1 2018-05-29 Richard Biener <rguenther@suse.de>
3 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
4 sure to use non-pattern stmts for get_earlier_stmt arguments.
5 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
6 called on pattern stmts.
7 (get_later_stmt): Likewise.
9 2018-05-29 Martin Liska <mliska@suse.cz>
12 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
15 2018-05-29 Jakub Jelinek <jakub@redhat.com>
17 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
19 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
23 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
24 VEC_PACK_FLOAT_EXPR): New tree codes.
25 * tree-pretty-print.c (op_code_prio): Handle
26 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
27 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
28 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
29 * tree-inline.c (estimate_operator_cost): Likewise.
30 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
31 * fold-const.c (const_binop): Likewise.
32 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
33 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
34 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
35 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
36 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
37 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
38 * expr.c (expand_expr_real_2): Likewise.
39 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
40 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
41 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
43 * optabs.c (expand_widen_pattern_expr): For
44 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
45 sign from result type rather than operand's type.
46 (expand_binop_directly): For vec_packu_float_optab and
47 vec_packs_float_optab allow result type to be different from operand's
49 * optabs-tree.c (optab_for_tree_code): Handle
50 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
51 VEC_PACK_FLOAT_EXPR. Formatting fixes.
52 * tree-vect-generic.c (expand_vector_operations_1): Handle
53 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
55 * tree-vect-stmts.c (supportable_widening_operation): Handle
57 (supportable_narrowing_operation): Handle FLOAT_EXPR.
58 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
59 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
60 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
61 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
63 (vec_pack<floatprefix>_float_<mode>): New expander.
64 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
66 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
67 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
68 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
69 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
70 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
72 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
73 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
74 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
75 VEC_PACK_FLOAT_EXPR): Document.
77 2018-05-29 Richard Biener <rguenther@suse.de>
79 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
81 (stmt_vec_info_vec): Make pointer.
82 (init_stmt_vec_info_vec): Remove.
83 (free_stmt_vec_info_vec): Likewise.
84 (set_stmt_vec_info_vec): New function.
85 (free_stmt_vec_infos): Likewise.
86 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
87 (set_vinfo_for_stmt): Likewise.
88 (get_earlier_stmt): Likewise.
89 (get_later_stmt): Likewise.
90 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
91 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
92 (vec_info::~vec_info): Free stmt_vec_infos.
93 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
94 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
95 (pass_slp_vectorize::execute): Likewise.
96 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
97 (free_stmt_vec_info_vec): Likewise.
98 (set_stmt_vec_info_vec): New function.
99 (free_stmt_vec_infos): Likewise.
100 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
101 the global stmt_vec_info_vec.
102 * tree-parloops.c (gather_scalar_reductions): Use
103 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
106 2018-05-29 Richard Biener <rguenther@suse.de>
108 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
110 2018-05-29 Martin Liska <mliska@suse.cz>
111 David Malcolm <dmalcolm@redhat.com>
113 * vec.c (test_reverse): New.
114 (vec_c_tests): Add new test.
115 * vec.h (vl_ptr>::reverse): New function.
117 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
119 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
121 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
124 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
126 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
128 2018-05-28 Richard Biener <rguenther@suse.de>
130 PR tree-optimization/85933
131 * tree-vect-data-refs.c (vect_record_base_alignments): Only
132 look at stmts marked as vectorizable.
134 2018-05-28 Richard Biener <rguenther@suse.de>
136 PR tree-optimization/85934
137 * tree-vect-generic.c (expand_vector_operations_1): Hoist
138 vector boolean check before scalar optimization.
140 2018-05-28 Jakub Jelinek <jakub@redhat.com>
142 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
145 2018-05-28 Mark Wielaard <mark@klomp.org>
147 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
148 if it is an expression containing a minus sign.
150 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
152 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
154 2018-05-27 Paul Koning <ni1d@arrl.net>
156 * config/pdp11/pdp11.md (truncsihi2): Remove.
158 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
159 Chung-Ju Wu <jasonwucj@gmail.com>
161 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
163 (unaligned_store_dw): Ditto.
164 * config/nds32/nds32-memory-manipulation.c
165 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
166 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
167 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
168 (emit_setmem_word_loop): Rename to ...
169 (emit_setmem_doubleword_loop): ... this.
170 (nds32_gen_dup_4_byte_to_word_value): New function.
171 (nds32_gen_dup_8_byte_to_double_word_value): New function.
172 (nds32_expand_setmem_loop): Refine implementation.
173 (nds32_expand_setmem_loop_v3m): Ditto.
174 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
177 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
179 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
181 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
183 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
184 (nds32_init_machine_status): Initialize machine->attr_naked_p and
185 machine->attr_no_prologue_p.
186 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
187 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
188 (nds32_expand_epilogue): Consider attr_naked_p.
189 (nds32_expand_epilogue_v3pop): Likewise.
190 (nds32_can_use_return_insn): Likewise.
191 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
192 attr_no_prologue_p fields.
193 * config/nds32/nds32.opt (mret-in-naked-func): New option.
195 2018-05-27 Jakub Jelinek <jakub@redhat.com>
198 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
201 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
203 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
205 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
207 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
209 (*<floatsuffix>floatv2div2sf2): Rename to ...
210 (*float<floatunssuffix>v2div2sf2): ... this.
211 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
212 (float<floatunssuffix>v2div2sf2_mask): ... this.
213 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
214 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
215 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
217 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
219 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
221 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
223 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
225 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
227 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
228 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
229 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
230 gen_ufix_truncv8dfv8si2.
231 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
232 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
233 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
234 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
235 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
236 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
237 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
238 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
240 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
244 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
246 2018-05-25 Jim Wilson <jimw@sifive.com>
248 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
249 * config/riscv/riscv.c (struct machine_function): Add
250 interrupt_handler_p and attribute_checked_p fields.
251 (riscv_attribute_table): Add interrupt.
252 (riscv_interrupt_type_p): New.
253 (riscv_save_reg_p): Save extra regs for interrupt handler.
254 (riscv_use_save_libcall): Return false for interrupt handler.
255 (riscv_first_stack_step): Add forward declaration.
256 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
257 for interrupt handler with large frame. Use it for saved reg list.
258 (riscv_expand_prologue): Move flag_stack_usage_info support to
259 eliminate duplication.
260 (riscv_expand_epilogue): Generate mret for interrupt handler.
261 (riscv_epilogue_uses): New.
262 (riscv_can_use_return_insn): Return false for interrupt handler.
263 (riscv_function_ok_for_sibcall): Likewise.
264 (riscv_set_current_function): Add interrupt handler support.
265 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
266 * config/riscv/riscv.md (UNSPECV_MRET): New.
268 (riscv_frflags, riscv_fsflags): Use tab after opcode.
270 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
272 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
274 PR tree-optimization/85712
275 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
276 this candidate has already been replaced in-situ by a copy.
278 2018-05-25 Jason Merrill <jason@redhat.com>
280 PR c++/80485 - inline function non-zero address.
281 * symtab.c (nonzero_address): Check DECL_COMDAT.
283 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
286 * config/alpha/alpha.md (ashlsi3): New insn pattern.
287 (*ashlsi_se): Rename from *ashldi_se. Define as sign
288 extension of SImode operation. Use const123_operand predicate.
290 (*saddl_se_1): Ditto.
292 (*ssubl_se_1): Ditto.
293 * config/alpha/predicates.md (const123_operand): New predicate.
294 * config/alpha/constraints.md (P): Use IN_RANGE.
296 2018-05-25 Richard Biener <rguenther@suse.de>
298 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
300 (ref_maybe_used_by_stmt_p): Likewise.
301 (stmt_may_clobber_ref_p): Likewise.
302 (stmt_may_clobber_ref_p_1): Likewise.
303 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
305 (ref_maybe_used_by_stmt_p): Likewise.
306 (stmt_may_clobber_ref_p): Likewise.
307 (stmt_may_clobber_ref_p_1): Likewise.
308 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
309 the alias oracle to disambiguate DRs with stmts DR analysis
311 (vect_analyze_data_refs): Do not give up on not analyzable
312 DRs for BB vectorization. Remove code truncating the dataref
315 2018-05-25 Jakub Jelinek <jakub@redhat.com>
318 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
319 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
320 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
322 2018-05-25 Richard Biener <rguenther@suse.de>
324 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
325 function, combining stmt data ref gathering and fatal analysis
327 (vect_analyze_data_refs): Remove now redudnant code and simplify.
328 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
329 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
330 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
331 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
333 2018-05-25 Bin Cheng <bin.cheng@arm.com>
335 PR tree-optimization/85720
336 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
337 SCC if all partitions are builtins.
338 (version_loop_by_alias_check): New parameter. Generate cancelable
339 runtime alias check if all partitions are builtins.
340 (distribute_loop): Update call to above function.
342 2018-05-25 Bin Cheng <bin.cheng@arm.com>
344 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
345 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
346 (parm_default_def_partition_arg): Ditto.
347 (set_parm_default_def_partition): Ditto.
348 (get_parm_default_def_partitions): Ditto and make it static.
349 (get_undefined_value_partitions): Ditto and make it static.
350 (remove_ssa_form): Refactor call to init_var_map here.
351 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
352 computation for loop region.
353 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
354 (register_default_def): Delete.
355 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
356 (parm_default_def_partition_arg): Ditto.
357 (set_parm_default_def_partition): Ditto.
358 (get_parm_default_def_partitions): Ditto and make it static.
359 (get_undefined_value_partitions): Ditto and make it static.
360 (coalesce_with_default, coalesce_with_default): Update comment.
361 (create_coalesce_list_for_region): New func factored out from
362 create_outofssa_var_map.
363 (populate_coalesce_list_for_outofssa): New func factored out from
364 create_outofssa_var_map and coalesce_ssa_name.
365 (create_outofssa_var_map): Delete.
366 (coalesce_ssa_name): Refactor to support live range computation.
367 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
368 (get_parm_default_def_partitions): Delete.
369 (get_undefined_value_partitions): Ditto.
370 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
371 computation for loop region.
372 (new_tree_live_info, loe_visit_block): Ditto.
373 (live_worklist, set_var_live_on_entry): Ditto.
374 (calculate_live_on_exit, verify_live_on_entry): Ditto.
375 * tree-ssa-live.h (struct _var_map): New fields.
376 (init_var_map): Change decl.
377 (region_contains_p): New.
379 2018-05-25 Bin Cheng <bin.cheng@arm.com>
381 * tree-ssa-live.h (live_merge_and_clear): Delete.
383 2018-05-25 Richard Biener <rguenther@suse.de>
386 * tree-dump.c (dequeue_and_dump): Remove access to removed
387 operand 2 of a SWITCH_EXPR.
389 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
391 * doc/sourcebuild.texi (vect_double_cond_arith): Include
392 multiplication and division.
393 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
394 (cond_udiv@var{m}, cond_umod@var{m}): Document.
395 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
396 (cond_udiv_optab, cond_umod_optab): New optabs.
397 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
398 (IFN_COND_RDIV): New internal functions.
399 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
400 TRUNC_MOD_EXPR and RDIV_EXPR.
401 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
402 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
404 (SVE_INT_BINARY): Include mult.
405 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
406 (optab, sve_int_op): Handle mult.
407 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
409 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
410 for SVE_INT_BINARY_SD.
412 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
414 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
415 (optab, sve_int_op): Handle div and udiv.
416 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
417 for SVE_INT_BINARY_SD.
418 (*<optab><mode>3): New insn for the same.
420 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
422 * tree-vect-patterns.c: Include predict.h.
423 (vect_recog_divmod_pattern): Restrict check for division support
424 to when optimizing for size.
426 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
428 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
429 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
430 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
431 (gimple_match_op::set_op): Likewise.
432 (gimple_resimplify4): Declare.
433 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
434 (expr::gen_transform): Likewise.
435 (decision_tree::gen): Generate a simplification routine for 4 operands.
436 * gimple-match-head.c (gimple_simplify): Add an overload for
437 4 operands. In the top-level function, handle up to 4 call
438 arguments and call gimple_resimplify4.
439 (gimple_resimplify4): New function.
440 (build_call_internal): Pass a fourth operand.
441 (maybe_push_to_seq): Likewise.
442 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
443 Fold VEC_COND_EXPRs of an operation and a default value into
444 an IFN_COND_* function if possible.
445 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
447 (SVE_COND_FP_BINARY): Include them.
448 (optab, sve_fp_op): Handle them.
449 (SVE_INT_BINARY_REV): New code iterator.
450 (SVE_COND_FP_BINARY_REV): New int iterator.
451 (commutative): New int attribute.
452 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
454 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
456 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
457 (*cond_<optab><mode>): New patterns for reversed operands.
459 2018-05-25 Richard Biener <rguenther@suse.de>
461 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
462 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
463 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
464 (STMT_VINFO_GROUPED_ACCESS): Adjust.
465 * tree-vect-data-refs.c (everywhere): Adjust users.
466 * tree-vect-loop.c (everywhere): Likewise.
467 * tree-vect-slp.c (everywhere): Likewise.
468 * tree-vect-stmts.c (everywhere): Likewise.
469 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
471 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
473 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
475 (gcc_cv_as_section_exclude): ... this.
476 Try Solaris as #exclude syntax.
477 * configure: Regenerate.
478 * config.in: Regenerate.
479 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
481 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
482 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
484 * varasm.c (default_elf_asm_named_section): Don't check if
485 HAVE_GAS_SECTION_EXCLUDE is defined.
487 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
489 * doc/md.texi: Update the documentation of the cond_* optabs
490 to mention the new final operand. Fix GET_MODE_NUNITS call.
491 Describe the scalar case too.
492 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
493 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
495 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
496 (get_conditional_internal_fn): Update comment.
497 * tree-vect-loop.c (vectorizable_reduction): Pass the original
498 accumulator value as a final argument to conditional functions.
499 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
500 a define_expand and add an "else" operand. Assert for now that
501 the else operand is equal to operand 2. Use SVE_INT_BINARY and
502 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
503 (*cond_<optab><mode>): New patterns.
504 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
505 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
506 (UNSPEC_COND_EOR): Delete.
507 (optab): Remove associated mappings.
508 (SVE_INT_BINARY): New code iterator.
509 (sve_int_op): Remove int attribute and add "minus" to the code
511 (SVE_COND_INT_OP): Delete.
512 (SVE_COND_FP_OP): Rename to...
513 (SVE_COND_FP_BINARY): ...this.
515 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
517 * optabs.c (can_reuse_operands_p): New function.
518 (maybe_legitimize_operands): Try to reuse the results for
521 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
523 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
524 Add {q} suffix to insn mnemonic.
526 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
528 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
529 (msp430_warn_func_return): New.
531 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
533 * fold-const.c (tree_nonzero_bits): New function.
534 * fold-const.h (tree_nonzero_bits): Likewise.
535 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
536 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
538 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
542 * varasm.c (assemble_alias): Check ifunc_resolver only on
545 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
548 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
549 when memory input operand is handled.
551 2018-05-24 Luis Machado <luis.machado@linaro.org>
553 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
555 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
557 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
559 * match.pd: Delay FMA folds until after vectorization.
561 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
564 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
565 address check not strict.
567 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
569 * gimple-match.h (gimple_match_op): New class.
570 (mprts_hook): Replace parameters with a gimple_match_op *.
571 (maybe_build_generic_op): Likewise.
572 (gimple_simplified_result_is_gimple_val): Replace parameters with
573 a const gimple_match_op *.
574 (gimple_simplify): Replace code_helper * and tree * parameters with
575 a gimple_match_op * parameter.
576 (gimple_resimplify1): Replace code_helper *, tree and tree *
577 parameters with a gimple_match_op * parameter.
578 (gimple_resimplify2): Likewise.
579 (gimple_resimplify3): Likewise.
580 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
581 parameters with a gimple_match_op * parameter.
582 * gimple-match-head.c (gimple_simplify): Change prototypes of
583 auto-generated functions to take a gimple_match_op * instead of
584 separate code_helper * and tree * parameters. Make the same
585 change in the top-level overload and update calls to the
586 gimple_resimplify routines. Update calls to the auto-generated
587 functions and to maybe_push_res_to_seq in the publicly-facing
588 operation-specific gimple_simplify overloads.
589 (gimple_match_op::MAX_NUM_OPS): Define.
590 (gimple_resimplify1): Replace rcode and ops with a single res_op
591 parameter. Update call to gimple_simplify.
592 (gimple_resimplify2): Likewise.
593 (gimple_resimplify3): Likewise.
594 (mprts_hook): Replace parameters with a gimple_match_op *.
595 (maybe_build_generic_op): Likewise.
596 (build_call_internal): Replace type, nargs and ops with
598 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
599 with a single gimple_match_op *. Update calls to mprts_hook,
600 build_call_internal and gimple_simplified_result_is_gimple_val.
601 Factor out code that is common to the tree_code and combined_fn cases.
602 * genmatch.c (expr::gen_transform): Replace tem_code and
603 tem_ops with a gimple_match_op called tem_op. Update calls
604 to the gimple_resimplify functions and maybe_push_res_to_seq.
605 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
606 res_ops. Update call to the gimple_resimplify functions.
607 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
608 (decision_tree::gen): Make the functions take a gimple_match_op *
609 called res_op instead of separate res_code and res_ops parameters.
610 Update call accordingly.
611 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
612 and ops with a single res_op parameter. Update calls to
613 maybe_build_generic_op and maybe_push_res_to_seq.
614 (fold_stmt_1): Update calls to gimple_simplify and
615 replace_stmt_with_simplification.
616 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
617 and gimple_simplified_result_is_gimple_val.
618 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
620 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
621 with a gimple_match_op *.
622 (vn_nary_build_or_lookup): Likewise. Update call to
623 vn_nary_build_or_lookup_1.
624 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
625 gimple_match_op *. Update calls to the gimple_resimplify routines
626 and to gimple_simplified_result_is_gimple_val.
627 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
628 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
629 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
630 (visit_nary_op): Likewise.
631 (visit_reference_op_load): Likewise.
633 2018-05-23 Luis Machado <luis.machado@linaro.org>
635 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
636 modifier for printing the step amount.
638 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
641 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
644 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
646 * doc/sourcebuild.texi (Endianness): New subsubsection.
648 2018-05-23 Luis Machado <luis.machado@linaro.org>
650 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
651 <prefetch_dynamic_strides>: New const bool field.
652 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
653 prefetch_dynamic_strides.
654 (exynosm1_prefetch_tune): Likewise.
655 (thunderxt88_prefetch_tune): Likewise.
656 (thunderx_prefetch_tune): Likewise.
657 (thunderx2t99_prefetch_tune): Likewise.
658 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
660 (aarch64_override_options_internal): Update to set
661 PARAM_PREFETCH_DYNAMIC_STRIDES.
662 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
663 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
664 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
665 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
666 prefetch-dynamic-strides setting.
668 2018-05-23 Luis Machado <luis.machado@linaro.org>
670 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
671 <minimum_stride>: New const int field.
672 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
673 minimum_stride field defaulting to -1.
674 (exynosm1_prefetch_tune): Likewise.
675 (thunderxt88_prefetch_tune): Likewise.
676 (thunderx_prefetch_tune): Likewise.
677 (thunderx2t99_prefetch_tune): Likewise.
678 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
679 <default_opt_level>: Set to 3.
680 (aarch64_override_options_internal): Update to set
681 PARAM_PREFETCH_MINIMUM_STRIDE.
682 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
683 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
684 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
685 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
686 stride is constant and is below the minimum stride threshold.
688 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
690 * config/arm/arm-cpus.in (mode26): Delete.
691 (armv4): Delete mode26 reference.
692 * config/arm/arm.c (arm_configure_build_target): Delete use of
695 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
697 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
699 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
700 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
702 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
703 (floatunsdidf2): Ditto.
705 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
707 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
708 (fixuns_trunc<mode>si2_avx512f): Ditto.
709 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
710 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
711 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
713 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
715 PR rtl-optimization/79985
716 * df-scan.c (df_insn_refs_collect): Remove special case for
717 global registers and asm statements.
719 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
721 * extend.texi (Global Register Variables): Rewrite the bullet list.
722 Note that the register is available for allocation. Note that access
723 via inline asm must use constraints. Add note about async-signal
724 handlers. Remove paragraph about automagic register selection.
726 2018-05-23 Richard Biener <rguenther@suse.de>
728 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
729 of fixed offset from memset VN.
731 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
733 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
735 (alloc_cand_and_find_basis): Initialize first_interp field.
736 (slsr_process_mul): Modify first_interp field.
737 (slsr_process_add): Likewise.
738 (slsr_process_cast): Modify first_interp field for each new
740 (slsr_process_copy): Likewise.
741 (dump_candidate): Dump first_interp field.
742 (replace_mult_candidate): Process all interpretations, not just
744 (replace_rhs_if_not_dup): Likewise.
745 (replace_one_candidate): Likewise.
747 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
749 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
751 (aarch64_needs_frame_chain): New function.
752 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
754 2018-05-23 Sudakshina Das <sudi.das@arm.com>
757 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
758 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
759 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
760 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
761 as true for strict-align.
762 (aarch64_can_inline_p): Perform checks even when callee has no
763 attributes to check for strict alignment.
764 * doc/extend.texi (AArch64 Function Attributes): Document
766 * doc/invoke.texi: (AArch64 Options): Likewise.
768 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
770 PR tree-optimization/85853
771 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
772 the handling of the root of the node to...
773 (vect_slp_analyze_node_operations_1): ...this new function,
774 and run the whole thing with the child nodes' def types
775 set according to their SLP node's def type.
777 2018-05-23 Richard Biener <rguenther@suse.de>
780 * tree-data-ref.c (create_runtime_alias_checks): Defer
781 and ignore overflow warnings.
783 2018-05-23 Yury Gribov <tetra2005@gmail.com>
785 PR tree-optimization/85822
786 * tree-vrp.c (is_masked_range_test): Fix handling of negative
789 2018-05-23 Richard Biener <rguenther@suse.de>
791 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
792 memset constants via native_interpret_expr.
794 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
797 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
799 (cgraph_node::create_alias): Likewise.
800 (cgraph_node::get_availability): Check ifunc_resolver instead
801 of looking up ifunc attribute.
802 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
803 * varasm.c (do_assemble_alias): Likewise.
804 (assemble_alias): Likewise.
805 (default_binds_local_p_3): Likewise.
806 * cgraph.h (cgraph_node): Add ifunc_resolver.
807 (cgraph_node::only_called_directly_or_aliased_p): Return false
809 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
811 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
812 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
813 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
814 instead of looking up ifunc attribute.
816 2018-05-22 Luis Machado <luis.machado@linaro.org>
818 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
820 2018-05-22 Martin Sebor <msebor@redhat.com>
823 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
824 only when expasion succeeds.
825 (expand_builtin_strcmp): Same.
826 (expand_builtin_strncmp): Same.
828 2018-05-22 Martin Sebor <msebor@redhat.com>
830 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
832 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
833 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
835 * config/aarch64/aarch64-ldpstp.md: Replace uses of
836 aarch64_mem_pair_operand with memory_operand and delete operand swapping
838 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
839 Add check for legitimate_address.
840 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
841 (aarch64_swap_ldrstr_operands): New.
842 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
845 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
846 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
848 * config/aarch64/aarch64.md: New patterns to generate stp
850 (store_pair_sw, store_pair_dw): New patterns to generate stp for
851 single words and double words.
852 (load_pair_sw, load_pair_dw): Likewise.
853 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
855 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
857 * config/aarch64/aarch64-ldpstp.md: Modify peephole
858 for different mode ldpstp and add peephole for merged zero stores.
860 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
862 (aarch64_gen_store_pair): Rename calls to match new patterns.
863 (aarch64_gen_load_pair): Rename calls to match new patterns.
864 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
865 (load_pair<DREG:mode><DREG2:mode>): ... This.
866 (store_pair<mode>): Rename to...
867 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
868 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
870 (V_INT_EQUIV): Handle SImode.
871 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
874 2018-05-22 Martin Sebor <msebor@redhat.com>
877 * calls.c (maybe_warn_nonstring_arg): Use string length to set
878 or ajust the presumed bound on an operation to avoid unnecessary
881 2018-05-22 Martin Sebor <msebor@redhat.com>
883 PR tree-optimization/85826
884 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
885 assuming that a DECL necesarily has a constant size.
887 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
890 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
892 2018-05-22 Richard Biener <rguenther@suse.de>
894 PR tree-optimization/85834
895 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
896 non-constant and non-zero memset arguments.
898 2018-05-22 Martin Liska <mliska@suse.cz>
901 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
903 2018-05-22 Richard Biener <rguenther@suse.de>
905 PR tree-optimization/85863
906 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
907 comparisons when vectype is specified.
908 (vectorizable_condition): Do not specify vectype for
909 vect_is_simple_cond when SLP vectorizing.
911 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
914 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
915 define __ibm128 as long double.
916 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
917 as a distinct type when IEEE 128-bit support is enabled.
918 (init_float128_ieee): Fix up conversions between IFmode and IEEE
919 128-bit types to use the correct functions.
920 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
921 convert between 128-bit floating point types that have different
922 modes but the same representation, instead of using gen_lowpart to
924 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
926 (IFKF_reg): New attributes to give the register constraints for
928 (extend<mode>tf2_internal): New insns to mark an explicit
929 conversion between 128-bit floating point types that have a
930 different mode but share the same representation.
932 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
934 PR tree-optimization/85814
935 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
936 a null return from get_strinfo when unsharing the next
937 strinfo in the chain.
939 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
942 * varasm.c (weak_finish): Clean up weak_decls.
944 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
946 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
947 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
948 UNSPEC_UADALP values.
949 * config/aarch64/iterators.md (ABAL): New int iterator.
952 (sur): Add mappings for the above.
953 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
955 (aarch64_<sur>abal<mode>_4): Likewise.
956 (aarch64_<sur>adalp<mode>_3): Likewise.
957 (<sur>sadv16qi): New define_expand.
959 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
961 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
962 (*movdf_internal): Ditto.
963 (*rcpsf2_sse): Ditto.
964 (*rsqrtsf2_sse): Ditto.
965 (*sqrt<mode>2_sse): Ditto.
967 2018-05-21 Tamar Christina <tamar.christina@arm.com>
969 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
971 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
972 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
973 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
974 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
976 * config/aarch64/arm_neon.h: Likewise.
977 * config/aarch64/iterators.md (VQ_I): New.
979 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
981 * config.gcc: Add arc/t-multilib-linux to tmake_file for
983 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
986 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
988 * config/nds32/constraints.md (S): New constraint.
989 * config/nds32/nds32.md (call_internal): Use constraint S.
990 (call_value_internal): Likewise.
991 (sibcall_internal): Likewise.
992 (sibcall_value_internal): Likewise.
994 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
995 Chung-Ju Wu <jasonwucj@gmail.com>
997 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
1000 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1001 Chung-Ju Wu <jasonwucj@gmail.com>
1003 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
1004 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
1005 (nds32_rtx_costs_impl): Simplify.
1006 (nds32_address_cost_impl): Simplify.
1007 (nds32_init_rtx_costs): New function.
1008 (nds32_rtx_costs_speed_prefer): Likewise.
1009 (nds32_rtx_costs_size_prefer): Likewise.
1010 (nds32_address_cost_speed_prefer): Likewise.
1011 (nds32_address_cost_speed_fwprop): Likewise.
1012 (nds32_address_cost_size_prefer): Likewise.
1013 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
1014 * config/nds32/nds32.c (nds32_option_override): Use
1015 nds32_init_rtx_costs function.
1017 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1019 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
1020 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
1021 (TARGET_PIPELINE_N8): Likewise.
1022 (TARGET_PIPELINE_N10): Likewise.
1023 (TARGET_PIPELINE_N13): Likewise.
1024 (TARGET_PIPELINE_GRAYWOLF): Likewise.
1026 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
1028 * config/nds32/nds32-fpu.md: Update copyright year.
1030 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1032 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
1034 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1036 * config/nds32/nds32.c
1037 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
1038 * config/nds32/nds32.opt (minline-asm-r15): New option.
1040 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1042 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
1044 * config/nds32/nds32.md (abssi2): New pattern.
1046 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
1048 * config/i386/i386.md (rex64namesuffix): New mode attribute.
1049 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
1050 Merge insn pattern from sse_cvtsi2ss<round_name> and
1051 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
1052 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
1053 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
1054 using SWI48 mode iterator.
1055 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
1056 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
1057 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
1058 pattern from sse_cvttss2si<round_saeonly_name>
1059 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
1060 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
1061 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
1062 using SWI48 mode iterator.
1063 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
1064 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
1065 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
1066 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
1067 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
1068 using SWI48 mode iterator.
1069 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
1070 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
1071 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
1072 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
1073 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
1074 SWI48 mode iterator.
1075 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
1076 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
1077 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
1078 pattern from sse_cvttsd2si<round_saeonly_name>
1079 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
1081 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1083 * config/nds32/nds32-md-auxiliary.c
1084 (nds32_valid_smw_lwm_base_p): Refine.
1085 (nds32_output_smw_single_word): Refine.
1086 (nds32_output_smw_double_word): New.
1087 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
1089 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1091 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
1092 (nds32_output_stack_pop): Refine.
1093 (nds32_expand_unaligned_load): Refine.
1094 (nds32_expand_unaligned_store): Refine.
1096 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1097 Chung-Ju Wu <jasonwucj@gmail.com>
1099 * config/nds32/constants.md: Add TP_REGNUM constant.
1100 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
1101 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
1103 * config/nds32/nds32-doubleword.md: Consider flag_pic.
1104 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
1105 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
1106 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
1107 and PIC code generation.
1108 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
1110 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
1112 * config/nds32/nds32.md: Support TLS and PIC.
1113 * config/nds32/nds32.c: Support TLS and PIC.
1114 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
1115 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
1118 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1120 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
1121 mode with E_ prefix.
1123 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1124 Chung-Ju Wu <jasonwucj@gmail.com>
1126 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
1127 * config/nds32/nds32-md-auxiliary.c
1128 (symbolic_reference_mentioned_p): New.
1129 (nds32_legitimize_ict_address): New.
1130 (nds32_expand_ict_move): New.
1131 (nds32_indirect_call_referenced_p): New.
1132 (nds32_symbol_binds_local_p): Delete.
1133 (nds32_long_call_p): Modify.
1134 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
1135 * config/nds32/nds32-protos.h
1136 (symbolic_reference_mentioned_p): Declare.
1137 (nds32_legitimize_ict_address): Declare.
1138 (nds32_expand_ict_move): Declare.
1139 (nds32_indirect_call_referenced_p): Declare.
1140 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
1141 (nds32_relax_group): Use nds32_ict_const_p as condition.
1142 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
1143 (nds32_asm_file_start): Output ict_model directive in asm code.
1144 (nds32_legitimate_address_p): Consider indirect call.
1145 (nds32_print_operand): Consider indirect call.
1146 (nds32_print_operand_address): Consider indirect call.
1147 (nds32_insert_attributes): Handle "indirect_call" attribute.
1148 (TARGET_LEGITIMATE_ADDRESS_P): Define.
1149 (TARGET_LEGITIMATE_CONSTANT_P): Define.
1150 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
1151 (TARGET_DELEGITIMIZE_ADDRESS): Define.
1152 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
1153 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
1154 (TARGET_ICT_MODEL_SMALL): Define.
1155 (TARGET_ICT_MODEL_LARGE): Define.
1156 * config/nds32/nds32.md (movsi): Consider ict model.
1157 (call, call_value): Consider ict model.
1158 (sibcall, sibcall_value): Consider ict model.
1159 * config/nds32/nds32.opt (mict-model): New option.
1160 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
1163 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
1164 Monk Chiang <sh.chiang04@gmail.com>
1165 Jim Wilson <jimw@sifive.com>
1167 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
1168 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
1169 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
1170 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
1171 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
1172 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
1173 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
1174 compute save_libcall_adjustment properly.
1175 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
1176 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
1177 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
1178 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
1179 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
1180 (ABI_SPEC): Handle mabi=ilp32e.
1181 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
1182 (RVE): Add RVE mask.
1183 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
1184 <-march>: Add rv32e as an example.
1186 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
1189 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
1190 (intra_create_variable_infos): Handle C++ constructors.
1192 2018-05-18 Martin Liska <mliska@suse.cz>
1194 * passes.def: Remove a redundant pass.
1196 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
1199 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
1201 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1203 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
1205 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
1207 (armv2, armv2a, armv3, armv3m): Delete architectures.
1208 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
1209 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
1210 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
1212 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
1213 (*mulsidi3adddi): Likewise.
1214 (mulsidi3): Likewise.
1215 (*mulsidi3_nov6): Likewise.
1216 (umulsidi3): Likewise.
1217 (umulsidi3_nov6): Likewise.
1218 (umaddsidi4): Likewise.
1219 (*umulsidi3adddi): Likewise.
1220 (smulsi3_highpart): Likewise.
1221 (*smulsi3_highpart_nov6): Likewise.
1222 (umulsi3_highpart): Likewise.
1223 (*umulsi3_highpart_nov6): Likewise.
1224 * config/arm/arm.h (arm_arch3m): Delete.
1225 * config/arm/arm.c (arm_arch3m): Delete.
1226 (arm_option_override_internal): Update armv3-related comment.
1227 (arm_configure_build_target): Delete use of isa_bit_mode32.
1228 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
1229 (arm_rtx_costs_internal): Delete check of arm_arch3m.
1230 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
1232 (mulusa3): Likewise.
1233 * config/arm/arm-protos.h (arm_arch3m): Delete.
1234 * config/arm/arm-tables.opt: Regenerate.
1235 * config/arm/arm-tune.md: Likewise.
1236 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
1237 deleted architectures.
1239 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1241 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
1242 (armv5t, armv5te): New features.
1243 (ARMv5, ARMv5e): Delete fgroups.
1244 (ARMv5t, ARMv5te): Adjust for above changes.
1246 (armv5, armv5e): Delete arches.
1247 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
1249 (*call_reg_arm): Likewise.
1250 (*call_value_reg_armv5): Likewise.
1251 (*call_value_reg_arm): Likewise.
1252 (*call_symbol): Likewise.
1253 (*call_value_symbol): Likewise.
1254 (*sibcall_insn): Likewise.
1255 (*sibcall_value_insn): Likewise.
1257 (prefetch): Likewise.
1258 (define_split and define_peephole2 dependent on arm_arch5):
1260 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
1262 (TARGET_ARM_QBIT): Likewise.
1263 (TARGET_DSP_MULTIPLY): Likewise.
1264 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
1265 (arm_arch5, arm_arch5e): Delete.
1266 (arm_arch5t, arm_arch5te): Declare.
1267 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
1268 (arm_arch5t): Declare.
1269 (arm_option_reconfigure_globals): Update for the above.
1270 (arm_options_perform_arch_sanity_checks): Update comment, replace
1271 use of arm_arch5 with arm_arch5t.
1272 (use_return_insn): Likewise.
1273 (arm_emit_call_insn): Likewise.
1274 (output_return_instruction): Likewise.
1275 (arm_final_prescan_insn): Likewise.
1276 (arm_coproc_builtin_available): Likewise.
1277 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
1278 arm_arch5e with arm_arch5t and arm_arch5te.
1279 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
1280 (arm_arch5t, arm_arch5te): Declare.
1281 * config/arm/arm-tables.opt: Regenerate.
1282 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
1283 * config/arm/t-multilib: Likewise.
1284 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
1285 instead of arm_arch5.
1286 (*call_reg_thumb1): Likewise.
1287 (*call_value_reg_thumb1_v5): Likewise.
1288 (*call_value_reg_thumb1): Likewise.
1289 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
1291 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
1293 2018-05-18 Martin Liska <mliska@suse.cz>
1295 PR gcov-profile/84846
1296 * doc/gcov.texi: Document -t option of gcov tool.
1298 2018-05-18 Martin Liska <mliska@suse.cz>
1300 PR gcov-profile/84846
1301 * gcov.c (print_usage): Add new -t option.
1302 (process_args): Handle the option.
1303 (generate_results): Use stdout as output when requested by
1306 2018-05-18 Martin Liska <mliska@suse.cz>
1308 PR gcov-profile/84846
1309 * coverage.c (coverage_init): Write PWD to .gcno file.
1310 * doc/gcov.texi: Document how working directory is printed.
1311 * gcov-dump.c (dump_gcov_file): Print PWD.
1312 * gcov.c (output_intermediate_file): Likewise.
1313 (read_graph_file): Read PWD string.
1314 (output_lines): Print PWD.
1316 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1319 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
1320 for retval and return false if all args to phi are zero.
1322 2018-05-18 Richard Biener <rguenther@suse.de>
1324 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
1326 (evrp_dom_walker::before_dom_children): Call it.
