c (convert_to_integer_1): Likewise.
11985 * cse.c (equiv_constant, cse_insn): Likewise.
11986 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
11987 (cselib_subst_to_values): Likewise.
11988 * dce.c (word_dce_process_block): Likewise.
11989 * df-problems.c (df_word_lr_mark_ref): Likewise.
11990 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
11991 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
11992 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
11993 (rtl_for_decl_location): Likewise.
11994 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
11995 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
11996 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
11997 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
11998 (expand_expr_real_1): Likewise.
11999 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
12000 (pad_below): Likewise.
12001 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12002 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
12003 * ira.c (get_subreg_tracking_sizes): Likewise.
12004 * ira-build.c (ira_create_allocno_objects): Likewise.
12005 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
12006 (ira_sort_regnos_for_alter_reg): Likewise.
12007 * ira-costs.c (record_operand_costs): Likewise.
12008 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
12009 (resolve_simple_move): Likewise.
12010 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
12011 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
12012 (lra_constraints): Likewise.
12013 (CONST_POOL_OK_P): Reject variable-sized modes.
12014 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
12015 (add_pseudo_to_slot, lra_spill): Likewise.
12016 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12017 * optabs-query.c (get_best_extraction_insn): Likewise.
12018 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12019 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
12020 (expand_mult_highpart, valid_multiword_target_p): Likewise.
12021 * recog.c (offsettable_address_addr_space_p): Likewise.
12022 * regcprop.c (maybe_mode_change): Likewise.
12023 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
12024 * regrename.c (build_def_use): Likewise.
12025 * regstat.c (dump_reg_info): Likewise.
12026 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
12027 (find_reloads, find_reloads_subreg_address): Likewise.
12028 * reload1.c (eliminate_regs_1): Likewise.
12029 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
12030 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
12031 (simplify_binary_operation_1, simplify_subreg): Likewise.
12032 * targhooks.c (default_function_arg_padding): Likewise.
12033 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
12034 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
12035 (verify_gimple_assign_ternary): Likewise.
12036 * tree-inline.c (estimate_move_cost): Likewise.
12037 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12038 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
12039 (get_address_cost_ainc): Likewise.
12040 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
12041 (vect_supportable_dr_alignment): Likewise.
12042 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12043 (vectorizable_reduction): Likewise.
12044 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
12045 (vectorizable_operation, vectorizable_load): Likewise.
12046 * tree.c (build_same_sized_truth_vector_type): Likewise.
12047 * valtrack.c (cleanup_auto_inc_dec): Likewise.
12048 * var-tracking.c (emit_note_insn_var_location): Likewise.
12049 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
12050 (ADDR_VEC_ALIGN): Likewise.
12052 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12053 Alan Hayward <alan.hayward@arm.com>
12054 David Sherwood <david.sherwood@arm.com>
12056 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
12058 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
12059 or if measurement_type is polynomial.
12060 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
12061 * combine.c (make_extraction): Likewise.
12062 * dse.c (find_shift_sequence): Likewise.
12063 * dwarf2out.c (mem_loc_descriptor): Likewise.
12064 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
12065 (extract_bit_field, extract_low_bits): Likewise.
12066 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
12067 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
12068 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
12069 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
12070 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
12071 * reload.c (find_reloads): Likewise.
12072 * reload1.c (alter_reg): Likewise.
12073 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
12074 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
12075 * tree-if-conv.c (predicate_mem_writes): Likewise.
12076 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
12077 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
12078 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
12079 * valtrack.c (dead_debug_insert_temp): Likewise.
12080 * varasm.c (mergeable_constant_section): Likewise.
12081 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
12083 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12084 Alan Hayward <alan.hayward@arm.com>
12085 David Sherwood <david.sherwood@arm.com>
12087 * expr.c (expand_assignment): Cope with polynomial mode sizes
12088 when assigning to a CONCAT.
12090 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12091 Alan Hayward <alan.hayward@arm.com>
12092 David Sherwood <david.sherwood@arm.com>
12094 * machmode.h (mode_precision): Change from unsigned short to
12096 (mode_to_precision): Return a poly_uint16 rather than an unsigned
12098 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
12099 or if measurement_type is not polynomial.
12100 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
12101 in which the mode is already known to be a scalar_int_mode.
12102 * genmodes.c (emit_mode_precision): Change the type of mode_precision
12103 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
12105 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12106 for GET_MODE_PRECISION.
12107 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12108 for GET_MODE_PRECISION.
12109 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
12111 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
12112 (expand_field_assignment, make_extraction): Likewise.
12113 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
12114 (get_last_value): Likewise.
12115 * convert.c (convert_to_integer_1): Likewise.
12116 * cse.c (cse_insn): Likewise.
12117 * expr.c (expand_expr_real_1): Likewise.
12118 * lra-constraints.c (simplify_operand_subreg): Likewise.
12119 * optabs-query.c (can_atomic_load_p): Likewise.
12120 * optabs.c (expand_atomic_load): Likewise.
12121 (expand_atomic_store): Likewise.
12122 * ree.c (combine_reaching_defs): Likewise.
12123 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
12124 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
12125 * tree.h (type_has_mode_precision_p): Likewise.
12126 * ubsan.c (instrument_si_overflow): Likewise.
12128 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12129 Alan Hayward <alan.hayward@arm.com>
12130 David Sherwood <david.sherwood@arm.com>
12132 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
12133 polynomial numbers of units.
12134 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
12135 (valid_vector_subparts_p): New function.
12136 (build_vector_type): Remove temporary shim and take the number
12137 of units as a poly_uint64 rather than an int.
12138 (build_opaque_vector_type): Take the number of units as a
12139 poly_uint64 rather than an int.
12140 * tree.c (build_vector_from_ctor): Handle polynomial
12141 TYPE_VECTOR_SUBPARTS.
12142 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
12143 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
12144 (build_vector_from_val): If the number of units is variable,
12145 use build_vec_duplicate_cst for constant operands and
12146 VEC_DUPLICATE_EXPR otherwise.
12147 (make_vector_type): Remove temporary is_constant ().