1328 2018-05-18 Richard Biener <rguenther@suse.de>
1330 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
1331 results when processing array refs with variable index.
1333 2018-05-18 Toon Moene <toon@moene.org>
1335 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
1336 directly after that of -floop-interchange. Indicate that both
1337 options are enabled by default when specifying -O3.
1339 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1341 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
1342 iterator. Delete separate integer-mode vec_set<mode> expander.
1343 (aarch64_simd_vec_setv2di): Delete.
1344 (vec_setv2di): Delete.
1345 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
1346 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
1347 the "w, r" alternative.
1349 2018-05-18 Martin Liska <mliska@suse.cz>
1351 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
1352 * tree-pass.h (make_pass_lower_switch_O0): New function.
1353 * tree-switch-conversion.c (node_has_low_bound): Remove.
1354 (node_has_high_bound): Likewise.
1355 (node_is_bounded): Likewise.
1356 (class pass_lower_switch): Make it a template type and create
1358 (pass_lower_switch::execute): Add template argument.
1359 (make_pass_lower_switch): New function.
1360 (make_pass_lower_switch_O0): New function.
1361 (do_jump_if_equal): Remove.
1362 (emit_case_nodes): Simplify to just handle all 3 cases and leave
1363 all the hard work to tree optimization passes.
1365 2018-05-18 Martin Liska <mliska@suse.cz>
1367 * dbgcnt.c (limit_low): Renamed from limit.
1368 (limit_high): New variable.
1369 (dbg_cnt_is_enabled): Check for upper limit.
1370 (dbg_cnt): Adjust dumping.
1371 (dbg_cnt_set_limit_by_index): Add new argument for high
1373 (dbg_cnt_set_limit_by_name): Likewise.
1374 (dbg_cnt_process_single_pair): Parse new format.
1375 (dbg_cnt_process_opt): Use strtok.
1376 (dbg_cnt_list_all_counters): Remove 'value' and add
1378 * doc/invoke.texi: Document changes.
1380 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
1382 * doc/sourcebuild.texi (scalar_all_fma): Document.
1383 * tree.def (FMA_EXPR): Delete.
1384 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
1385 * internal-fn.c (ternary_direct): New macro.
1386 (expand_ternary_optab_fn): Likewise.
1387 (direct_ternary_optab_supported_p): Likewise.
1388 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
1389 * builtins.c (fold_builtin_fma): Delete.
1390 (fold_builtin_3): Don't call it.
1391 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
1392 * expr.c (expand_expr_real_2): Likewise.
1393 * fold-const.c (operand_equal_p): Likewise.
1394 (fold_ternary_loc): Likewise.
1395 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
1396 * gimple.c (DEFTREECODE): Likewise.
1397 * gimplify.c (gimplify_expr): Likewise.
1398 * optabs-tree.c (optab_for_tree_code): Likewise.
1399 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
1400 * tree-eh.c (operation_could_trap_p): Likewise.
1401 (stmt_could_throw_1_p): Likewise.
1402 * tree-inline.c (estimate_operator_cost): Likewise.
1403 * tree-pretty-print.c (dump_generic_node): Likewise.
1404 (op_code_prio): Likewise.
1405 * tree-ssa-loop-im.c (stmt_cost): Likewise.
1406 * tree-ssa-operands.c (get_expr_operands): Likewise.
1407 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
1408 * fold-const-call.h (fold_fma): Delete.
1409 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
1410 CFN_FNMA and CFN_FNMS.
1412 * genmatch.c (combined_fn): New enum.
1413 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
1414 (commutative_op): New function.
1415 (commutate): Use it. Handle more than 2 operands.
1416 (dt_operand::gen_gimple_expr): Use commutative_op.
1417 (parser::parse_expr): Allow :c to be used with non-binary
1418 operators if the commutative operand is known.
1419 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
1420 CFN_FMS, CFN_FNMA and CFN_FNMS.
1421 (backprop::process_assign_use): Remove FMA_EXPR handling.
1422 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
1423 (gen_hsa_fma): New function.
1424 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
1425 IFN_FNMA and IFN_FNMS.
1426 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
1427 * gimple-fold.h (follow_all_ssa_edges): Declare.
1428 * gimple-fold.c (follow_all_ssa_edges): New function.
1429 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
1430 gimple_build interface and use follow_all_ssa_edges to fold the result.
1431 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
1432 instead of checking for optabs directly.
1433 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
1434 rather than FMA_EXPRs.
1435 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
1436 call to IFN_FMA instead of an FMA_EXPR.
1438 2018-05-17 Jim Wilson <jimw@sifive.com>
1440 * expr.c (do_tablejump): When converting index to Pmode, if we have a
1441 sign extended promoted subreg, and the range does not have the sign bit
1442 set, then do a sign extend.
1444 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
1445 test, check for sign extended subreg and/or constant operands, and
1446 do a sign extend in that case.
1448 2018-05-17 Steve Ellcey <sellcey@cavium.com>
1450 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
1451 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
1453 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
1454 Change logics_shift_reg to logics_shift_imm.
1455 (thunderx2t99_fp_loadpair_basic): Delete.
1456 (thunderx2t99_fp_storepair_basic): Delete.
1457 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
1458 (thunderx2t99_asimd_polynomial): Delete.
1459 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
1460 and neon_fp_mul_d_scalar_q.
1461 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
1462 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
1463 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
1464 (thunderx2t99_asimd_lut): Add missing tbl types.
1465 (thunderx2t99_asimd_ext): Delete.
1466 (thunderx2t99_asimd_load1_1_mult): Delete.
1467 (thunderx2t99_asimd_load1_2_mult): Delete.
1468 (thunderx2t99_asimd_load1_ldp): New.
1469 (thunderx2t99_asimd_load1): New.
1470 (thunderx2t99_asimd_load2): Add missing *load2* types.
1471 (thunderx2t99_asimd_load3): New.
1472 (thunderx2t99_asimd_load4): New.
1473 (thunderx2t99_asimd_store1_1_mult): Delete.
1474 (thunderx2t99_asimd_store1_2_mult): Delete.
1475 (thunderx2t99_asimd_store2_mult): Delete.
1476 (thunderx2t99_asimd_store2_onelane): Delete.
1477 (thunderx2t99_asimd_store_stp): New.
1478 (thunderx2t99_asimd_store1): New.
1479 (thunderx2t99_asimd_store2): New.
1480 (thunderx2t99_asimd_store3): New.
1481 (thunderx2t99_asimd_store4): New.
1483 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
1485 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
1486 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
1488 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
1489 Segher Boessenkool <segher@kernel.crashing.org>
1492 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
1495 2018-05-17 Richard Biener <rguenther@suse.de>
1497 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
1498 for pruning loop and prune defs feeding only already visited PHIs.
1500 2018-05-17 Richard Biener <rguenther@suse.de>
1502 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
1504 2018-05-17 Bin Cheng <bin.cheng@arm.com>
1505 Richard Biener <rguenther@suse.de>
1507 PR tree-optimization/85793
1508 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
1509 for VMAT_ELEMENTWISE.
1511 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1513 * internal-fn.h (lookup_internal_fn): Declare
1514 * internal-fn.c (lookup_internal_fn): New function.
1515 * gimple.c (gimple_build_call_from_tree): Handle calls to
1517 * gimple-pretty-print.c (dump_gimple_call): Print "." before
1518 internal function names.
1519 * tree-pretty-print.c (dump_generic_node): Likewise.
1520 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
1522 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1524 * gimple-fold.h (gimple_build): Make the function forms take
1525 combined_fn rather than built_in_function.
1526 (gimple_simplify): Likewise.
1527 * gimple-match-head.c (gimple_simplify): Likewise.
1528 * gimple-fold.c (gimple_build): Likewise.
1529 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
1530 rather than gimple_build_call_internal.
1531 (get_initial_defs_for_reduction): Likewise.
1532 (vect_create_epilog_for_reduction): Likewise.
1533 (vectorizable_live_operation): Likewise.
1535 2018-05-17 Martin Liska <mliska@suse.cz>
1537 * gimple-ssa-sprintf.c (format_directive): Do not use
1538 space in between 'G_' and '('.
1540 2018-05-17 Jakub Jelinek <jakub@redhat.com>
1543 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
1544 even if the mask is not all ones.
1547 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
1549 (ix86_gimple_fold_builtin): Likewise.
1552 * config/i386/i386.c: Include tree-vector-builder.h.
1553 (ix86_vector_shift_count): New function.
1554 (ix86_fold_builtin): Fold shift builtins by scalar count.
1555 (ix86_gimple_fold_builtin): Likewise.
1557 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
1558 _mm512_setzero): New intrinsics.
1560 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
1561 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1563 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
1564 code generation for cases where splatting a value is not useful.
1565 * simplify-rtx.c (simplify_ternary_operation): Simplify
1566 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
1567 mode to a vec_concat.
1569 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
1571 * config.gcc: Support "goldmont-plus".
1572 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1574 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1575 PROCESSOR_GOLDMONT_PLUS.
1576 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
1577 (processor_target_table): Add "goldmont-plus".
1578 (PTA_GOLDMONT_PLUS): Define.
1579 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
1580 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
1581 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
1582 (fold_builtin_cpu): Add "goldmont-plus".
1583 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
1584 (ix86_option_override_internal): Add "goldmont-plus".
1585 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
1586 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
1587 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
1588 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
1590 2018-05-17 Richard Biener <rguenther@suse.de>
1592 PR tree-optimization/85757
1593 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
1594 remove defs that only feed that PHI from further processing.
1596 2018-05-16 Jim Wilson <jimw@sifive.com>
1598 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
1600 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
1602 2018-05-16 Mark Wielaard <mark@klomp.org>
1604 * dwarf2out.c (count_index_strings): New function.
1605 (output_indirect_strings): Call count_index_strings and generate
1606 header for dwarf_version >= 5.
1608 2018-05-16 Mark Wielaard <mark@klomp.org>
1610 * dwarf2out.c (dwarf_FORM): New function.
1611 (set_indirect_string): Use dwarf_FORM.
1612 (reset_indirect_string): Likewise.
1613 (size_of_die): Likewise.
1614 (value_format): Likewise.
1615 (output_die): Likewise.
1616 (add_skeleton_AT_string): Likewise.
1617 (output_macinfo_op): Likewise.
1618 (index_string): Likewise.
1619 (output_index_string_offset): Likewise.
1620 (output_index_string): Likewise.
1621 (count_index_strings): Likewise.
1623 2018-05-16 Carl Love <cel@us.ibm.com>
1625 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1626 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
1628 2018-05-16 Martin Jambor <mjambor@suse.cz>
1630 * ipa-prop.c (ipa_free_all_edge_args): Remove.
1631 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
1633 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
1635 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
1636 (fnma<mode>4): Likewise.
1637 (fms<mode>4): Likewise.
1638 (fnms<mode>4): Likewise.
1639 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
1640 (aarch64_fnma<mode>4): Likewise.
1641 (aarch64_fms<mode>4): Likewise.
1642 (aarch64_fnms<mode>4): Likewise.
1643 (aarch64_fnmadd<mode>4): Likewise.
1645 2018-05-16 Jason Merrill <jason@redhat.com>
1647 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
1649 2018-05-16 Richard Biener <rguenther@suse.de>
1651 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
1652 (dump_stmt_cost): Declare.
1653 (add_stmt_cost): Dump cost we add.
1654 (add_stmt_costs): New function.
1655 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
1657 (vect_analyze_stmt): Adjust prototype.
1658 (vectorizable_condition): Likewise.
1659 (vectorizable_live_operation): Likewise.
1660 (vectorizable_reduction): Likewise.
1661 (vectorizable_induction): Likewise.
1662 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
1663 cost vector to pass to vectorizable_ and record afterwards.
1664 (vect_model_reduction_cost): Take cost vector argument and adjust.
1665 (vect_model_induction_cost): Likewise.
1666 (vectorizable_reduction): Likewise.
1667 (vectorizable_induction): Likewise.
1668 (vectorizable_live_operation): Likewise.
1669 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
1670 SLP_TREE_NUMBER_OF_VEC_STMTS.
1671 (vect_analyze_slp_cost_1): Remove.
1672 (vect_analyze_slp_cost): Likewise.
1673 (vect_slp_analyze_node_operations): Take visited args and
1674 a target cost vector. Avoid processing already visited stmt sets.
1675 (vect_slp_analyze_operations): Use a local cost vector to gather
1676 costs and register those of non-discarded instances.
1677 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
1678 (vect_schedule_slp_instance): Remove copying of
1679 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
1681 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
1682 adding cost. Record cost entry location.
1683 (vect_prologue_cost_for_slp_op): Function to compute cost of
1684 a constant or invariant generated for SLP vect in the prologue,
1685 split out from vect_analyze_slp_cost_1.
1686 (vect_model_simple_cost): Make static. Adjust for SLP costing.
1687 (vect_model_promotion_demotion_cost): Likewise.
1688 (vect_model_store_cost): Likewise, make static.
1689 (vect_model_load_cost): Likewise.
1690 (vectorizable_bswap): Add cost vector arg and adjust.
1691 (vectorizable_call): Likewise.
1692 (vectorizable_simd_clone_call): Likewise.
1693 (vectorizable_conversion): Likewise.
1694 (vectorizable_assignment): Likewise.
1695 (vectorizable_shift): Likewise.
1696 (vectorizable_operation): Likewise.
1697 (vectorizable_store): Likewise.
1698 (vectorizable_load): Likewise.
1699 (vectorizable_condition): Likewise.
1700 (vectorizable_comparison): Likewise.
1701 (can_vectorize_live_stmts): Likewise.
1702 (vect_analyze_stmt): Likewise.
1703 (vect_transform_stmt): Adjust calls to vectorizable_*.
1704 * tree-vectorizer.c: Include gimple-pretty-print.h.
1705 (dump_stmt_cost): New function.
1707 2018-05-16 Richard Biener <rguenther@suse.de>
1709 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
1710 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
1711 * tree-ssa-dse.c: Include tree-ssa-loop.h.
1712 (check_name): New callback.
1713 (dse_classify_store): Track cycles via a visited bitmap of PHI
1714 defs and simplify handling of in-loop and across loop dead stores
1715 and properly fail for loop-variant refs. Handle byte-tracking with
1716 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
1719 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
1721 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
1722 (vect_get_mask_type_for_stmt): Likewise.
1723 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
1725 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
1726 to determine the statement's vector type and the vector type that
1727 should be used for calculating nunits. Deal with cases in which
1728 the type has to be deferred.
1729 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
1730 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
1731 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
1732 (vect_determine_vf_for_stmt): New functions, split out from...
1733 (vect_determine_vectorization_factor): ...here.
1734 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
1735 (vect_get_mask_type_for_stmt): New functions, split out from
1736 vect_determine_vectorization_factor.
1738 2018-05-16 Richard Biener <rguenther@suse.de>
1740 * tree-cfg.c (verify_gimple_assign_ternary): Properly
1741 verify the [VEC_]COND_EXPR embedded comparison.
1743 2018-05-15 Martin Sebor <msebor@redhat.com>
1745 PR tree-optimization/85753
1746 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
1747 RECORD_TYPE in addition to ARRAY_TYPE.
1749 2018-05-15 Martin Sebor <msebor@redhat.com>
1752 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
1754 2018-05-15 Richard Biener <rguenther@suse.de>
1756 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
1757 add by_clobber_p one. Change algorithm to collect all defs
1758 representing uses we need to walk and try reducing them to
1759 a single one before failing.
1760 (dse_dom_walker::dse_optimize_stmt): Adjust.
1762 2018-05-13 Mark Wielaard <mark@klomp.org>
1764 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
1765 (size_of_loc_descr): Likewise.
1766 (output_loc_operands): Likewise.
1767 (output_loc_operands_raw): Likewise.
1768 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
1769 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
1770 (hash_loc_operands): Likewise.
1771 (compare_loc_operands): Likewise.
1773 2018-05-14 Mark Wielaard <mark@klomp.org>
1775 * dwarf2out.c (count_index_addrs): New function.
1776 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
1778 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1780 PR tree-optimization/83648
1781 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
1782 return value as malloc candidate.
1784 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1787 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
1788 as true in call to suggest_attribute.
1790 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
1792 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
1793 -mreadonly-in-sdata.
1795 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1797 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
1799 (aarch64_crypto_aesd_fused): Likewise.
1801 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
1803 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
1804 (movsi_aarch64): Likewise.
1805 (load_pairsi): Likewise.
1806 (load_pairdi): Likewise.
1807 (store_pairsi): Likewise.
1808 (store_pairdi): Likewise.
1809 (load_pairsf): Likewise.
1810 (load_pairdf): Likewise.
1811 (store_pairsf): Likewise.
1812 (store_pairdf): Likewise.
1813 (zero_extend): Likewise.
1814 (trunc): Swap alternatives.
1815 (fcvt_target): Add '?' to prefer w over r.
1817 2018-05-14 Jakub Jelinek <jakub@redhat.com>
1820 * config/i386/i386.md: Disallow non-commutative arithmetics in
1821 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
1822 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
1823 in the peephole2 before it.
1825 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
1827 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
1828 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
1829 (ix86_handle_option): Handle -mcldemote.
1830 * config.gcc: New header.
1831 * config/i386/cldemoteintrin.h: New file.
1832 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
1833 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1835 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1836 OPTION_MASK_ISA_CLDEMOTE.
1837 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
1838 (ix86_valid_target_attribute_inner_p): Ditto.
1839 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
1840 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
1841 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
1842 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
1843 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
1845 * config/i386/i386.opt: Add -mcldemote.
1846 * config/i386/x86intrin.h: New header.
1847 * doc/invoke.texi: Add -mcldemote.
1849 2018-05-14 Richard Biener <rguenther@suse.de>
1851 * doc/match-and-simplify.texi: Adjust :s documentation.
1853 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
1855 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
1856 intended memcpy size.
1857 (REORDER_45): Likewise.
1859 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
1861 * sort.cc: New file.
1862 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
1863 * vec.c (qsort_chk): Use gcc_qsort.
1864 * Makefile.in (OBJS-libcommon): Add sort.o.
1865 (build/sort.o): New target. Use it...
1866 (BUILD_RTL): ... here, and...
1867 (build/gencfn-macros): ... here, and...
1868 (build/genmatch): ... here.
1870 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1871 Chung-Ju Wu <jasonwucj@gmail.com>
1873 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
1874 * config/nds32/nds32-graywolf.md: New file.
1875 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
1876 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
1878 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
1879 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
1880 * config/nds32/nds32.md (pipeline_model): Add graywolf.
1881 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
1882 * config/nds32/pipelines.md: Include n15 settings.
1884 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1885 Chung-Ju Wu <jasonwucj@gmail.com>
1887 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
1888 * config/nds32/nds32-n13.md: New file.
1889 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
1890 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
1892 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
1893 * config/nds32/nds32.md (pipeline_model): Add n13.
1894 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
1895 * config/nds32/pipelines.md: Include n13 settings.
1897 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1898 Chung-Ju Wu <jasonwucj@gmail.com>
1900 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
1901 * config/nds32/nds32-n10.md: New file.
1902 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
1903 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
1905 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
1906 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
1907 * config/nds32/nds32.md (pipeline_model): Add n10.
1908 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
1909 * config/nds32/pipelines.md: Include n10 settings.
1911 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
1912 Kito Cheng <kito.cheng@gmail.com>
1913 Chung-Ju Wu <jasonwucj@gmail.com>
1915 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
1916 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1917 Add enum values for DSP extension instructions.
1918 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
1920 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
1921 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
1923 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
1924 * config/nds32/nds32-dspext.md: New file for DSP implementation.
1925 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
1926 * config/nds32/nds32-intrinsic.md: Likewise.
1927 * config/nds32/nds32_intrinsic.h: Likewise.
1928 * config/nds32/nds32-md-auxiliary.c: Likewise.
1929 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
1930 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
1931 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
1932 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
1933 * config/nds32/nds32-protos.h: New declarations for DSP extension.
1934 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
1935 TYPE_DMAC in switch statement.
1936 * config/nds32/nds32.c: New checking and implementation for DSP
1937 extension instructions.
1938 * config/nds32/nds32.h: Likewise.
1939 * config/nds32/nds32.md: Likewise.
1940 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
1941 * config/nds32/predicates.md: Implement new predicates for DSP
1944 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
1946 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
1947 Reformat alternatives and attributes so it is easier to identify
1948 which constraints/attributes go with which instruction.
1949 (mov<mode>_hardfloat32, FMOVE64): Likewise.
1950 (mov<mode>_softfloat32, FMOVE64): Likewise.
1951 (mov<mode>_hardfloat64, FMOVE64): Likewise.
1952 (mov<mode>_softfloat64, FMOVE64): Likewise.
1954 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
1956 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1958 (Basic PowerPC Built-in Functions): The new name of the
1959 subsection previously known as "PowerPC Built-in Functions".
1960 (Basic PowerPC Built-in Functions Available on all Configurations):
1962 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
1963 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
1964 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
1965 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
1967 2018-05-11 Martin Jambor <mjambor@suse.cz>
1970 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
1973 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1976 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
1978 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
1980 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
1981 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
1982 (ix86_handle_option): Handle -mwaitpkg.
1983 * config.gcc: New header.
1984 * config/i386/cpuid.h (bit_WAITPKG): New bit.
1985 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
1986 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
1988 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1989 OPTION_MASK_ISA_WAITPKG.
1990 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
1991 (ix86_option_override_internal): Add PTA_WAITPKG.
1992 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
1993 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
1994 IX86_BUILTIN_TPAUSE.
1995 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
1996 __builtin_ia32_umwait and __builtin_ia32_tpause.
1997 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
1998 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
1999 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
2000 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
2001 UNSPECV_TPAUSE): New.
2002 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
2003 * config/i386/i386.opt: Add -mwaitpkg.
2004 * config/i386/waitpkgintrin.h: New file.
2005 * config/i386/x86intrin.h: New header.
2006 * doc/invoke.texi: Add -mwaitpkg.
2008 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2011 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
2013 (cortex-m0): Use armv6s-m isa.
2014 (cortex-m0plus): Likewise.
2015 (cortex-m1): Likewise.
2016 (cortex-m0.small-multiply): Likewise.
2017 (cortex-m0plus.small-multiply): Likewise.
2018 (cortex-m1.small-multiply): Likewise.
2020 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
2021 Jakub Jelinek <jakub@redhat.com>
2023 PR tree-optimization/85692
2024 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
2025 source permute as well.
2027 2018-05-11 Martin Liska <mliska@suse.cz>
2030 * doc/extend.texi: Document LLVM style format for no_sanitize
2033 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
2035 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
2036 mode_supports_vsx_dform_quad to mode_supports_dq_form.
2037 (mode_supports_vsx_dform_quad): Likewise.
2038 (mode_supports_vmx_dform): Move these functions to be next to the
2039 other mode_supports functions.
2040 (mode_supports_dq_form): Likewise.
2041 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
2042 mode_supports_dq_form.
2043 (reg_offset_addressing_ok_p): Likewise.
2044 (offsettable_ok_by_alignment): Likewise.
2045 (rs6000_legitimate_offset_address_p): Likewise.
2046 (legitimate_lo_sum_address_p): Likewise.
2047 (rs6000_legitimize_address): Likewise.
2048 (rs6000_legitimize_reload_address): Likewise.
2049 (rs6000_secondary_reload_inner): Likewise.
2050 (rs6000_preferred_reload_class): Likewise.
2051 (rs6000_output_move_128bit): Likewise.
2053 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2055 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
2056 Generate SImode target register for null target.
2057 <case IX86_BUILTIN_XGETBV>: Ditto.
2058 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
2059 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
2061 2018-05-10 Carl Love <cel@us.ibm.com>
2063 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2064 dcbtt and dcbtstt if operands[2] is 0.
2066 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2069 * config/i386/sse.md (usadv64qi): New expander.
2071 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
2073 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
2074 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
2075 -maltivec=be support.
2076 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
2077 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
2078 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
2079 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
2080 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
2081 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
2082 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
2083 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
2084 altivec_vsumsws): Adjust.
2085 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
2086 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
2087 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
2089 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
2090 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
2091 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
2092 (altivec_lve<VI_char>x): Delete expand.
2093 (*altivec_lve<VI_char>x_internal): Rename to...
2094 (altivec_lve<VI_char>x): ... this.
2095 (altivec_lvxl_<mode>): Delete expand.
2096 (*altivec_lvxl_<mode>_internal): Rename to ...
2097 (altivec_lvxl_<mode>): ... this.
2098 (altivec_stvxl_<mode>): Delete expand.
2099 (*altivec_stvxl_<mode>_internal): Rename to ...
2100 (altivec_stvxl_<mode>): ... this.
2101 (altivec_stve<VI_char>x): Delete expand.
2102 (*altivec_stve<VI_char>x_internal): Rename to ...
2103 (altivec_stve<VI_char>x): ... this.
2104 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
2105 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
2106 reduc_plus_scal_<mode>): Adjust.
2107 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
2109 (rs6000_cpu_cpp_builtins): Adjust.
2110 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
2111 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
2112 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
2113 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
2114 -maltivec=be support.
2115 (rs6000_split_vec_extract_var): Adjust.
2116 (rs6000_split_v4si_init): Adjust.
2117 (swap_selector_for_mode): Delete.
2118 (altivec_expand_lvx_be, altivec_expand_stvx_be,
2119 altivec_expand_stvex_be): Delete.
2120 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
2121 -maltivec=be support.
2122 (rs6000_gimple_fold_builtin): Ditto.
2123 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
2125 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
2126 (TARGET_DIRECT_MOVE_64BIT): Adjust.
2127 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
2128 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
2129 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
2130 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
2131 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
2132 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
2133 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
2134 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
2135 anonymous split): Adjust.
2136 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
2137 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
2139 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
2141 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
2142 when --with-gxx-include-dir is also specified.
2143 * configure: Regenerate.
2145 2018-05-09 Jim Wilson <jimw@sifive.com>
2148 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
2149 * config/riscv/t-withmultilib: New.
2150 * config/riscv/withmultilib.h: New.
2151 * doc/install.texi: Document RISC-V --with-multilib-list support.
2153 2018-05-09 Richard Biener <rguenther@suse.de>
2155 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
2157 (vect_bb_vectorization_profitable_p): Adjust. Compute
2158 actual scalar cost using the cost vector and the add_stmt_cost
2161 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2163 PR rtl-optimization/85645
2164 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
2165 in the REG_CFA_REGISTER note for LR, don't leave it empty.
2167 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2169 PR rtl-optimization/85645
2170 * shrink-wrap.c (spread_components): Return a boolean saying if
2171 anything was changed.
2172 (try_shrink_wrapping_separate): Iterate spread_components until
2173 nothing changes anymore.
2175 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2177 PR rtl-optimization/85645
2178 * regrename.c (build_def_use): Also kill the chains that include the
2179 destination of a REG_CFA_REGISTER note.
2181 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2183 PR rtl-optimization/85645
2184 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
2185 insn that has a REG_CFA_REGISTER note.
2187 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
2189 * cfgexpand.c (expand_clobber): New function.
2190 (expand_gimple_stmt_1): Use it.
2191 * tree-vect-stmts.c (vect_clobber_variable): New function,
2193 (vectorizable_simd_clone_call): ...here.
2194 (vectorizable_store): Emit a clobber either side of an
2195 IFN_STORE_LANES sequence.
2196 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
2198 2018-05-09 Tom de Vries <tom@codesourcery.com>
2201 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
2202 (define_insn "trap_if_false"): Add exit after trap.
2204 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
2206 PR rtl-optimization/85638
2207 * bb-reorder.c: Include common/common-target.h.
2208 (create_forwarder_block): New function extracted from...
2209 (fix_up_crossing_landing_pad): ...here. Rename into...
2210 (dw2_fix_up_crossing_landing_pad): ...this.
2211 (sjlj_fix_up_crossing_landing_pad): New function.
2212 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
2213 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
2214 from both partitions and exit the loop after one iteration.
2216 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2219 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2221 (Basic PowerPC Built-in Functions): The new name of the
2222 subsection previously known as "PowerPC Built-in Functions".
2223 (Basic PowerPC Built-in Functions Available on all Configurations):
2225 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2227 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2229 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2231 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2234 2018-05-08 Jim Wilson <jimw@sifive.com>
2236 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
2237 (LD_EMUL_SUFFIX): New.
2238 (LINK_SPEC): Use it.
2240 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2242 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2244 (Basic PowerPC Built-in Functions): The new name of the
2245 subsection previously known as "PowerPC Built-in Functions".
2246 (Basic PowerPC Built-in Functions Available on all Configurations):
2248 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2250 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2252 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2254 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2257 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2260 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
2261 after cmpelim optimization.
2263 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
2265 * config.gcc: Support "goldmont".
2266 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
2267 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2269 * config/i386/i386.c (m_GOLDMONT): Define.
2270 (processor_target_table): Add "goldmont".
2271 (PTA_GOLDMONT): Define.
2272 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
2273 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
2274 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
2275 (fold_builtin_cpu): Add "goldmont".
2276 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
2277 (ix86_option_override_internal): Add "goldmont".
2278 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
2279 (processor_type): Add PROCESSOR_GOLDMONT.
2280 * config/i386/i386.md: Add CPU "glm".
2281 * config/i386/glm.md: New file.
2282 * config/i386/x86-tune.def: Add m_GOLDMONT.
2283 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
2285 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2288 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
2290 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
2291 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
2292 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
2295 * config/i386/i386.c (ix86_fold_builtin): Handle
2296 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
2299 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
2300 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
2302 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
2305 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
2306 (check_arch): Likewise.
2307 (check_fpu): Return the result rather than printing it.
2308 (end arch): Fix operator precedence.
2309 (end cpu): Likewise.
2310 (END): Print the result from check_fpu.
2312 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2313 Alan Hayward <alan.hayward@arm.com>
2314 David Sherwood <david.sherwood@arm.com>
2316 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
2317 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
2318 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
2319 (*fcmuo<mode>_and): New patterns.
2321 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2323 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
2324 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
2325 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
2326 (cmp_op, sve_imm_con): New code attributes.
2327 (SVE_COND_INT_CMP, imm_con): Delete.
2328 (cmp_op): Remove above unspecs from int attribute.
2329 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
2331 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
2332 comparison-specific unspecs.
2333 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
2334 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
2335 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
2336 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
2337 (*vec_fcm<cmp_op><mode>): Rename to...
2338 (*fcm<cmp_op><mode>): ...this and adjust likewise.
2339 (*vec_fcmuo<mode>): Rename to...
2340 (*fcmuo<mode>): ...this and adjust likewise.
2341 (*pred_fcm<cmp_op><mode>): New pattern.
2342 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
2343 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
2345 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
2347 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
2348 (aarch64_emit_sve_predicated_cond): New function.
2349 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
2350 (aarch64_emit_unspec_cond_or): Replace with...
2351 (aarch64_emit_sve_or_conds): ...this new function. Use
2352 aarch64_emit_sve_ptrue_op for the individual comparisons and
2353 aarch64_emit_binop to OR them together.
2354 (aarch64_emit_inverted_unspec_cond): Replace with...
2355 (aarch64_emit_sve_inverted_cond): ...this new function. Use
2356 aarch64_emit_sve_ptrue_op for the comparison and
2357 aarch64_emit_unop to invert the result.
2358 (aarch64_expand_sve_vec_cmp_float): Update after the above
2359 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
2361 2018-05-07 Nathan Sidwell <nathan@acm.org>
2363 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
2364 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
2365 (Backwards Compatibility): Likewise.
2367 2018-05-07 Luis Machado <luis.machado@linaro.org>
2371 2018-05-07 Luis Machado <luis.machado@linaro.org>
2373 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2374 <prefetch_dynamic_strides>: New const bool field.
2375 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2376 prefetch_dynamic_strides.
2377 (exynosm1_prefetch_tune): Likewise.
2378 (thunderxt88_prefetch_tune): Likewise.
2379 (thunderx_prefetch_tune): Likewise.
2380 (thunderx2t99_prefetch_tune): Likewise.
2381 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2382 (aarch64_override_options_internal): Update to set
2383 PARAM_PREFETCH_DYNAMIC_STRIDES.
2384 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2385 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2386 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2387 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2388 prefetch-dynamic-strides setting.
2390 2018-05-07 Luis Machado <luis.machado@linaro.org>
2392 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2393 <minimum_stride>: New const int field.
2394 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2395 minimum_stride field.
2396 (exynosm1_prefetch_tune): Likewise.
2397 (thunderxt88_prefetch_tune): Likewise.
2398 (thunderx_prefetch_tune): Likewise.
2399 (thunderx2t99_prefetch_tune): Likewise.
2400 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2401 (aarch64_override_options_internal): Update to set
2402 PARAM_PREFETCH_MINIMUM_STRIDE.
2403 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2404 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2405 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2406 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2407 stride is constant and is below the minimum stride threshold.
2409 2018-05-07 Luis Machado <luis.machado@linaro.org>
2411 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
2414 2018-05-07 Luis Machado <luis.machado@linaro.org>
2416 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2417 <prefetch_dynamic_strides>: New const bool field.
2418 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2419 prefetch_dynamic_strides.
2420 (exynosm1_prefetch_tune): Likewise.
2421 (thunderxt88_prefetch_tune): Likewise.
2422 (thunderx_prefetch_tune): Likewise.
2423 (thunderx2t99_prefetch_tune): Likewise.
2424 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2425 (aarch64_override_options_internal): Update to set
2426 PARAM_PREFETCH_DYNAMIC_STRIDES.
2427 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2428 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2429 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2430 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2431 prefetch-dynamic-strides setting.
2433 2018-05-07 Luis Machado <luis.machado@linaro.org>
2435 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2436 <minimum_stride>: New const int field.
2437 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2438 minimum_stride field.
2439 (exynosm1_prefetch_tune): Likewise.
2440 (thunderxt88_prefetch_tune): Likewise.
2441 (thunderx_prefetch_tune): Likewise.
2442 (thunderx2t99_prefetch_tune): Likewise.
2443 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2444 (aarch64_override_options_internal): Update to set
2445 PARAM_PREFETCH_MINIMUM_STRIDE.
2446 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2447 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2448 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2449 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2450 stride is constant and is below the minimum stride threshold.
2452 2018-05-06 Jakub Jelinek <jakub@redhat.com>
2455 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
2456 the type is addressable. Don't force op into register if it has
2459 2018-05-05 Roland McGrath <mcgrathr@google.com>
2462 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
2463 any section for which we don't know a specific type it should have,
2464 regardless of name. Previously this was done only for the exact
2465 names ".init_array", ".fini_array", and ".preinit_array".
2466 (default_elf_asm_named_section): Add comment about
2467 relationship with default_section_type_flags and SECTION_NOTYPE.
2468 (get_section): Don't consider it a type conflict if one side has
2469 SECTION_NOTYPE and the other doesn't, as long as neither has the
2470 SECTION_BSS et al used in the default_section_type_flags logic.
2472 2018-05-05 Tom de Vries <tom@codesourcery.com>
2475 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
2476 (workaround_barsyncs): New function.
2477 (nvptx_reorg): Use workaround_barsyncs.
2478 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2479 (define_expand "nvptx_membar_cta"): New define_expand.
2480 (define_insn "*nvptx_membar_cta"): New insn.
2482 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
2484 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
2485 To improve optimization opportunities.
2486 * builtin-types.def: The new needed builtin types for the above.
2488 2018-05-04 Richard Biener <rguenther@suse.de>
2490 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
2491 * gimple-ssa-store-merging.c
2492 (imm_store_chain_info::output_merged_store): Remove redundant create,
2493 release split_store vector contents on failure.
2494 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
2495 scalar stmt vector on cache hit.
2497 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
2499 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
2501 * config.gcc (powerpc-xilinx-eabi*): Remove.
2502 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
2504 (fusion_addis_mem_combo_load): Ditto.
2505 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
2507 (rs6000_cpu_cpp_builtins): Ditto.
2508 * config/rs6000/rs6000-linux.c
2509 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
2510 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
2511 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
2513 (rs6000_setup_reg_addr_masks): Ditto.
2514 (rs6000_init_hard_regno_mode_ok): Ditto.
2515 (rs6000_option_override_internal): Ditto.
2516 (legitimate_lo_sum_address_p): Ditto.
2517 (rs6000_legitimize_address): Ditto.
2518 (rs6000_legitimize_reload_address): Ditto.
2519 (rs6000_legitimate_address_p): Ditto.
2520 (abi_v4_pass_in_fpr): Ditto.
2521 (setup_incoming_varargs): Ditto.
2522 (rs6000_gimplify_va_arg): Ditto.
2523 (rs6000_split_multireg_move): Ditto.
2524 (rs6000_savres_strategy): Ditto.
2525 (rs6000_emit_prologue_components): Ditto.
2526 (rs6000_emit_epilogue_components): Ditto.
2527 (rs6000_emit_prologue): Ditto.
2528 (rs6000_emit_epilogue): Ditto.
2529 (rs6000_elf_file_end): Ditto.
2530 (rs6000_function_value): Ditto.
2531 (rs6000_libcall_value): Ditto.
2532 * config/rs6000/rs6000.h: Ditto.
2533 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
2534 (TARGET_MINMAX): ... this. New.
2535 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
2536 * config/rs6000/rs6000.md: Remove Xilinx FP support.
2537 (*movsi_internal1_single): Delete.
2538 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
2539 mfpu=, mxilinx-fpu): Delete.
2540 * config/rs6000/singlefp.h: Delete.
2541 * config/rs6000/sysv4.h: Remove Xilinx FP support.
2542 * config/rs6000/t-rs6000: Ditto.
2543 * config/rs6000/t-xilinx: Delete.
2544 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
2545 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
2546 (VStype_simple): Delete.
2547 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
2548 * config/rs6000/xfpu.h: Delete.
2549 * config/rs6000/xfpu.md: Delete.
2550 * config/rs6000/xilinx.h: Delete.
2551 * config/rs6000/xilinx.opt: Delete.
2552 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
2553 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
2555 2018-05-04 Tom de Vries <tom@codesourcery.com>
2558 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
2561 2018-05-04 Richard Biener <rguenther@suse.de>
2564 * tree-complex.c (update_complex_assignment): We are always in SSA form.
2565 (expand_complex_div_wide): Likewise.
2566 (expand_complex_operations_1): Likewise.
2567 (expand_complex_libcall): Preserve EH info of the original stmt.
2568 (tree_lower_complex): Handle removed blocks.
2569 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
2570 on complex multiplication and division libcall builtins.
2572 2018-05-04 Richard Biener <rguenther@suse.de>
2575 * fold-const.c (negate_expr_p): Restrict negation of operand
2576 zero of a division to when we know that can happen without
2578 (fold_negate_expr_1): Likewise.
2580 2018-05-04 Jakub Jelinek <jakub@redhat.com>
2583 * real.h (real_nextafter): Declare.
2584 * real.c (real_nextafter): New function.
2585 * fold-const-call.c (fold_const_nextafter): New function.
2586 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
2587 CASE_CFN_NEXTTOWARD.
2588 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
2589 even when arg1_mode is different from arg0_mode.
2591 2018-05-03 Nathan Sidwell <nathan@acm.org>
2593 * doc/extend.texi (Deprecated Features): Remove
2595 (Backwards Compatibility): Likewise.
2596 * doc/invoke.texi (C++ Language Options): Likewise.
2597 (C++ Dialect Options): Likewise.
2599 2018-05-03 Jakub Jelinek <jakub@redhat.com>
2602 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
2603 _mm512_mask_mullox_epi64): New intrinsics.
2605 2018-05-03 Tom de Vries <tom@codesourcery.com>
2608 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2609 dump files): Add offload-tree.
2611 2018-05-03 Richard Biener <rguenther@suse.de>
2613 PR tree-optimization/85615
2614 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
2615 to loops not nested in BBs loop father to avoid creating multi-entry
2618 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2620 PR tree-optimization/70291
2621 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
2622 arguments. Change return type to tree. Emit libcall as a new
2623 statement rather than replacing existing one when inplace_p is true.
2624 (expand_complex_multiplication_components): New function.
2625 (expand_complex_multiplication): Expand floating-point complex
2626 multiplication using the above.
2627 (expand_complex_division): Rename inner_type parameter to type.
2628 Update expand_complex_libcall call-site.
2629 (expand_complex_operations_1): Update expand_complex_multiplication
2630 and expand_complex_division call-sites.
2632 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2635 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2636 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2637 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
2638 the highest significant bit of the shift count mask is clear. In
2639 check whether and[sq]i3 is needed verify that all significant bits
2640 of the shift count other than the highest are set.
2642 2018-05-02 Tom de Vries <tom@codesourcery.com>
2645 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
2646 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
2647 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
2648 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
2649 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
2650 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
2651 __builtin_goacc_parlevel_size.
2653 2018-05-02 Richard Biener <rguenther@suse.de>
2655 PR tree-optimization/85597
2656 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
2657 do not use split vect_get_vec_defs call but call vect_get_slp_defs
2660 2018-05-02 Tom de Vries <tom@codesourcery.com>
2663 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2664 dump files): Add ltrans-tree.
2666 2018-05-02 Tom de Vries <tom@codesourcery.com>
2669 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2670 dump files): Add wpa-ipa.
2672 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
2674 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
2675 powerpc*-*-linux*paired* target.
2676 * config/rs6000/750cl.h: Delete.
2677 * config/rs6000/paired.h: Delete.
2678 * config/rs6000/paired.md: Delete.
2679 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
2681 * config/rs6000/rs6000-builtin.def: Remove paired float support.
2682 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
2683 comment. Remove paired float support.
2684 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
2685 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
2687 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
2688 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
2690 * config/rs6000/rs6000.c: Remove paired float support.
2691 (paired_expand_vector_init, paired_expand_vector_move,
2692 paired_emit_vector_compare, paired_emit_vector_cond_expr,
2693 (paired_expand_lv_builtin, paired_expand_stv_builtin,
2694 paired_expand_builtin, paired_expand_predicate_builtin,
2695 paired_init_builtins): Delete.
2696 * config/rs6000/rs6000.h: Remove paired float support.
2697 * config/rs6000/rs6000.md: Remove paired float support.
2698 (move_from_CR_ov_bit): Delete.
2699 * config/rs6000/rs6000.opt (mpaired): Delete.
2700 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
2701 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
2703 2018-05-02 Richard Biener <rguenther@suse.de>
2706 * gimplify.c (gimplify_save_expr): When in SSA form allow
2707 SAVE_EXPRs to compute to SSA vars.
2709 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2712 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2713 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2714 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
2715 clobber operands[2], instead use a new pseudo. Formatting fixes.
2717 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
2719 PR tree-optimization/85586
2720 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
2721 exit early for statements in the same group if the accesses are
2724 2018-05-02 Tom de Vries <tom@codesourcery.com>
2727 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
2730 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
2732 PR tree-optimization/85143
2733 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
2735 2018-05-01 Tom de Vries <tom@codesourcery.com>
2738 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
2739 not found" error message.
2741 2018-05-01 Tom de Vries <tom@codesourcery.com>
2744 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
2745 * vec.c (test_ordered_remove_if): New function.
2746 (vec_c_tests): Call test_ordered_remove_if.
2747 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
2748 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
2749 * tree-vect-patterns.c (vect_pattern_recog_1): Use
2750 VEC_ORDERED_REMOVE_IF.
2752 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2754 PR tree-optimization/82665
2755 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
2756 pointer subtraction where arguments come from a memchr call.
2758 2018-05-01 Jakub Jelinek <jakub@redhat.com>
2760 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
2761 --push-state --as-needed and --pop-state instead of --as-needed and
2762 --no-as-needed if ld supports it.
2763 * configure: Regenerated.
2766 * doc/install.texi2html: Replace _002d with - and _002a with * in
2767 generated html files using sed.
2769 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2772 * gcc-rich-location.c (blank_line_before_p): New function.
2773 (use_new_line): New function.
2774 (gcc_rich_location::add_fixit_insert_formatted): New function.
2775 * gcc-rich-location.h
2776 (gcc_rich_location::add_fixit_insert_formatted): New function.
2778 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2780 * selftest.c (assert_streq): Rename "expected" and "actual" to
2781 "val1" and "val2". Extend NULL-handling to cover both inputs
2782 symmetrically, while still requiring both to be non-NULL for a pass.
2783 * selftest.h (assert_streq): Rename "expected" and "actual" to
2785 (ASSERT_EQ): Likewise.
2786 (ASSERT_EQ_AT): Likewise.
2787 (ASSERT_KNOWN_EQ): Likewise.
2788 (ASSERT_KNOWN_EQ_AT): Likewise.
2789 (ASSERT_NE): Likewise.
2790 (ASSERT_MAYBE_NE): Likewise.
2791 (ASSERT_MAYBE_NE_AT): Likewise.
2792 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
2793 the assertion to pass.
2794 (ASSERT_STREQ_AT): Likewise.
2796 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2798 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
2799 interaction with -pie.
2801 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2803 * selftest.h: Fix alphabetization of per-source-file selftest
2806 2018-04-30 Jason Merrill <jason@redhat.com>
2808 PR c++/61982 - dead stores to destroyed objects.
2809 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
2812 2018-04-30 Jason Merrill <jason@redhat.com>
2814 * tree.c (build_clobber): New.
2815 * tree.h: Declare it.
2816 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
2818 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2820 * diagnostic-show-locus.c (layout::layout): Update for
2821 location_get_source_line returning a char_span.
2822 (struct char_span): Move to input.h.
2823 (struct correction): Update for fields in char_span becoming
2825 (struct source_line): Update for location_get_source_line
2826 returning a char_span.
2827 (layout::print_line): Likewise.
2828 * edit-context.c (edited_file::print_content): Likewise.
2829 (edited_file::print_diff_hunk): Likewise.
2830 (edited_file::print_run_of_changed_lines): Likewise.
2831 (edited_file::get_num_lines): Likewise.
2832 (edited_line::edited_line): Likewise.
2833 * final.c (asm_show_source): Likewise.
2834 * input.c (location_get_source_line): Convert return type
2835 from const char * to char_span, losing the final "line_len"
2837 (dump_location_info): Update for the above.
2838 (get_substring_ranges_for_loc): Likewise. Use a char_span
2839 when handling the literal within the line.
2840 (test_reading_source_line): Update for location_get_source_line
2841 returning a char_span.
2842 * input.h (class char_span): Move here from
2843 diagnostic-show-locus.c, converting from a struct to a class.
2844 Make data members private.
2845 (char_span::operator bool): New.
2846 (char_span::length): New.
2847 (char_span::get_buffer): New.
2848 (char_span::operator[]): New.
2849 (char_span::subspan): Make const.
2850 (char_span::xstrdup): New.
2851 (location_get_source_line): Convert return type from const char *
2852 to char_span, losing the final "line_size" param.
2854 2018-04-30 Jan Hubicka <jh@suse.cz>
2856 * lto-wrapper.c (ltrans_priorities): New static var.
2857 (cmp_priority): New.
2858 (run_gcc): Read priorities and if doing parallel build order
2859 the Makefile by them.
2861 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2863 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
2865 2018-04-30 Richard Biener <rguenther@suse.de>
2867 * tree-cfg.c (verify_address): Remove base argument, add
2868 flag whether to check TREE_ADDRESSABLE and do that.
2869 (verify_expr): Remove.
2870 (verify_types_in_gimple_reference): Add pieces from verify_expr.
2871 (verify_gimple_assign_single): Likewise.
2872 (verify_gimple_switch): Likewise.
2873 (verify_expr_location_1): Dereference tp once. Add (disabled)
2874 piece from verify_expr.
2875 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
2877 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2879 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
2881 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2883 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
2884 (small_data_pattern): Likewise.
2885 (arc_rewrite_small_data): Likewise.
2886 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
2887 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
2888 (get_symbol_alignment): New function.
2889 (legitimate_small_data_address_p): Likewise.
2890 (legitimate_scaled_address): Update, call
2891 legitimate_small_data_address_p.
2892 (output_sdata): New static variable.
2893 (arc_print_operand): Update how we handle small data operands.
2894 (arc_print_operand_address): Likewise.
2895 (arc_legitimate_address_p): Update, use
2896 legitimate_small_data_address_p.
2897 (arc_rewrite_small_data_p): Remove.
2898 (arc_rewrite_small_data_1): Likewise.
2899 (arc_rewrite_small_data): Likewise.
2900 (small_data_pattern): Likewise.
2901 (compact_sda_memory_operand): Update to use
2902 legitimate_small_data_address_p and get_symbol_alignment.
2903 (prepare_move_operands): Don't rewite sdata pattern.
2904 (prepare_extend_operands): Remove.
2905 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
2907 (zero_extendqisi2): Likewise.
2908 (zero_extendhisi2): Likewise.
2909 (extendqihi2): Likewise.
2910 (extendqisi2): Likewise.
2911 (extendhisi2): Likewise.
2915 * config/arc/constraints.md (Usd): Change it to memory constraint.
2917 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2919 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
2920 as source of std instructions.
2921 * config/arc/arc.md (movsi_insn): Update pattern predicate to
2922 allow 6-bit constants as source for store instructions.
2923 (movdi_insn): Update instruction pattern to allow 6-bit constants
2924 as source for store instructions.
2926 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2928 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
2930 2018-04-30 Nathan Sidwell <nathan@acm.org>
2931 Sandra Loosemore <sandra@codesourcery.com>
2933 * dumpfile.c (dump_open): Allow '-' for stdout.
2934 * doc/invoke.texi (Developer Options): Document dump filename
2935 determination early. Document stdin/stdout selection.
2937 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
2939 Microblaze Target: PIC data text relative
2941 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
2942 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
2944 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
2945 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
2946 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
2947 New addressing mode for data-text relative position indepenedent code.
2948 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
2949 'ADDRESS_SYMBOLIC_TXT_REL'.
2950 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
2951 (microblaze_legitimate_pic_operand): Exclude function calls from
2952 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
2953 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
2955 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2956 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2957 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
2958 for 'address + offset'.
2959 (microblaze_expand_prologue): Add new function prologue call for
2961 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
2962 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
2963 table in case of TARGET_PIC_DATA_TEXT_REL.
2964 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
2965 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
2966 Add new macros 'UNSPEC_TEXT',
2967 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
2968 + exclude function calls from 'UNSPEC_PLT' in case of data text
2970 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2971 new target hook for generating address diff vector tables in case of
2973 * doc/tm.texi : Regenerate.
2974 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
2975 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
2976 of addr diff vector generation.
2977 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2978 target hook definition.
2979 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
2980 Add default function for generate_pic_addr_diff_vec -> flag_pic.
2981 * doc/invoke.texi (Add new pic option): Add new microblaze pic
2982 option for data text relative.
2984 2018-04-30 Richard Biener <rguenther@suse.de>
2986 * tree-chrec.h (evolution_function_is_constant_p): Remove
2988 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
2990 2018-04-30 Richard Biener <rguenther@suse.de>
2993 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
2995 2018-04-30 Richard Biener <rguenther@suse.de>
2997 PR tree-optimization/28364
2998 PR tree-optimization/85275
2999 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
3000 copying first exit test.
3002 2018-04-28 Mark Wielaard <mark@klomp.org>
3004 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
3006 (dwarf_AT): Handle DW_AT_addr_base.
3007 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
3009 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
3012 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
3013 (*ashl<dwi>3_doubleword_mask_1): Ditto.
3014 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
3015 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
3017 2018-04-28 Richard Biener <rguenther@suse.de>
3019 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
3020 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
3021 to reflect use. Only add interesting stmts.
3023 2018-04-27 Martin Jambor <mjambor@suse.cz>
3026 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
3027 the jump function allows for passing through aggregate values.
3029 2018-04-27 David Malcolm <dmalcolm@redhat.com>
3031 * input.h (in_system_header_at): Convert from macro to inline
3033 (from_macro_expansion_at): Likewise.
3034 (from_macro_definition_at): Likewise.
3036 2018-04-27 Jeff Law <law@redhat.com>
3038 * config.gcc: Mark tile* targets as deprecated/obsolete.
3040 2018-04-27 Richard Biener <rguenther@suse.de>
3042 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
3045 2018-04-27 Richard Biener <rguenther@suse.de>
3047 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
3049 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
3051 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
3052 with Yd constraint. Set "preferred_for_speed" attribute from
3053 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
3055 (*movdi_internal): Ditto.
3056 (movti_interunit splitters): Remove
3057 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
3058 (movdi_interunit splitters): Ditto.
3059 * config/i386/constraints.md (Ye): Remove.
3060 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
3062 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3065 * config/aarch64/constraints.md (Usg): Limit to 31.
3068 2018-04-27 Jakub Jelinek <jakub@redhat.com>
3070 PR tree-optimization/85529
3071 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
3072 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
3073 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
3074 zero extension or masking of the MSB bit.
3075 (optimize_range_tests): Add FIRST_BB argument, pass it through
3076 to optimize_range_tests_var_bound.
3077 (maybe_optimize_range_tests, reassociate_bb): Adjust
3078 optimize_range_tests callers.
3080 2018-04-26 Richard Biener <rguenther@suse.de>
3081 Jakub Jelinek <jakub@redhat.com>
3083 * cgraph.h (symbol_table): Just declare debug method here.
3084 * symtab.c (symbol_table::debug): Define.
3086 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
3088 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
3090 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
3092 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
3093 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
3094 (*movdi_internal): Substitute Yi and Yj constraint with x
3095 and Ym and Yn constraint with y constraint. Update "isa"
3096 attribute and set "preferred_for_speed" attribute from
3097 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
3098 (*movsi_internal): Ditto.
3099 (*movdf_internal): Ditto.
3100 (*movsf_internal): Ditto.
3101 (*zero_extendsidi2): Ditto.
3102 * config/i386/sse.md (vec_set<mode>_0): Ditto.
3103 (sse2_loadld): Ditto.
3104 (*vec_extract<ssevecmodelower>_0): Ditto.
3105 (*vec_extractv4si_0_zext_sse4): Ditto.
3106 (vec_concatv2di): Ditto.
3107 (*vec_dup<mode>): Ditto.
3108 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
3109 * config/i386/constraints.md (Yi): Remove.
3114 2018-04-26 Nathan Sidwell <nathan@acm.org>
3116 * dumpfile.c (dump_open): New.
3117 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
3118 (dump_finish): Detect stdio/stderr by value not name.
3120 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
3122 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
3124 2018-04-26 Tom de Vries <tom@codesourcery.com>
3127 * config/nvptx/nvptx.c (verify_neutering_jumps)
3128 (verify_neutering_labels): New function
3129 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
3131 2018-04-26 Tom de Vries <tom@codesourcery.com>
3134 * config/nvptx/nvptx.c (needs_neutering_p): New function.
3135 (nvptx_single): Use needs_neutering_p to skip over insns that do not
3138 2018-04-26 Richard Biener <rguenther@suse.de>
3139 Tom de Vries <tom@codesourcery.com>
3142 * lto-streamer-out.c (output_function): Fixup loops if required to match
3143 discovery done in the reader.
3145 2018-04-26 Richard Biener <rguenther@suse.de>
3147 PR tree-optimization/85116
3148 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
3149 have a loop exit from the single latch predecessor. Remove
3150 case of header with just condition.
3151 (ch_base::copy_headers): Exclude infinite loops from any
3153 (pass_ch::execute): Record exits.
3155 2018-04-26 Richard Biener <rguenther@suse.de>
3157 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
3158 prologue cost vector and pass it to vect_get_load_cost.
3159 (vect_get_peeling_costs_all_drs): Likewise.
3160 (vect_peeling_hash_get_lowest_cost): Likewise.
3161 (vect_enhance_data_refs_alignment): Likewise.
3163 2018-04-26 Richard Biener <rguenther@suse.de>
3166 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
3167 checking of integer<->pointer conversions.
3168 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
3169 sign-/zero-extending pointer types.
3170 (expand_omp_for_static_chunk): Likewise.
3172 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
3173 Jean Lee <xiaoyur347@gmail.com>
3175 * config/mips/mips.c (mips_asan_shadow_offset): New function.
3176 (TARGET_ASAN_SHADOW_OFFSET): Define.
3177 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
3178 true for -fsanitize=address.
3180 2018-04-25 Mark Wielaard <mark@klomp.org>
3182 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
3185 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3187 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
3188 than "alu", remove explicit "memory" and "imm_disp" attributes.
3189 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
3192 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
3193 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
3194 gen_lowpart_no_emit.
3196 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
3199 * config/i386/i386.c (ix86_expand_builtin): Change memory
3200 operand to XI, extend p0 to Pmode.
3201 * config/i386/i386.md: Change unspec volatile and operand
3202 1 mode to XI, change operand 0 mode to P.
3204 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3206 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
3207 GET_MODE_MASK before any checking.
3208 (nds32_can_use_bset_p): Likewise.
3209 (nds32_can_use_btgl_p): Likewise.
3211 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3213 * config/nds32/nds32-doubleword.md: New define_split pattern for
3214 illegal register number.
3216 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3218 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
3220 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3222 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
3224 2018-04-25 Richard Biener <rguenther@suse.de>
3226 * lto-streamer.h (LTO_major_version): Bump to 8.
3228 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3230 * BASE-VER: Set to 9.0.0.
3232 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
3234 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
3235 in __abskf2 and __powikf2.
3237 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3240 * config/aarch64/constraints.md (Usg, Usj): New constraints.
3241 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
3242 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
3243 Use the above on operand 2. Reindent.
3244 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
3246 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3249 * common/config/i386/i386-common.c (ix86_handle_option): Don't
3251 * config/i386/i386.opt (mcet): Removed.
3252 * doc/install.texi: Remove -mcet documentation.
3253 * doc/invoke.texi: Likewise.
3255 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3258 * doc/install.texi: Remove -mcet from bootstrap-cet.
3260 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3263 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
3264 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
3268 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
3269 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
3270 containing a CONST_VECTOR.
3272 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
3274 * doc/install.texi: Update newlib dependency for nvptx.
3276 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3279 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
3280 instead of INTVAL when shifting x left.
3282 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
3284 PR tree-optimization/85478
3285 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
3286 vect_grouped_store_supported for single element vectors.
3288 2018-04-24 Richard Biener <rguenther@suse.de>
3291 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
3292 load cost increase to the case of non-constant step.
3294 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3297 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
3298 destination if any_malformed_asm.
3300 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
3303 * expr.c (store_field): In the bitfield case, if the value comes from
3304 a function call and is returned in registers by means of a PARALLEL,
3305 do not change the mode of the temporary unless BLKmode and VOIDmode.
3307 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
3309 PR rtl-optimization/85423
3310 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
3311 dependencies to debug insns when the previous insn is non-debug.
3313 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3315 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
3316 enums into a single definition.
3317 (fls): Fix predicates and printing.
3320 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3322 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
3323 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
3324 and short u6 immediate.
3325 (check_if_valid_sleep_operand): Remove.
3326 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
3328 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3330 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
3331 flag_always_save_lp condition.
3332 * config/nds32/nds32.opt (malways-save-lp): New option.
3334 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3336 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
3337 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
3338 * config/nds32/nds32.h
3339 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
3340 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
3342 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3344 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
3345 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
3347 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3348 Chung-Ju Wu <jasonwucj@gmail.com>
3350 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
3352 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
3353 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
3355 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3357 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
3359 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3361 * config/nds32/nds32-protos.h (nds32_data_alignment,
3362 nds32_local_alignment): Declare.
3363 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
3364 nds32_local_alignment): New functions.
3365 (TARGET_CONSTANT_ALIGNMENT): Define.
3366 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
3368 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3370 * config/nds32/nds32.c
3371 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
3372 (TARGET_MODES_TIEABLE_P): Likewise.
3374 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3376 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
3379 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
3380 Chung-Ju Wu <jasonwucj@gmail.com>
3382 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3383 for unaligned access.
3384 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3386 * config/nds32/nds32-intrinsic.md: Likewise.
3387 * config/nds32/nds32_intrinsic.h: Likewise.
3388 * config/nds32/nds32.h (nds32_builtins): Likewise.
3389 * config/nds32/nds32.opt (munaligned-access): New option.
3390 * config/nds32/nds32.c (nds32_asm_file_start): Display
3391 flag_unaligned_access status.
3393 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3395 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
3396 -mno-relax is present.
3397 * config/riscv/linux.h (LINK_SPEC): Ditto.
3399 2018-04-20 Martin Sebor <msebor@redhat.com>
3402 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
3404 (gimple_fold_builtin_stxcpy_chk): Same.
3405 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
3407 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
3410 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
3411 __powikf2 when long double is IEEE 128-bit.
3413 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3415 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
3416 step to make sure stack always aligned.
3418 2018-04-20 Carl Love <cel@us.ibm.com>
3421 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
3422 size check for arg0.
3424 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
3425 Tom de Vries <tom@codesourcery.com>
3428 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
3429 Emit insns for calls too.
3430 (nvptx_find_par): Always look for worker-level predecessor insn.
3431 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
3433 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
3434 (nvptx_process_pars): Propagate frames for calls.
3436 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
3439 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
3441 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
3442 (ix86_handle_option): Don't handle OPT_mibt.
3443 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
3445 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
3447 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
3448 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
3449 (ix86_target_macros): Define __CET__ with flag_cf_protection
3450 for -fcf-protection.
3451 * config/i386/i386.c (isa2_opts): Remove -mibt.
3452 * config/i386/i386.h (TARGET_IBT): Removed.
3453 (TARGET_IBT_P): Likewise.
3454 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
3455 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
3456 * config/i386/i386.opt (mcet): Update help message.
3459 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
3460 -mcet as an alias for -mshstk.
3462 2018-04-20 Richard Biener <rguenther@suse.de>
3465 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
3466 complexity by forcing a single use of the multiply operand.
3468 2018-04-20 Martin Jambor <mjambor@suse.cz>
3471 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
3472 recursion dependency to only apply to non-clones.
3474 2018-04-20 Martin Jambor <mjambor@suse.cz>
3477 * ipa-cp.c (create_specialized_node): Check that clones of
3478 self-recursive edges exist during IPA-CP.
3480 2018-04-19 Toon Moene <toon@moene.org>
3482 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
3485 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3487 PR tree-optimization/85467
3488 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
3489 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
3490 VECTOR_CST element to type.
3492 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3495 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
3496 * config/i386/i386.md (builtin_setjmp_setup): Removed.
3497 (builtin_longjmp): Likewise.
3498 (save_stack_nonlocal): New pattern.
3499 (restore_stack_nonlocal): Likewise.
3501 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3504 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3505 Replace ASM_OUTPUT_LABEL with fprintf.
3507 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3510 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3511 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
3512 * config/i386/i386-c.c (ix86_target_macros_internal): Also
3513 define __IBT__ and __SHSTK__ for -fcf-protection.
3514 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
3516 (ix86_trampoline_init): Likewise.
3517 (x86_output_mi_thunk): Likewise.
3518 (ix86_notrack_prefixed_insn_p): Likewise.
3519 (ix86_option_override_internal): Don't disallow -fcf-protection.
3520 * config/i386/i386.md (rdssp<mode>): Also enable for
3522 (incssp<mode>): Likewise.
3523 (nop_endbr): Likewise.
3524 * config/i386/i386.opt (mcet): Change help message to built-in
3528 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
3529 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
3530 enable CET built-in functions.
3532 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
3534 * common/config/i386/i386-common.c
3535 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
3536 OPTION_MASK_ISA_MOVDIRI_UNSET,
3537 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
3538 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
3539 * config.gcc (movdirintrin.h): New header.
3540 * config/i386/cpuid.h (bit_MOVDIRI,
3541 bit_MOVDIR64B): New bits.
3542 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
3544 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
3545 (VOID, PVOID, PCVOID)): New function types.
3546 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
3547 __builtin_ia32_directstoreu_u64,
3548 __builtin_ia32_movdir64b): New builtins.
3549 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
3550 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
3552 (ix86_valid_target_attribute_inner_p): Ditto.
3553 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
3554 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
3555 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
3556 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
3557 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
3558 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
3559 (movdiri<mode>, movdir64b_<mode>): New.
3560 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
3561 * config/i386/immintrin.h: Include movdirintrin.h.
3562 * config/i386/movdirintrin.h: New file.
3563 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
3565 2018-04-19 Richard Biener <rguenther@suse.de>
3568 * cfg.c (clear_bb_flags): When loop state says we have
3569 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
3571 2018-04-19 Richard Biener <rguenther@suse.de>
3573 PR tree-optimization/84737
3574 * tree-vect-data-refs.c (vect_copy_ref_info): New function
3575 copying restrict info.
3576 (vect_setup_realignment): Use it.
3577 * tree-vectorizer.h (vect_copy_ref_info): Declare.
3578 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
3579 the first DR to all generated stores.
3580 (vectorizable_load): Likewise for loads.
3582 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3584 PR tree-optimization/85446
3585 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
3586 the integral and pointer types to have the same precision.
3588 * doc/install.texi: Document --disable-cet being the default and
3591 2018-04-18 Martin Liska <mliska@suse.cz>
3593 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
3596 2018-04-18 Martin Liska <mliska@suse.cz>
3599 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3602 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3603 arguments if they are comparable.
3605 2018-04-18 Martin Liska <mliska@suse.cz>
3608 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3611 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3614 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
3617 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
3618 ENDBR after calling __morestack.
3620 2018-04-18 David Malcolm <dmalcolm@redhat.com>
3623 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
3624 by using gcc_base_ver to generate a gcc_driver_version, and use
3625 it when generating GCC_DRIVER_NAME.
3626 * configure: Regenerate.
3628 2018-04-18 Jakub Jelinek <jakub@redhat.com>
3631 * config.gcc: Obsolete powerpc*-*-*spe*.
3633 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3636 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
3637 (stabstr_D): Change type of unum from unsigned int to
3638 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
3641 2018-04-17 Jim Wilson <jimw@sifive.com>
3644 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
3645 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
3646 Set arg_pointer_offset after using pretend_args_size.
3648 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3650 PR rtl-optimization/85431
3651 * dse.c (record_store): Ignore zero width stores.
3654 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
3655 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
3656 __builtin_stack_restore rather than after it.
3657 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
3658 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
3659 argument instead of virtual_dynamic_stack_rtx.
3661 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
3663 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
3665 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
3666 Add note to error message to explain internal mapping of overloaded
3667 built-in function name to non-overloaded built-in function name.
3668 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
3671 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3674 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
3675 where the inputs overlap with the output.
3677 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3679 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
3680 (=v, v) alternative and explicit "memory" attribute.
3681 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
3682 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3684 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
3685 "sselog1" type instead of "sselog".
3686 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
3687 "sselog". Remove explicit "memory" attribute.
3688 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
3689 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3691 (vec_extract_hi_v32hi): Merge all alternatives into one, use
3692 "sselog1" type instead of "sselog". Remove explicit "memory"
3694 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
3695 use "sselog1" type instead of "sselog". Remove explicit "memory"
3697 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
3698 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3700 (vec_extract_hi_v64qi): Merge all alternatives into one, use
3701 "sselog1" type instead of "sselog". Remove explicit "memory"
3703 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
3704 use "sselog1" type instead of "sselog". Remove explicit "memory"
3708 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
3711 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
3714 2018-04-17 Martin Jambor <mjambor@suse.cz>
3717 * ipa-cp.c (create_specialized_node): Call
3718 expand_all_artificial_thunks if necessary.
3720 2018-04-17 Martin Liska <mliska@suse.cz>
3723 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
3724 in message, remote space in between '_G' and '('.
3726 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3729 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
3730 avx512f_vmcmp<mode>3<round_saeonly_name>,
3731 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
3732 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
3733 avx512f_rndscale<mode><round_saeonly_name>,
3734 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
3735 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
3736 Use %<iptr>2 instead of %2 for -masm=intel.
3737 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
3738 avx512f_vcvttss2usi<round_saeonly_name>,
3739 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
3741 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
3742 avx512f_vcvttsd2usi<round_saeonly_name>,
3743 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
3744 Use %q1 instead of %1 for -masm=intel.
3745 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
3746 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
3747 of %3 for -masm=intel.
3748 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
3750 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
3752 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
3754 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
3755 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
3757 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
3759 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
3760 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
3761 %g1 and one with %0 and %1.
3762 (avx512er_vmrcp28<mode><round_saeonly_name>,
3763 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
3765 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
3766 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
3767 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
3768 of %0 and %{%4%} for -masm=intel.
3769 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
3770 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
3771 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
3772 order of %0 and %{%5%}%{z%} for -masm=intel.
3774 2018-04-17 Jan Hubicka <jh@suse.cz>
3777 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
3779 2018-04-17 Martin Liska <mliska@suse.cz>
3782 * multiple_target.c (create_dispatcher_calls): Set apostrophes
3783 for target_clone error message. Make default implementation
3784 clone to be a local declaration.
3785 (separate_attrs): Add new argument and check for an empty
3787 (expand_target_clones): Handle it.
3788 (ipa_target_clone): Make redirection just for target_clones
3791 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
3792 Tom de Vries <tom@codesourcery.com>
3795 * omp-expand.c (expand_oacc_for): Add dummy false branch for
3796 tiled basic blocks without omp continue statements.
3798 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
3801 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
3802 vec_extract expression as having side effects to make sure it gets
3805 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
3808 * config/i386/i386.c (get_builtin_code_for_version): Check
3811 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
3814 * gcc/config.gcc: Support "skylake".
3815 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
3817 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
3818 (processor_target_table): Add "skylake".
3819 (ix86_option_override_internal): Add "skylake".
3820 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
3821 PROCESSOR_CANNONLAKE.
3822 (get_builtin_code_for_version): Fix priority for
3823 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
3824 PROCESSOR_SKYLAKE-AVX512.
3825 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
3826 (processor_type): Add PROCESSOR_SKYLAKE.
3828 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
3829 Jason Merrill <jason@redhat.com>
3832 * convert.c (convert_to_integer_1): Use direct recursion for
3833 enumeral types and types with a precision less than the number
3834 of bits in their mode.
3836 2018-04-16 Julia Koval <julia.koval@intel.com>
3839 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
3840 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
3842 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
3845 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
3846 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
3847 and -mno-direct-move.
3849 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
3852 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
3853 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
3854 Ensure negative shifts result in {0}.
3856 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
3858 PR rtl-optimization/79916
3859 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
3860 regs (if any) to define how to gnerate SD moves when LRA is in
3863 2018-04-13 Jakub Jelinek <jakub@redhat.com>
3865 PR rtl-optimization/85393
3866 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
3867 * except.c (expand_dw2_landing_pad_for_region): Make static.
3868 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
3869 a label and unconditional jump to old_bb, rather than
3870 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
3873 PR rtl-optimization/85376
3874 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
3875 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
3876 instead of a specific value.
3878 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3879 Bin Cheng <bin.cheng@arm.com>
3881 PR tree-optimization/82965
3882 PR tree-optimization/83991
3883 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
3884 by_profile_only parameter.
3885 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
3886 information if the loop was predicted to iterate too many times.
3887 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
3889 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3892 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
3895 2018-04-13 Martin Liska <mliska@suse.cz>
3896 Jakub Jelinek <jakub@redhat.com>
3899 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
3900 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
3901 * builtins.c (expand_builtin_memory_copy_args): Use
3902 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
3903 handle dest_addr == pc_rtx.
3905 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
3908 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
3909 asked to not generate direct moves.
3910 (fix_trunc<mode>si2_stfiwx): Similar.
3911 (fix_trunc<mode>si2_internal): Similar.
3913 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3916 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
3917 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
3918 lookup if dest in some wider mode is known to be const0_rtx and
3919 if so, record permanent equivalence for it to be ZERO_EXTEND of
3920 the narrower mode destination.
3922 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3924 * lto-streamer-out.c (output_function): Revert 259346.
3925 * omp-expand.c (expand_oacc_for): Likewise.
3927 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
3929 PR rtl-optimization/85354
3930 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
3931 * sel-sched.c (sel_global_init): ... here.
3933 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
3936 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
3937 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
3938 mode for PE-COFF targets.
3939 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
3940 (i386_pe_asm_lto_end): Likewise.
3941 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
3942 (TARGET_ASM_LTO_END): Likewise.
3943 * config/i386/winnt.c (saved_debug_info_level): New static variable.
3944 (i386_pe_asm_lto_start): New function.
3945 (i386_pe_asm_lto_end): Likewise.
3947 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3948 Richard Biener <rguenther@suse.de>
3951 * lto-streamer-out.c (output_function): Fix CFG loop state before
3953 * omp-expand.c (expand_oacc_for): Handle calls to internal
3954 functions like regular functions.
3956 2018-04-12 Richard Biener <rguenther@suse.de>
3959 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
3960 for the early LTO debug to properly generate references to it
3961 during DIE emission. Do not re-use that for the skeleton for
3963 (dwarf2out_early_finish): Likewise.
3965 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3968 * config/i386/sse.md
3969 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
3970 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
3971 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
3972 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
3973 and output is a reg, avoid creating invalid lowpart subreg, but
3974 instead split into a 512-bit move. Don't split if not AVX512VL,
3975 input is xmm16+ reg and output is a mem.