12148 (build_vector_type, build_opaque_vector_type): Take the number of
12149 units as a poly_uint64 rather than an int.
12150 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
12152 * cfgexpand.c (expand_debug_expr): Likewise.
12153 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
12154 (store_constructor, expand_expr_real_1): Likewise.
12155 (const_scalar_mask_from_tree): Likewise.
12156 * fold-const-call.c (fold_const_reduction): Likewise.
12157 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
12158 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
12159 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
12160 (fold_relational_const): Likewise.
12161 (native_interpret_vector): Likewise. Change the size from an
12162 int to an unsigned int.
12163 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
12164 TYPE_VECTOR_SUBPARTS.
12165 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
12166 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
12167 duplicating a non-constant operand into a variable-length vector.
12168 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
12169 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
12170 * ipa-icf.c (sem_variable::equals): Likewise.
12171 * match.pd: Likewise.
12172 * omp-simd-clone.c (simd_clone_subparts): Likewise.
12173 * print-tree.c (print_node): Likewise.
12174 * stor-layout.c (layout_type): Likewise.
12175 * targhooks.c (default_builtin_vectorization_cost): Likewise.
12176 * tree-cfg.c (verify_gimple_comparison): Likewise.
12177 (verify_gimple_assign_binary): Likewise.
12178 (verify_gimple_assign_ternary): Likewise.
12179 (verify_gimple_assign_single): Likewise.
12180 * tree-pretty-print.c (dump_generic_node): Likewise.
12181 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12182 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
12183 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
12184 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
12185 (vect_shift_permute_load_chain): Likewise.
12186 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
12187 (expand_vector_condition, optimize_vector_constructor): Likewise.
12188 (lower_vec_perm, get_compute_type): Likewise.
12189 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
12190 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
12191 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
12192 (vect_recog_mask_conversion_pattern): Likewise.
12193 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
12194 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
12195 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12196 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
12197 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
12198 (vectorizable_shift, vectorizable_operation, vectorizable_store)
12199 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
12200 (supportable_widening_operation): Likewise.
12201 (supportable_narrowing_operation): Likewise.
12202 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
12204 * varasm.c (output_constant): Likewise.
12206 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12207 Alan Hayward <alan.hayward@arm.com>
12208 David Sherwood <david.sherwood@arm.com>
12210 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
12211 so that both the length == 3 and length != 3 cases set up their
12212 own permute vectors. Add comments explaining why we know the
12213 number of elements is constant.
12214 (vect_permute_load_chain): Likewise.
12216 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12217 Alan Hayward <alan.hayward@arm.com>
12218 David Sherwood <david.sherwood@arm.com>
12220 * machmode.h (mode_nunits): Change from unsigned char to
12222 (ONLY_FIXED_SIZE_MODES): New macro.
12223 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
12224 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
12225 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
12227 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
12228 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
12229 or if measurement_type is not polynomial.
12230 * genmodes.c (ZERO_COEFFS): New macro.
12231 (emit_mode_nunits_inline): Make mode_nunits_inline return a
12233 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
12234 Use ZERO_COEFFS when emitting initializers.
12235 * data-streamer.h (bp_pack_poly_value): New function.
12236 (bp_unpack_poly_value): Likewise.
12237 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
12238 for GET_MODE_NUNITS.
12239 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
12240 for GET_MODE_NUNITS.
12241 * tree.c (make_vector_type): Remove temporary shim and make
12242 the real function take the number of units as a poly_uint64
12243 rather than an int.
12244 (build_vector_type_for_mode): Handle polynomial nunits.
12245 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
12246 * emit-rtl.c (const_vec_series_p_1): Likewise.
12247 (gen_rtx_CONST_VECTOR): Likewise.
12248 * fold-const.c (test_vec_duplicate_folding): Likewise.
12249 * genrecog.c (validate_pattern): Likewise.
12250 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
12251 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
12252 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
12253 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
12254 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
12255 * rtlanal.c (subreg_get_info): Likewise.
12256 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12257 (vect_grouped_load_supported): Likewise.
12258 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
12259 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
12260 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12261 (simplify_const_unary_operation, simplify_binary_operation_1)
12262 (simplify_const_binary_operation, simplify_ternary_operation)
12263 (test_vector_ops_duplicate, test_vector_ops): Likewise.
12264 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
12265 instead of CONST_VECTOR_NUNITS.
12266 * varasm.c (output_constant_pool_2): Likewise.
12267 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
12268 explicit-encoded elements in the XVEC for variable-length vectors.
12270 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12272 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
12274 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12275 Alan Hayward <alan.hayward@arm.com>
12276 David Sherwood <david.sherwood@arm.com>
12278 * coretypes.h (fixed_size_mode): Declare.
12279 (fixed_size_mode_pod): New typedef.
12280 * builtins.h (target_builtins::x_apply_args_mode)
12281 (target_builtins::x_apply_result_mode): Change type to
12282 fixed_size_mode_pod.
12283 * builtins.c (apply_args_size, apply_result_size, result_vector)
12284 (expand_builtin_apply_args_1, expand_builtin_apply)
12285 (expand_builtin_return): Update accordingly.
12287 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12289 * cse.c (hash_rtx_cb): Hash only the encoded elements.
12290 * cselib.c (cselib_hash_rtx): Likewise.
12291 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
12292 CONST_VECTOR encoding.
12294 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12295 Jeff Law <law@redhat.com>
12298 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
12299 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
12300 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
12301 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
12304 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
12305 explicitly probe *sp in a noreturn function if there were any callee
12306 register saves or frame pointer is needed.
12308 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12311 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
12312 BLKmode for ternary, binary or unary expressions.
12315 * var-tracking.c (delete_vta_debug_insn): New inline function.
12316 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
12317 insns from get_insns () to NULL instead of each bb separately.
12318 Use delete_vta_debug_insn. No longer static.
12319 (vt_debug_insns_local, variable_tracking_main_1): Adjust
12320 delete_vta_debug_insns callers.
12321 * rtl.h (delete_vta_debug_insns): Declare.
12322 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
12323 instead of variable_tracking_main.