3976 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
3977 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
3978 xmm16+ reg and output is a mem.
3980 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3982 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
3983 also for flag_dwarf2_cfi_asm.
3985 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3987 PR rtl-optimization/85342
3988 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
3989 a bool scalar var inside of the loop instead. Don't try to update
3990 recog_data.operand after failed apply_change_group.
3992 2018-04-12 Tom de Vries <tom@codesourcery.com>
3995 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
3996 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
3997 array with flexible array member as array without given dimension.
3998 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
3999 argument for undefined param to true.
4001 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
4004 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4005 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
4006 from PowerPC section.
4007 * config/rs6000/sysv4.opt (mcall-): Improve help text.
4008 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
4009 help text that is too long.
4010 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
4011 help text that is too long.
4012 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
4013 help text that is too long.
4015 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
4017 * config/alpha/alpha.md (stack_probe_internal): Rename
4018 from "probe_stack". Update all callers.
4020 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4022 PR rtl-optimization/84566
4023 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
4024 sched_macro_fuse_insns.
4026 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4029 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
4030 (compute_block_dependences): ... from here.
4032 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4034 PR tree-optimization/85331
4035 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
4036 from int to HOST_WIDE_INT.
4038 2018-04-11 Martin Jambor <mjambor@suse.cz>
4041 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
4042 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
4043 not the same as the source val.
4044 (cgraph_edge_brings_value_p): New parameter.
4045 (gather_edges_for_value): Pass destination value to
4046 cgraph_edge_brings_value_p.
4047 (perhaps_add_new_callers): Likewise.
4048 (get_info_about_necessary_edges): Likewise and exclude values brought
4049 only by self-recursive edges.
4050 (create_specialized_node): Redirect only clones of self-calling edges.
4051 (+self_recursive_pass_through_p): New function.
4052 (find_more_scalar_values_for_callers_subset): Use it.
4053 (find_aggregate_values_for_callers_subset): Likewise.
4054 (known_aggs_to_agg_replacement_list): Removed.
4055 (decide_whether_version_node): Re-calculate known constants for all
4056 remaining context clones.
4058 2018-04-11 Richard Biener <rguenther@suse.de>
4061 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
4062 from early DWARF output.
4063 (dwarf2out_early_finish): Output line info unconditionally into
4064 early DWARF and add reference to it.
4066 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4069 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
4070 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
4071 other than V2DFmode using iptr mode attribute.
4072 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
4074 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4076 PR rtl-optimization/84659
4077 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
4079 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4082 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
4084 (output_loc_list): Pass address of a dummy size variable even in the
4085 locview handling loop.
4086 (index_location_lists): Add comment on why skip_loc_list_entry can't
4089 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
4092 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
4095 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
4098 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4099 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
4100 and -mstring-compare-inline-limit.
4102 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4105 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
4106 for stack clash protection in a register whenever we need it to be in
4109 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4111 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
4112 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
4114 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4117 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
4119 (mlong-double-): Ditto.
4120 * config/rs6000/sysv4.opt (msdata=): Ditto.
4121 (mtls-size=): Ditto.
4123 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4125 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4126 erroneous entries for
4127 "vector int vec_ldl (int, long int *)", and
4128 "vector unsigned int vec_ldl (int, unsigned long int *)".
4129 Add comments and entries for
4130 "vector bool char vec_ldl (int, bool char *)",
4131 "vector bool short vec_ldl (int, bool short *)",
4132 "vector bool int vec_ldl (int, bool int *)",
4133 "vector bool long long vec_ldl (int, bool long long *)",
4134 "vector pixel vec_ldl (int, pixel *)",
4135 "vector long long vec_ldl (int, long long *)",
4136 "vector unsigned long long vec_ldl (int, unsigned long long *)".
4137 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
4138 type tree bool_long_long_type_node and correct definition of
4139 bool_V2DI_type_node to make reference to this new type tree.
4140 (rs6000_mangle_type): Replace erroneous reference to
4141 bool_long_type_node with bool_long_long_type_node.
4142 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
4143 comments to emphasize sign distinctions for char and int types and
4144 replace RS6000_BTI_bool_long constant with
4145 RS6000_BTI_bool_long_long constant. Also add comment to restrict
4146 use of RS6000_BTI_pixel.
4147 (bool_long_type_node): Remove this macro definition.
4148 (bool_long_long_type_node): New macro definition
4150 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4152 PR rtl-optimization/85300
4153 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
4154 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
4155 simplify_unary_operation fails.
4157 2018-04-10 Martin Liska <mliska@suse.cz>
4159 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
4160 cgraph_edge and ipa_ref.
4162 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4166 * config/i386/sse.md
4167 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
4168 computation of the VEC_MERGE selector from mask.
4169 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
4170 Fix decoding of the VEC_MERGE selector into mask.
4172 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4174 PR tree-optimization/85286
4175 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
4177 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4179 * final.c (final_1): Set insn_last_address as well as
4180 insn_current_address.
4182 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4185 * explow.c (emit_stack_probe): Call validize_mem on memory location
4186 before passing it to gen_probe_stack. Create address operand and
4187 legitimize it for the probe_stack_address case.
4189 2018-04-09 Jan Hubicka <jh@suse.cz>
4192 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
4193 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
4194 * tree.c (free_lang_data_in_type): Fix handling of binfos;
4196 (free_lang_data): Rebuild type inheritance graph.
4198 2018-04-09 Martin Sebor <msebor@redhat.com>
4200 * invoke.texi (-finline-small-functions): Mention other optimization
4202 (-findirect-inlining, -fpartial-inlining): Same.
4203 (-finline-functions-called-once): Same.
4204 (-freorder-blocks-and-partition): Same.
4206 2018-04-09 Jan Hubicka <jh@suse.cz>
4209 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
4210 jumps; choose last target that matches the criteria (i.e.
4211 no partition changes for non-crossing jumps).
4212 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
4213 support for redirecting crossing jumps to non-crossing.
4215 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
4217 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
4218 also for naked functions.
4220 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
4222 * config/arc/arc.md (add_shift): New pattern.
4223 (add_shift2): Likewise.
4224 (sub_shift): Likewise.
4225 (sub_shift_cmp0_noout): Likewise.
4226 (compare_si_ashiftsi): Likewise.
4227 (xbfu_cmp0_noout): New combine pattern.
4228 (xbfu_cmp0"): Likewise.
4229 (movsi_set_cc_insn): Place the predicable variant first.
4230 (commutative_binary_cmp0_noout): Remove clobber.
4231 (commutative_binary_cmp0): New pattern.
4232 (noncommutative_binary_cmp0): Likewise.
4233 (noncommutative_binary_cmp0_noout): Likewise.
4234 (noncommutative_binary_comparison_result_used): Removed.
4235 (rsub_cmp0): New pattern.
4236 (rsub_cmp0_noout): Likewise.
4237 (extzvsi): Changed, keep only meaningful variants.
4238 (SQH, SEZ): New iterators.
4239 (SQH_postfix): New mode attribute.
4240 (SEZ_prefix): New code attribute.
4241 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
4242 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
4243 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
4245 (noncommutative_operator): Check the availability of barrel
4248 2018-04-09 Richard Biener <rguenther@suse.de>
4250 PR tree-optimization/85284
4251 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
4252 Only use the niter constraining form of simple_iv when the exit
4255 2018-04-09 Tom de Vries <tom@codesourcery.com>
4258 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4259 (define_expand "*memory_barrier"): New define_expand.
4260 (define_insn "memory_barrier"): New insn.
4262 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4264 PR rtl-optimization/80463
4265 PR rtl-optimization/83972
4266 PR rtl-optimization/83480
4268 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
4269 correct producer for the insn.
4270 (tidy_control_flow): Fixup seqnos in case of debug insns.
4272 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4274 PR rtl-optimization/83913
4276 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
4277 different sched-times when merging exprs.
4279 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4281 PR rtl-optimization/83962
4283 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
4284 tidy_fallthru_edge and tidy_control_flow.
4286 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4288 PR rtl-optimization/83530
4290 * sel-sched.c (force_next_insn): New global variable.
4291 (remove_insn_for_debug): When force_next_insn is true, also leave only
4292 next insn in the ready list.
4293 (sel_sched_region): When the region wasn't scheduled, make another pass
4294 over it with force_next_insn set to 1.
4296 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4298 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
4300 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4301 for interrupt control.
4302 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4303 functions for interrupt control.
4304 * config/nds32/nds32-intrinsic.md: Likewise.
4305 * config/nds32/nds32_intrinsic.h: Likewise.
4306 * config/nds32/nds32.h (nds32_builtins): Likewise.
4308 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4310 * config/nds32/nds32.c (nds32_init_machine_status,
4311 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
4312 strict_aligned_p field.
4313 (nds32_expand_to_rtl_hook): New function.
4314 (TARGET_EXPAND_TO_RTL_HOOK): Define.
4315 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
4317 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4318 Chung-Ju Wu <jasonwucj@gmail.com>
4320 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
4321 * config/nds32/nds32-n7.md: New file.
4322 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
4323 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
4325 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
4326 * config/nds32/nds32.md (pipeline_model): Add n7.
4327 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
4328 * config/nds32/pipelines.md: Include n7 settings.
4330 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4331 Chung-Ju Wu <jasonwucj@gmail.com>
4333 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
4334 * config/nds32/nds32-e8.md: New file.
4335 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
4336 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
4338 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
4339 * config/nds32/nds32.md (pipeline_model): Add e8.
4340 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
4341 * config/nds32/pipelines.md: Include e8 settings.
4343 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4344 Chung-Ju Wu <jasonwucj@gmail.com>
4346 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
4347 * config/nds32/nds32-n8.md: New file.
4348 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
4349 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
4351 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
4352 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
4353 * config/nds32/nds32.md (pipeline_model): Add n8.
4354 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
4355 * config/nds32/pipelines.md: Include n8 settings.
4357 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4358 Chung-Ju Wu <jasonwucj@gmail.com>
4360 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
4361 * config/nds32/nds32-n9-2r1w.md: New file.
4362 * config/nds32/nds32-n9-3r2w.md: New file.
4363 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
4364 nds32_register_ports): New or modify for cpu n9.
4365 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
4367 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
4368 * config/nds32/nds32-utils.c: New file.
4369 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
4370 TARGET_MUL_SLOW): Define.
4371 * config/nds32/nds32.md (pipeline_model): New attribute.
4372 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
4373 New options that support cpu n9.
4374 * config/nds32/pipelines.md: Include n9 settings.
4375 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
4377 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4379 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
4380 information if necessary.
4381 (output_cond_branch_compare_zero): Likewise.
4382 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
4383 (nds32_target_alignment): Refine for alignment.
4384 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
4385 (FUNCTION_BOUNDARY): Modify.
4386 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
4388 * config/nds32/nds32.opt (malways-align, malign-functions): New.
4390 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4392 * config/nds32/constants.md (unspec_volatile_element): Add values for
4393 TLB operation and data prefetch.
4394 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4395 functions for TLB operation and data prefetch.
4396 * config/nds32/nds32-intrinsic.md: Likewise.
4397 * config/nds32/nds32_intrinsic.h: Likewise.
4398 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
4399 (nds32_print_operand): Likewise.
4400 * config/nds32/nds32.h (nds32_builtins): Likewise.
4402 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
4403 Andrew Pinski <pinsika@gcc.gnu.org>
4406 * match.pd: Use constant_boolean_node of correct type instead of
4407 boolean_true_node or boolean_false_node for simplifying
4408 pointer comparisons to zero.
4410 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4412 PR tree-optimization/80021
4413 * tree.c (verify_type_variant): Make error call in verify_variant_match
4414 translatable and remove final full stop.
4416 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4418 * config/nds32/constants.md (unspec_volatile_element): Add
4419 UNSPEC_VOLATILE_EH_RETURN.
4420 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
4421 nds32_output_stack_pop): Support dwarf exception handling process.
4422 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
4423 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
4424 exception handling process.
4425 (nds32_compute_stack_frame): Likewise.
4426 (nds32_return_addr_rtx): Likewise.
4427 (nds32_initial_elimination_offset): Likewise.
4428 (nds32_expand_prologue): Likewise.
4429 (nds32_expand_epilogue): Likewise.
4430 (nds32_dynamic_chain_address): New function.
4431 * config/nds32/nds32.h (machine_function): Add fields for dwarf
4433 (DYNAMIC_CHAIN_ADDRESS): Define.
4434 (EH_RETURN_DATA_REGNO): Define.
4435 (EH_RETURN_STACKADJ_RTX): Define.
4436 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
4437 patterns for dwarf exception handling.
4439 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4441 * config/nds32/nds32.h: Clean up obsolete macros.
4443 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4445 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4446 Add enum values for particular instructions.
4447 * config/nds32/nds32-intrinsic.c: Implementation of expanding
4448 particular intrinsic functions.
4449 * config/nds32/nds32-intrinsic.md: Likewise.
4450 * config/nds32/nds32_intrinsic.h: Likewise.
4451 * config/nds32/nds32.h (nds32_builtins): Likewise.
4452 * config/nds32/nds32.md (type): Add pbsad and pbsada.
4453 (btst, ave): New patterns for particular instructions.
4455 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4457 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4458 Add enum values for atomic load/store and memory sync.
4459 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
4461 * config/nds32/nds32-intrinsic.md: Likewise.
4462 * config/nds32/nds32_intrinsic.h: Likewise.
4463 * config/nds32/nds32.h (nds32_builtins): Likewise.
4465 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4467 PR tree-optimization/85257
4468 * fold-const.c (native_encode_vector): If not all elts could fit
4469 and off is -1, return 0 rather than offset.
4470 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
4471 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
4472 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
4473 adjust buffer in native_interpret_expr call.
4475 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4477 * config/nds32/constants.md (unspec_volatile_element): Add cache
4478 control enum values.
4479 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
4480 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
4481 * config/nds32/nds32.c (nds32_cctl_names): New.
4482 (nds32_print_operand): Handle cache control register names.
4483 * config/nds32/nds32.h (nds32_builtins): New enum values.
4484 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
4486 * config/nds32/nds32.md (type): Add mmu.
4487 * config/nds32/pipelines.md (simple_insn): Add mmu.
4489 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4491 * config/nds32/nds32.md (type): Remove call.
4492 * config/nds32/pipelines.md (simple_insn): Likewise.
4494 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4496 * config/nds32/constants.md (unspec_volatile_element): Add
4497 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
4498 UNSPEC_VOLATILE_FMFCFG.
4499 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
4500 description for fmfcfg and fmfcsr.
4501 (bdesc_1arg): Add fmtcsr.
4502 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
4503 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
4504 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
4505 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
4506 unspec_fmfcfg): New patterns.
4507 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
4508 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
4509 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
4510 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
4511 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
4512 __nds32__fmfcfg): Define.
4514 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4516 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
4517 intrinsic register names.
4518 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
4519 intrinsic register enum values and macros.
4521 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4523 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
4524 for load/store addressing form.
4525 (nds32_print_operand_address): Likewise.
4527 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
4530 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
4531 based on LABEL_REF. Remove useless assertion.
4532 (pic_address_needs_scratch): Fix formatting.
4533 (sparc_legitimize_pic_address): Minor tweaks.
4534 (sparc_delegitimize_address): Adjust assertion accordingly.
4535 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
4536 into symbolic_operand.
4537 (movsi_high_pic_label_ref): Likewise.
4538 (movsi_lo_sum_pic_label_ref): Likewise.
4539 (movdi_pic_label_ref): Likewise.
4540 (movdi_high_pic_label_ref): Likewise.
4541 (movdi_lo_sum_pic_label_ref): Likewise.
4543 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
4545 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
4546 custom LIB_SPEC setup.
4548 2018-04-06 Ruslan Bukin <br@bsdpad.com>
4549 Kito Cheng <kito.cheng@gmail.com>
4551 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
4552 * config/riscv/freebsd.h: New.
4554 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4556 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
4557 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
4560 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4561 Kito Cheng <kito.cheng@gmail.com>
4563 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
4564 nds32_output_call, nds32_symbol_binds_local_p): New functions.
4565 * config/nds32/nds32-protos.h (nds32_output_call,
4566 nds32_output_return): Declare.
4567 * config/nds32/nds32.md: Refine all the call and return patterns.
4569 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4572 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
4573 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
4575 PR rtl-optimization/84872
4576 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
4577 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
4580 2018-04-06 Tamar Christina <tamar.christina@arm.com>
4582 * expr.c (copy_blkmode_to_reg): Revert 254862.
4583 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
4585 2018-04-06 Richard Biener <rguenther@suse.de>
4588 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
4589 after seeing a component reference with an adjacent field. Treat
4590 refs to arrays at struct end of external decls similar to
4591 refs to unconstrained commons.
4593 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4596 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
4597 look through SAVE_EXPRs with non-side-effects argument. Adjust
4599 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
4602 2018-04-06 Richard Biener <rguenther@suse.de>
4605 * alias.c (find_base_term): New wrapper around find_base_term
4606 unwinding CSELIB_VAL_PTR changes.
4607 (find_base_term): Do not restore CSELIB_VAL_PTR during the
4610 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4612 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
4614 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
4615 constant definitions.
4616 ("nop"): lr 0,0 -> nopr r0
4617 ("nop_lr0", "nop_lr1"): New insn definitions.
4619 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4621 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
4622 NDS32_V3PUSH_AVAILABLE_P macro.
4624 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
4625 Chung-Ju Wu <jasonwucj@gmail.com>
4627 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
4628 (nds32*-*-*): Add float and fpu_config into supported_defaults.
4629 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
4630 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
4631 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
4632 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
4633 * config/nds32/constraints.md: New constraints and checking for hard
4634 float configuration.
4635 * config/nds32/iterators.md: New mode iterator and attribute for hard
4636 float configuration.
4637 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
4639 * config/nds32/nds32-fpu.md: New file.
4640 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
4641 deal with hard float code generation.
4642 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
4644 (abi_type, float_reg_number): New enum type.
4645 * config/nds32/nds32-predicates.c: New predicates for hard float.
4646 * config/nds32/nds32-protos.h: Declare functions for hard float.
4647 * config/nds32/nds32.c: Implementation for hard float configuration.
4648 * config/nds32/nds32.h: Definitions for hard float configuration.
4649 * config/nds32/nds32.md: Include hard float machine description and
4650 modify patterns for hard float configuration.
4651 * config/nds32/nds32.opt: New options for hard float configuration.
4652 * config/nds32/predicates.md: New predicates for hard float
4655 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4657 * common/config/nds32/nds32-common.c
4658 (nds32_option_optimization_table): Enable -mreleax-hint by default.
4660 2018-04-05 Jakub Jelinek <jakub@redhat.com>
4663 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
4664 CONSTRUCTOR_ELT (ctor, ...)->value.
4666 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
4669 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
4671 2018-04-05 Tom de Vries <tom@codesourcery.com>
4674 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
4677 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4678 Kito Cheng <kito.cheng@gmail.com>
4680 * config/nds32/constraints.md (U33): Fine-tune checking condition.
4681 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
4682 * config/nds32/nds32.h (nds32_16bit_address_type): Add
4683 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
4685 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4686 Kito Cheng <kito.cheng@gmail.com>
4688 * config/nds32/constraints.md (Ufe): New memory constraint.
4689 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
4690 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
4691 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
4693 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
4694 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
4696 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4698 * config/nds32/nds32.md: Use optimize_size in the condition for
4699 alu-shift instructions.
4701 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4703 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
4705 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4707 * config/nds32/nds32.md (negsi2): Refine pattern.
4709 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4710 Chung-Ju Wu <jasonwucj@gmail.com>
4712 * config/nds32/iterators.md (shift_rotate): New code iterator.
4713 (shift): New code attribute.
4714 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
4715 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
4716 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
4717 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
4718 bit-wise operations.
4719 (andsi3, *andsi3): Ditto.
4720 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
4721 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
4722 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
4723 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
4724 nds32_ior_operand, nds32_xor_operand): New predicates.
4726 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4728 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
4729 (addsi3, subsi3): ... this.
4731 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4733 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
4735 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4737 * config/nds32/nds32.md: Adjust indention.
4739 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4741 * config/nds32/nds32.md (feature): New attribute.
4743 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4745 * config/nds32/nds32.md (subtype): New attribute.
4747 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
4750 * config/arm/arm-builtins.c (arm_expand_builtin): Change
4751 expansion to perform a bitwise AND of the argument followed by a
4752 boolean negation of the result.
4754 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
4756 PR rtl-optimization/84878
4757 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
4758 the basic block. Assert the use reference is not artificial and that
4759 it has an associated insn.
4761 2018-04-04 Michael Matz <matz@suse.de>
4763 * builtins.c (compute_objsize): Pass correct operand
4764 to array_at_struct_end_p.
4766 2018-04-04 Richard Biener <rguenther@suse.de>
4769 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
4770 from contexts for DINFO_LEVEL_TERSE and below.
4772 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4774 * config/nds32/nds32-doubleword.md (move_<mode>): Require
4775 resiter_operand condition.
4776 * config/nds32/nds32.md (*move<mode>): Ditto.
4778 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4779 Monk Chiang <sh.chiang04@gmail.com>
4781 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
4783 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4785 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
4787 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4788 Kito Cheng <kito.cheng@gmail.com>
4790 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
4791 nds32_cond_code_str, output_cond_branch,
4792 output_cond_branch_compare_zero, nds32_expand_cbranch,
4793 nds32_expand_cstore, nds32_expand_movcc,
4794 nds32_output_cbranchsi4_equality_zero,
4795 nds32_output_cbranchsi4_equality_reg,
4796 nds32_output_cbranchsi4_equality_reg_or_const_int,
4797 nds32_output_cbranchsi4_greater_less_zero: New functions.
4798 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
4799 nds32_expand_cstore, nds32_expand_movcc,
4800 nds32_output_cbranchsi4_equality_zero,
4801 nds32_output_cbranchsi4_equality_reg,
4802 nds32_output_cbranchsi4_equality_reg_or_const_int,
4803 nds32_output_cbranchsi4_greater_less_zero): Declare.
4804 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
4805 nds32_rimm11s_operand): New predicates.
4806 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
4807 * config/nds32/nds32.md: Rewrite all the branch and conditional move
4810 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4812 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
4813 * config/nds32/nds32.md: Ditto.
4814 * config/nds32/pipelines.md: Ditto.
4816 2018-04-04 Richard Biener <rguenther@suse.de>
4818 PR tree-optimization/85168
4819 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
4820 propagating abnormals.
4822 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4824 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
4826 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4827 Kito Cheng <kito.cheng@gmail.com>
4829 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
4830 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
4831 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
4832 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
4833 * config/nds32/nds32.md (sibcall_internal): New.
4834 (sibcall_register): Remove.
4835 (sibcall_immediate): Remove.
4836 (sibcall_value_internal): New.
4837 (sibcall_value_register): Remove.
4838 (sibcall_value_immediate): Remove.
4839 * config/nds32/predicates.md (nds32_general_register_operand): New.
4840 (nds32_call_address_operand): New.
4842 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4844 PR rtl-optimization/85167
4845 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
4846 bb_defs if *split_p, instead preinitialize it to NULL.
4848 PR tree-optimization/85156
4849 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
4850 evaluating the argument multiple times.
4852 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
4854 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
4856 (_mm_cvtpd_ps): Likewise.
4857 (_mm_cvttpd_epi32): Likewise.
4858 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
4859 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
4860 vector, pixel, and bool following altivec.h include.
4862 2018-04-03 Martin Sebor <msebor@redhat.com>
4864 * doc/extend.texi (Common Function Attributes): Clarify.
4865 (const attribute): Likewise.
4866 (pure attribute): Likewise.
4868 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4871 * config/i386/i386.c (ix86_expand_vector_set): Use
4872 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
4874 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
4876 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
4877 instructions when changing rounding bits to preserve precision bits
4878 in the x87 control word.
4880 2018-04-03 Martin Liska <mliska@suse.cz>
4882 PR tree-optimization/82491
4883 * rtl.h (strip_offset_and_add): Replace += suboffset with
4884 poly_uint64 () + suboffset.
4886 2018-03-29 Martin Liska <mliska@suse.cz>
4887 Martin Jambor <mjambor@suse.cz>
4890 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
4891 param_type is not an integral or pointer type.
4893 2018-04-03 Richard Biener <rguenther@suse.de>
4895 * sese.h (recompute_all_dominators): Remove.
4897 2018-04-02 Martin Sebor <msebor@redhat.com>
4899 * doc/invoke.texi (-Wrestrict): Fix typos.
4901 2018-04-02 Jim Wilson <jimw@sifive.com>
4903 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
4904 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
4905 (<optab>di3, <optab>si3_extend): Likewise.
4906 (<optab>si3_mask, <optab>si3_mask_1): New.
4907 (<optab>di3_mask, <optab>di3_mask_1): New.
4908 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
4909 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
4910 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
4912 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
4914 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
4917 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
4919 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
4920 (nds32_canonicalize_comparison): New function.
4922 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4923 Kito Cheng <kito.cheng@gmail.com>
4924 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4926 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
4927 * config/nds32/constants.md (unspec_volatile_element): Add
4928 UNSPEC_VOLATILE_RELAX_GROUP.
4929 * config/nds32/nds32-relax-opt.c: New file.
4930 * config/nds32/nds32-predicates.c
4931 (nds32_symbol_load_store_p): New function.
4932 * config/nds32/nds32-protos.h
4933 (nds32_symbol_load_store_p): Declare function.
4934 (make_pass_nds32_relax_opt): Declare new rtl pass function.
4935 * config/nds32/nds32.c
4936 (nds32_register_pass): New function to register pass.
4937 (nds32_register_passes): New function to register passes.
4938 * config/nds32/nds32.md (relax_group): New pattern.
4939 * config/nds32/nds32.opt (mrelax-hint): New option.
4940 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
4942 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
4944 * config/nds32/t-nds32: Modify files dependency.
4946 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4948 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
4949 (PROFILE_HOOK): Define its implementation.
4951 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4953 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
4954 type and 32-bit size.
4956 2018-04-01 Jakub Jelinek <jakub@redhat.com>
4959 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
4960 (V_128_256): New mode iterator.
4961 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
4962 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
4963 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
4965 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
4966 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
4968 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
4971 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
4972 NaN inputs correctly.
4974 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
4977 * config/rs6000/vsx.md (??r): New mode attribute.
4978 (*vsx_mov<mode>_64bit): Use it.
4979 (*vsx_mov<mode>_32bit): Likewise.
4981 2018-03-30 Martin Sebor <msebor@redhat.com>
4983 PR tree-optimization/84818
4984 * builtins.c (check_access): Use warning_n.
4986 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4989 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
4991 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
4994 2018-03-30 Julia Koval <julia.koval@intel.com>
4997 * x86-tune.def (movx, partial_reg_dependency): Enable for
5000 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
5003 * lra-constraints.c (process_alt_operands): Move setting
5004 this_alternative_matches below.
5006 2018-03-29 Martin Liska <mliska@suse.cz>
5009 * doc/invoke.texi: Document how LTO works with debug info.
5010 Describe auto-load support of binutils. Mention 'x86-64'
5011 as valid option value of -march option.
5013 2018-03-29 Jakub Jelinek <jakub@redhat.com>
5015 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
5018 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
5019 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
5020 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
5023 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5026 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
5027 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
5028 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
5029 for RS6000_BTM_POWERPC64.
5030 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
5031 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
5032 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
5037 2018-03-28 Carl Love <cel@us.ibm.com>
5040 2017-09-27 Carl Love <cel@us.ibm.com>
5042 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
5043 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
5044 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
5047 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5049 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
5050 instead of __vector bool.
5051 (_mm_max_pu8): Likewise.
5052 (_mm_min_pi16): Likewise.
5054 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5057 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
5058 (DIVWEUO): Likewise.
5060 (DIVDEUO): Likewise.
5061 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
5062 DIVWEUO and DIVDEUO.
5063 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
5064 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
5065 (div_extend): Likewise.
5066 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
5068 (__builtin_divweuo): Likewise.
5069 (__builtin_divdeo): Likewise.
5070 (__builtin_divdeuo): Likewise.
5072 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5075 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
5076 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
5078 PR tree-optimization/82004
5079 * gimple-match-head.c (optimize_pow_to_exp): New function.
5080 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
5081 Don't fold to exp if optimize_pow_to_exp is false.
5083 2018-03-28 Martin Liska <mliska@suse.cz>
5086 * calls.c (initialize_argument_information): Fix trailing space.
5087 * common.opt: Fix typo and provide better explanation for
5088 -fsanitize-coverage option.
5089 * config/i386/i386.opt: Fix typo.
5091 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5092 Martin Liska <mliska@suse.cz>
5095 * gimplify.c (asan_poison_variable): Don't do the check for
5096 gimplify_omp_ctxp here.
5097 (gimplify_decl_expr): Do it here.
5098 (gimplify_target_expr): Likewise.
5100 2018-03-28 Martin Liska <mliska@suse.cz>
5103 * config/i386/i386.c (ix86_function_arg_advance): Do not call
5104 chkp_type_bounds_count if MPX is not enabled.
5106 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
5108 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
5110 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
5113 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
5114 function to create the function decl for complex long double
5115 multiply and divide for -mabi=ieeelongdouble.
5116 (init_float128_ieee): Call it.
5118 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
5121 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
5122 -fcf-protection=branch -mibt.
5123 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
5125 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5128 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
5130 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
5133 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
5134 extern array declarations.
5136 2018-03-27 Richard Biener <rguenther@suse.de>
5139 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
5140 explicit single_use checks.
5142 2018-03-27 Richard Biener <rguenther@suse.de>
5144 PR tree-optimization/85082
5145 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
5148 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5150 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
5151 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
5152 Turn on fasynchronous-unwind-tables and funwind-tables.
5154 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
5157 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
5158 (*bmi_blsr_<mode>_ccz): Ditto.
5160 2018-03-26 Tom de Vries <tom@codesourcery.com>
5162 PR tree-optimization/85063
5163 * omp-general.c (offloading_function_p): New function. Factor out
5165 * omp-offload.c (pass_omp_target_link::gate): ... here.
5166 * omp-general.h (offloading_function_p): Declare.
5167 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
5168 with attribute omp declare target for offloading functions.
5170 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
5172 PR tree-optimization/84005
5173 * tree-data-ref.h (get_base_for_alignment): Declare.
5174 * tree-data-ref.c (get_base_for_alignment_1): New function.
5175 (get_base_for_alignment): Likewise.
5176 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
5177 get_base_for_alignment to find a suitable base object, instead
5178 of always using drb->base_address.
5180 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5183 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
5184 known size by default.
5186 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
5189 * lra-constraints.c (process_alt_operands): Don't match BLKmode
5190 and non BLKmode operands.
5192 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5195 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
5196 Clean up attributes.
5198 2018-03-23 Richard Biener <rguenther@suse.de>
5201 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
5202 we are going to emit early debug for LTO.
5204 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5207 * function.c (match_asm_constraints_1): Don't optimize if input
5208 doesn't satisfy general_operand predicate for output's mode.
5211 * alias.c (write_dependence_p): Don't require for x_canonicalized
5212 non-VOIDmode if x has VOIDmode.
5215 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
5216 just don't try to optimize it rather than assert it never happens.
5218 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5220 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
5221 macro expansions for definition of ST_INTERNAL_<mode> and
5222 LD_INTERNAL_<mode> builtins.
5223 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
5225 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
5227 (altivec_expand_st_builtin): Likewise.
5228 (altivec_expand_builtin): Remove calls to deleted functions.
5229 (rs6000_address_for_altivec): Delete this function.
5230 * config/rs6000/vector.md: Remove expands for
5231 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
5233 2018-03-22 Sudakshina Das <sudi.das@arm.com>
5236 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
5237 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
5238 re-computing once computed.
5239 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
5240 (arm_init_machine_status): Initialize
5241 machine->static_chain_stack_bytes.
5243 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5246 * doc/extend.texi: Add four new prototypes for vec_ld.
5247 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
5248 definitions for more logical presentation.
5249 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
5250 entries for V1TI variants of __builtin_altivec_ld builtin.
5251 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
5252 handling of V1TI variant of LVX icode pattern.
5253 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
5254 (rs6000_gimple_fold_builtin): Likewise.
5255 (altivec_init_builtins): Add code to define
5256 __builtin_altivec_lvx_v1ti function.
5258 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5261 * function.c (match_asm_constraints_1): Don't do the optimization
5262 if input isn't a REG, SUBREG, MEM or constant.
5264 2018-03-22 Tom de Vries <tom@codesourcery.com>
5266 PR tree-optimization/84956
5267 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
5268 bb_has_abnormal_pred.
5270 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5273 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
5274 DECL_INITIAL (decl) to decl at the end.
5275 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
5278 2018-03-21 Joseph Myers <joseph@codesourcery.com>
5280 * doc/extend.texi (__builtin_tgmath): Document when complex
5281 integer types are treated as _Complex _Float64.
5283 2018-03-21 Tom de Vries <tom@codesourcery.com>
5285 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
5287 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5289 PR tree-optimization/84960
5290 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
5291 if it is ENTRY block, move them into single succ of ENTRY in that case.
5293 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
5295 PR tree-optimization/84811
5296 * poly-int.h (poly_span_traits): Remove the T3 parameter and
5297 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
5298 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
5299 (known_subrange_p): Update accordingly. Cast each value involved
5300 in the size comparison, rather than casting the result of the
5303 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5305 PR tree-optimization/84982
5306 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
5307 by flipping the least significant bit rather than all bits from
5308 bitpos to bitpos + bitsize - 1.
5310 2018-03-21 Nathan Sidwell <nathan@acm.org>
5312 * doc/extend.texi (Deprecated Features): Remove mention of
5313 long-deleted deprecations.
5315 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5318 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
5319 * configure: Regenerate.
5321 2018-03-21 Tom de Vries <tom@codesourcery.com>
5323 PR tree-optimization/83126
5324 * tree-parloops.c (num_phis): New function.
5325 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
5327 2018-03-21 Nathan Sidwell <nathan@acm.org>
5329 * doc/extend.texi (Deprecated Features): Update deprecated flags,
5330 mention anon-struct/union members and trailing attributes.
5332 2018-03-21 Bin Cheng <bin.cheng@arm.com>
5334 PR tree-optimization/84969
5335 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
5336 builtin memset partitions if they set different rhs values.
5338 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5340 PR rtl-optimization/84989
5341 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
5342 VEC_DUPLICATE with scalar result mode.
5344 2018-03-21 Martin Liska <mliska@suse.cz>
5347 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
5348 not intended return statement.
5350 2018-03-21 Martin Liska <mliska@suse.cz>
5353 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
5354 (chkp_find_bound_slots_1): Limit number of iterations.
5356 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
5359 * Minor grammar fixes for x86 options.
5361 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5364 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
5365 holding REG_CFA_RESTORE notes, instead turn them into a USE.
5367 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
5370 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
5371 (altivec_lvx_<mode>_1op): Likewise.
5372 (altivec_stvx_<mode>_2op): Likewise.
5373 (altivec_stvx_<mode>_1op): Likewise.
5374 (altivec_lvx_<VM2:mode>): New define_expand.
5375 (altivec_stvx_<VM2:mode>): Likewise.
5376 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
5377 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5378 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
5379 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5380 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
5381 (rs6000_gen_lvx): Likewise.
5382 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
5383 (altivec_expand_stv_builtin): Likewise.
5384 (altivec_expand_builtin): Likewise.
5385 * config/rs6000/vector.md: Likewise.
5387 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5390 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
5393 2018-03-20 Richard Biener <rguenther@suse.de>
5396 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
5397 sign-conversions as zero, fall back to standard scalar_stmt
5400 2018-03-20 Martin Liska <mliska@suse.cz>
5403 * predict.c (rebuild_frequencies): Handle case when we have
5404 PROFILE_ABSENT, but flag_guess_branch_prob is false.
5406 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5409 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
5410 flag_section_anchors.
5411 * varasm.c (use_blocks_for_decl_p): Remove hack for
5412 dw2_force_const_mem.
5415 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
5417 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
5418 be created, use lowpart_subreg of operands[0] rather than operands[0]
5420 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
5421 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
5422 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
5423 and n constraint instead of aarch64_shift_imm_di and Usd.
5424 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
5425 (*aarch64_<optab>_reg_minus<mode>3): ... this.
5427 2018-03-20 Sudakshina Das <sudi.das@arm.com>
5430 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
5431 to favor GPR over NEON registers.
5432 (<shift>di3_neon): Likewise.
5434 2018-03-20 Tom de Vries <tom@codesourcery.com>
5437 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
5438 (nvptx_process_pars): Emit bar.sync asap and alap.
5440 2018-03-20 Tom de Vries <tom@codesourcery.com>
5443 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
5444 seen_label if seen_label is already set.
5446 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5449 * config/i386/i386.c (fold_builtin_cpu): For features above 31
5450 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
5451 Use 1U instead of 1. Formatting fixes.
5454 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
5455 instead of TREE_TYPE (s1) for the return value.
5457 2018-03-19 Jakub Jelinek <jakub@redhat.com>
5459 PR tree-optimization/84946
5460 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
5461 bitsize + bitsize in poly_uint64 rather than poly_int64.
5464 * dwarf2asm.c: Include fold-const.c.
5465 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
5466 of decl rather than decl itself.
5468 PR rtl-optimization/84643
5469 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
5471 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
5474 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
5475 calling assemble_variable.
5477 2018-03-19 Sudakshina Das <sudi.das@arm.com>
5480 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
5481 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
5483 2018-03-19 Jim Wilson <jimw@sifive.com>
5486 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
5487 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
5488 (riscv_first_stack_step): Likewise.
5489 (riscv_option_override): Use STACK_BOUNDARY instead of
5491 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
5493 (BIGGEST_ALIGNMENT): Set to 128.
5494 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
5495 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
5498 2018-03-19 Richard Biener <rguenther@suse.de>
5500 PR tree-optimization/84933
5501 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
5502 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
5504 2018-03-19 Richard Biener <rguenther@suse.de>
5506 PR tree-optimization/84859
5507 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
5508 (cond_if_else_store_replacement): Perform sinking operation on
5509 single-store BBs regardless of MAX_STORES_TO_SINK setting.
5510 Generalize what a BB with a single eligible store is.
5512 2018-03-19 Richard Biener <rguenther@suse.de>
5514 PR tree-optimization/84929
5515 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
5516 chrec_is_positive against non-chrec arg.
5518 2018-03-19 Tamar Christina <tamar.christina@arm.com>
5521 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
5523 2018-03-18 Martin Liska <mliska@suse.cz>
5525 PR rtl-optimization/84635
5526 * regrename.c (build_def_use): Use matches_mode only when
5529 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
5531 PR tree-optimization/84913
5532 * tree-vect-loop.c (vectorizable_reduction): Don't try to
5533 vectorize chains of COND_EXPRs.
5535 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5537 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
5539 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5541 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
5543 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5545 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
5547 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
5548 Kito Cheng <kito.cheng@gmail.com>
5550 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
5551 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
5552 (nds32_adjust_reg_alloc_order): New function.
5553 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
5555 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
5557 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
5558 nds32_print_operand, nds32_print_operand_address): Use
5559 HOST_WIDE_INT_PRINT_DEC instead.
5561 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
5563 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
5565 2018-03-17 Jakub Jelinek <jakub@redhat.com>
5568 * config/i386/i386.c (initial_ix86_tune_features,
5569 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
5571 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
5572 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
5573 rather than 1u << ix86_tune. Formatting fix.
5574 (ix86_option_override_internal): Change ix86_arch_mask from
5575 unsigned int to unsigned HOST_WIDE_INT, initialize to
5576 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
5577 (ix86_function_specific_restore): Likewise.
5579 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5582 * postreload.c (reload_combine_recognize_pattern): Perform
5583 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
5584 truncate_int_for_mode the result for the destination's mode.
5587 * hsa-gen.c (mem_type_for_type): Fix comment typo.
5588 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
5590 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5593 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
5596 * lra-assigns.c (lra_split_hard_reg_for): Don't use
5597 regno_allocno_class_array and sorted_pseudos.
5598 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
5599 insns where regno is used.
5601 2018-03-16 Martin Liska <mliska@suse.cz>
5604 * multiple_target.c (create_dispatcher_calls): Redirect
5605 reference in the symbol table.
5607 2018-03-16 Martin Liska <mliska@suse.cz>
5610 * multiple_target.c (create_dispatcher_calls): Redirect also
5613 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5617 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
5618 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
5619 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
5621 2018-03-16 Julia Koval <julia.koval@intel.com>
5623 * doc/invoke.texi (Skylake Server): Add CLWB.
5624 Cannonlake): Remove CLWB.
5626 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5628 PR tree-optimization/84841
5629 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
5631 (FLOAT_ONE_CONST_TYPE): Define.
5632 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
5633 (sort_by_operand_rank): Put entries with higher constant_type last
5634 rather than first to match comments.
5636 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
5638 * config/nios2/nios2.md (movsi_internal): Fix thinko in
5641 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5644 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
5645 check and use address of target always.
5647 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
5650 * config/i386/i386.c (indirect_thunk_needed): Update comments.
5651 (indirect_thunk_bnd_needed): Likewise.
5652 (indirect_thunks_used): Likewise.
5653 (indirect_thunks_bnd_used): Likewise.
5654 (indirect_return_needed): New.
5655 (indirect_return_bnd_needed): Likewise.
5656 (output_indirect_thunk_function): Add a bool argument for
5658 (output_indirect_thunk_function): Don't generate alias for
5659 function return thunk.
5660 (ix86_code_end): Call output_indirect_thunk_function to generate
5661 function return thunks.
5662 (ix86_output_function_return): Set indirect_return_bnd_needed
5663 and indirect_return_needed instead of indirect_thunk_bnd_needed
5664 and indirect_thunk_needed.
5666 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
5668 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
5669 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
5670 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
5672 2018-03-15 David Malcolm <dmalcolm@redhat.com>
5673 Paul Hua <paul.hua.gm@gmail.com>
5676 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
5678 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
5680 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
5681 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
5684 2018-03-15 Tamar Christina <tamar.christina@arm.com>
5687 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
5688 instead of GET_MODE_SIZE when comparing Units.
5690 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
5693 * varasm.c (hash_section): Return an unchangeble hash value
5694 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
5695 Return !aarch64_can_use_per_function_literal_pools_p ().
5697 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5700 * optabs.c (emit_conditional_move): Pass address of cmode's copy
5701 rather than address of cmode as last argument to prepare_cmp_insn.
5703 2018-03-15 Julia Koval <julia.koval@intel.com>
5705 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
5706 F_AVX512VNNI, F_AVX512BITALG): New.
5708 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
5711 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
5712 insn for floating-point loads and stores.
5714 2018-03-14 Carl Love <cel@us.ibm.com>
5716 * config/rs6000/rs6000-c.c: Add macro definitions for
5717 ALTIVEC_BUILTIN_VEC_PERMXOR.
5718 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
5719 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
5720 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
5721 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
5723 * config/doc/extend.texi: Add prototypes for vec_permxor.
5725 2018-03-14 David Malcolm <dmalcolm@redhat.com>
5728 * diagnostic-show-locus.c (class layout_point): Convert m_line
5729 from int to linenum_type.
5730 (line_span::comparator): Use linenum "compare" function when
5731 comparing line numbers.
5732 (test_line_span): New function.
5733 (layout_range::contains_point): Convert param "row" from int to
5735 (layout_range::intersects_line_p): Likewise.
5736 (layout::will_show_line_p): Likewise.
5737 (layout::print_source_line): Likewise.
5738 (layout::should_print_annotation_line_p): Likewise.
5739 (layout::print_annotation_line): Likewise.
5740 (layout::print_leading_fixits): Likewise.
5741 (layout::annotation_line_showed_range_p): Likewise.
5742 (struct line_corrections): Likewise for field m_row.
5743 (line_corrections::line_corrections): Likewise for param "row".
5744 (layout::print_trailing_fixits): Likewise.
5745 (layout::get_state_at_point): Likewise.
5746 (layout::get_x_bound_for_row): Likewise.
5747 (layout::print_line): Likewise.
5748 (diagnostic_show_locus): Likewise for locals "last_line" and
5750 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
5751 * input.c (selftest::test_linenum_comparisons): New function.
5752 (selftest::input_c_tests): Call it.
5753 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
5754 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
5755 * selftest.h (ASSERT_GT): New macro.
5756 (ASSERT_GT_AT): New macro.
5757 (ASSERT_LT): New macro.
5758 (ASSERT_LT_AT): New macro.
5760 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
5762 PR rtl-optimization/84780
5763 * combine.c (distribute_links): Don't make a link based on pc_rtx.
5765 2018-03-14 Martin Liska <mliska@suse.cz>
5767 * tree.c (record_node_allocation_statistics): Use
5768 get_stats_node_kind.
5769 (get_stats_node_kind): New function extracted from
5770 record_node_allocation_statistics.
5771 (free_node): Use get_stats_node_kind.
5773 2018-03-14 Richard Biener <rguenther@suse.de>
5775 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
5776 that the value-set of ANTIC_IN doesn't grow.
5779 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5781 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5782 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5784 2018-03-14 Julia Koval <julia.koval@intel.com>
5786 * config.gcc (icelake-client, icelake-server): New.
5788 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
5789 (initial_ix86_arch_features): Ditto.
5790 (PTA_SKYLAKE): Add SGX.
5791 (PTA_ICELAKE): Remove.
5792 (PTA_ICELAKE_CLIENT): New.
5793 (PTA_ICELAKE_SERVER): New.
5794 (ix86_option_override_internal): Split up icelake on icelake client and
5796 (get_builtin_code_for_version): Ditto.
5797 (fold_builtin_cpu): Ditto.
5798 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
5799 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
5800 * config/i386/i386.h (processor_type): Ditto.
5801 * doc/invoke.texi: Ditto.
5803 2018-03-14 Jakub Jelinek <jakub@redhat.com>
5806 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
5807 INTEGER_CST offset, add it together with bitpos / 8 and
5808 sign extend based on POINTER_SIZE.
5812 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
5815 * config/i386/constraints.md (Yc): New register constraint.
5816 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
5817 Use Yc constraint for alternative 2 of operand 0. Remove
5818 preferred_for_speed attribute.
5820 2018-03-14 Richard Biener <rguenther@suse.de>
5822 PR tree-optimization/84830
5823 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
5824 with the old one to avoid oscillations.
5826 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
5829 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
5831 (assign_by_spills): Return a flag of reload assignment failure.
5832 Do not process the reload assignment failures. Do not spill other
5833 reload pseudos if they has the same reg class. Update n if
5835 (lra_assign): Add a return arg. Set up from the result of
5836 assign_by_spills call.
5837 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5838 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5839 usage_insns if it is not NULL.
5840 (spill_hard_reg_in_range): New function.
5841 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5842 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5843 function prototypes.
5844 (lra_assign): Change prototype.
5845 * lra.c (lra): Add code to deal with fails by splitting hard reg
5848 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
5850 * config/riscv/riscv.opt (mrelax): New option.
5851 * config/riscv/riscv.c (riscv_file_start): Emit ".option
5852 "norelax" when riscv_mrelax is disabled.
5853 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
5855 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5858 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
5859 reassociation for int modes.
5861 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5863 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
5864 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
5866 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
5867 * config/aarch64/aarch64-sve.md
5868 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
5869 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
5870 (*extend<mode><Vwide>2): Rename to...
5871 (aarch64_sve_extend<mode><Vwide>2): ...this.
5872 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
5873 renaming the old pattern to...
5874 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
5876 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
5877 define_expand, renaming the old pattern to...
5878 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
5879 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
5880 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
5881 account when deciding which SVE instruction the optab should use.
5882 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
5884 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5886 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
5887 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
5888 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
5889 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
5890 (tlsdesc_small_<mode>): Turn a define_expand and use
5891 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
5892 (tlsdesc_small_advsimd_<mode>): ...this.
5893 (tlsdesc_small_sve_<mode>): New pattern.
5895 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5897 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
5898 (UNSPEC_UMUL_HIGHPART): New constants.
5899 (MUL_HIGHPART): New int iteraor.
5900 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
5901 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
5903 (*<su>mul<mode>3_highpart): New define_insn.
5905 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
5908 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
5911 2018-03-13 Martin Liska <mliska@suse.cz>
5914 * (sem_item_optimizer::sem_item_optimizer): Initialize new
5916 (sem_item_optimizer::~sem_item_optimizer): Release it.
5917 (sem_item_optimizer::merge_classes): Register variable aliases.
5918 (sem_item_optimizer::fixup_pt_set): New function.
5919 (sem_item_optimizer::fixup_points_to_sets): Likewise.
5920 * ipa-icf.h: Declare new variables and functions.
5922 2018-03-13 Jakub Jelinek <jakub@redhat.com>
5925 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
5926 integer_pow2p@2 and test integer_pow2p in condition.
5927 (A < 0 ? C : 0): Similarly for @1.
5930 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
5931 characters starting at p contain '\0' character, don't look beyond
5935 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
5936 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
5939 * reg-stack.c (change_stack): Change update_end var from int to
5940 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
5941 also call set_block_for_insn on the newly added insns and rescan.
5944 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
5945 on the last operand.
5948 * tree.c (stabilize_reference_1): Return save_expr (e) for
5949 STATEMENT_LIST even if it doesn't have side-effects.
5951 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
5953 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
5955 2018-03-12 Renlin Li <renlin.li@arm.com>
5957 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
5958 aarch64_output_scalar_simd_mov_immediate.
5960 2018-03-12 Martin Sebor <msebor@redhat.com>
5962 PR tree-optimization/83456
5963 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
5964 for perfectly overlapping calls to memcpy.
5965 (gimple_fold_builtin_memory_chk): Same.
5966 (gimple_fold_builtin_strcpy): Handle no-warning.
5967 (gimple_fold_builtin_stxcpy_chk): Same.
5968 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
5970 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5972 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
5973 parameter. Use it for SFmode.
5974 (rs6000_function_arg_advance_1): Adjust.
5975 (rs6000_function_arg): Adjust.
5976 (rs6000_gimplify_va_arg): Pass false for that new parameter.
5978 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5980 PR rtl-optimization/84169
5981 PR rtl-optimization/84780
5982 * combine.c (can_combine_p): Check for a 2-insn combination whether
5983 the destination register is used between the two insns, too.
5985 2018-03-12 Richard Biener <rguenther@suse.de>
5987 PR tree-optimization/84803
5988 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
5989 for refs DR analysis didn't process.
5991 2018-03-12 Richard Biener <rguenther@suse.de>
5993 PR tree-optimization/84777
5994 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
5995 force-vectorize loops ignore whether we are optimizing for size.
5997 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
5999 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
6000 (TARGET_MD_ASM_ADJUST): Define.
6002 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
6003 Kito Cheng <kito.cheng@gmail.com>
6004 Chung-Ju Wu <jasonwucj@gmail.com>
6006 * config/nds32/nds32.c (nds32_compute_stack_frame,
6007 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
6008 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
6009 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
6010 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
6011 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
6012 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
6013 * config/nds32/nds32.md (prologue, epilogue): Use macro
6014 NDS32_V3PUSH_AVAILABLE_P to do checking.
6016 2018-03-11 Jakub Jelinek <jakub@redhat.com>
6019 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
6020 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
6021 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
6022 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
6023 addition of most attributes on !orig_type_die or the attribute not
6024 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
6026 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6027 Chung-Ju Wu <jasonwucj@gmail.com>
6029 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
6031 * config/nds32/nds32.opt (mvh): New option.
6033 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6034 Chung-Ju Wu <jasonwucj@gmail.com>
6036 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
6038 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
6039 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
6042 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6043 Chung-Ju Wu <jasonwucj@gmail.com>
6045 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
6047 * config/nds32/nds32-multiple.md (strlensi): New pattern.
6048 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
6050 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
6051 Kito Cheng <kito.cheng@gmail.com>
6052 Chung-Ju Wu <jasonwucj@gmail.com>
6054 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
6055 UNSPEC_FFMISM and UNSPEC_FLMISM.
6056 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
6057 for ffb, ffmism and flmism.
6058 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
6059 (unspec_ffmism): Ditto.
6060 (unspec_flmism): Ditto.
6061 (nds32_expand_builtin_impl): Check if string extension is available.
6062 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
6063 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
6065 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
6068 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6071 * lra-assigns.c (assign_by_spills): Return a flag of reload
6072 assignment failure. Do not process the reload assignment
6073 failures. Do not spill other reload pseudos if they has the same
6075 (lra_assign): Add a return arg. Set up from the result of
6076 assign_by_spills call.
6077 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6078 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6079 usage_insns if it is not NULL.
6080 (spill_hard_reg_in_range): New function.
6081 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6082 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6083 function prototypes.
6084 (lra_assign): Change prototype.
6085 * lra.c (lra): Add code to deal with fails by splitting hard reg
6088 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
6091 * config/i386/i386.opt: Replace Enforcment with Enforcement.
6093 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
6096 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
6097 (dw_val_node): Add val_symbolic_view.
6098 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
6099 (symview_upper_bound): New.
6100 (new_line_info_table): Initialize symviews_since_reset.
6101 (dwarf2out_source_line): Count symviews_since_reset and set
6102 symview_upper_bound.
6103 (dw_val_equal_p): Handle symview.
6104 (add_AT_symview): New.
6105 (print_dw_val): Handle symview.
6106 (attr_checksum, attr_checksum_ordered): Likewise.
6107 (same_dw_val_p, size_of_die): Likewise.
6108 (value_format, output_die): Likewise.
6109 (add_high_low_attributes): Use add_AT_symview for entry_view.
6110 (dwarf2out_finish): Reset symview_upper_bound, clear
6113 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
6116 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
6117 Add strict argument and use it.
6118 (rs6000_split_multireg_move): Update for new strict argument.
6119 (mem_operand_gpr): Disallow all non-offsettable addresses.
6120 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
6122 2018-03-09 Jakub Jelinek <jakub@redhat.com>
6125 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
6126 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
6127 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
6130 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
6131 decl, use remap_type if we want to use the type.
6133 2018-03-09 Martin Sebor <msebor@redhat.com>
6135 PR tree-optimization/84526
6136 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6138 (builtin_access::generic_overlap): Be prepared to handle non-array
6141 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
6143 PR rtl-optimization/84682
6144 * lra-constraints.c (process_address_1): Check is_address flag
6145 for address constraints.
6146 (process_alt_operands): Likewise.
6147 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
6148 preprocess_constraints.
6149 * recog.h (preprocess_constraints): Add oploc parameter.
6151 * recog.c (preprocess_constraints): Test address_operand for
6152 CT_ADDRESS constraints.
6154 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6157 * lra-assigns.c (assign_by_spills): Return a flag of reload
6158 assignment failure. Do not process the reload assignment
6159 failures. Do not spill other reload pseudos if they has the same
6161 (lra_assign): Add a return arg. Set up from the result of
6162 assign_by_spills call.
6163 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6164 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6165 usage_insns if it is not NULL.
6166 (spill_hard_reg_in_range): New function.
6167 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6168 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6169 function prototypes.
6170 (lra_assign): Change prototype.
6171 * lra.c (lra): Add code to deal with fails by splitting hard reg
6174 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6177 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
6178 Accept complain bool parameter. Only emit errors if it is true.
6179 (arm_parse_cpu_option_name): Likewise.
6180 (arm_target_thumb_only): Adjust callers of the above.
6181 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
6182 prototype to take a default true bool parameter.
6183 (arm_parse_arch_option_name): Likewise.
6185 2018-03-09 David Malcolm <dmalcolm@redhat.com>
6186 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
6190 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
6191 * configure: Regenerate.
6192 * configure.ac ("linker --version-script option"): New.
6193 ("linker soname option"): New.
6195 2018-03-09 Richard Biener <rguenther@suse.de>
6197 PR tree-optimization/84775
6198 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
6199 immediate uses of predicate stmts and mark them modified.
6202 PR tree-optimization/84178
6203 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6205 (version_loop_for_if_conversion): Delay update_ssa call.
6206 (tree_if_conversion): Delay update_ssa until after predicate
6209 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
6212 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
6213 when the function accesses prior frames.
6215 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6218 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
6219 gen_llsym, otherwise call maybe_gen_llsym.
6222 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
6223 has ',' character inside of it.
6225 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6228 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
6229 as clobbering CC_REGNUM.
6231 2018-03-08 Richard Biener <rguenther@suse.de>
6234 * tree-scalar-evolution.c: Include tree-into-ssa.h.
6235 (follow_copies_to_constant): Do not follow SSA names registered
6238 2018-03-08 Richard Biener <rguenther@suse.de>
6240 PR tree-optimization/84178
6241 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6243 (version_loop_for_if_conversion): Delay update_ssa call.
6244 (tree_if_conversion): Delay update_ssa until after predicate
6247 2018-03-08 David Malcolm <dmalcolm@redhat.com>
6249 PR tree-optimization/84178
6250 * tree-if-conv.c (release_bb_predicate): Remove the
6251 the assertion that the stmts have NULL use_ops.
6252 Discard the statements, asserting that they haven't
6253 yet been added to a BB.
6255 2018-03-08 Richard Biener <rguenther@suse.de>
6257 PR tree-optimization/84746
6258 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
6259 (phi_translate): Pass in destination ANTIC_OUT set.
6260 (phi_translate_1): Likewise. For a simplified result lookup
6261 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
6262 (phi_translate_set): Adjust.
6263 (do_pre_regular_insertion): Likewise.
6264 (do_pre_partial_partial_insertion): Likewise.
6266 2018-03-08 Martin Liska <mliska@suse.cz>
6268 PR gcov-profile/84735
6269 * doc/gcov.texi: Document usage of profile files.
6270 * gcov-io.h: Document changes in the format.
6272 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
6276 * dwarf2out.c (struct dw_line_info_table): Update comments for
6278 (FORCE_RESET_NEXT_VIEW): New.
6279 (FORCE_RESETTING_VIEW_P): New.
6280 (RESETTING_VIEW_P): Check for -1 too.
6281 (ZERO_VIEW_P): Likewise.
6282 (new_line_info_table): Force-reset next view.
6283 (dwarf2out_begin_function): Likewise.
6284 (dwarf2out_source_line): Simplify zero_view_p initialization.
6285 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
6286 view directly. Omit view when omitting .loc at line 0.
6288 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6290 PR tree-optimization/84740
6291 * tree-switch-conversion.c (process_switch): Call build_constructors
6292 only if info.phi_count is non-zero.
6294 PR tree-optimization/84739
6295 * tree-tailcall.c (find_tail_calls): Check call arguments against
6296 DECL_ARGUMENTS (current_function_decl) rather than
6297 DECL_ARGUMENTS (func) when checking for tail recursion.
6299 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6301 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
6302 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
6303 Volker Reichelt's entry and add entries for people that perform
6304 GCC fuzzy testing and report numerous bugs.
6306 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
6309 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
6310 readonly data in sdata, if that is disabled.
6311 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
6312 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
6313 -mreadonly-in-sdata option.
6315 2018-03-07 Martin Sebor <msebor@redhat.com>
6317 PR tree-optimization/84468
6318 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
6319 basic block when looking for nul assignment.
6321 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
6324 * except.h (output_function_exception_table): Adjust prototype.
6325 * except.c (output_function_exception_table): Remove FNNAME parameter
6326 and add SECTION parameter. Ouput one part of the table at a time.
6327 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
6328 the first part of the exception table and emit unwind directives.
6329 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
6330 (i386_pe_seh_cold_init): Likewise.
6331 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
6332 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
6333 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
6334 (ix86_output_call_insn): Emit a nop in one more case for SEH.
6335 * config/i386/winnt.c: Include except.h.
6336 (struct seh_frame_state): Add reg_offset, after_prologue and
6337 in_cold_section fields.
6338 (i386_pe_seh_end_prologue): Set seh->after_prologue.
6339 (i386_pe_seh_cold_init): New function.
6340 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
6341 to seh->in_cold_section.
6342 (seh_emit_push): Record the offset of the push.
6343 (seh_emit_save): Record the offet of the save.
6344 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
6345 Test seh->after_prologue to disregard the epilogue.
6346 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
6347 (i386_pe_end_cold_function): New function.
6349 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6352 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
6353 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
6356 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
6358 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
6359 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
6362 * multiple_target.c: Include tree-inline.h and intl.h.
6363 (expand_target_clones): Diagnose and fail if node->definition and
6364 !tree_versionable_function_p (node->decl).
6366 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
6368 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
6370 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
6371 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
6372 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
6374 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6377 * combine.c (try_combine): Use reg_or_subregno instead of handling
6378 just paradoxical SUBREGs and REGs.
6380 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
6382 * config/arc/arc.c (arc_finalize_pic): Remove function.
6383 (arc_must_save_register): We use single base PIC register, remove
6384 checks to save/restore the PIC register.
6385 (arc_expand_prologue): Likewise.
6386 * config/arc/arc-protos.h (arc_set_default_type_attributes):
6388 (arc_verify_short): Likewise.
6389 (arc_attr_type): Likewise.
6390 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
6391 (walk_stores): Likewise.
6392 (arc_address_cost): Make it static.
6393 (arc_verify_short): Likewise.
6394 (branch_dest): Likewise.
6395 (arc_attr_type): Likewise.
6396 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
6397 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
6398 (arc_final_prescan_insn): Remove inserting the nops due to
6399 hardware hazards. It is done in reorg step.
6400 (insn_length_variant_t): Remove.
6401 (insn_length_parameters_t): Likewise.
6402 (arc_insn_length_parameters): Likewise.
6403 (arc_get_insn_variants): Likewise.
6404 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
6406 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6409 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
6412 PR tree-optimization/84687
6413 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
6415 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
6417 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6419 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
6420 Rename to ppc_speculation_barrier.
6421 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
6422 __builtin_ppc_speculation_barrier.
6424 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6427 * combine.c (combine_simplify_rtx): Don't try to simplify if
6428 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
6431 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
6433 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
6434 to 32 bytes when compiling for POWER9.
6436 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6439 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
6440 regparm >= 3 with no arg reg available also for calls with
6441 flag_force_indirect_call. Pass decl to ix86_function_regparm.
6444 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
6446 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
6448 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
6451 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
6453 2018-03-05 Richard Biener <rguenther@suse.de>
6455 PR tree-optimization/84486
6456 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
6457 When inserting a __builtin_assume_aligned call set the LHS
6458 SSA name alignment info accordingly.
6460 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
6462 PR tree-optimization/84114
6463 * config/aarch64/aarch64.c (aarch64_reassociation_width)
6464 Avoid reassociation of FLOAT_MODE addition.
6466 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
6468 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
6469 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
6470 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
6471 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
6472 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
6473 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
6474 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
6476 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
6477 __builtin_ia32_wbinvd): New builtins.
6478 (SPECIAL_ARGS2): New.
6479 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
6480 (SPECIAL_ARGS2): New.
6481 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
6482 (ix86_valid_target_attribute_inner_p): Ditto.
6483 (ix86_init_mmx_sse_builtins): Add special_args2.
6484 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
6485 TARGET_WBNOINVD_P): New.
6486 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
6487 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
6488 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
6489 * config/i386/immintrin.h (_wbinvd): New intrinsic.
6490 * config/i386/pconfigintrin.h: New file.
6491 * config/i386/wbnoinvdintrin.h: Ditto.
6492 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
6493 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
6495 2018-03-05 Richard Biener <rguenther@suse.de>
6497 PR tree-optimization/84670
6498 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6500 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6501 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6502 (compute_antic_aux): Only assert the number of values in ANTIC_IN
6503 doesn't grow if all successors (recursively) were visited at least
6506 2018-03-05 Richard Biener <rguenther@suse.de>
6508 PR tree-optimization/84650
6509 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
6510 if executed in the loop pipeline.
6512 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
6514 * doc/configfiles.texi (Configuration Files): Move info about
6515 conditionalizing $target-protos.h to...
6516 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
6517 differs from $target-protos.h.
6519 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
6520 Chung-Ju Wu <jasonwucj@gmail.com>
6522 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
6523 * config/nds32/nds32-multiple.md (setmemsi): Define.
6524 * config/nds32/nds32-memory-manipulation.c
6525 (nds32_gen_dup_4_byte_to_word_value): New.
6526 (emit_setmem_word_loop): New.
6527 (emit_setmem_byte_loop): New.
6528 (nds32_expand_setmem_loop): New.
6529 (nds32_expand_setmem_loop_v3m): New.
6530 (nds32_expand_setmem_unroll): New.
6531 (nds32_expand_setmem): New.
6533 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6534 Chung-Ju Wu <jasonwucj@gmail.com>
6536 * config/nds32/nds32-memory-manipulation.c
6537 (nds32_emit_load_store): New.
6538 (nds32_emit_post_inc_load_store): New.
6539 (nds32_emit_mem_move): New.
6540 (nds32_emit_mem_move_block): New.
6541 (nds32_expand_movmemsi_loop_unknown_size): New.
6542 (nds32_expand_movmemsi_loop_known_size): New.
6543 (nds32_expand_movmemsi_loop): New.
6544 (nds32_expand_movmemsi_unroll): New.
6545 (nds32_expand_movmemqi): Rename ...
6546 (nds32_expand_movmemsi): ... to this.
6547 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
6548 (movmemsi): ... to this.
6549 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
6550 (nds32_expand_movmemsi): ... to this.
6552 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6553 Monk Chiang <sh.chiang04@gmail.com>
6554 Chung-Ju Wu <jasonwucj@gmail.com>
6556 * config/nds32/nds32-protos.h
6557 (nds32_expand_load_multiple): New arguments.
6558 (nds32_expand_store_multiple): Ditto.
6559 (nds32_valid_multiple_load_store): Rename ...
6560 (nds32_valid_multiple_load_store_p): ... to this.
6561 * config/nds32/nds32-memory-manipulation.c
6562 (nds32_expand_load_multiple): Refine implementation.
6563 (nds32_expand_store_multiple): Ditto.
6564 * config/nds32/nds32-multiple.md
6565 (load_multiple): Update nds32_expand_load_multiple interface.
6566 (store_multiple): Update nds32_expand_store_multiple interface.
6567 * config/nds32/nds32-predicates.c
6568 (nds32_valid_multiple_load_store): Rename ...
6569 (nds32_valid_multiple_load_store_p): ... to this and refine
6571 * config/nds32/predicates.md
6572 (nds32_load_multiple_and_update_address_operation): New predicate.
6573 (nds32_store_multiple_and_update_address_operation): New predicate.
6575 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6576 Chung-Ju Wu <jasonwucj@gmail.com>
6578 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
6579 (combo): New attribute.
6580 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
6582 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
6584 * config/nds32/nds32.opt: Change -mcmodel= default value.
6586 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
6587 Monk Chiang <sh.chiang04@gmail.com>
6588 Chung-Ju Wu <jasonwucj@gmail.com>
6590 * config/nds32/constants.md (unspec_element): New enum.
6591 * config/nds32/constraints.md (Umw): New constraint.
6592 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
6593 * config/nds32/nds32-intrinsic.md: Likewise.
6594 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
6595 (nds32_valid_smw_lwm_base_p): New.
6596 (nds32_output_smw_single_word): New.
6597 (nds32_output_lmw_single_word): New.
6598 (nds32_expand_unaligned_load): New.
6599 (nds32_expand_unaligned_store): New.
6600 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
6601 (nds32_output_smw_single_word): Declare.
6602 (nds32_output_lmw_single_word): Declare.
6603 (nds32_expand_unaligned_load): Declare.
6604 (nds32_expand_unaligned_store): Declare.
6605 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
6606 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
6607 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
6608 NDS32_BUILTIN_UASTORE_DW.
6609 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
6612 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
6613 Kito Cheng <kito.cheng@gmail.com>
6614 Chung-Ju Wu <jasonwucj@gmail.com>
6616 * config/nds32/nds32-intrinsic.c
6617 (nds32_expand_builtin_null_ftype_reg): Delete.
6618 (nds32_expand_builtin_reg_ftype_imm): Ditto.
6619 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
6620 (nds32_read_argument): New.
6621 (nds32_legitimize_target): Ditto.
6622 (nds32_legitimize_argument): Ditto.
6623 (nds32_check_constant_argument): Ditto.
6624 (nds32_expand_unop_builtin): Ditto.
6625 (nds32_expand_unopimm_builtin): Ditto.
6626 (nds32_expand_binop_builtin): Ditto.
6627 (nds32_builtin_decl_impl): Ditto.
6628 (builtin_description): Ditto.
6629 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
6630 (nds32_init_builtins_impl): Ditto.
6631 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
6632 (nds32_builtin_decl): New.
6633 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
6634 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
6636 2018-03-02 Jeff Law <law@redhat.com>
6638 * reorg.c (stop_search_p): Handle DEBUG_INSN.
6639 (redundant_insn, fill_simple_delay_slots): Likewise.
6640 (fill_slots_from_thread): Likewise.
6641 * resource.c (mark_referenced_resources): Likewise.
6642 (mark_set_resources, find_dead_or_set_registers): Likewise.
6644 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6646 * substring-locations.h (format_warning_va): Formatting fix for
6648 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
6650 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
6651 * substring-locations.c: Include intl.h.
6652 (format_warning_va): Turned into small wrapper around
6653 format_warning_n_va, renamed to ...
6654 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
6655 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
6657 (format_warning_at_substring_n): New function.
6658 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
6659 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
6660 format_warning_at_substring with just a shorter name instead of
6661 const function pointer.
6662 (fmtwarn_n): New function.
6663 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
6664 appropriate, get rid of all the fmtstr temporaries, move conditionals
6665 with G_() wrapped string literals directly into fmtwarn arguments,
6666 cast dir.len to (int), formatting fixes.
6668 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
6670 * doc/invoke.texi: Remove "Cilk Plus" references.
6672 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6673 Richard Biener <rguenther@suse.de>
6676 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
6677 for error or warning attributes if CALL_FROM_THUNK_P is set.
6680 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6683 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
6684 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
6687 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
6688 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
6690 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
6691 instead of -1U in last predictors element's probability member.
6693 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
6696 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
6697 arguments if they are comparable.
6699 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6701 PR tree-optimization/84634
6702 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
6703 masks and masked_loop_p with a single loop_masks, making sure it's
6704 null for bb vectorization.
6706 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6708 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
6709 (vect_analyze_data_ref_access): Use loop->safe_len rather than
6710 loop->force_vectorize to check whether there is no alias.
6712 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6715 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
6717 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
6719 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
6720 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
6721 instead of a loop around prev_real_insn.
6722 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
6726 * config/i386/i386.c (ix86_print_operand): Use conditional
6727 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
6730 2018-03-02 Richard Biener <rguenther@suse.de>
6732 PR tree-optimization/84427
6733 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
6734 (bitmap_set_subtract_values): Rewrite to handle multiple
6737 (prune_clobbered_mems): Likewise.
6738 (phi_translate): Take edge instead of pred/phiblock.
6739 (phi_translate_1): Likewise.
6740 (phi_translate_set): Likewise. Insert all translated
6741 exprs for a value into the set, keeping possibly multiple
6742 expressions per value.
6743 (compute_antic_aux): Adjust for phi_translate changes.
6744 When intersecting union the expressions and prune those
6745 not in the final value set, keeping possibly multiple
6746 expressions per value. Do not use value-insertion
6747 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
6748 all expressions. Add verification that the value-sets
6749 only shrink during iteration.
6750 (compute_partial_antic_aux): Adjust for the phi_translate changes.
6751 (do_pre_regular_insertion): Likewise.
6752 (do_pre_partial_partial_insertion): Likewise.
6754 2018-03-02 Richard Biener <rguenther@suse.de>
6757 * config/darwin.c (saved_debug_info_level): New static global.
6758 (darwin_asm_lto_start): Disable debug info generation for LTO out.
6759 (darwin_asm_lto_end): Restore debug info generation settings.
6761 2018-03-01 Martin Liska <mliska@suse.cz>
6764 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
6767 2018-03-01 Richard Biener <rguenther@suse.de>
6770 * dwarf2out.c (gen_variable_die): Properly handle late VLA
6771 type annotation with LTO when debug was disabled at compile-time.
6773 2018-03-01 Matthew Fortune <mfortune@gmail.com>
6775 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
6777 (mips_final_postscan_insn): Likewise.
6779 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
6781 PR rtl-optimization/84528
6782 * alias.c (init_alias_target): Add commentary.
6783 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
6784 a unique base value if the frame pointer is not eliminated
6785 to the stack pointer.
6787 2018-03-01 Tom de Vries <tom@codesourcery.com>
6789 PR rtl-optimization/83327
6790 * lra-int.h (hard_regs_spilled_into): Declare.
6791 * lra.c (hard_regs_spilled_into): Define.
6792 (init_reg_info): Init hard_regs_spilled_into.
6793 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
6794 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
6795 (process_bb_lives): Handle hard_regs_spilled_into.
6796 (lra_create_live_ranges_1): Before doing liveness propagation, clear
6797 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
6799 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
6801 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
6802 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
6803 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
6804 * config/rs6000/aix72.h: New file.