12325 2018-01-03 Martin Sebor <msebor@redhat.com>
12327 PR tree-optimization/83603
12328 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
12329 arguments past the endof the argument list in functions declared
12330 without a prototype.
12331 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
12332 Avoid checking when arguments are null.
12334 2018-01-03 Martin Sebor <msebor@redhat.com>
12337 * doc/extend.texi (attribute const): Fix a typo.
12338 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
12339 issuing -Wsuggest-attribute for void functions.
12341 2018-01-03 Martin Sebor <msebor@redhat.com>
12343 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
12344 offset_int::from instead of wide_int::to_shwi.
12345 (maybe_diag_overlap): Remove assertion.
12346 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
12347 * gimple-ssa-sprintf.c (format_directive): Same.
12348 (parse_directive): Same.
12349 (sprintf_dom_walker::compute_format_length): Same.
12350 (try_substitute_return_value): Same.
12352 2018-01-03 Jeff Law <law@redhat.com>
12354 PR middle-end/83654
12355 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
12356 non-constant residual for zero at runtime and avoid probing in
12357 that case. Reorganize code for trailing problem to mirror handling
12360 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12362 PR tree-optimization/83501
12363 * tree-ssa-strlen.c (get_string_cst): New.
12364 (handle_char_store): Call get_string_cst.
12366 2018-01-03 Martin Liska <mliska@suse.cz>
12368 PR tree-optimization/83593
12369 * tree-ssa-strlen.c: Include tree-cfg.h.
12370 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
12371 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
12372 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
12374 (strlen_dom_walker::before_dom_children): Call
12375 gimple_purge_dead_eh_edges. Dump tranformation with details
12377 (strlen_dom_walker::before_dom_children): Update call by adding
12378 new argument cleanup_eh.
12379 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
12381 2018-01-03 Martin Liska <mliska@suse.cz>
12384 * cif-code.def (VARIADIC_THUNK): New enum value.
12385 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
12388 2018-01-03 Jan Beulich <jbeulich@suse.com>
12390 * sse.md (mov<mode>_internal): Tighten condition for when to use
12391 vmovdqu<ssescalarsize> for TI and OI modes.
12393 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12395 Update copyright years.
12397 2018-01-03 Martin Liska <mliska@suse.cz>
12400 * ipa-visibility.c (function_and_variable_visibility): Skip
12401 functions with noipa attribure.
12403 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12405 * gcc.c (process_command): Update copyright notice dates.
12406 * gcov-dump.c (print_version): Ditto.
12407 * gcov.c (print_version): Ditto.
12408 * gcov-tool.c (print_version): Ditto.
12409 * gengtype.c (create_file): Ditto.
12410 * doc/cpp.texi: Bump @copying's copyright year.
12411 * doc/cppinternals.texi: Ditto.
12412 * doc/gcc.texi: Ditto.
12413 * doc/gccint.texi: Ditto.
12414 * doc/gcov.texi: Ditto.
12415 * doc/install.texi: Ditto.
12416 * doc/invoke.texi: Ditto.
12418 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12420 * vector-builder.h (vector_builder::m_full_nelts): Change from
12421 unsigned int to poly_uint64.
12422 (vector_builder::full_nelts): Update prototype accordingly.
12423 (vector_builder::new_vector): Likewise.
12424 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12425 (vector_builder::operator ==): Likewise.
12426 (vector_builder::finalize): Likewise.
12427 * int-vector-builder.h (int_vector_builder::int_vector_builder):
12428 Take the number of elements as a poly_uint64 rather than an
12430 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12431 from unsigned int to poly_uint64.
12432 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12433 (vec_perm_indices::new_vector): Likewise.
12434 (vec_perm_indices::length): Likewise.
12435 (vec_perm_indices::nelts_per_input): Likewise.
12436 (vec_perm_indices::input_nelts): Likewise.
12437 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12438 number of elements per input as a poly_uint64 rather than an
12439 unsigned int. Use the original encoding for variable-length
12440 vectors, rather than clamping each individual element.
12441 For the second and subsequent elements in each pattern,
12442 clamp the step and base before clamping their sum.
12443 (vec_perm_indices::series_p): Handle polynomial element counts.
12444 (vec_perm_indices::all_in_range_p): Likewise.
12445 (vec_perm_indices_to_tree): Likewise.
12446 (vec_perm_indices_to_rtx): Likewise.
12447 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12448 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12449 (tree_vector_builder::new_binary_operation): Handle polynomial
12450 element counts. Return false if we need to know the number
12451 of elements at compile time.
12452 * fold-const.c (fold_vec_perm): Punt if the number of elements
12453 isn't known at compile time.
12455 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12457 * vec-perm-indices.h (vec_perm_builder): Change element type
12458 from HOST_WIDE_INT to poly_int64.
12459 (vec_perm_indices::element_type): Update accordingly.
12460 (vec_perm_indices::clamp): Handle polynomial element_types.
12461 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12462 (vec_perm_indices::all_in_range_p): Likewise.
12463 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12465 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12466 polynomial vec_perm_indices element types.
12467 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12468 * fold-const.c (fold_vec_perm): Likewise.
12469 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12470 * tree-vect-generic.c (lower_vec_perm): Likewise.
12471 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12472 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12473 element type to HOST_WIDE_INT.
12475 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12476 Alan Hayward <alan.hayward@arm.com>
12477 David Sherwood <david.sherwood@arm.com>
12479 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12480 rather than an int. Use plus_constant.
12481 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12482 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12484 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12485 Alan Hayward <alan.hayward@arm.com>
12486 David Sherwood <david.sherwood@arm.com>
12488 * calls.c (emit_call_1, expand_call): Change struct_value_size from
12489 a HOST_WIDE_INT to a poly_int64.
12491 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12492 Alan Hayward <alan.hayward@arm.com>
12493 David Sherwood <david.sherwood@arm.com>
12495 * calls.c (load_register_parameters): Cope with polynomial
12496 mode sizes. Require a constant size for BLKmode parameters
12497 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
12498 forces a parameter to be padded at the lsb end in order to
12499 fill a complete number of words, require the parameter size
12500 to be ordered wrt UNITS_PER_WORD.
12502 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12503 Alan Hayward <alan.hayward@arm.com>
12504 David Sherwood <david.sherwood@arm.com>
12506 * reload1.c (spill_stack_slot_width): Change element type
12507 from unsigned int to poly_uint64_pod.
12508 (alter_reg): Treat mode sizes as polynomial.
12510 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12511 Alan Hayward <alan.hayward@arm.com>
12512 David Sherwood <david.sherwood@arm.com>
12514 * reload.c (complex_word_subreg_p): New function.
12515 (reload_inner_reg_of_subreg, push_reload): Use it.
12517 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12518 Alan Hayward <alan.hayward@arm.com>
12519 David Sherwood <david.sherwood@arm.com>
12521 * lra-constraints.c (process_alt_operands): Reject matched
12522 operands whose sizes aren't ordered.
12523 (match_reload): Refer to this check here.
12525 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12526 Alan Hayward <alan.hayward@arm.com>
12527 David Sherwood <david.sherwood@arm.com>
12529 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12530 that the mode size is in the set {1, 2, 4, 8, 16}.
12532 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12533 Alan Hayward <alan.hayward@arm.com>
12534 David Sherwood <david.sherwood@arm.com>
12536 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12537 Use plus_constant instead of gen_rtx_PLUS.
12539 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12540 Alan Hayward <alan.hayward@arm.com>
12541 David Sherwood <david.sherwood@arm.com>
12543 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12544 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12545 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12546 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12547 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12548 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12549 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12550 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12551 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12552 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12554 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12555 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12556 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12557 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12558 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12559 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12560 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12561 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12562 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12563 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12565 * expr.c (emit_move_resolve_push): Treat the input and result
12566 of PUSH_ROUNDING as a poly_int64.
12567 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12568 (emit_push_insn): Likewise.
12569 * lra-eliminations.c (mark_not_eliminable): Likewise.
12570 * recog.c (push_operand): Likewise.
12571 * reload1.c (elimination_effects): Likewise.
12572 * rtlanal.c (nonzero_bits1): Likewise.
12573 * calls.c (store_one_arg): Likewise. Require the padding to be
12574 known at compile time.
12576 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12577 Alan Hayward <alan.hayward@arm.com>
12578 David Sherwood <david.sherwood@arm.com>
12580 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12581 Use plus_constant instead of gen_rtx_PLUS.
12583 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12584 Alan Hayward <alan.hayward@arm.com>
12585 David Sherwood <david.sherwood@arm.com>
12587 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12588 rather than an int.
12590 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12591 Alan Hayward <alan.hayward@arm.com>
12592 David Sherwood <david.sherwood@arm.com>
12594 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12595 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12596 via stack temporaries. Treat the mode size as polynomial too.
12598 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12599 Alan Hayward <alan.hayward@arm.com>
12600 David Sherwood <david.sherwood@arm.com>
12602 * expr.c (expand_expr_real_2): When handling conversions involving
12603 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12604 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12605 as a poly_uint64 too.
12607 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12608 Alan Hayward <alan.hayward@arm.com>
12609 David Sherwood <david.sherwood@arm.com>
12611 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
12613 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12614 Alan Hayward <alan.hayward@arm.com>
12615 David Sherwood <david.sherwood@arm.com>
12617 * combine.c (can_change_dest_mode): Handle polynomial
12618 REGMODE_NATURAL_SIZE.
12619 * expmed.c (store_bit_field_1): Likewise.
12620 * expr.c (store_constructor): Likewise.
12621 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12622 and polynomial REGMODE_NATURAL_SIZE.
12623 (gen_lowpart_common): Likewise.
12624 * reginfo.c (record_subregs_of_mode): Likewise.
12625 * rtlanal.c (read_modify_subreg_p): Likewise.
12627 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12628 Alan Hayward <alan.hayward@arm.com>
12629 David Sherwood <david.sherwood@arm.com>
12631 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12632 numbers of elements.
12634 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12635 Alan Hayward <alan.hayward@arm.com>
12636 David Sherwood <david.sherwood@arm.com>
12638 * match.pd: Cope with polynomial numbers of vector elements.
12640 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12641 Alan Hayward <alan.hayward@arm.com>
12642 David Sherwood <david.sherwood@arm.com>
12644 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12645 in a POINTER_PLUS_EXPR.
12647 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12648 Alan Hayward <alan.hayward@arm.com>
12649 David Sherwood <david.sherwood@arm.com>
12651 * omp-simd-clone.c (simd_clone_subparts): New function.
12652 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12653 (ipa_simd_modify_function_body): Likewise.
12655 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12656 Alan Hayward <alan.hayward@arm.com>
12657 David Sherwood <david.sherwood@arm.com>
12659 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
12660 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
12661 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
12662 (expand_vector_condition, vector_element): Likewise.
12663 (subparts_gt): New function.
12664 (get_compute_type): Use subparts_gt.
12665 (count_type_subparts): Delete.
12666 (expand_vector_operations_1): Use subparts_gt instead of
12667 count_type_subparts.
12669 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12670 Alan Hayward <alan.hayward@arm.com>
12671 David Sherwood <david.sherwood@arm.com>
12673 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
12674 (vect_compile_time_alias): ...this new function. Do the calculation
12675 on poly_ints rather than trees.
12676 (vect_prune_runtime_alias_test_list): Update call accordingly.
12678 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12679 Alan Hayward <alan.hayward@arm.com>
12680 David Sherwood <david.sherwood@arm.com>
12682 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
12684 (vect_schedule_slp_instance): Likewise.
12686 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12687 Alan Hayward <alan.hayward@arm.com>
12688 David Sherwood <david.sherwood@arm.com>
12690 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12691 constant and extern definitions for variable-length vectors.
12692 (vect_get_constant_vectors): Note that the number of units
12693 is known to be constant.
12695 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12696 Alan Hayward <alan.hayward@arm.com>
12697 David Sherwood <david.sherwood@arm.com>
12699 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
12700 of units as polynomial. Choose between WIDE and NARROW based
12703 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12704 Alan Hayward <alan.hayward@arm.com>
12705 David Sherwood <david.sherwood@arm.com>
12707 * tree-vect-stmts.c (simd_clone_subparts): New function.