6806 2018-02-28 Jakub Jelinek <jakub@redhat.com>
6808 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
6809 instead of warning_at with conditional singular and plural messages
6813 * stor-layout.c (update_alignment_for_field): For
6814 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
6815 && !DECL_PACKED (field), do the alignment update, just use
6816 only desired_align instead of MAX (type_align, desired_align)
6818 (place_field): Don't do known_align < desired_align handling
6819 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
6820 is non-NULL, instead do it after rli->prev_field handling and
6821 only if not within a bitfield word. For DECL_PACKED (field)
6822 use type_align of BITS_PER_UNIT.
6824 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
6826 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
6827 superfluous parentheses and trailing spaces.
6829 2018-02-28 Richard Biener <rguenther@suse.de>
6831 PR tree-optimization/84584
6832 * graphite-scop-detection.c (scop_detection::add_scop): Discard
6833 SCoPs with fake exit edge.
6835 2018-02-28 Martin Liska <mliska@suse.cz>
6838 * timevar.c (timer::print): Fix format to properly print 100%
6841 2018-02-28 Richard Biener <rguenther@suse.de>
6844 * genmatch.c (capture_info::walk_match): Do not mark
6845 captured expressions without operands as expr_p given
6846 they act more like predicates and should be subject to
6847 "lost tail" side-effect preserving.
6849 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
6851 PR rtl-optimization/81611
6852 * auto-inc-dec.c (attempt_change): Move dead note from
6853 mem_insn if it's the next use of regno
6854 (find_address): Take address use of reg holding
6855 non-incremented value. Add parm to limit search to the named
6857 (merge_in_block): Attempt to use a mem insn that is the next
6858 use of the original regno.
6860 2018-02-27 Martin Sebor <msebor@redhat.com>
6863 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
6864 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
6866 2018-02-27 Martin Sebor <msebor@redhat.com>
6868 PR translation/84207
6869 * diagnostic-core.h (warning_n, error_n, inform_n): Change
6870 n argument to unsigned HOST_WIDE_INT.
6871 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
6872 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
6873 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
6874 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
6876 2018-02-27 Richard Biener <rguenther@suse.de>
6878 PR tree-optimization/84512
6879 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
6880 Do not use the estimate returned from record_stmt_cost for
6881 the scalar iteration cost but sum properly using add_stmt_cost.
6883 2018-02-27 Richard Biener <rguenther@suse.de>
6885 PR tree-optimization/84466
6886 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6887 Adjust last change to less strictly validate use operands.
6889 2018-02-27 Martin Liska <mliska@suse.cz>
6891 PR gcov-profile/84548
6892 * gcov.c (process_file): Allow partial overlap and consider it
6893 also as group functions.
6894 (output_lines): Properly calculate range of lines for a group.
6896 2018-02-27 Martin Liska <mliska@suse.cz>
6898 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
6899 'ggc' suffixes. Change first column width.
6900 (timer::print): Fix formatting of the column.
6902 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
6904 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
6905 preserve inline entry blocks for the sake of debug inline
6906 entry point markers alone.
6907 (remove_unused_locals): Suggest in comments a better place to
6908 force the preservation of inline entry blocks that are
6909 otherwise unused, but do not preserve them.
6911 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6913 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
6915 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6918 * config/i386/constraints.md (Bs): Replace
6919 ix86_indirect_branch_register with
6920 TARGET_INDIRECT_BRANCH_REGISTER.
6922 * config/i386/i386.md (indirect_jump): Likewise.
6923 (tablejump): Likewise.
6924 (*sibcall_memory): Likewise.
6925 (*sibcall_value_memory): Likewise.
6926 Peepholes of indirect call and jump via memory: Likewise.
6927 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
6928 (*sibcall_value_GOT_32): Likewise.
6929 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6930 (GOT_memory_operand): Likewise.
6931 (call_insn_operand): Likewise.
6932 (sibcall_insn_operand): Likewise.
6933 (GOT32_symbol_operand): Likewise.
6934 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
6936 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6938 PR rtl-optimization/83496
6939 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
6940 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
6941 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
6942 redundant insn, if any.
6943 (relax_delay_slots): Likewise.
6944 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
6946 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
6948 PR tree-optimization/83965
6949 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
6950 that grouped statements are part of a reduction chain. Return
6951 true if the statement is not marked as a reduction itself but
6953 (vect_recog_dot_prod_pattern): Don't check whether the statement
6954 is part of a group here.
6955 (vect_recog_sad_pattern): Likewise.
6956 (vect_recog_widen_sum_pattern): Likewise.
6958 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6961 * final.c (rest_of_clean_state): Also look for calls inside sequences.
6963 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6966 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
6968 (ix86_output_indirect_function_return): New prototype.
6969 (ix86_split_simple_return_pop_internal): Likewise.
6970 * config/i386/i386.c (indirect_return_via_cx): New.
6971 (indirect_return_via_cx_bnd): Likewise.
6972 (indirect_thunk_name): Handle return va CX_REG.
6973 (output_indirect_thunk_function): Create alias for
6974 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
6975 (ix86_output_indirect_jmp): Remove the bool argument.
6976 (ix86_output_indirect_function_return): New function.
6977 (ix86_split_simple_return_pop_internal): Likewise.
6978 * config/i386/i386.md (*indirect_jump): Don't pass false
6979 to ix86_output_indirect_jmp.
6980 (*tablejump_1): Likewise.
6981 (simple_return_pop_internal): Change it to define_insn_and_split.
6982 Call ix86_split_simple_return_pop_internal to split it for
6984 (simple_return_indirect_internal): Call
6985 ix86_output_indirect_function_return instead of
6986 ix86_output_indirect_jmp.
6988 2018-02-26 Jakub Jelinek <jakub@redhat.com>
6991 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
6992 memset and value initialization afterwards.
6994 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
6996 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
6998 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7001 * common/config/aarch64/aarch64-common.c
7002 (aarch_option_optimization_table[]): Switch
7003 off fomit-frame-pointer
7005 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7006 Chung-Ju Wu <jasonwucj@gmail.com>
7008 * config/nds32/nds32-multiple.md (load_multiple): Disallow
7010 (store_multiple): Ditto.
7012 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7014 * config.gcc: Add --with-cpu support for nds32 target.
7015 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
7016 * config/nds32/nds32.opt: Add -mcpu= option.
7018 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
7020 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
7021 isel=yes): Warn for these deprecated options.
7023 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
7025 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
7026 ISA_2_5_MASKS_EMBEDDED.
7028 2018-02-23 Jakub Jelinek <jakub@redhat.com>
7030 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
7031 p->max as pointers rather than using iterative_hash_expr.
7033 2018-02-23 Carl Love <cel@us.ibm.com>
7035 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
7036 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
7038 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
7039 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
7040 P8V_BUILTIN_VEC_VUNSIGNED2.
7042 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
7045 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
7046 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
7048 * lra-constraints.c (curr_insn_transform): Set up
7049 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
7051 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
7052 * lra-eliminations.c (spill_pseudos): Ditto.
7053 (process_insn_for_elimination): Ditto.
7054 * lra-lives.c (reg_early_clobber_p): Use the new macros.
7055 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
7056 LRA_NON_CLOBBERED_ALT.
7058 2018-02-22 Martin Sebor <msebor@redhat.com>
7060 PR tree-optimization/84480
7061 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
7062 to maybe_diag_stxncpy_trunc. Call it.
7063 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
7064 from gimple_fold_builtin_strcpy. Print inlining stack.
7065 (handle_builtin_stxncpy): Print inlining stack.
7066 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
7068 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
7071 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
7072 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
7073 and -fcheck-pointer-bounds are used together.
7074 (indirect_thunk_prefix): New enum.
7075 (indirect_thunk_need_prefix): New function.
7076 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
7077 "_nt" instead of "_bnd" for NOTRACK prefix.
7078 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
7079 (output_indirect_thunk_function): Likewise.
7081 (ix86_code_end): Update output_indirect_thunk_function calls.
7082 (ix86_output_indirect_branch_via_reg): Replace
7083 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
7084 (ix86_output_indirect_branch_via_push): Likewise.
7085 (ix86_output_function_return): Likewise.
7086 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
7087 incompatible with -fcf-protection=branch and
7088 -fcheck-pointer-bounds.
7090 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7093 * config/aarch64/aarch64.c (aarch64_print_address_internal):
7094 Change gcc_assert call to output_operand_lossage.
7096 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7098 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
7100 2018-02-22 DJ Delorie <dj@redhat.com>
7101 Sebastian Perta <sebastian.perta@renesas.com>
7102 Oleg Endo <olegendo@gcc.gnu.org>
7104 * config/rx/rx.c (rx_rtx_costs): New function.
7105 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
7107 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
7109 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
7111 2018-02-22 Martin Liska <mliska@suse.cz>
7114 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
7115 Add "native" as a possible value.
7117 2018-02-22 Martin Liska <mliska@suse.cz>
7120 * config/i386/i386.c (ix86_option_override_internal):
7121 Add "native" as a possible value for -march and -mtune.
7123 2018-02-22 Jakub Jelinek <jakub@redhat.com>
7126 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
7127 to all type variants.
7129 PR tree-optimization/84503
7130 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
7131 width as info->bitpos + info->bitsize - start.
7132 (merged_store_group::merge_overlapping): Simplify width computation.
7133 (check_no_overlap): New function.
7134 (imm_store_chain_info::try_coalesce_bswap): Compute expected
7135 start + width and last_order of the group, fail if check_no_overlap
7137 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
7138 to group if check_no_overlap fails.
7140 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7142 * config/rs6000/altivec.md: Delete contraint arguments to
7143 define_expand, define_split, and define_peephole2, and in
7144 define_insn_and_split if always unused.
7145 * config/rs6000/darwin.md: Ditto.
7146 * config/rs6000/dfp.md: Ditto.
7147 * config/rs6000/rs6000.md: Ditto.
7148 * config/rs6000/sync.md: Ditto.
7149 * config/rs6000/vector.md: Ditto.
7150 * config/rs6000/vsx.md: Ditto.
7152 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7154 * config/rs6000/altivec.md: Write output control strings as braced
7155 blocks instead of double-quoted strings.
7156 * config/rs6000/darwin.md: Ditto.
7157 * config/rs6000/rs6000.md: Ditto.
7158 * config/rs6000/vector.md: Ditto.
7159 * config/rs6000/vsx.md: Ditto.
7161 2018-02-21 Jason Merrill <jason@redhat.com>
7163 PR c++/84314 - ICE with templates and fastcall attribute.
7164 * attribs.c (build_type_attribute_qual_variant): Remove assert.
7166 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7168 * ipa-cp.c (determine_versionability): Fix comment typos.
7170 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7173 * ipa-cp.c (determine_versionability): Do not version functions caling
7176 2018-02-21 Martin Liska <mliska@suse.cz>
7179 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
7180 Add "native" as a possible value.
7181 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
7182 the macro when native cpu detection is available.
7184 2018-02-21 Martin Liska <mliska@suse.cz>
7187 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
7188 Add "native" as a possible value.
7189 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
7190 when native cpu detection is available.
7192 2018-02-21 Jakub Jelinek <jakub@redhat.com>
7193 Martin Sebor <msebor@redhat.com>
7195 PR tree-optimization/84478
7196 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
7198 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
7199 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
7200 support which is conservatively correct, for 2 only stay conservative
7201 for maxlen. Formatting and comment capitalization fixes. Add STRICT
7202 argument to the 2 argument get_range_strlen, adjust 6 arg
7203 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
7205 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
7206 (gimple_fold_builtin_strlen): Pass true as last argument to
7209 2018-02-20 Martin Sebor <msebor@redhat.com>
7212 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
7213 (builtin_memref::set_base_and_offset): Same. Handle inner references.
7214 (builtin_memref::builtin_memref): Factor out parts into
7215 set_base_and_offset and call it.
7217 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7220 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
7221 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
7222 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
7223 search at the associated MODE_INT.
7225 2018-02-20 Jeff Law <law@redhat.com>
7228 PR tree-optimization/81592
7230 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
7231 for range data rather than using global data.
7232 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
7233 range data rather than using global data.
7234 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
7235 pass it to children as needed.
7236 (struct directive::fmtresult): Similarly.
7237 (struct directive::set_width): Similarly.
7238 (struct directive::set_precision): Similarly.
7239 (format_integer, format_directive, parse_directive): Similarly.
7240 (format_none): Accept unnamed vr_values parameter.
7241 (format_percent, format_floating, format_character): Similarly.
7242 (format_string, format_plain): Similarly.
7243 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
7244 the EVRP range analyzer for range data rather than using global data.
7245 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
7246 gimple-ssa-evrp-analyze.h
7247 (class sprintf_dom_walker): Add after_dom_children member function.
7248 Add evrp_range_analyzer member.
7249 (sprintf_dom_walker::before_dom_children): Call into the EVRP
7250 range analyzer as needed.
7251 (sprintf_dom_walker::after_dom_children): New member function.
7252 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
7254 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
7255 (evrp_range_analyzer::pop_to_marker): Likewise.
7257 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7259 PR tree-optimization/84419
7260 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
7261 with the required type if its current type is compatible but
7264 2018-02-20 Jakub Jelinek <jakub@redhat.com>
7267 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
7268 after vectorization.
7270 2018-02-20 Martin Liska <mliska@suse.cz>
7273 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
7274 possible values if we don't have a hint.
7276 2018-02-20 Martin Liska <mliska@suse.cz>
7280 * final.c (shorten_branches): Build align_tab array with one
7282 * opts.c (finish_options): Add alignment option limit check.
7283 (MAX_CODE_ALIGN): Likewise.
7284 (MAX_CODE_ALIGN_VALUE): Likewise.
7285 * doc/invoke.texi: Document maximum allowed option value for
7286 all -falign-* options.
7288 2018-02-19 Jakub Jelinek <jakub@redhat.com>
7291 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
7292 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
7293 * var-tracking.c (emit_note_insn_var_location): Remove all references
7294 to NOTE_INSN_CALL_ARG_LOCATION.
7295 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
7296 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
7297 Use copy_rtx_if_shared.
7298 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
7299 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
7300 (dwarf2out_var_location): Remove handling of
7301 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
7303 * final.c (final_scan_insn): Remove all references to
7304 NOTE_INSN_CALL_ARG_LOCATION.
7305 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
7306 before dumping final insns.
7307 * except.c (emit_note_eh_region_end): Remove all references to
7308 NOTE_INSN_CALL_ARG_LOCATION.
7309 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
7310 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
7311 * config/arc/arc.c (hwloop_optimize): Likewise.
7312 * config/arm/arm.c (create_fix_barrier): Likewise.
7313 * config/s390/s390.c (s390_chunkify_start): Likewise.
7314 * config/sh/sh.c (find_barrier): Likewise.
7315 * config/i386/i386.c (rest_of_insert_endbranch,
7316 ix86_seh_fixup_eh_fallthru): Likewise.
7317 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
7318 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
7319 * config/frv/frv.c (frv_function_prologue): Likewise.
7320 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
7322 (note_outside_basic_block_p): Remove all references to
7323 NOTE_INSN_CALL_ARG_LOCATION.
7324 * gengtype.c (adjust_field_rtx_def): Likewise.
7325 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
7327 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
7328 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
7331 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
7334 PR tree-optimization/84452
7335 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
7336 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
7339 2018-02-19 Martin Liska <mliska@suse.cz>
7342 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
7344 2018-02-19 Martin Liska <mliska@suse.cz>
7345 Richard Sandiford <richard.sandiford@linaro.org>
7347 PR tree-optimization/82491
7348 * gimple-fold.c (get_base_constructor): Make earlier bail out
7351 2018-02-19 Carl Love <cel@us.ibm.com>
7353 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
7354 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
7356 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
7357 P8V_BUILTIN_VEC_NEG.
7359 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
7361 * config/rl78/rl78.md (movdf): New define expand.
7363 2018-02-19 Martin Liska <mliska@suse.cz>
7366 * doc/invoke.texi: Fix typo.
7367 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
7369 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
7371 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
7372 handle rs6000_single_float and rs6000_double_float specially for
7375 2018-02-16 Jeff Law <law@redhat.com>
7377 * config/rx/rx.c (add_pop_cfi_notes): New function.;
7380 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7383 * ipa-inline.c (inline_small_functions): Fix a typo.
7385 2018-02-16 Nathan Sidwell <nathan@acm.org>
7387 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
7389 2018-02-16 Carl Love <cel@us.ibm.com>
7391 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
7392 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
7393 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
7394 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
7395 expansion to P8V_BUILTIN_VEC_FLOAT2.
7397 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
7399 PR rtl-optimization/70023
7400 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
7401 src_regno into account.
7403 2018-02-16 Carl Love <cel@us.ibm.com>
7405 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
7406 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
7407 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
7408 * config/rs6000/rs6000.c: Remove case statements for
7409 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
7410 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
7411 and P9V_BUILTIN_VEC_VINSERT4B.
7412 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
7413 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
7414 * config/rs6000/vsx.md:
7415 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
7418 2018-02-16 Carl Love <cel@us.ibm.com>
7420 * config/rs6000/altivec.h: Add builtin names vec_extract4b
7422 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
7424 * config/rs6000/rs6000-c.c: Add the definitions for
7425 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
7426 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
7427 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
7428 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
7429 definition for insert4b and define insn *insert3b_internal.
7430 * doc/extend.texi: Add documentation for vec_extract4b.
7432 2018-02-16 Nathan Sidwell <nathan@acm.org>
7434 * doc/extend.texi (Backwards Compatibility): Mention friend
7435 injection. Note for-scope is deprecated.
7436 * doc/invoke.texi (-ffriend-injection): Deprecate.
7438 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
7440 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
7441 that moved to I2, also allow destinations that are a paradoxical
7442 subreg (instead of a normal reg).
7444 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
7447 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
7450 2018-02-16 Richard Biener <rguenther@suse.de>
7452 PR tree-optimization/84037
7453 PR tree-optimization/84016
7455 * config/i386/i386.c (ix86_builtin_vectorization_cost):
7456 Adjust vec_construct for the fact we need additional higher latency
7457 128bit inserts for AVX256 and AVX512 vector builds.
7458 (ix86_add_stmt_cost): Scale vector construction cost for
7461 2018-02-16 Richard Biener <rguenther@suse.de>
7463 PR tree-optimization/84417
7464 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
7465 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
7466 (non_rewritable_lvalue_p): Likewise, use poly-ints.
7468 2018-02-16 Martin Liska <mliska@suse.cz>
7471 * internal-fn.def (ASAN_CHECK): Set proper flags.
7472 (ASAN_MARK): Likewise.
7474 2018-02-16 Julia Koval <julia.koval@intel.com>
7476 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
7477 from PTA_CANNONLAKE.
7479 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7482 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
7483 Use ++iter rather than iter++ for std::list iterators.
7484 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
7485 defer deleting them until all nodes in the forest are processed. Do
7486 free even leaf nodes. Change to_process into auto_vec.
7489 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
7490 * vec.h (vec_default_construct): Use memset instead of placement new
7491 if BROKEN_VALUE_INITIALIZATION is defined.
7492 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
7493 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
7496 PR rtl-optimization/83723
7497 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
7498 * lra.c (lra_substitute_pseudo): Likewise. If true, use
7499 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
7501 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
7503 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
7505 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
7507 PR rtl-optimization/81443
7508 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
7509 from inner REGs to paradoxical SUBREGs.
7511 2018-02-16 Richard Biener <rguenther@suse.de>
7513 PR tree-optimization/84399
7514 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7515 For operands we can analyze at their definition make sure we can
7516 analyze them at each use as well.
7518 2018-02-16 Richard Biener <rguenther@suse.de>
7520 PR tree-optimization/84190
7521 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
7522 volatile accesses if the decl isn't volatile.
7524 2018-02-15 Jason Merrill <jason@redhat.com>
7526 PR c++/84314 - ICE with templates and fastcall attribute.
7527 * attribs.c (build_type_attribute_qual_variant): Don't clobber
7528 TYPE_CANONICAL on an existing type.
7530 2018-02-15 Jakub Jelinek <jakub@redhat.com>
7532 PR tree-optimization/84383
7533 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
7534 dstoff nor call operand_equal_p if dstbase is NULL.
7536 PR tree-optimization/84334
7537 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
7538 also a CONSTANT_CLASS_P, punt.
7540 2018-02-14 Jim Wilson <jimw@sifive.com>
7542 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
7543 first SMALL_OPERAND check. New local min_second_step. Move assert
7544 to where locals are set. Add TARGET_RVC support.
7545 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
7547 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
7549 * doc/invoke.texi: Correct -Wformat-overflow code sample.
7551 2018-02-14 Martin Sebor <msebor@redhat.com>
7553 PR tree-optimization/83698
7554 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
7555 arrays constrain the offset range to their bounds.
7556 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
7557 (builtin_access::overlap): Avoid setting the size of overlap if it's
7559 (maybe_diag_overlap): Also consider arrays when deciding what values
7560 of offsets to include in diagnostics.
7562 2018-02-14 Martin Sebor <msebor@redhat.com>
7565 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
7566 that correspond to the kind of a declaration.
7568 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
7571 * config/pa/pa.md: Load address of PIC label using the linkage table
7572 if the label is nonlocal.
7574 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
7576 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
7577 warning message if user requests -maltivec=be.
7578 * doc/invoke.texi: Document deprecation of -maltivec=be.
7580 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
7583 * config/rs6000/rs6000-c.c: Update definitions for
7584 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
7585 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
7587 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7590 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
7591 add _get_ssp intrinsics. Remove argument from
7592 __builtin_ia32_rdssp[d|q].
7593 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
7594 * config/i386/i386-builtin.def: Remove argument from
7595 __builtin_ia32_rdssp[d|q].
7596 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
7597 ix86_expand_special_args_builtin for _rdssp[d|q].
7598 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
7599 Clear register before usage.
7600 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
7601 Add documentation for new _get_ssp and _inc_ssp intrinsics.
7603 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
7605 PR tree-optimization/84357
7606 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
7607 operand 1 of an ARRAY_REF too.
7609 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
7612 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
7613 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7615 (set_of_reg): New struct.
7616 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
7617 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
7618 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7620 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
7621 Split into bitclr, bitset, bitinvert patterns if appropriate.
7622 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
7623 use rx_fuse_in_memory_bitop.
7624 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
7625 to named insn, correct maximum insn length.
7627 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
7630 * machmode.def: Define a complex mode for PARTIAL_INT.
7631 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
7633 * doc/rtl.texi: Document CSPImode.
7634 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
7636 (msp430_hard_regno_nregs_with_padding): Likewise.
7638 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
7641 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
7643 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
7645 PR rtl-optimization/84169
7646 * combine.c (try_combine): New variable split_i2i3. Set it to true if
7647 we generated a parallel as new i3 and we split that to new i2 and i3
7648 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
7649 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
7650 those to i2, not i1. Partially rewrite this scan code.
7652 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7655 * stor-layout.c (place_field): For variable length fields, adjust
7656 offset_align afterwards not just based on the field's alignment,
7657 but also on the size.
7660 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
7661 of exps and logs in the use_exp2 case.
7663 2018-02-13 Jeff Law <law@redhat.com>
7665 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
7668 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
7671 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
7675 * common.opt (gas-loc-support, gas-locview-support): New.
7676 (ginline-points, ginternal-reset-location-views): New.
7677 * doc/invoke.texi: Document them. Use @itemx where intended.
7678 (gvariable-location-views): Adjust.
7679 * target.def (reset_location_view): New.
7680 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
7681 (TARGET_RESET_LOCATION_VIEW): New.
7682 * doc/tm.texi: Rebuilt.
7683 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
7684 (dwarf2out_default_as_locview_support): New.
7685 (output_asm_line_debug_info): Use option variables.
7686 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
7687 (output_loc_list): Likewise.
7688 (add_high_low_attributes): Check option variables.
7689 Don't output entry view attribute in strict mode.
7690 (gen_inlined_subroutine_die): Check option variables.
7691 (dwarf2out_inline_entry): Likewise.
7692 (init_sections_and_labels): Likewise.
7693 (dwarf2out_early_finish): Likewise.
7694 (maybe_reset_location_view): New, from...
7695 (dwarf2out_var_location): ... here. Call it.
7696 * debug.h (dwarf2out_default_as_loc_support): Declare.
7697 (dwarf2out_default_as_locview_support): Declare.
7698 * hooks.c (hook_int_rtx_insn_0): New.
7699 * hooks.h (hook_int_rtx_insn_0): Declare.
7700 * toplev.c (process_options): Take -gas-loc-support and
7701 -gas-locview-support from dwarf2out. Enable
7702 -gvariable-location-views by default only with locview
7703 assembler support. Enable -ginternal-reset-location-views by
7704 default only if the target defines the corresponding hook.
7705 Enable -ginline-points by default if location views are
7706 enabled; force it disabled if statement frontiers are
7708 * tree-inline.c (expand_call_inline): Check option variables.
7709 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
7711 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
7713 PR tree-optimization/84321
7714 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
7715 handling. Also check whether the anti-range contains any values
7716 that satisfy the mask; switch to a VR_RANGE if not.
7718 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
7721 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
7723 2018-02-13 Martin Jambor <mjambor@suse.cz>
7726 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
7727 of call statements, also set location of a load to a temporary.
7729 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
7731 * config/rl78/rl78.c (add_vector_labels): New function.
7732 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
7733 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
7734 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
7735 which checks that no arguments are passed.
7736 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
7737 * doc/extend.texi: Documentation for the new attribute.
7739 2018-02-13 Andreas Schwab <schwab@suse.de>
7741 * config/riscv/linux.h (CPP_SPEC): Define.
7743 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7746 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
7747 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
7748 OPTION_MASK_ISA_AES as first argument to def_builtin_const
7749 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
7750 instead of OPTION_MASK_ISA_PCLMUL as first argument to
7751 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
7752 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
7753 temporarily for AES and PCLMUL builtins.
7755 PR tree-optimization/84339
7756 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
7757 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
7761 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
7762 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
7763 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
7765 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
7767 * omp-simd-clone.h: New file.
7768 * omp-simd-clone.c: Include omp-simd-clone.h.
7769 (expand_simd_clones): No longer static.
7770 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
7771 cgraph.h and omp-simd-clone.h.
7772 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
7773 (vect_recog_widen_shift_pattern): Formatting fix.
7774 (vect_pattern_recog_1): Don't check optab for calls.
7777 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
7778 operands[2] into a REG before using gen_lowpart on it.
7780 2018-02-12 Jeff Law <law@redhat.com>
7783 * config/sh/sh.c (find_barrier): Consider a sibling call
7786 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
7787 successfully back substituting a reg.
7789 2018-02-12 Richard Biener <rguenther@suse.de>
7791 PR tree-optimization/84037
7792 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
7793 parameter, move visited init to caller.
7794 (vect_slp_analyze_operations): Separate cost from validity
7795 check, initialize visited once for all instances.
7796 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
7798 * tree-vect-stmts.c (vect_model_simple_cost): Make early
7800 (vect_model_promotion_demotion_cost): Likewise.
7801 (vectorizable_bswap): Guard cost modeling with !slp_node
7802 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
7804 (vectorizable_call): Likewise.
7805 (vectorizable_conversion): Likewise.
7806 (vectorizable_assignment): Likewise.
7807 (vectorizable_shift): Likewise.
7808 (vectorizable_operation): Likewise.
7809 (vectorizable_store): Likewise.
7810 (vectorizable_load): Likewise.
7811 (vectorizable_condition): Likewise.
7812 (vectorizable_comparison): Likewise.
7814 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
7817 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
7818 (ASAN_MARK): Fix fnspec to account for return value, change pointer
7819 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
7821 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7824 * params.def (inline-min-speedup): Increase from 8 to 15.
7825 (max-inline-insns-auto): Decrease from 40 to 30.
7826 * ipa-split.c (consider_split): Add some buffer for function to
7827 be considered inlining candidate.
7828 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
7831 2018-02-12 Richard Biener <rguenther@suse.de>
7833 PR tree-optimization/84037
7834 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
7835 matched stmts if we cannot swap the non-matched ones.
7837 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7839 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
7840 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
7841 _mm_maskz_scalef_round_ss): New intrinsics.
7842 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
7843 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
7844 __builtin_ia32_scalefss_round): Remove.
7845 (__builtin_ia32_scalefsd_mask_round,
7846 __builtin_ia32_scalefss_mask_round): New intrinsics.
7847 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
7848 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
7849 ((match_operand:VF_128 2 "<round_nimm_predicate>"
7850 "<round_constraint>")): Changed to ...
7851 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
7852 "<round_scalar_constraint>")): ... this.
7853 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
7854 %0, %1, %2<round_op3>}"): Changed to ...
7855 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
7856 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
7857 %2<round_scalar_mask_op3>}"): ... this.
7858 * config/i386/subst.md (round_scalar_nimm_predicate): New.
7860 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7862 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
7863 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
7864 (_mm_maskz_sqrt_round_ss): New intrinsics.
7865 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
7866 (__builtin_ia32_sqrtsd_mask_round)
7867 (__builtin_ia32_sqrtss_mask_round): New builtins.
7868 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
7869 (__builtin_ia32_sqrtss_round): Remove.
7870 (__builtin_ia32_sqrtsd_mask_round)
7871 (__builtin_ia32_sqrtss_mask_round): New builtins.
7872 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
7873 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
7874 ((match_operand:VF_128 1 "vector_operand"
7875 "xBm,<round_constraint>")): Changed to ...
7876 ((match_operand:VF_128 1 "vector_operand"
7877 "xBm,<round_scalar_constraint>")): ... this.
7878 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
7879 %0, %2, %<iptr>1<round_op3>}): Changed to ...
7880 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
7881 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
7882 %<iptr>1<round_scalar_mask_op3>}): ... this.
7883 ((set_attr "prefix" "<round_prefix>")): Changed to ...
7884 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
7886 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
7889 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
7890 Cast vec_cmpeq result to correct type.
7891 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
7892 Cast vec_cmpgt result to correct type.
7894 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
7896 * final.c (final_scan_insn_1): Renamed from...
7897 (final_scan_insn): ... this. New wrapper, to recover
7898 seen from the outermost call in recursive ones.
7899 * config/sparc/sparc.c (output_return): Drop seen from call.
7900 (output_sibcall): Likewise.
7901 * config/visium/visium.c (output_branch): Likewise.
7903 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
7905 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
7908 2018-02-10 Alan Modra <amodra@gmail.com>
7911 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
7912 Specify LR as an input.
7914 2018-02-10 Jakub Jelinek <jakub@redhat.com>
7917 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
7918 remove_member_access_dummy_vars): New functions.
7919 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
7920 lower_omp_1, execute_lower_omp): Use them.
7922 PR rtl-optimization/84308
7923 * shrink-wrap.c (spread_components): Release todo vector.
7925 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
7927 PR rtl-optimization/57193
7928 * ira-color.c (struct allocno_color_data): Add member
7929 conflict_allocno_hard_prefs.
7930 (update_conflict_allocno_hard_prefs): New.
7931 (bucket_allocno_compare_func): Add a preference based on
7932 conflict_allocno_hard_prefs.
7933 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
7934 (color_allocnos): Remove a dead code. Initiate
7935 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
7937 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7940 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
7941 constraint from =wa to wa. Avoid a subreg on the output operand,
7942 instead use a pseudo and subreg it in a move.
7943 (p9_xxbrd_<mode>): Changed to ...
7944 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
7945 (p9_xxbrd_v2df): New expander.
7946 (p9_xxbrw_<mode>): Changed to ...
7947 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
7948 (p9_xxbrw_v4sf): New expander.
7950 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7952 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
7954 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
7957 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
7958 multiply in 32-bit mode.
7959 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
7960 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
7963 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7965 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
7966 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
7967 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
7968 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
7970 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
7973 * dwarf2out.c (is_trivial_indirect_ref): New function.
7974 (dwarf2out_late_global_decl): Do not generate a location
7975 attribute for variables that have a non-trivial DECL_VALUE_EXPR
7976 and that are not defined in the current unit.
7978 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7980 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
7981 instead of a libcall for UNORDERED.
7983 2018-02-09 Tamar Christina <tamar.christina@arm.com>
7986 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
7987 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
7989 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7992 * config/s390/s390.c (s390_set_current_function): Invoke
7993 s390_indirect_branch_settings also if fndecl didn't change.
7995 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7997 * config/rs6000/rs6000.md (blockage): Set length to zero.
7999 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8001 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
8003 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8006 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
8007 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
8011 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
8012 PARALLEL incoming that failed vt_get_decl_and_offset check.
8015 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
8016 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
8018 (get_variable_section): For decls in named .bss* sections pass true as
8019 second argument to bss_initializer_p.
8021 2018-02-09 Marek Polacek <polacek@redhat.com>
8022 Jakub Jelinek <jakub@redhat.com>
8025 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
8026 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
8027 Sync some changes from cxx_fold_indirect_ref.
8029 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8031 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
8033 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
8034 (BLOCK_INLINE_ENTRY_LABEL): New.
8035 (dwarf2out_var_location): Disregard inline entry markers.
8036 (inline_entry_data): New struct.
8037 (inline_entry_data_hasher): New hashtable type.
8038 (inline_entry_data_hasher::hash): New.
8039 (inline_entry_data_hasher::equal): New.
8040 (inline_entry_data_table): New variable.
8041 (add_high_low_attributes): Add DW_AT_entry_pc and
8042 DW_AT_GNU_entry_view attributes if a pending entry is found
8043 in inline_entry_data_table. Add old entry_pc attribute only
8044 if debug nonbinding markers are disabled.
8045 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
8046 markers are enabled.
8047 (block_within_block_p, dwarf2out_inline_entry): New.
8048 (dwarf2out_finish): Check that no entries remained in
8049 inline_entry_data_table.
8050 * final.c (reemit_insn_block_notes): Handle inline entry notes.
8051 (final_scan_insn, notice_source_line): Likewise.
8052 (rest_of_clean_state): Skip inline entry markers.
8053 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
8055 * gimple.c (gimple_build_debug_inline_entry): New.
8056 * gimple.h (enum gimple_debug_subcode): Add
8057 GIMPLE_DEBUG_INLINE_ENTRY.
8058 (gimple_build_debug_inline_entry): Declare.
8059 (gimple_debug_inline_entry_p): New.
8060 (gimple_debug_nonbind_marker_p): Adjust.
8061 * insn-notes.def (INLINE_ENTRY): New.
8062 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
8063 inline entry marker notes.
8064 (print_insn): Likewise.
8065 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
8066 (INSN_DEBUG_MARKER_KIND): Likewise.
8067 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
8068 * tree-inline.c (expand_call_inline): Build and insert
8069 debug_inline_entry stmt.
8070 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
8071 inline entry blocks early, if nonbind markers are enabled.
8072 (dump_scope_block): Dump fragment info.
8073 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
8074 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
8075 (gimple_build_debug_inline_entry): New.
8076 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
8077 Enable/disable inline entry points too.
8078 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
8079 (DEBUG_INSN): Describe inline entry markers.
8081 * common.opt (gvariable-location-views): New.
8082 (gvariable-location-views=incompat5): New.
8083 * config.in: Rebuilt.
8084 * configure: Rebuilt.
8085 * configure.ac: Test assembler for view support.
8086 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
8087 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
8088 * dwarf2out.c (var_loc_view): New typedef.
8089 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
8090 (dwarf2out_locviews_in_attribute): New.
8091 (dwarf2out_locviews_in_loclist): New.
8092 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
8093 (enum dw_line_info_opcode): Add LI_adv_address.
8094 (struct dw_line_info_table): Add view.
8095 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
8096 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
8097 (zero_view_p): New variable.
8098 (ZERO_VIEW_P): New macro.
8099 (output_asm_line_debug_info): New.
8100 (struct var_loc_node): Add view.
8101 (add_AT_view_list, AT_loc_list): New.
8102 (add_var_loc_to_decl): Add view param. Test it against last.
8103 (new_loc_list): Add view params. Record them.
8104 (AT_loc_list_ptr): Handle loc and view lists.
8105 (view_list_to_loc_list_val_node): New.
8106 (print_dw_val): Handle dw_val_class_view_list.
8107 (size_of_die): Likewise.
8108 (value_format): Likewise.
8109 (loc_list_has_views): New.
8110 (gen_llsym): Set vl_symbol too.
8111 (maybe_gen_llsym, skip_loc_list_entry): New.
8112 (dwarf2out_maybe_output_loclist_view_pair): New.
8113 (output_loc_list): Output view list or entries too.
8114 (output_view_list_offset): New.
8115 (output_die): Handle dw_val_class_view_list.
8116 (output_dwarf_version): New.
8117 (output_compilation_unit_header): Use it.
8118 (output_skeleton_debug_sections): Likewise.