12708 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
12710 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12711 Alan Hayward <alan.hayward@arm.com>
12712 David Sherwood <david.sherwood@arm.com>
12714 * tree-vect-stmts.c (vectorizable_call): Treat the number of
12715 vectors as polynomial. Use build_index_vector for
12716 IFN_GOMP_SIMD_LANE.
12718 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12719 Alan Hayward <alan.hayward@arm.com>
12720 David Sherwood <david.sherwood@arm.com>
12722 * tree-vect-stmts.c (get_load_store_type): Treat the number of
12723 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
12724 for variable-length vectors.
12725 (vectorizable_mask_load_store): Treat the number of units as
12726 polynomial, asserting that it is constant if the condition has
12727 already been enforced.
12728 (vectorizable_store, vectorizable_load): Likewise.
12730 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12731 Alan Hayward <alan.hayward@arm.com>
12732 David Sherwood <david.sherwood@arm.com>
12734 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
12735 of units as polynomial. Punt if we can't tell at compile time
12736 which vector contains the final result.
12738 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12739 Alan Hayward <alan.hayward@arm.com>
12740 David Sherwood <david.sherwood@arm.com>
12742 * tree-vect-loop.c (vectorizable_induction): Treat the number
12743 of units as polynomial. Punt on SLP inductions. Use an integer
12744 VEC_SERIES_EXPR for variable-length integer reductions. Use a
12745 cast of such a series for variable-length floating-point
12748 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12749 Alan Hayward <alan.hayward@arm.com>
12750 David Sherwood <david.sherwood@arm.com>
12752 * tree.h (build_index_vector): Declare.
12753 * tree.c (build_index_vector): New function.
12754 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
12755 of units as polynomial, forcibly converting it to a constant if
12756 vectorizable_reduction has already enforced the condition.
12757 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
12758 to create a {1,2,3,...} vector.
12759 (vectorizable_reduction): Treat the number of units as polynomial.
12760 Choose vectype_in based on the largest scalar element size rather
12761 than the smallest number of units. Enforce the restrictions
12764 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12765 Alan Hayward <alan.hayward@arm.com>
12766 David Sherwood <david.sherwood@arm.com>
12768 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
12769 number of units as polynomial.
12771 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12772 Alan Hayward <alan.hayward@arm.com>
12773 David Sherwood <david.sherwood@arm.com>
12775 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
12776 * target.def (autovectorize_vector_sizes): Return the vector sizes
12777 by pointer, using vector_sizes rather than a bitmask.
12778 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
12779 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
12780 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
12782 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
12783 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
12784 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
12785 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
12786 * omp-general.c (omp_max_vf): Likewise.
12787 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12788 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
12789 * tree-vect-loop.c (vect_analyze_loop): Likewise.
12790 * tree-vect-slp.c (vect_slp_bb): Likewise.
12791 * doc/tm.texi: Regenerate.
12792 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
12794 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
12795 the vector size as a poly_uint64 rather than an unsigned int.
12796 (current_vector_size): Change from an unsigned int to a poly_uint64.
12797 (get_vectype_for_scalar_type): Update accordingly.
12798 * tree.h (build_truth_vector_type): Take the size and number of
12799 units as a poly_uint64 rather than an unsigned int.
12800 (build_vector_type): Add a temporary overload that takes
12801 the number of units as a poly_uint64 rather than an unsigned int.
12802 * tree.c (make_vector_type): Likewise.
12803 (build_truth_vector_type): Take the number of units as a poly_uint64
12804 rather than an unsigned int.
12806 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12807 Alan Hayward <alan.hayward@arm.com>
12808 David Sherwood <david.sherwood@arm.com>
12810 * target.def (get_mask_mode): Take the number of units and length
12811 as poly_uint64s rather than unsigned ints.
12812 * targhooks.h (default_get_mask_mode): Update accordingly.
12813 * targhooks.c (default_get_mask_mode): Likewise.
12814 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
12815 * doc/tm.texi: Regenerate.
12817 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12818 Alan Hayward <alan.hayward@arm.com>
12819 David Sherwood <david.sherwood@arm.com>
12821 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
12822 * omp-general.c (omp_max_vf): Likewise.
12823 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
12824 (expand_omp_simd): Handle polynomial safelen.
12825 * omp-low.c (omplow_simd_context): Add a default constructor.
12826 (omplow_simd_context::max_vf): Change from int to poly_uint64.
12827 (lower_rec_simd_input_clauses): Update accordingly.
12828 (lower_rec_input_clauses): Likewise.
12830 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12831 Alan Hayward <alan.hayward@arm.com>
12832 David Sherwood <david.sherwood@arm.com>
12834 * tree-vectorizer.h (vect_nunits_for_cost): New function.
12835 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
12836 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
12837 (vect_analyze_slp_cost): Likewise.
12838 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
12839 (vect_model_load_cost): Likewise.
12841 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12842 Alan Hayward <alan.hayward@arm.com>
12843 David Sherwood <david.sherwood@arm.com>
12845 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
12846 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
12847 from an unsigned int * to a poly_uint64_pod *.
12848 (calculate_unrolling_factor): New function.
12849 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
12851 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12852 Alan Hayward <alan.hayward@arm.com>
12853 David Sherwood <david.sherwood@arm.com>
12855 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
12856 from an unsigned int to a poly_uint64.
12857 (_loop_vec_info::slp_unrolling_factor): Likewise.
12858 (_loop_vec_info::vectorization_factor): Change from an int
12860 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
12861 (vect_get_num_vectors): New function.
12862 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
12863 (vect_get_num_copies): Use vect_get_num_vectors.
12864 (vect_analyze_data_ref_dependences): Change max_vf from an int *
12865 to an unsigned int *.
12866 (vect_analyze_data_refs): Change min_vf from an int * to a
12868 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12869 than an unsigned HOST_WIDE_INT.
12870 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
12871 (vect_analyze_data_ref_dependence): Change max_vf from an int *
12872 to an unsigned int *.
12873 (vect_analyze_data_ref_dependences): Likewise.
12874 (vect_compute_data_ref_alignment): Handle polynomial vf.