8119 (output_rnglists, output_line_info): Likewise.
8120 (output_pubnames, output_aranges): Update version comments.
8121 (output_one_line_info_table): Output view numbers in asm comments.
8122 (dw_loc_list): Determine current endview, pass it to new_loc_list.
8123 Call maybe_gen_llsym.
8124 (loc_list_from_tree_1): Adjust.
8125 (add_AT_location_description): Create view list attribute if
8126 needed, check it's absent otherwise.
8127 (convert_cfa_to_fb_loc_list): Adjust.
8128 (maybe_emit_file): Call output_asm_line_debug_info for test.
8129 (dwarf2out_var_location): Reset views as needed. Precompute
8130 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
8131 attribute. Set view.
8132 (new_line_info_table): Reset next view.
8133 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
8134 (dwarf2out_source_line): Likewise. Output view resets and labels to
8135 the assembler, or select appropriate line info opcodes.
8136 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
8137 (optimize_string_length): Catch it. Adjust.
8138 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
8139 dw_val_class_view_list, and remove it if no longer needed.
8140 (hash_loc_list): Hash view numbers.
8141 (loc_list_hasher::equal): Compare them.
8142 (optimize_location_lists): Check whether a view list symbol is
8143 needed, and whether the locview attribute is present, and
8144 whether they match. Remove the locview attribute if no longer
8146 (index_location_lists): Call skip_loc_list_entry for test.
8147 (dwarf2out_finish): Call output_asm_line_debug_info for test.
8148 Use output_dwarf_version.
8149 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
8150 (struct dw_val_node): Add val_view_list.
8151 * final.c (SEEN_NEXT_VIEW): New.
8152 (set_next_view_needed): New.
8153 (clear_next_view_needed): New.
8154 (maybe_output_next_view): New.
8155 (final_start_function): Rename to...
8156 (final_start_function_1): ... this. Take pointer to FIRST,
8157 add SEEN parameter. Emit param bindings in the initial view.
8158 (final_start_function): Reintroduce SEEN-less interface.
8159 (final): Rename to...
8160 (final_1): ... this. Take SEEN parameter. Output final pending
8161 next view at the end.
8162 (final): Reintroduce seen-less interface.
8163 (final_scan_insn): Output pending next view before switching
8164 sections or ending a block. Mark the next view as needed when
8165 outputting variable locations. Notify debug backend of section
8166 changes, and of location view changes.
8167 (rest_of_handle_final): Adjust.
8168 * toplev.c (process_options): Autodetect value for debug variable
8169 location views option. Warn on incompat5 without -gdwarf-5.
8170 * doc/invoke.texi (gvariable-location-views): New.
8171 (gvariable-location-views=incompat5): New.
8172 (gno-variable-location-views): New.
8174 2018-02-08 David Malcolm <dmalcolm@redhat.com>
8176 PR tree-optimization/84136
8177 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
8178 that the result of find_edge is non-NULL.
8180 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8183 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
8184 storing integer register in SImode. Fix cost of 256 and 512
8185 byte aligned SSE register store.
8187 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8189 * config/i386/i386.c (ix86_multiplication_cost): Fix
8190 multiplication cost for TARGET_AVX512DQ.
8192 2018-02-08 Marek Polacek <polacek@redhat.com>
8194 PR tree-optimization/84238
8195 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
8198 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8200 PR tree-optimization/84265
8201 * tree-vect-stmts.c (vectorizable_store): Don't treat
8202 VMAT_CONTIGUOUS accesses as grouped.
8203 (vectorizable_load): Likewise.
8205 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8207 PR tree-optimization/81635
8208 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
8209 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
8210 (test_round_for_mask): New functions.
8211 (wide_int_cc_tests): Call test_round_for_mask.
8212 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
8213 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
8214 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
8215 range returned by get_range_info.
8217 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8220 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
8221 * symtab.c: Include builtins.h
8222 (symtab_node::output_to_lto_symbol_table_p): Move here
8223 from lto-streamer-out.c:output_symbol_p.
8224 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
8225 (output_symbol_p): Move all logic to symtab.c
8226 (produce_symtab): Update.
8228 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8230 * config/s390/s390-opts.h (enum indirect_branch): Define.
8231 * config/s390/s390-protos.h (s390_return_addr_from_memory)
8232 (s390_indirect_branch_via_thunk)
8233 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
8234 (enum s390_indirect_branch_type): Define.
8235 * config/s390/s390.c (struct s390_frame_layout, struct
8236 machine_function): Remove.
8237 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
8238 (indirect_branch_table_label_no, indirect_branch_table_name):
8240 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
8241 (enum s390_indirect_branch_option): Define.
8242 (s390_return_addr_from_memory): New function.
8243 (s390_handle_string_attribute): New function.
8244 (s390_attribute_table): Add new attribute handler.
8245 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
8246 (s390_indirect_branch_via_thunk): New function.
8247 (s390_indirect_branch_via_inline_thunk): New function.
8248 (s390_function_ok_for_sibcall): When jumping via thunk disallow
8249 sibling call optimization for non z10 compiles.
8250 (s390_emit_call): Force indirect branch target to be a single
8251 register. Add r1 clobber for non-z10 compiles.
8252 (s390_emit_epilogue): Emit return jump via return_use expander.
8253 (s390_reorg): Handle JUMP_INSNs as execute targets.
8254 (s390_option_override_internal): Perform validity checks for the
8255 new command line options.
8256 (s390_indirect_branch_attrvalue): New function.
8257 (s390_indirect_branch_settings): New function.
8258 (s390_set_current_function): Invoke s390_indirect_branch_settings.
8259 (s390_output_indirect_thunk_function): New function.
8260 (s390_code_end): Implement target hook.
8261 (s390_case_values_threshold): Implement target hook.
8262 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
8264 * config/s390/s390.h (struct s390_frame_layout)
8265 (struct machine_function): Move here from s390.c.
8266 (TARGET_INDIRECT_BRANCH_NOBP_RET)
8267 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
8268 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
8269 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
8270 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
8271 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
8272 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
8273 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
8274 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
8275 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
8276 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
8277 (mnemonic attribute): Add values which aren't recognized
8279 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
8280 pattern for branch conversion. Fix mnemonic attribute.
8281 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
8282 indirect branch via thunk if requested.
8283 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
8284 ("*indirect_jump"): Disable for branch conversion using out of
8286 ("indirect_jump_via_thunk<mode>_z10")
8287 ("indirect_jump_via_thunk<mode>")
8288 ("indirect_jump_via_inlinethunk<mode>_z10")
8289 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
8290 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
8291 ("casesi_jump_via_inlinethunk<mode>_z10")
8292 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
8293 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
8294 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
8295 ("*indirect2_jump"): Disable for branch conversion.
8296 ("casesi_jump"): Turn into expander and expand patterns for branch
8298 ("return_use"): New expander.
8299 ("*return"): Emit return via thunk and rename it to ...
8300 ("*return<mode>"): ... this one.
8301 * config/s390/s390.opt: Add new options and and enum for the
8304 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8306 * lra-constraints.c (match_reload): Unconditionally use
8307 gen_lowpart_SUBREG, rather than selecting between that
8308 and equivalent gen_rtx_SUBREG code.
8310 2018-02-08 Richard Biener <rguenther@suse.de>
8312 PR tree-optimization/84233
8313 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
8314 changed flag instead of boguously re-using phi_inserted.
8316 2018-02-08 Martin Jambor <mjambor@suse.cz>
8318 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
8319 static local variables.
8321 2018-02-08 Richard Biener <rguenther@suse.de>
8323 PR tree-optimization/84278
8324 * tree-vect-stmts.c (vectorizable_store): When looking for
8325 smaller vector types to perform grouped strided loads/stores
8326 make sure the mode is supported by the target.
8327 (vectorizable_load): Likewise.
8329 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8331 * config/aarch64/aarch64.c (aarch64_components_for_bb):
8332 Increase LDP/STP opportunities by adding adjacent callee-saves.
8334 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8336 PR rtl-optimization/84068
8337 PR rtl-optimization/83459
8338 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
8340 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
8342 PR tree-optimization/84224
8343 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
8344 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
8347 2018-02-07 Iain Sandoe <iain@codesourcery.com>
8350 * config/rs6000/altivec.md (*restore_world): Remove LR use.
8351 * config/rs6000/predicates.md (restore_world_operation): Adjust op
8352 count, remove one USE.
8354 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
8356 * doc/install.texi (Configuration): Document the
8357 --with-long-double-format={ibm,ieee} PowerPC configuration
8361 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
8362 Convert from define_expand to be define_insn_and_split. Rework
8363 float/double/_Float128 conversions to QI/HI/SImode to work with
8364 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
8365 conversions to QI/HImode types did a store and then a load to
8366 truncate the value. For conversions to VSX registers, don't split
8367 the insn, instead emit the code directly. Use the code iterator
8368 any_fix to combine signed and unsigned conversions.
8369 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
8370 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
8371 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8372 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
8373 (fix_<mode>di2_hw): Likewise.
8374 (fixuns_<mode>di2_hw): Likewise.
8375 (fix_<mode>si2_hw): Likewise.
8376 (fixuns_<mode>si2_hw): Likewise.
8377 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
8378 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8379 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
8380 fix<uns>_trunc<SFDF:mode>si2_p8.
8381 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
8383 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
8384 (fix<uns>_<mode>_mem): Likewise.
8385 (fctiw<u>z_<mode>_mem): Likewise.
8386 (fix<uns>_<mode>_mem): Likewise.
8387 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
8388 the register allocator from doing a direct move to the GPRs to do
8389 a store, and instead use the ISA 3.0 store byte/half-word from
8390 vector register instruction. For IEEE 128-bit floating point,
8391 also optimize stores of 32-bit ints.
8392 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
8394 2018-02-07 Alan Hayward <alan.hayward@arm.com>
8396 * genextract.c (push_pathstr_operand): New function to support
8398 (walk_rtx): Call push_pathstr_operand.
8399 (print_path): Support [a-zA-Z].
8401 2018-02-07 Richard Biener <rguenther@suse.de>
8403 PR tree-optimization/84037
8404 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
8405 (cse_and_gimplify_to_preheader): Declare.
8406 (vect_get_place_in_interleaving_chain): Likewise.
8407 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8409 (_loop_vec_info::~_loop_vec_info): Delete it.
8410 (cse_and_gimplify_to_preheader): New function.
8411 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
8412 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
8413 (vectorizable_load): Likewise. For grouped stores always base
8414 the IV on the first element.
8415 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
8416 condition before gimplifying.
8418 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8420 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
8421 *DIV_EXPR and *MOD_EXPR.
8423 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
8426 * config/i386/i386.c (ix86_option_override_internal): Mask out
8427 the CF_SET bit when checking -fcf-protection.
8429 2018-02-07 Tom de Vries <tom@codesourcery.com>
8432 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
8435 2018-02-07 Richard Biener <rguenther@suse.de>
8437 PR tree-optimization/84204
8438 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
8441 PR tree-optimization/84205
8442 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
8443 special-case isl_ast_op_zdiv_r.
8445 PR tree-optimization/84223
8446 * graphite-scop-detection.c (gather_bbs::before_dom_children):
8447 Only add conditions from within the region.
8448 (gather_bbs::after_dom_children): Adjust.
8450 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
8453 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
8454 * config/avr/avr.md: Only post-reload split REG-REG moves if
8455 either register is GENERAL_REG_P.
8457 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8459 PR tree-optimization/84235
8460 * tree-ssa-scopedtables.c
8461 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
8462 if the subtraction is performed in floating point type where NaNs are
8463 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
8464 build 1. Formatting fix.
8466 2018-02-06 Jakub Jelinek <jakub@redhat.com>
8469 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
8470 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
8471 and skip it regardless of bb boundaries. Use CALL_P macro,
8472 don't test INSN_P (insn) together with CALL_P or JUMP_P check
8473 unnecessarily, formatting fix.
8475 2018-02-06 Michael Collison <michael.collison@arm.com>
8477 * config/arm/thumb2.md:
8478 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
8479 (*thumb_mov_notscc): Ditto.
8481 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
8484 * config/rs6000/rs6000.md (su code attribute): Use "u" for
8485 unsigned_fix, not "s".
8487 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8489 * configure.ac (gcc_fn_eh_frame_ro): New function.
8490 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
8491 correct .eh_frame permissions.
8492 * configure: Regenerate.
8494 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
8496 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
8499 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8501 * config/rs6000/rs6000.c (rs6000_option_override_internal):
8502 Display warning message for -mno-speculate-indirect-jumps.
8504 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
8506 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
8508 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
8510 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
8512 PR tree-optimization/84225
8513 * tree-eh.c (find_trapping_overflow): Only call
8514 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
8516 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8519 * config/i386/i386.c: Reimplement the check of possible options
8520 -mibt/-mshstk conbination. Change error messages.
8521 * doc/invoke.texi: Fix a typo: remove extra '='.
8523 2018-02-06 Marek Polacek <polacek@redhat.com>
8525 PR tree-optimization/84228
8526 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
8528 2018-02-06 Tamar Christina <tamar.christina@arm.com>
8531 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
8532 emitted arch directives.
8533 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
8534 __ARM_FEATURE_COPROC before changing architectures.
8536 2018-02-06 Richard Biener <rguenther@suse.de>
8538 * config/i386/i386.c (print_reg): Fix typo.
8539 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
8541 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
8543 * configure: Regenerate.
8545 2018-02-05 Martin Sebor <msebor@redhat.com>
8547 PR tree-optimization/83369
8548 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
8551 2018-02-05 Martin Liska <mliska@suse.cz>
8553 * doc/invoke.texi: Cherry-pick upstream r323995.
8555 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
8557 * ira.c (ira_init_register_move_cost): Adjust comment.
8559 2018-02-05 Martin Liska <mliska@suse.cz>
8561 PR gcov-profile/84137
8562 * doc/gcov.texi: Fix typo in documentation.
8564 2018-02-05 Martin Liska <mliska@suse.cz>
8566 PR gcov-profile/83879
8567 * doc/gcov.texi: Document necessity of --dynamic-list-data when
8568 using dlopen functionality.
8570 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
8572 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
8573 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
8574 _mm_maskz_range_ss, _mm_mask_range_round_ss,
8575 _mm_maskz_range_round_ss): New intrinsics.
8576 (__builtin_ia32_rangesd128_round)
8577 (__builtin_ia32_rangess128_round): Remove.
8578 (__builtin_ia32_rangesd128_mask_round,
8579 __builtin_ia32_rangess128_mask_round): New builtins.
8580 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
8581 __builtin_ia32_rangess128_round): Remove.
8582 (__builtin_ia32_rangesd128_mask_round,
8583 __builtin_ia32_rangess128_mask_round): New builtins.
8584 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
8585 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
8586 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
8587 "<round_saeonly_constraint>")): Changed to ...
8588 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
8589 "<round_saeonly_scalar_constraint>")): ... this.
8590 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
8591 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
8592 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
8593 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
8594 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
8596 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
8598 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
8600 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
8601 Remove all values except native, 8540 and 8548.
8603 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
8605 * config/i386/i386.c (ix86_output_function_return): Pass
8606 INVALID_REGNUM, instead of -1, as invalid register number to
8607 indirect_thunk_name and output_indirect_thunk.
8609 2018-02-02 Julia Koval <julia.koval@intel.com>
8611 * config.gcc: Add -march=icelake.
8612 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
8613 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
8614 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
8615 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
8616 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
8617 (processor_target_table): Add icelake.
8618 (ix86_option_override_internal): Handle new PTAs.
8619 (get_builtin_code_for_version): Handle icelake.
8620 (M_INTEL_COREI7_ICELAKE): New.
8621 (fold_builtin_cpu): Handle icelake.
8622 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
8623 * doc/invoke.texi: Add -march=icelake.
8625 2018-02-02 Julia Koval <julia.koval@intel.com>
8627 * config/i386/i386.c (ix86_option_override_internal): Change flags type
8628 to wide_int_bitmask.
8629 * wide-int-bitmask.h: New.
8631 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8634 * config/i386/i386.md: Replace Pmode with word_mode in
8635 builtin_setjmp_setup and builtin_longjmp to support x32.
8637 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
8641 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
8643 (rs6000_supported_cpu_names): New static variable.
8644 (linux_cpu_translation_table): Likewise.
8645 (elf_platform) <cpu>: Define new static variable and use it.
8646 Translate kernel AT_PLATFORM name to canonical name if needed.
8647 Error if platform name is unknown.
8649 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
8652 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
8654 2018-02-01 Jeff Law <law@redhat.com>
8657 * config/i386/i386.c (release_scratch_register_on_entry): Add new
8658 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
8659 the scratch if RELEASE_VIA_POP is false.
8660 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
8661 If we have to save a temporary register, decrement SIZE appropriately.
8662 Pass new arguments to release_scratch_register_on_entry.
8663 (ix86_adjust_stack_and_probe): Likewise.
8664 (ix86_emit_probe_stack_range): Pass new arguments to
8665 release_scratch_register_on_entry.
8667 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
8669 PR rtl-optimization/84157
8670 * combine.c (change_zero_ext): Use REG_P predicate in
8671 front of HARD_REGISTER_P predicate.
8673 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
8675 * config/avr/avr.c (avr_option_override): Move disabling of
8676 -fdelete-null-pointer-checks to...
8677 * common/config/avr/avr-common.c (avr_option_optimization_table):
8680 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8682 PR tree-optimization/81635
8683 * tree-data-ref.c (split_constant_offset_1): For types that
8684 wrap on overflow, try to use range info to prove that wrapping
8687 2018-02-01 Renlin Li <renlin.li@arm.com>
8690 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
8692 (aarch64_register_move_cost): Likewise.
8693 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
8695 (REG_CLASS_NAMES): Likewise.
8696 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
8697 TAILCALL_ADDR_REGS. Remove IP registers.
8698 * config/aarch64/aarch64.md (Ucs): Update register constraint.
8700 2018-02-01 Richard Biener <rguenther@suse.de>
8702 * domwalk.h (dom_walker::dom_walker): Add additional constructor
8703 for specifying RPO order and allow NULL for that.
8704 * domwalk.c (dom_walker::dom_walker): Likewise.
8705 (dom_walker::walk): Handle NULL RPO order.
8706 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
8708 (rewrite_update_dom_walker): Likewise.
8709 (mark_def_dom_walker): Likewise.
8711 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8713 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
8714 (aarch64_maybe_expand_sve_subreg_move): Declare.
8715 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
8716 * config/aarch64/predicates.md (aarch64_any_register_operand): New
8718 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
8719 that are semantically a reverse operation.
8720 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
8721 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
8722 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
8724 (aarch64_can_change_mode_class): For big-endian, forbid changes
8725 between two SVE modes if they have different element sizes.
8727 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8729 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
8730 the TImode handling for big-endian targets.
8732 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8734 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
8735 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
8737 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
8738 Remove BSWAP handing for big-endian targets and use the form of
8739 LD1RQ appropariate for the mode.
8741 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8743 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
8744 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
8747 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8750 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
8751 check for operands that need to go through aarch64_sve_reload_be.
8753 2018-02-01 Jakub Jelinek <jakub@redhat.com>
8755 PR tree-optimization/81661
8756 PR tree-optimization/84117
8757 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
8758 * tree-eh.c: Include gimplify.h.
8759 (find_trapping_overflow, replace_trapping_overflow,
8760 rewrite_to_non_trapping_overflow): New functions.
8761 * tree-vect-loop.c: Include tree-eh.h.
8762 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
8763 * tree-data-ref.c: Include tree-eh.h.
8764 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
8766 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
8768 PR rtl-optimization/84123
8769 * combine.c (change_zero_ext): Check if hard register satisfies
8770 can_change_dest_mode before calling gen_lowpart_SUBREG.
8772 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
8775 * ira.c (ira_init_register_move_cost): Remove assert.
8777 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8779 PR rtl-optimization/84071
8780 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
8781 * doc/tm.texi: Regenerate.
8783 2018-01-31 Richard Biener <rguenther@suse.de>
8785 PR tree-optimization/84132
8786 * tree-data-ref.c (analyze_miv_subscript): Properly
8787 check whether evolution_function_is_affine_multivariate_p
8788 before calling gcd_of_steps_may_divide_p.
8790 2018-01-31 Julia Koval <julia.koval@intel.com>
8793 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
8794 * config/i386/i386.md (rdpid_rex64) New.
8795 (rdpid): Make 32bit only.
8797 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
8800 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
8801 an IDENTIFIER_NODE for FUNCTION_TYPE's.
8803 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8806 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8808 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8810 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8812 PR rtl-optimization/84071
8813 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
8814 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
8816 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8818 * config/arc/arc.c (arc_handle_aux_attribute): New function.
8819 (arc_attribute_table): Add 'aux' attribute.
8820 (arc_in_small_data_p): Consider aux like variables.
8821 (arc_is_aux_reg_p): New function.
8822 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
8823 (arc_get_aux_arg): New function.
8824 (prepare_move_operands): Handle aux-register access.
8825 (arc_handle_aux_attribute): New function.
8826 * doc/extend.texi (ARC Variable attributes): Add subsection.
8828 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8830 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
8831 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
8832 (arc_attribute_table): Add 'uncached' attribute.
8833 (arc_print_operand): Print '.di' flag for uncached memory
8835 (arc_in_small_data_p): Do not consider for small data the uncached
8837 (arc_is_uncached_mem_p): New function.
8838 * config/arc/predicates.md (compact_store_memory_operand): Check
8839 for uncached memory accesses.
8840 (nonvol_nonimm_operand): Likewise.
8841 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
8843 2018-01-31 Jakub Jelinek <jakub@redhat.com>
8846 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
8847 falign-loops=): Add Optimization flag.
8849 2018-01-30 Jeff Law <law@redhat.com>
8852 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
8853 INT_REGISTERS_SAVED. Check it prior to calling
8854 get_scratch_register_on_entry.
8855 (ix86_adjust_stack_and_probe): Similarly.
8856 (ix86_emit_probe_stack_range): Similarly.
8857 (ix86_expand_prologue): Corresponding changes.
8859 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8862 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
8863 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
8865 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
8868 * lra-constraints.c (curr_insn_transform): Process AND in the
8871 2018-01-30 Jakub Jelinek <jakub@redhat.com>
8873 PR rtl-optimization/83986
8874 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
8875 dependence against last_pending_memory_flush in addition to
8878 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
8880 PR tree-optimization/81611
8881 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
8884 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8887 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
8890 2018-01-30 Richard Biener <rguenther@suse.de>
8891 Jakub Jelinek <jakub@redhat.com>
8893 PR tree-optimization/84111
8894 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
8895 inner loops added during recursion, as they don't have up-to-date
8898 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8901 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
8902 (can_inline_edge_by_limits_p): ... here.
8903 (can_early_inline_edge_p, check_callers,
8904 update_caller_keys, update_callee_keys, recursive_inlining,
8905 add_new_edges_to_heap, speculation_useful_p,
8906 inline_small_functions,
8907 inline_small_functions, flatten_function,
8908 inline_to_all_callers_1): Update.
8910 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8912 * profile-count.c (profile_count::combine_with_ipa_count): Handle
8915 2018-01-30 Richard Biener <rguenther@suse.de>
8917 PR tree-optimization/83008
8918 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
8919 invariant and constant vector uses in stmts when they need
8922 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8925 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
8926 * configure: Regenerate.
8928 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8930 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
8932 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
8933 Use gen_rtx_REG rather than gen_lowpart.
8935 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8937 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
8938 rather than 0 when creating partial subregs.
8940 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8942 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
8945 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
8948 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
8949 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
8950 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
8951 flags. This restores the settings used before the 2017-07-24.
8952 Turning off pre increment/decrement/modify allows IVOPTS to
8953 optimize DF/SF loops where the index is an int.
8955 2018-01-29 Richard Biener <rguenther@suse.de>
8956 Kelvin Nilsen <kelvin@gcc.gnu.org>
8959 * tree-vect-stmts.c (vectorizable_call): Don't call
8960 targetm.vectorize_builtin_md_vectorized_function if callee is
8963 2018-01-22 Carl Love <cel@us.ibm.com>
8965 * doc/extend.tex: Fix typo in second arg in
8966 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
8968 2018-01-29 Richard Biener <rguenther@suse.de>
8970 PR tree-optimization/84086
8971 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
8972 (flush_ssaname_freelist): When SSA names were released reset
8973 the SCEV hash table.
8975 2018-01-29 Richard Biener <rguenther@suse.de>
8977 PR tree-optimization/84057
8978 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
8979 removed paths when removing edges.
8981 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
8983 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
8984 -mfunction-return=@var{choice}.
8986 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
8989 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
8991 (layout::print_source_line): Likewise.
8992 (test_get_line_width_without_trailing_whitespace): Add test cases.
8994 2018-01-27 Jakub Jelinek <jakub@redhat.com>
8997 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
9000 2018-01-26 Jim Wilson <jimw@sifive.com>
9002 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
9004 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
9007 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9009 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
9010 and CMP + SUB-immediate -> SUBS.
9012 2018-01-26 Martin Sebor <msebor@redhat.com>
9014 PR tree-optimization/83896
9015 * tree-ssa-strlen.c (get_string_len): Rename...
9016 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
9017 Avoid assuming length is constant.
9018 (handle_char_store): Use HOST_WIDE_INT for string length.
9020 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
9023 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
9024 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
9026 2018-01-26 Richard Biener <rguenther@suse.de>
9028 PR rtl-optimization/84003
9029 * dse.c (record_store): Only record redundant stores when
9030 the earlier store aliases at least all accesses the later one does.
9032 2018-01-26 Jakub Jelinek <jakub@redhat.com>
9034 PR rtl-optimization/83985
9035 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
9036 REG_CFA_RESTORE insns.
9037 (delete_unmarked_insns): Don't ignore separate shrink wrapping
9038 REG_CFA_RESTORE insns here.
9041 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
9042 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
9044 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9046 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
9047 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
9048 (arc_init): Likewise.
9049 (arc_override_options): Likewise.
9050 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
9052 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
9054 * config/arc/arc.h (TARGET_DBNZ): Define.
9055 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
9056 properly set the tune attribute.
9057 (dbnz): Use TARGET_DBNZ guard.
9058 * config/arc/arc.opt (mtune): Add core3 option.
9060 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9062 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
9063 recognize new pic like addresses.
9064 (arc_delegitimize_address): Clean up.
9066 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9068 * config/arc/arc-arches.def: Option mrf16 valid for all
9070 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
9071 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
9072 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
9073 * config/arc/arc-tables.opt: Regenerate.
9074 * config/arc/arc.c (arc_conditional_register_usage): Handle
9075 reduced register file case.
9076 (arc_file_start): Set must have build attributes.
9077 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
9079 * config/arc/arc.opt (mrf16): Add new option.
9080 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
9081 * config/arc/genmultilib.awk: Handle new mrf16 option.
9082 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
9083 * config/arc/t-multilib: Regenerate.
9084 * doc/invoke.texi (ARC Options): Document mrf16 option.
9086 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9088 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
9089 * config/arc/arc.c (arc_handle_secure_attribute): New function.
9090 (arc_attribute_table): Add 'secure_call' attribute.
9091 (arc_print_operand): Print secure call operand.
9092 (arc_function_ok_for_sibcall): Don't optimize tail calls when
9094 (arc_is_secure_call_p): New function. * config/arc/arc.md
9095 (call_i): Add support for sjli instruction.
9096 (call_value_i): Likewise.
9097 * config/arc/constraints.md (Csc): New constraint.
9099 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9100 John Eric Martin <John.Martin@emmicro-us.com>
9102 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
9103 * config/arc/arc.c (_arc_jli_section): New struct.
9104 (arc_jli_section): New type.
9105 (rc_jli_sections): New static variable.
9106 (arc_handle_jli_attribute): New function.
9107 (arc_attribute_table): Add jli_always and jli_fixed attribute.
9108 (arc_file_end): New function.
9109 (TARGET_ASM_FILE_END): Define.
9110 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
9111 (arc_add_jli_section): New function.
9112 (jli_call_scan): Likewise.
9113 (arc_reorg): Call jli_call_scan.
9114 (arc_output_addsi): Remove 'S' from printing asm operand.
9115 (arc_is_jli_call_p): New function.
9116 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
9118 (movhi_insn): Likewise.
9119 (movsi_insn): Likewise.
9120 (movsi_set_cc_insn): Likewise.
9121 (loadqi_update): Likewise.
9122 (load_zeroextendqisi_update): Likewise.
9123 (load_signextendqisi_update): Likewise.
9124 (loadhi_update): Likewise.
9125 (load_zeroextendhisi_update): Likewise.
9126 (load_signextendhisi_update): Likewise.
9127 (loadsi_update): Likewise.
9128 (loadsf_update): Likewise.
9129 (movsicc_insn): Likewise.
9130 (bset_insn): Likewise.
9131 (bxor_insn): Likewise.
9132 (bclr_insn): Likewise.
9133 (bmsk_insn): Likewise.
9134 (bicsi3_insn): Likewise.
9135 (cmpsi_cc_c_insn): Likewise.
9136 (movsi_ne): Likewise.
9137 (movsi_cond_exec): Likewise.
9138 (clrsbsi2): Likewise.
9149 (call_i): Remove 'S' asm letter, add jli instruction.
9150 (call_value_i): Likewise.
9151 * config/arc/arc.op (mjli-always): New option.
9152 * config/arc/constraints.md (Cji): New constraint.
9153 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
9155 (subsf3_fpx): Likewise.
9156 (mulsf3_fpx): Likewise.
9157 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
9159 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
9161 * doc/invoke.texi (ARC): Document mjli-always option.
9163 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
9165 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
9166 avoid addition with 0 and use incw and decw where possible.
9168 2018-01-26 Richard Biener <rguenther@suse.de>
9170 PR tree-optimization/81082
9171 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
9172 association if it requires casting to unsigned.
9173 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
9174 from fold_plusminus_mult_expr to catch important cases late when
9175 range info is available.
9177 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9179 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
9180 * configure.ac (hidden_linkonce): New test.
9181 * configure: Regenerate.
9182 * config.in: Regenerate.
9184 2018-01-26 Julia Koval <julia.koval@intel.com>
9186 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
9187 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
9188 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
9189 _mm_mask_bitshuffle_epi64_mask): Fix type.
9190 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
9191 USI_FTYPE_V4DI_V4DI_USI): Remove.
9192 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
9193 __builtin_ia32_vpshufbitqmb256_mask,
9194 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
9195 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
9196 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
9198 2018-01-26 Alan Modra <amodra@gmail.com>
9201 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
9202 UNSPEC_VBPERMQ. Sort other unspecs.
9204 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
9206 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
9208 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
9211 * predict.c (drop_profile): Do not push/pop cfun; update also
9213 (handle_missing_profiles): Fix logic looking for zero profiles.
9215 2018-01-25 Jakub Jelinek <jakub@redhat.com>
9218 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
9219 on functions with #pragma omp declare simd or functions with simd
9221 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
9222 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
9223 Remove trailing \n from warning_at calls.
9225 2018-01-25 Tom de Vries <tom@codesourcery.com>
9228 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9229 for neutered workers.
9231 2018-01-24 Joseph Myers <joseph@codesourcery.com>
9234 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
9235 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
9237 2018-01-24 Jeff Law <law@redhat.com>
9240 * i386.c (get_probe_interval): Move to earlier point.
9241 (ix86_compute_frame_layout): If -fstack-clash-protection and
9242 the frame is larger than the probe interval, then use pushes
9243 to save registers rather than reg->mem moves.
9244 (ix86_expand_prologue): Remove conditional for int_registers_saved
9247 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
9250 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
9251 min/max for never referenced object.
9253 2018-01-24 Jakub Jelinek <jakub@redhat.com>
9256 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
9258 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
9259 attributes from DECL_ATTRIBUTES (decl) without affecting
9260 DECL_ATTRIBUTES (current_function_decl).
9261 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
9262 functions with non-NULL DECL_ABSTRACT_ORIGIN.
9264 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
9266 PR tree-optimization/83979
9267 * fold-const.c (fold_comparison): Use constant_boolean_node
9268 instead of boolean_{true,false}_node.
9270 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
9272 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
9275 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9277 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9278 Simplify the clause that sets the length attribute.
9279 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9280 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
9281 clause that sets the length attribute.
9282 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9284 2018-01-24 Tom de Vries <tom@codesourcery.com>
9287 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
9288 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
9289 Add strict parameter.
9290 (prevent_branch_around_nothing): Insert dummy insn between branch to
9291 label and label with no ptx insn inbetween.
9292 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
9294 2018-01-24 Tom de Vries <tom@codesourcery.com>
9297 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9298 for neutered threads in warp.
9299 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
9301 2018-01-24 Richard Biener <rguenther@suse.de>
9303 PR tree-optimization/83176
9304 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
9307 2018-01-24 Richard Biener <rguenther@suse.de>
9309 PR tree-optimization/82819
9310 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
9311 code generating pluses that are no-ops in the target precision.
9313 2018-01-24 Richard Biener <rguenther@suse.de>
9316 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
9318 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9320 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
9321 to merge probabilities.
9322 * predict.c (probably_never_executed): Also mark as cold functions
9323 with global 0 profile and guessed local profile.
9324 * profile-count.c (profile_probability::combine_with_count): New
9326 * profile-count.h (profile_probability::operator*,
9327 profile_probability::operator*=, profile_probability::operator/,
9328 profile_probability::operator/=): Reduce precision to adjusted
9329 and set value to guessed on contradictory divisions.
9330 (profile_probability::combine_with_freq): Remove.
9331 (profile_probability::combine_wiht_count): Declare.
9332 (profile_count::force_nonzero):: Set to adjusted.
9333 (profile_count::probability_in):: Set quality to adjusted.
9334 * tree-ssa-tail-merge.c (replace_block_by): Use
9337 2018-01-23 Andrew Waterman <andrew@sifive.com>
9338 Jim Wilson <jimw@sifive.com>
9340 * config/riscv/riscv.c (riscv_stack_boundary): New.
9341 (riscv_option_override): Set riscv_stack_boundary. Handle
9342 riscv_preferred_stack_boundary_arg.
9343 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
9344 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
9345 (STACK_BOUNDARY): Set to riscv_stack_boundary.
9346 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
9347 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
9348 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
9350 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
9353 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
9354 of struct ix86_frame.
9355 (ix86_expand_epilogue): Likewise. Add a local variable for
9356 the reg_save_offset field in struct ix86_frame.
9358 2018-01-23 Bin Cheng <bin.cheng@arm.com>
9360 PR tree-optimization/82604
9361 * tree-loop-distribution.c (enum partition_kind): New enum item
9362 PKIND_PARTIAL_MEMSET.
9363 (partition_builtin_p): Support above new enum item.
9364 (generate_code_for_partition): Ditto.
9365 (compute_access_range): Differentiate cases that equality can be
9366 proven at all loops, the innermost loops or no loops.
9367 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
9368 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
9369 (finalize_partitions, distribute_loop): Don't fuse partition of
9370 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
9371 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
9374 2018-01-23 Martin Liska <mliska@suse.cz>
9376 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
9377 order to ignore the predictor.
9378 (PRED_POLYMORPHIC_CALL): Likewise.
9379 (PRED_RECURSIVE_CALL): Likewise.
9381 2018-01-23 Martin Liska <mliska@suse.cz>
9383 * tree-profile.c (tree_profiling): Print function header to
9384 aware reader which function we are working on.
9385 * value-prof.c (gimple_find_values_to_profile): Do not print
9386 not interesting value histograms.
9388 2018-01-23 Martin Liska <mliska@suse.cz>
9390 * profile-count.h (enum profile_quality): Add
9391 profile_uninitialized as the first value. Do not number values
9392 as they are zero based.
9393 (profile_count::verify): Update sanity check.
9394 (profile_probability::verify): Likewise.
9396 2018-01-23 Nathan Sidwell <nathan@acm.org>
9398 * doc/invoke.texi (ffor-scope): Deprecate.
9400 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9402 PR tree-optimization/83510
9403 * domwalk.c (set_all_edges_as_executable): New function.
9404 (dom_walker::dom_walker): Convert bool param
9405 "skip_unreachable_blocks" to enum reachability. Move setup of
9406 edge flags to set_all_edges_as_executable and only do it when
9407 reachability is REACHABLE_BLOCKS.
9408 * domwalk.h (enum dom_walker::reachability): New enum.
9409 (dom_walker::dom_walker): Convert bool param
9410 "skip_unreachable_blocks" to enum reachability.
9411 (set_all_edges_as_executable): New decl.
9412 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
9413 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
9415 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
9416 but converting true to REACHABLE_BLOCKS.
9417 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
9419 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
9420 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
9421 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
9423 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
9424 if check_all_array_refs will be called.