12875 (vect_enhance_data_refs_alignment): Likewise.
12876 (vect_prune_runtime_alias_test_list): Likewise.
12877 (vect_shift_permute_load_chain): Likewise.
12878 (vect_supportable_dr_alignment): Likewise.
12879 (dependence_distance_ge_vf): Take the vectorization factor as a
12880 poly_uint64 rather than an unsigned HOST_WIDE_INT.
12881 (vect_analyze_data_refs): Change min_vf from an int * to a
12883 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
12884 vfm1 as a poly_uint64 rather than an int. Make the same change
12885 for the returned bound_scalar.
12886 (vect_gen_vector_loop_niters): Handle polynomial vf.
12887 (vect_do_peeling): Likewise. Update call to
12888 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
12889 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
12891 * tree-vect-loop.c (vect_determine_vectorization_factor)
12892 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
12893 (vect_get_known_peeling_cost): Likewise.
12894 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
12895 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
12896 (vect_transform_loop): Likewise. Use the lowest possible VF when
12897 updating the upper bounds of the loop.
12898 (vect_min_worthwhile_factor): Make static. Return an unsigned int
12899 rather than an int.
12900 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
12901 polynomial unroll factors.
12902 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
12903 (vect_make_slp_decision): Likewise.
12904 (vect_supported_load_permutation_p): Likewise, and polynomial
12906 (vect_analyze_slp_cost): Handle polynomial vf.
12907 (vect_slp_analyze_node_operations): Likewise.
12908 (vect_slp_analyze_bb_1): Likewise.
12909 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12910 than an unsigned HOST_WIDE_INT.
12911 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
12912 (vectorizable_load): Handle polynomial vf.
12913 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
12915 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
12917 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12918 Alan Hayward <alan.hayward@arm.com>
12919 David Sherwood <david.sherwood@arm.com>
12921 * match.pd: Handle bit operations involving three constants
12922 and try to fold one pair.
12924 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12926 * tree-vect-loop-manip.c: Include gimple-fold.h.
12927 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
12928 niters_maybe_zero parameters. Handle other cases besides a step of 1.
12929 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
12930 Add a path that uses a step of VF instead of 1, but disable it
12932 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
12933 and niters_no_overflow parameters. Update calls to
12934 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
12935 Create a new SSA name if the latter choses to use a ste other
12936 than zero, and return it via niters_vector_mult_vf_var.
12937 * tree-vect-loop.c (vect_transform_loop): Update calls to
12938 vect_do_peeling, vect_gen_vector_loop_niters and
12939 slpeel_make_loop_iterate_ntimes.
12940 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
12941 (vect_gen_vector_loop_niters): Update declarations after above changes.
12943 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
12945 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
12946 128-bit round to integer instructions.
12947 (ceil<mode>2): Likewise.
12948 (btrunc<mode>2): Likewise.
12949 (round<mode>2): Likewise.
12951 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12953 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
12954 unaligned VSX load/store on P8/P9.
12955 (expand_block_clear): Allow the use of unaligned VSX
12956 load/store on P8/P9.
12958 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12960 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
12962 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
12963 swap associated with both a load and a store.
12965 2018-01-02 Andrew Waterman <andrew@sifive.com>
12967 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
12968 * config/riscv/riscv.md (clear_cache): Use it.
12970 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
12972 * web.c: Remove out-of-date comment.
12974 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12976 * expr.c (fixup_args_size_notes): Check that any existing
12977 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
12978 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
12979 (emit_single_push_insn): ...here.
12981 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12983 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
12984 (const_vector_encoded_nelts): New function.
12985 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
12986 (const_vector_int_elt, const_vector_elt): Declare.
12987 * emit-rtl.c (const_vector_int_elt_1): New function.
12988 (const_vector_elt): Likewise.
12989 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
12990 of CONST_VECTOR_ELT.
12992 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12994 * expr.c: Include rtx-vector-builder.h.
12995 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
12996 directly on the tree encoding.
12997 (const_vector_from_tree): Likewise.
12998 * optabs.c: Include rtx-vector-builder.h.
12999 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
13000 sequence of "u" values.
13001 * vec-perm-indices.c: Include rtx-vector-builder.h.
13002 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
13003 directly on the vec_perm_indices encoding.
13005 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13007 * doc/rtl.texi (const_vector): Describe new encoding scheme.
13008 * Makefile.in (OBJS): Add rtx-vector-builder.o.
13009 * rtx-vector-builder.h: New file.
13010 * rtx-vector-builder.c: Likewise.
13011 * rtl.h (rtx_def::u2): Add a const_vector field.
13012 (CONST_VECTOR_NPATTERNS): New macro.
13013 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
13014 (CONST_VECTOR_DUPLICATE_P): Likewise.
13015 (CONST_VECTOR_STEPPED_P): Likewise.
13016 (CONST_VECTOR_ENCODED_ELT): Likewise.
13017 (const_vec_duplicate_p): Check for a duplicated vector encoding.
13018 (unwrap_const_vec_duplicate): Likewise.
13019 (const_vec_series_p): Check for a non-duplicated vector encoding.
13020 Say that the function only returns true for integer vectors.
13021 * emit-rtl.c: Include rtx-vector-builder.h.
13022 (gen_const_vec_duplicate_1): Delete.
13023 (gen_const_vector): Call gen_const_vec_duplicate instead of
13024 gen_const_vec_duplicate_1.
13025 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
13026 (gen_const_vec_duplicate): Use rtx_vector_builder.
13027 (gen_const_vec_series): Likewise.
13028 (gen_rtx_CONST_VECTOR): Likewise.
13029 * config/powerpcspe/powerpcspe.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.
13033 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
13034 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
13035 Build a new vector rather than modifying a CONST_VECTOR in-place.
13036 (handle_special_swappables): Update call accordingly.
13038 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13040 * simplify-rtx.c (simplify_const_binary_operation): Use
13041 CONST_VECTOR_ELT instead of XVECEXP.
13043 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13045 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
13046 the selector elements to be different from the data elements
13047 if the selector is a VECTOR_CST.
13048 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
13049 ssizetype for the selector.
13051 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13053 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
13054 before testing each element individually.
13055 * tree-vect-generic.c (lower_vec_perm): Likewise.
13057 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13059 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
13060 * selftest-run-tests.c (selftest::run_tests): Call it.
13061 * vector-builder.h (vector_builder::operator ==): New function.
13062 (vector_builder::operator !=): Likewise.
13063 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
13064 (vec_perm_indices::all_from_input_p): New function.
13065 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13066 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
13067 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
13068 instead of reading the VECTOR_CST directly. Detect whether both
13069 vector inputs are the same before constructing the vec_perm_indices,
13070 and update the number of inputs argument accordingly. Use the
13071 utility functions added above. Only construct sel2 if we need to.
13073 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13075 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
13076 the broadcast of the low byte.
13077 (expand_mult_highpart): Use an explicit encoding for the permutes.
13078 * optabs-query.c (can_mult_highpart_p): Likewise.
13079 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
13080 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13081 (vectorizable_bswap): Likewise.
13082 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
13083 explicit encoding for the power-of-2 permutes.
13084 (vect_permute_store_chain): Likewise.
13085 (vect_grouped_load_supported): Likewise.
13086 (vect_permute_load_chain): Likewise.
13088 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13090 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
13091 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
13092 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
13093 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13094 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
13095 (vect_gen_perm_mask_any): Likewise.
13097 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13099 * int-vector-builder.h: New file.
13100 * vec-perm-indices.h: Include int-vector-builder.h.
13101 (vec_perm_indices): Redefine as an int_vector_builder.
13102 (auto_vec_perm_indices): Delete.
13103 (vec_perm_builder): Redefine as a stand-alone class.
13104 (vec_perm_indices::vec_perm_indices): New function.
13105 (vec_perm_indices::clamp): Likewise.
13106 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
13107 (vec_perm_indices::new_vector): New function.
13108 (vec_perm_indices::new_expanded_vector): Update for new
13109 vec_perm_indices class.
13110 (vec_perm_indices::rotate_inputs): New function.
13111 (vec_perm_indices::all_in_range_p): Operate directly on the
13112 encoded form, without computing elided elements.
13113 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
13114 encoding. Update for new vec_perm_indices class.
13115 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
13116 the given vec_perm_builder.
13117 (expand_vec_perm_var): Update vec_perm_builder constructor.
13118 (expand_mult_highpart): Use vec_perm_builder instead of
13119 auto_vec_perm_indices.
13120 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
13121 vec_perm_indices instead of auto_vec_perm_indices. Use a single
13122 or double series encoding as appropriate.
13123 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
13124 vec_perm_indices instead of auto_vec_perm_indices.
13125 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13126 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13127 (vect_permute_store_chain): Likewise.
13128 (vect_grouped_load_supported): Likewise.
13129 (vect_permute_load_chain): Likewise.
13130 (vect_shift_permute_load_chain): Likewise.
13131 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13132 (vect_transform_slp_perm_load): Likewise.
13133 (vect_schedule_slp_instance): Likewise.
13134 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13135 (vectorizable_mask_load_store): Likewise.
13136 (vectorizable_bswap): Likewise.
13137 (vectorizable_store): Likewise.
13138 (vectorizable_load): Likewise.
13139 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
13140 vec_perm_indices instead of auto_vec_perm_indices. Use
13141 tree_to_vec_perm_builder to read the vector from a tree.
13142 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
13143 vec_perm_builder instead of a vec_perm_indices.
13144 (have_whole_vector_shift): Use vec_perm_builder and
13145 vec_perm_indices instead of auto_vec_perm_indices. Leave the
13146 truncation to calc_vec_perm_mask_for_shift.
13147 (vect_create_epilog_for_reduction): Likewise.
13148 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
13149 from auto_vec_perm_indices to vec_perm_indices.
13150 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13151 instead of changing individual elements.
13152 (aarch64_vectorize_vec_perm_const): Use new_vector to install
13153 the vector in d.perm.
13154 * config/arm/arm.c (expand_vec_perm_d::perm): Change
13155 from auto_vec_perm_indices to vec_perm_indices.
13156 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
13157 instead of changing individual elements.
13158 (arm_vectorize_vec_perm_const): Use new_vector to install
13159 the vector in d.perm.
13160 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
13161 Update vec_perm_builder constructor.
13162 (rs6000_expand_interleave): Likewise.
13163 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
13164 (rs6000_expand_interleave): Likewise.
13166 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13168 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
13169 to qimode could truncate the indices.
13170 * optabs.c (expand_vec_perm_var): Likewise.
13172 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13174 * Makefile.in (OBJS): Add vec-perm-indices.o.
13175 * vec-perm-indices.h: New file.
13176 * vec-perm-indices.c: Likewise.
13177 * target.h (vec_perm_indices): Replace with a forward class
13179 (auto_vec_perm_indices): Move to vec-perm-indices.h.
13180 * optabs.h: Include vec-perm-indices.h.
13181 (expand_vec_perm): Delete.
13182 (selector_fits_mode_p, expand_vec_perm_var): Declare.
13183 (expand_vec_perm_const): Declare.
13184 * target.def (vec_perm_const_ok): Replace with...
13185 (vec_perm_const): ...this new hook.
13186 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
13187 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
13188 * doc/tm.texi: Regenerate.
13189 * optabs.def (vec_perm_const): Delete.
13190 * doc/md.texi (vec_perm_const): Likewise.
13191 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
13192 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
13193 expand_vec_perm for constant permutation vectors. Assert that
13194 the mode of variable permutation vectors is the integer equivalent
13195 of the mode that is being permuted.
13196 * optabs-query.h (selector_fits_mode_p): Declare.
13197 * optabs-query.c: Include vec-perm-indices.h.
13198 (selector_fits_mode_p): New function.
13199 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
13200 is defined, instead of checking whether the vec_perm_const_optab
13201 exists. Use targetm.vectorize.vec_perm_const instead of
13202 targetm.vectorize.vec_perm_const_ok. Check whether the indices
13203 fit in the vector mode before using a variable permute.
13204 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
13205 vec_perm_indices instead of an rtx.
13206 (expand_vec_perm): Replace with...
13207 (expand_vec_perm_const): ...this new function. Take the selector
13208 as a vec_perm_indices rather than an rtx. Also take the mode of
13209 the selector. Update call to shift_amt_for_vec_perm_mask.
13210 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
13211 Use vec_perm_indices::new_expanded_vector to expand the original
13212 selector into bytes. Check whether the indices fit in the vector
13213 mode before using a variable permute.
13214 (expand_vec_perm_var): Make global.
13215 (expand_mult_highpart): Use expand_vec_perm_const.
13216 * fold-const.c: Includes vec-perm-indices.h.
13217 * tree-ssa-forwprop.c: Likewise.
13218 * tree-vect-data-refs.c: Likewise.
13219 * tree-vect-generic.c: Likewise.
13220 * tree-vect-loop.c: Likewise.
13221 * tree-vect-slp.c: Likewise.
13222 * tree-vect-stmts.c: Likewise.
13223 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
13225 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
13226 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
13227 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
13228 (aarch64_vectorize_vec_perm_const): ...this new function.
13229 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13230 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13231 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
13232 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
13233 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13234 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13235 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
13237 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
13238 check for NEON modes.
13239 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
13240 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
13241 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
13242 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
13244 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
13245 the old VEC_PERM_CONST conditions.
13246 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
13247 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
13248 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
13249 (ia64_vectorize_vec_perm_const_ok): Merge into...
13250 (ia64_vectorize_vec_perm_const): ...this new function.
13251 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
13252 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
13253 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
13254 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
13255 * config/mips/mips.c (mips_expand_vec_perm_const)
13256 (mips_vectorize_vec_perm_const_ok): Merge into...
13257 (mips_vectorize_vec_perm_const): ...this new function.
13258 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
13259 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
13260 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
13261 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
13262 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
13263 (rs6000_expand_vec_perm_const): Delete.
13264 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
13266 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13267 (altivec_expand_vec_perm_const_le): Take each operand individually.
13268 Operate on constant selectors rather than rtxes.
13269 (altivec_expand_vec_perm_const): Likewise. Update call to
13270 altivec_expand_vec_perm_const_le.
13271 (rs6000_expand_vec_perm_const): Delete.
13272 (rs6000_vectorize_vec_perm_const_ok): Delete.
13273 (rs6000_vectorize_vec_perm_const): New function.
13274 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13275 an element count and rtx array.
13276 (rs6000_expand_extract_even): Update call accordingly.
13277 (rs6000_expand_interleave): Likewise.
13278 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
13279 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
13280 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
13281 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
13282 (rs6000_expand_vec_perm_const): Delete.
13283 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
13284 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13285 (altivec_expand_vec_perm_const_le): Take each operand individually.
13286 Operate on constant selectors rather than rtxes.
13287 (altivec_expand_vec_perm_const): Likewise. Update call to
13288 altivec_expand_vec_perm_const_le.
13289 (rs6000_expand_vec_perm_const): Delete.
13290 (rs6000_vectorize_vec_perm_const_ok): Delete.
13291 (rs6000_vectorize_vec_perm_const): New function. Remove stray
13292 reference to the SPE evmerge intructions.
13293 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
13294 an element count and rtx array.
13295 (rs6000_expand_extract_even): Update call accordingly.
13296 (rs6000_expand_interleave): Likewise.
13297 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
13298 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
13300 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
13302 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13304 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
13305 vector mode and that that mode matches the mode of the data
13307 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
13308 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
13309 directly using expand_vec_perm_1 when forcing selectors into
13311 (expand_vec_perm_var): New function, split out from expand_vec_perm.
13313 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13315 * optabs-query.h (can_vec_perm_p): Delete.
13316 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
13317 * optabs-query.c (can_vec_perm_p): Split into...
13318 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
13319 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
13320 particular selector is valid.
13321 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13322 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13323 (vect_grouped_load_supported): Likewise.
13324 (vect_shift_permute_load_chain): Likewise.
13325 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13326 (vect_transform_slp_perm_load): Likewise.
13327 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13328 (vectorizable_bswap): Likewise.
13329 (vect_gen_perm_mask_checked): Likewise.
13330 * fold-const.c (fold_ternary_loc): Likewise. Don't take
13331 implementations of variable permutation vectors into account
13332 when deciding which selector to use.
13333 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
13334 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
13335 with a false third argument.
13336 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
13337 to test whether the constant selector is valid and can_vec_perm_var_p
13338 to test whether a variable selector is valid.
13340 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13342 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
13343 * optabs-query.c (can_vec_perm_p): Likewise.
13344 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
13345 instead of vec_perm_indices.
13346 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
13347 (vect_gen_perm_mask_checked): Likewise,
13348 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
13349 (vect_gen_perm_mask_checked): Likewise,
13351 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13353 * optabs-query.h (qimode_for_vec_perm): Declare.
13354 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
13355 (qimode_for_vec_perm): ...this new function.
13356 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
13358 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13360 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
13361 does not have a conditional at the top.
13363 2018-01-02 Richard Biener <rguenther@suse.de>
13365 * ipa-inline.c (big_speedup_p): Fix expression.
13367 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13370 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
13373 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13377 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
13378 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
13379 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
13380 cond_taken_branch_cost 3->4.
13382 2018-01-01 Jakub Jelinek <jakub@redhat.com>
13384 PR tree-optimization/83581
13385 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
13386 TODO_cleanup_cfg if any changes have been made.
13388 PR middle-end/83608
13389 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
13390 convert_modes if target mode has the right side, but different mode
13393 PR middle-end/83609
13394 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
13395 last argument when extracting from CONCAT. If either from_real or
13396 from_imag is NULL, use expansion through memory. If result is not
13397 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
13398 the parts directly to inner mode, if even that fails, use expansion
13401 PR middle-end/83623
13402 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
13403 check for bswap in mode rather than HImode and use that in expand_unop
13406 Copyright (C) 2018 Free Software Foundation, Inc.
13408 Copying and distribution of this file, with or without modification,
13409 are permitted in any medium without royalty provided the copyright
13410 notice and this notice are preserved.