9426 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9428 * tree.c (selftest::test_location_wrappers): Add more test
9431 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9433 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
9434 (selftest::test_bit_in_range): Likewise.
9436 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9439 * doc/sourcebuild.texi (vect_float): Say that the selector
9440 only describes the situation when -funsafe-math-optimizations is on.
9441 (vect_float_strict): Document.
9443 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9445 PR tree-optimization/83965
9446 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
9447 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
9448 instead of checking only for a reduction.
9449 (vect_recog_widen_sum_pattern): Likewise.
9451 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9453 * predict.c (probably_never_executed): Only use precise profile info.
9454 (compute_function_frequency): Skip after inlining hack since we now
9455 have quality checking.
9457 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9459 * profile-count.h (profile_probability::very_unlikely,
9460 profile_probability::unlikely, profile_probability::even): Set
9461 precision to guessed.
9463 2018-01-23 Richard Biener <rguenther@suse.de>
9465 PR tree-optimization/83963
9466 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
9467 Properly terminate dominator walk when crossing the exit edge not
9468 when visiting its source block.
9470 2018-01-23 Jakub Jelinek <jakub@redhat.com>
9473 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
9474 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
9476 2018-01-22 Jakub Jelinek <jakub@redhat.com>
9478 PR tree-optimization/83957
9479 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
9480 semicolon after for body surrounded by braces.
9482 PR tree-optimization/83081
9483 * profile-count.h (profile_probability::split): New method.
9484 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
9485 Use profile_probability::split.
9486 (do_compare_rtx_and_jump): Fix adjustment of probabilities
9487 when splitting a single conditional jump into 2.
9489 2018-01-22 David Malcolm <dmalcolm@redhat.com>
9491 PR tree-optimization/69452
9492 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
9495 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9497 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
9498 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
9499 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
9501 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9503 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
9504 * config/rl78/rl78.md (movdi): New define_expand.
9505 * config/rl78/rl78.c (rl78_split_movdi): New function.
9507 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
9510 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
9512 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
9513 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
9514 128-bit to produce an UNSPEC move to get the double word with the
9515 signbit and then a shift directly to do signbit.
9516 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
9517 implementation with a new version that just does either a direct
9518 move or a regular move. Move memory interface to separate insns.
9519 Move insns so they are next to the expander.
9520 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
9521 with signbit move. Split big and little endian case.
9522 (signbit<mode>2_dm_mem_le): Likewise.
9523 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
9524 (signbit<mode>2_dm2): Likewise.
9526 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9528 * config/rl78/rl78.md (anddi3): New define_expand.
9530 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9532 * config/rl78/rl78.md (umindi3): New define_expand.
9534 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9536 * config/rl78/rl78.md (smindi3): New define_expand.
9538 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9540 * config/rl78/rl78.md (smaxdi3): New define_expand.
9542 2018-01-22 Carl Love <cel@us.ibm.com>
9544 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
9545 LVX_V1TI): Add macro expansion.
9546 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
9547 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
9548 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
9549 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
9550 Change check to determine if the instruction is a byte reversing
9551 entry. Fix typo in comment.
9552 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
9553 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
9554 Add def_builtin calls for new builtins.
9555 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
9556 Add define_insn expansion.
9558 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9560 * config/rl78/rl78.md (umaxdi3): New define_expand.
9562 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9564 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
9565 for non-QImode registers.
9567 2018-01-22 Richard Biener <rguenther@suse.de>
9569 PR tree-optimization/83963
9570 * graphite-scop-detection.c (scop_detection::get_sese): Delay
9571 including the loop exit block.
9572 (scop_detection::merge_sese): Likewise.
9573 (scop_detection::add_scop): Do it here instead.
9575 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9577 * doc/sourcebuild.texi (arm_softfloat): Document.
9579 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
9582 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
9583 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
9584 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
9586 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9587 David Edelsohn <dje.gcc@gmail.com>
9590 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9591 Change "crset eq" to "crset 2".
9592 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9593 (*call_indirect_aix<mode>_nospec): Likewise.
9594 (*call_value_indirect_aix<mode>_nospec): Likewise.
9595 (*call_indirect_elfv2<mode>_nospec): Likewise.
9596 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
9597 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
9598 change assembly output from . to $.
9599 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9600 (indirect_jump<mode>_nospec): Change assembly output from . to $.
9601 (*tablejump<mode>_internal1_nospec): Likewise.
9603 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
9606 * config/sh/sh_optimize_sett_clrt.cc:
9607 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
9609 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
9611 PR tree-optimization/83940
9612 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
9613 offset_dt to vect_constant_def rather than vect_unknown_def_type.
9614 (vect_check_load_store_mask): Add a mask_dt_out parameter and
9615 use it to pass back the definition type.
9616 (vect_check_store_rhs): Likewise rhs_dt_out.
9617 (vect_build_gather_load_calls): Add a mask_dt argument and use
9618 it instead of a call to vect_is_simple_use.
9619 (vectorizable_store): Update calls to vect_check_load_store_mask
9620 and vect_check_store_rhs. Use the dt returned by the latter instead
9621 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
9622 instead of calls to vect_is_simple_use. Pass the scalar rather
9623 than the vector operand to vect_is_simple_use when handling
9624 second and subsequent copies of an rhs value.
9625 (vectorizable_load): Update calls to vect_check_load_store_mask
9626 and vect_build_gather_load_calls. Use the cached mask_dt and
9627 gs_info.offset_dt instead of calls to vect_is_simple_use.
9629 2018-01-20 Jakub Jelinek <jakub@redhat.com>
9632 * tree-emutls.c: Include gimplify.h.
9633 (lower_emutls_2): New function.
9634 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
9635 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
9636 it before further processing.
9639 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
9640 UINTVAL (trueop1) instead of INTVAL (op1).
9642 2018-01-19 Jakub Jelinek <jakub@redhat.com>
9646 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
9647 INCOMING_FRAME_SP_OFFSET if not defined.
9648 (scan_trace): Add ENTRY argument. If true and
9649 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
9650 emit a note to adjust the CFA offset.
9651 (create_cfi_notes): Adjust scan_trace callers.
9652 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
9653 INCOMING_FRAME_SP_OFFSET in the CIE.
9654 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
9655 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
9657 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
9658 * doc/tm.texi: Regenerated.
9660 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9662 PR rtl-optimization/83147
9663 * lra-constraints.c (remove_inheritance_pseudos): Use
9664 lra_substitute_pseudo_within_insn.
9666 2018-01-19 Tom de Vries <tom@codesourcery.com>
9667 Cesar Philippidis <cesar@codesourcery.com>
9670 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
9672 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
9675 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
9676 spaces for function labels.
9678 2018-01-19 Martin Liska <mliska@suse.cz>
9680 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
9681 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
9682 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
9683 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
9684 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
9685 (PRED_CONST_RETURN): Change from 69 to 65.
9686 (PRED_NULL_RETURN): Change from 91 to 71.
9687 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
9688 (PRED_LOOP_GUARD): Change from 66 to 73.
9690 2018-01-19 Martin Liska <mliska@suse.cz>
9692 * predict.c (predict_insn_def): Add new assert.
9693 (struct branch_predictor): Change type to signed integer.
9694 (test_prediction_value_range): Amend test to cover
9696 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
9697 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
9698 (PRED_LOOP_ITERATIONS_MAX): Likewise.
9699 (PRED_LOOP_IV_COMPARE): Likewise.
9700 * predict.h (PROB_UNINITIALIZED): Define new constant.
9702 2018-01-19 Martin Liska <mliska@suse.cz>
9704 * predict.c (dump_prediction): Add new format for
9705 analyze_brprob.py script which is enabled with -details
9707 * profile-count.h (precise_p): New function.
9709 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9711 PR tree-optimization/83922
9712 * tree-vect-loop.c (vect_verify_full_masking): Return false if
9713 there are no statements that need masking.
9714 (vect_active_double_reduction_p): New function.
9715 (vect_analyze_loop_operations): Use it when handling phis that
9716 are not in the loop header.
9718 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9720 PR tree-optimization/83914
9721 * tree-vect-loop.c (vectorizable_induction): Don't convert
9722 init_expr or apply the peeling adjustment for inductions
9723 that are nested within the vectorized loop.
9725 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9727 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
9730 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9734 * function.h (gimplify_parameters): Add gimple_seq * argument.
9735 * function.c: Include gimple.h and options.h.
9736 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
9737 for the added local temporaries if needed.
9738 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
9739 if there are any parameter cleanups, wrap whole body into a
9740 try/finally with the cleanups.
9742 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
9745 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
9746 Use GET_MODE_CLASS for scalar floating point.
9748 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9752 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
9753 Fix call of call_cgraph_insertion_hooks.
9755 2018-01-18 Martin Sebor <msebor@redhat.com>
9757 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
9759 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9762 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
9765 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
9768 * common.opt: (-ffile-prefix-map): New option.
9769 * opts.c (common_handle_option): Defer it.
9770 * opts-global.c (handle_common_deferred_options): Handle it.
9771 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
9772 * file-prefix-map.h: New file.
9773 (remap_debug_filename, add_debug_prefix_map): ...here.
9774 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
9775 * final.c (debug_prefix_map, add_debug_prefix_map
9776 remap_debug_filename): Move to...
9777 * file-prefix-map.c: New file.
9778 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
9779 generalize, get rid of alloca(), use strrchr() instead of strchr().
9780 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
9781 Implement in terms of add_prefix_map().
9782 (remap_macro_filename, remap_debug_filename): Implement in term of
9784 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
9785 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
9786 * dbxout.c: Include file-prefix-map.h.
9787 * varasm.c: Likewise.
9788 * vmsdbgout.c: Likewise.
9789 * xcoffout.c: Likewise.
9790 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
9791 * doc/cppopts.texi (-fmacro-prefix-map): Document.
9792 * doc/invoke.texi (-ffile-prefix-map): Document.
9793 (-fdebug-prefix-map): Update description.
9795 2018-01-18 Martin Liska <mliska@suse.cz>
9797 * config/i386/i386.c (indirect_thunk_name): Document that also
9799 (output_indirect_thunk): Document why both instructions
9800 (pause and lfence) are generated.
9802 2018-01-18 Richard Biener <rguenther@suse.de>
9804 PR tree-optimization/83887
9805 * graphite-scop-detection.c
9806 (scop_detection::get_nearest_dom_with_single_entry): Remove.
9807 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
9808 (scop_detection::merge_sese): Re-implement with a flood-fill
9809 algorithm that properly finds a SESE region if it exists.
9811 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9814 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
9815 pointer_diff optimizations use view_convert instead of convert.
9817 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9819 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9820 Generate different code for -mno-speculate-indirect-jumps.
9821 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9822 (*call_indirect_aix<mode>): Disable for
9823 -mno-speculate-indirect-jumps.
9824 (*call_indirect_aix<mode>_nospec): New define_insn.
9825 (*call_value_indirect_aix<mode>): Disable for
9826 -mno-speculate-indirect-jumps.
9827 (*call_value_indirect_aix<mode>_nospec): New define_insn.
9828 (*sibcall_nonlocal_sysv<mode>): Generate different code for
9829 -mno-speculate-indirect-jumps.
9830 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9832 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
9834 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
9835 long double type, set the flags for noting the default long double
9836 type, even if we don't pass or return a long double type.
9838 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
9841 * ipa-inline.c (flatten_function): Do not overwrite final inlining
9844 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
9846 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
9847 support for merge[hl].
9848 (fold_mergehl_helper): New helper function.
9849 (tree-vector-builder.h): New #include for tree_vector_builder usage.
9850 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
9851 (altivec_vmrglw_direct): Add xxmrglw insn.
9853 2018-01-17 Andrew Waterman <andrew@sifive.com>
9855 * config/riscv/riscv.c (riscv_conditional_register_usage): If
9856 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
9858 2018-01-17 David Malcolm <dmalcolm@redhat.com>
9861 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
9862 call the lto_location_cache before reading the
9863 DECL_SOURCE_LOCATION of the types.
9865 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
9866 Richard Sandiford <richard.sandiford@linaro.org>
9868 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
9869 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
9870 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
9871 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
9872 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
9874 * config/aarch64/constraints.md (aarch64_movti_operand):
9876 * config/aarch64/predicates.md (Uti): Add new constraint.
9878 2018-01-17 Carl Love <cel@us.ibm.com>
9880 * config/rs6000/vsx.md (define_expand xl_len_r,
9881 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
9882 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
9884 (define_expand, define_insn): Move the shift left from the
9885 define_insn to the define_expand for lxvl and stxvl instructions.
9886 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
9887 and XL_LEN_R definitions to PURE.
9889 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
9891 * config/i386/i386.c (indirect_thunk_name): Declare regno
9892 as unsigned int. Compare regno with INVALID_REGNUM.
9893 (output_indirect_thunk): Ditto.
9894 (output_indirect_thunk_function): Ditto.
9895 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
9896 in the call to output_indirect_thunk_function.
9898 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
9901 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
9902 rather than the size of inner_type to determine the stack slot size
9903 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
9905 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
9908 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
9911 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
9913 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
9914 endian Linux systems to optionally enable multilibs for selecting
9915 the long double type if the user configured an explicit type.
9916 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
9917 have no long double multilibs if not defined.
9918 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
9919 warn if the user used -mabi={ieee,ibm}longdouble and we built
9920 multilibs for long double.
9921 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
9922 appropriate multilib option.
9923 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
9925 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
9926 for building long double multilibs.
9927 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
9929 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
9931 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
9934 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
9936 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
9939 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
9942 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
9945 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9947 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
9948 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
9950 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
9952 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
9953 different rtl trees depending on TARGET_64BIT.
9954 (rs6000_gen_lvx): Likewise.
9956 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9958 * config/visium/visium.md (nop): Tweak comment.
9959 (hazard_nop): Likewise.
9961 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9963 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
9964 -mspeculate-indirect-jumps.
9965 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
9966 for -mno-speculate-indirect-jumps.
9967 (*call_indirect_elfv2<mode>_nospec): New define_insn.
9968 (*call_value_indirect_elfv2<mode>): Disable for
9969 -mno-speculate-indirect-jumps.
9970 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
9971 (indirect_jump): Emit different RTL for
9972 -mno-speculate-indirect-jumps.
9973 (*indirect_jump<mode>): Disable for
9974 -mno-speculate-indirect-jumps.
9975 (*indirect_jump<mode>_nospec): New define_insn.
9976 (tablejump): Emit different RTL for
9977 -mno-speculate-indirect-jumps.
9978 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
9979 (tablejumpsi_nospec): New define_expand.
9980 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
9981 (tablejumpdi_nospec): New define_expand.
9982 (*tablejump<mode>_internal1): Disable for
9983 -mno-speculate-indirect-jumps.
9984 (*tablejump<mode>_internal1_nospec): New define_insn.
9985 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
9988 2018-01-16 Artyom Skrobov tyomitch@gmail.com
9990 * caller-save.c (insert_save): Drop unnecessary parameter. All
9993 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9994 Richard Biener <rguenth@suse.de>
9997 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
9998 return early, inline manually is_gimple_sizepos. Make sure if we
9999 call gimplify_expr we don't end up with a gimple constant.
10000 * tree.c (variably_modified_type_p): Don't return true for
10001 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
10002 * gimplify.h (is_gimple_sizepos): Remove.
10004 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10006 PR tree-optimization/83857
10007 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
10008 vectorizable_live_operation for pure SLP statements.
10009 (vectorizable_live_operation): Handle PHIs.
10011 2018-01-16 Richard Biener <rguenther@suse.de>
10013 PR tree-optimization/83867
10014 * tree-vect-stmts.c (vect_transform_stmt): Precompute
10015 nested_in_vect_loop_p since the scalar stmt may get invalidated.
10017 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10020 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
10021 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
10022 If off is not INTEGER_CST, issue a may not be aligned warning
10023 rather than isn't aligned. Use isn%'t rather than isn't.
10024 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
10026 <case MULT_EXPR>: Improve the case when bottom and one of the
10027 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
10028 operand, in that case check if the other operand is multiple of
10029 bottom divided by the INTEGER_CST operand.
10031 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10034 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
10035 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
10036 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
10037 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
10038 * config/pa/pa.c (pa_function_arg_advance): Likewise.
10039 (pa_function_arg, pa_arg_partial_bytes): Likewise.
10040 (pa_function_arg_size): New function.
10042 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10044 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
10045 in a separate statement.
10047 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10049 PR tree-optimization/83847
10050 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
10051 group gathers and scatters.
10053 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10055 PR rtl-optimization/86620
10056 * params.def (max-sched-ready-insns): Bump minimum value to 1.
10058 PR rtl-optimization/83213
10059 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
10060 to last if both are JUMP_INSNs.
10062 PR tree-optimization/83843
10063 * gimple-ssa-store-merging.c
10064 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
10065 store_immediate_info for bswap/nop orig_stores.
10067 2018-01-15 Andrew Waterman <andrew@sifive.com>
10069 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
10071 <UDIV>: Increase cost if !TARGET_DIV.
10073 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
10075 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
10076 (define_attr "cr_logical_3op"): New.
10077 (cceq_ior_compare): Adjust.
10078 (cceq_ior_compare_complement): Adjust.
10079 (*cceq_rev_compare): Adjust.
10080 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
10081 (is_cracked_insn): Adjust.
10082 (insn_must_be_first_in_group): Adjust.
10083 * config/rs6000/40x.md: Adjust.
10084 * config/rs6000/440.md: Adjust.
10085 * config/rs6000/476.md: Adjust.
10086 * config/rs6000/601.md: Adjust.
10087 * config/rs6000/603.md: Adjust.
10088 * config/rs6000/6xx.md: Adjust.
10089 * config/rs6000/7450.md: Adjust.
10090 * config/rs6000/7xx.md: Adjust.
10091 * config/rs6000/8540.md: Adjust.
10092 * config/rs6000/cell.md: Adjust.
10093 * config/rs6000/e300c2c3.md: Adjust.
10094 * config/rs6000/e500mc.md: Adjust.
10095 * config/rs6000/e500mc64.md: Adjust.
10096 * config/rs6000/e5500.md: Adjust.
10097 * config/rs6000/e6500.md: Adjust.
10098 * config/rs6000/mpc.md: Adjust.
10099 * config/rs6000/power4.md: Adjust.
10100 * config/rs6000/power5.md: Adjust.
10101 * config/rs6000/power6.md: Adjust.
10102 * config/rs6000/power7.md: Adjust.
10103 * config/rs6000/power8.md: Adjust.
10104 * config/rs6000/power9.md: Adjust.
10105 * config/rs6000/rs64.md: Adjust.
10106 * config/rs6000/titan.md: Adjust.
10108 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10110 * config/i386/predicates.md (indirect_branch_operand): Rewrite
10111 ix86_indirect_branch_register logic.
10113 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10115 * config/i386/constraints.md (Bs): Update
10116 ix86_indirect_branch_register check. Don't check
10117 ix86_indirect_branch_register with GOT_memory_operand.
10119 * config/i386/predicates.md (GOT_memory_operand): Don't check
10120 ix86_indirect_branch_register here.
10121 (GOT32_symbol_operand): Likewise.
10123 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10125 * config/i386/predicates.md (constant_call_address_operand):
10126 Rewrite ix86_indirect_branch_register logic.
10127 (sibcall_insn_operand): Likewise.
10129 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10131 * config/i386/constraints.md (Bs): Replace
10132 ix86_indirect_branch_thunk_register with
10133 ix86_indirect_branch_register.
10135 * config/i386/i386.md (indirect_jump): Likewise.
10136 (tablejump): Likewise.
10137 (*sibcall_memory): Likewise.
10138 (*sibcall_value_memory): Likewise.
10139 Peepholes of indirect call and jump via memory: Likewise.
10140 * config/i386/i386.opt: Likewise.
10141 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10142 (GOT_memory_operand): Likewise.
10143 (call_insn_operand): Likewise.
10144 (sibcall_insn_operand): Likewise.
10145 (GOT32_symbol_operand): Likewise.
10147 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10149 PR middle-end/83837
10150 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
10151 type rather than type addr's type points to.
10152 (expand_omp_atomic_mutex): Likewise.
10153 (expand_omp_atomic): Likewise.
10155 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10158 * config/i386/i386.c (output_indirect_thunk_function): Use
10159 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
10160 for __x86_return_thunk.
10162 2018-01-15 Richard Biener <rguenther@suse.de>
10164 PR middle-end/83850
10165 * expmed.c (extract_bit_field_1): Fix typo.
10167 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10170 * config/arm/iterators.md (VF): New mode iterator.
10171 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
10172 Remove integer-related logic from pattern.
10173 (neon_vabd<mode>_3): Likewise.
10175 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10177 PR middle-end/82694
10178 * common.opt (fstrict-overflow): No longer an alias.
10179 (fwrapv-pointer): New option.
10180 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
10181 also for pointer types based on flag_wrapv_pointer.
10182 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
10183 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
10184 opts->x_flag_wrapv got set.
10185 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
10186 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
10187 POINTER_TYPE_OVERFLOW_UNDEFINED.
10188 * match.pd: Likewise in address comparison pattern.
10189 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
10191 2018-01-15 Richard Biener <rguenther@suse.de>
10194 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
10195 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
10196 Reset type names to their identifier if their TYPE_DECL doesn't
10197 have linkage (and thus is used for ODR and devirt).
10198 (save_debug_info_for_decl): Remove.
10199 (save_debug_info_for_type): Likewise.
10200 (add_tree_to_fld_list): Adjust.
10201 * tree-pretty-print.c (dump_generic_node): Make dumping of
10202 type names more robust.
10204 2018-01-15 Richard Biener <rguenther@suse.de>
10206 * BASE-VER: Bump to 8.0.1.
10208 2018-01-14 Martin Sebor <msebor@redhat.com>
10211 * builtins.c (check_access): Avoid warning when the no-warning bit
10214 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
10216 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
10217 * ira-color (allocno_hard_regs_compare): Likewise.
10219 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
10222 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
10223 Use .pushsection/.popsection.
10225 2018-01-14 Martin Sebor <msebor@redhat.com>
10228 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
10230 2018-01-14 Jakub Jelinek <jakub@redhat.com>
10232 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
10233 entry from extra_headers.
10234 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
10235 extra_headers, make the list bitwise identical to the i?86-*-* one.
10237 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10239 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
10240 -mcmodel=large with -mindirect-branch=thunk,
10241 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
10242 -mfunction-return=thunk-extern.
10243 * doc/invoke.texi: Document -mcmodel=large is incompatible with
10244 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
10245 -mfunction-return=thunk and -mfunction-return=thunk-extern.
10247 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10249 * config/i386/i386.c (print_reg): Print the name of the full
10250 integer register without '%'.
10251 (ix86_print_operand): Handle 'V'.
10252 * doc/extend.texi: Document 'V' modifier.
10254 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10256 * config/i386/constraints.md (Bs): Disallow memory operand for
10257 -mindirect-branch-register.
10259 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10260 (GOT_memory_operand): Likewise.
10261 (call_insn_operand): Likewise.
10262 (sibcall_insn_operand): Likewise.
10263 (GOT32_symbol_operand): Likewise.
10264 * config/i386/i386.md (indirect_jump): Call convert_memory_address
10265 for -mindirect-branch-register.
10266 (tablejump): Likewise.
10267 (*sibcall_memory): Likewise.
10268 (*sibcall_value_memory): Likewise.
10269 Disallow peepholes of indirect call and jump via memory for
10270 -mindirect-branch-register.
10271 (*call_pop): Replace m with Bw.
10272 (*call_value_pop): Likewise.
10273 (*sibcall_pop_memory): Replace m with Bs.
10274 * config/i386/i386.opt (mindirect-branch-register): New option.
10275 * doc/invoke.texi: Document -mindirect-branch-register option.
10277 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10279 * config/i386/i386-protos.h (ix86_output_function_return): New.
10280 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
10281 set function_return_type.
10282 (indirect_thunk_name): Add ret_p to indicate thunk for function
10284 (output_indirect_thunk_function): Pass false to
10285 indirect_thunk_name.
10286 (ix86_output_indirect_branch_via_reg): Likewise.
10287 (ix86_output_indirect_branch_via_push): Likewise.
10288 (output_indirect_thunk_function): Create alias for function
10289 return thunk if regno < 0.
10290 (ix86_output_function_return): New function.
10291 (ix86_handle_fndecl_attribute): Handle function_return.
10292 (ix86_attribute_table): Add function_return.
10293 * config/i386/i386.h (machine_function): Add
10294 function_return_type.
10295 * config/i386/i386.md (simple_return_internal): Use
10296 ix86_output_function_return.
10297 (simple_return_internal_long): Likewise.
10298 * config/i386/i386.opt (mfunction-return=): New option.
10299 (indirect_branch): Mention -mfunction-return=.
10300 * doc/extend.texi: Document function_return function attribute.
10301 * doc/invoke.texi: Document -mfunction-return= option.
10303 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10305 * config/i386/i386-opts.h (indirect_branch): New.
10306 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
10307 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
10308 with local indirect jump when converting indirect call and jump.
10309 (ix86_set_indirect_branch_type): New.
10310 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
10311 (indirectlabelno): New.
10312 (indirect_thunk_needed): Likewise.
10313 (indirect_thunk_bnd_needed): Likewise.
10314 (indirect_thunks_used): Likewise.
10315 (indirect_thunks_bnd_used): Likewise.
10316 (INDIRECT_LABEL): Likewise.
10317 (indirect_thunk_name): Likewise.
10318 (output_indirect_thunk): Likewise.
10319 (output_indirect_thunk_function): Likewise.
10320 (ix86_output_indirect_branch_via_reg): Likewise.
10321 (ix86_output_indirect_branch_via_push): Likewise.
10322 (ix86_output_indirect_branch): Likewise.
10323 (ix86_output_indirect_jmp): Likewise.
10324 (ix86_code_end): Call output_indirect_thunk_function if needed.
10325 (ix86_output_call_insn): Call ix86_output_indirect_branch if
10327 (ix86_handle_fndecl_attribute): Handle indirect_branch.
10328 (ix86_attribute_table): Add indirect_branch.
10329 * config/i386/i386.h (machine_function): Add indirect_branch_type
10330 and has_local_indirect_jump.
10331 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
10333 (tablejump): Likewise.
10334 (*indirect_jump): Use ix86_output_indirect_jmp.
10335 (*tablejump_1): Likewise.
10336 (simple_return_indirect_internal): Likewise.
10337 * config/i386/i386.opt (mindirect-branch=): New option.
10338 (indirect_branch): New.
10341 (thunk-inline): Likewise.
10342 (thunk-extern): Likewise.
10343 * doc/extend.texi: Document indirect_branch function attribute.
10344 * doc/invoke.texi: Document -mindirect-branch= option.
10346 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
10349 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
10351 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
10353 * ipa-inline.c (want_inline_small_function_p): Return false if
10354 inlining has already failed with CIF_FINAL_ERROR.
10355 (update_caller_keys): Call want_inline_small_function_p before
10357 (update_callee_keys): Likewise.
10359 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
10361 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
10363 (rs6000_quadword_masked_address_p): Likewise.
10364 (quad_aligned_load_p): Likewise.
10365 (quad_aligned_store_p): Likewise.
10366 (const_load_sequence_p): Add comment to describe the outer-most loop.
10367 (mimic_memory_attributes_and_flags): New function.
10368 (rs6000_gen_stvx): Likewise.
10369 (replace_swapped_aligned_store): Likewise.
10370 (rs6000_gen_lvx): Likewise.
10371 (replace_swapped_aligned_load): Likewise.
10372 (replace_swapped_load_constant): Capitalize argument name in
10373 comment describing this function.
10374 (rs6000_analyze_swaps): Add a third pass to search for vector loads
10375 and stores that access quad-word aligned addresses and replace
10376 with stvx or lvx instructions when appropriate.
10377 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
10378 New function prototype.
10379 (rs6000_quadword_masked_address_p): Likewise.
10380 (rs6000_gen_lvx): Likewise.
10381 (rs6000_gen_stvx): Likewise.
10382 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
10383 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
10384 when memory address is aligned.
10385 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
10386 this split to select lvx instruction when memory address is aligned.
10387 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
10388 instruction when memory address is aligned.
10389 (*vsx_le_perm_load_v16qi): Likewise.
10390 (four unnamed splitters): Modify to select the stvx instruction
10391 when memory is aligned.
10393 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
10395 * predict.c (determine_unlikely_bbs): Handle correctly BBs
10396 which appears in the queue multiple times.
10398 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10399 Alan Hayward <alan.hayward@arm.com>
10400 David Sherwood <david.sherwood@arm.com>
10402 * tree-vectorizer.h (vec_lower_bound): New structure.
10403 (_loop_vec_info): Add check_nonzero and lower_bounds.
10404 (LOOP_VINFO_CHECK_NONZERO): New macro.
10405 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
10406 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
10407 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
10408 fields. Make seg_len the distance travelled, not including the
10410 (dr_direction_indicator): Declare.
10411 (dr_zero_step_indicator): Likewise.
10412 (dr_known_forward_stride_p): Likewise.
10413 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
10415 (runtime_alias_check_p): Allow runtime alias checks with
10417 (operator ==): Compare access_size and align.
10418 (prune_runtime_alias_test_list): Rework for new distinction between
10419 the access_size and seg_len.
10420 (create_intersect_range_checks_index): Likewise. Cope with polynomial
10422 (get_segment_min_max): New function.
10423 (create_intersect_range_checks): Use it.
10424 (dr_step_indicator): New function.
10425 (dr_direction_indicator): Likewise.
10426 (dr_zero_step_indicator): Likewise.
10427 (dr_known_forward_stride_p): Likewise.
10428 * tree-loop-distribution.c (data_ref_segment_size): Return
10429 DR_STEP * (niters - 1).
10430 (compute_alias_check_pairs): Update call to the dr_with_seg_len
10432 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
10433 (vect_preserves_scalar_order_p): New function, split out from...
10434 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
10435 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
10436 (vect_vfa_access_size): New function.
10437 (vect_vfa_align): Likewise.
10438 (vect_compile_time_alias): Take access_size_a and access_b arguments.
10439 (dump_lower_bound): New function.
10440 (vect_check_lower_bound): Likewise.
10441 (vect_small_gap_p): Likewise.
10442 (vectorizable_with_step_bound_p): Likewise.
10443 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
10444 depencies if the vectorization factor is 1. Convert the checks
10445 for nonzero steps into checks on the bounds of DR_STEP. Try using
10446 a bunds check for variable steps if the minimum required step is
10447 relatively small. Update calls to the dr_with_seg_len
10448 constructor and to vect_compile_time_alias.
10449 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
10451 (vect_loop_versioning): Call it.
10452 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
10454 (vect_estimate_min_profitable_iters): Account for any bounds checks.
10456 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10457 Alan Hayward <alan.hayward@arm.com>
10458 David Sherwood <david.sherwood@arm.com>
10460 * doc/sourcebuild.texi (vect_scatter_store): Document.
10461 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
10463 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
10465 * genopinit.c (main): Add supports_vec_scatter_store and
10466 supports_vec_scatter_store_cached to target_optabs.
10467 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
10468 IFN_MASK_SCATTER_STORE.
10469 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
10471 * internal-fn.h (internal_store_fn_p): Declare.
10472 (internal_fn_stored_value_index): Likewise.
10473 * internal-fn.c (scatter_store_direct): New macro.
10474 (expand_scatter_store_optab_fn): New function.
10475 (direct_scatter_store_optab_supported_p): New macro.
10476 (internal_store_fn_p): New function.
10477 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
10478 IFN_MASK_SCATTER_STORE.
10479 (internal_fn_mask_index): Likewise.
10480 (internal_fn_stored_value_index): New function.
10481 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
10482 for scatter stores.
10483 * optabs-query.h (supports_vec_scatter_store_p): Declare.
10484 * optabs-query.c (supports_vec_scatter_store_p): New function.
10485 * tree-vectorizer.h (vect_get_store_rhs): Declare.
10486 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
10487 true for scatter stores.
10488 (vect_gather_scatter_fn_p): Handle scatter stores too.
10489 (vect_check_gather_scatter): Consider using scatter stores if
10490 supports_vec_scatter_store_p.
10491 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
10492 scatter stores too.
10493 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10494 internal_fn_stored_value_index.
10495 (check_load_store_masking): Handle scatter stores too.
10496 (vect_get_store_rhs): Make public.
10497 (vectorizable_call): Use internal_store_fn_p.
10498 (vectorizable_store): Handle scatter store internal functions.
10499 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
10500 when deciding whether the end of the group has been reached.
10501 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
10502 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
10503 (mask_scatter_store<mode>): New insns.
10505 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10506 Alan Hayward <alan.hayward@arm.com>
10507 David Sherwood <david.sherwood@arm.com>
10509 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
10510 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
10511 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
10513 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
10514 Use vect_truncate_gather_scatter_offset if we can't treat the
10515 operation as a normal gather load or scatter store.
10516 (get_group_load_store_type): Take the gather_scatter_info
10517 as argument. Try using a gather load or scatter store for
10518 single-element groups.
10519 (get_load_store_type): Update calls to get_group_load_store_type
10520 and vect_use_strided_gather_scatters_p.
10522 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10523 Alan Hayward <alan.hayward@arm.com>
10524 David Sherwood <david.sherwood@arm.com>
10526 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
10527 optional tree argument.
10528 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
10530 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
10531 but continue to use the current value as a fallback.
10532 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
10533 to compare the updates.
10534 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
10535 (get_load_store_type): Use it when handling a strided access.
10536 (vect_get_strided_load_store_ops): New function.
10537 (vect_get_data_ptr_increment): Likewise.
10538 (vectorizable_load): Handle strided gather loads. Always pass
10539 a step to vect_create_data_ref_ptr and bump_vector_ptr.
10541 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10542 Alan Hayward <alan.hayward@arm.com>
10543 David Sherwood <david.sherwood@arm.com>
10545 * doc/md.texi (gather_load@var{m}): Document.
10546 (mask_gather_load@var{m}): Likewise.
10547 * genopinit.c (main): Add supports_vec_gather_load and
10548 supports_vec_gather_load_cached to target_optabs.
10549 * optabs-tree.c (init_tree_optimization_optabs): Use
10550 ggc_cleared_alloc to allocate target_optabs.
10551 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
10552 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
10554 * internal-fn.h (internal_load_fn_p): Declare.
10555 (internal_gather_scatter_fn_p): Likewise.
10556 (internal_fn_mask_index): Likewise.
10557 (internal_gather_scatter_fn_supported_p): Likewise.
10558 * internal-fn.c (gather_load_direct): New macro.
10559 (expand_gather_load_optab_fn): New function.
10560 (direct_gather_load_optab_supported_p): New macro.
10561 (direct_internal_fn_optab): New function.
10562 (internal_load_fn_p): Likewise.
10563 (internal_gather_scatter_fn_p): Likewise.
10564 (internal_fn_mask_index): Likewise.
10565 (internal_gather_scatter_fn_supported_p): Likewise.
10566 * optabs-query.c (supports_at_least_one_mode_p): New function.
10567 (supports_vec_gather_load_p): Likewise.
10568 * optabs-query.h (supports_vec_gather_load_p): Declare.
10569 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
10570 and memory_type field.
10571 (NUM_PATTERNS): Bump to 15.
10572 * tree-vect-data-refs.c: Include internal-fn.h.
10573 (vect_gather_scatter_fn_p): New function.
10574 (vect_describe_gather_scatter_call): Likewise.
10575 (vect_check_gather_scatter): Try using internal functions for
10576 gather loads. Recognize existing calls to a gather load function.
10577 (vect_analyze_data_refs): Consider using gather loads if
10578 supports_vec_gather_load_p.
10579 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
10580 (vect_get_gather_scatter_offset_type): Likewise.
10581 (vect_convert_mask_for_vectype): Likewise.
10582 (vect_add_conversion_to_patterm): Likewise.
10583 (vect_try_gather_scatter_pattern): Likewise.
10584 (vect_recog_gather_scatter_pattern): New pattern recognizer.
10585 (vect_vect_recog_func_ptrs): Add it.
10586 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10587 internal_fn_mask_index and internal_gather_scatter_fn_p.
10588 (check_load_store_masking): Take the gather_scatter_info as an
10589 argument and handle gather loads.
10590 (vect_get_gather_scatter_ops): New function.
10591 (vectorizable_call): Check internal_load_fn_p.
10592 (vectorizable_load): Likewise. Handle gather load internal
10594 (vectorizable_store): Update call to check_load_store_masking.
10595 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
10596 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
10597 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
10598 (aarch64_gather_scale_operand_d): New predicates.
10599 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
10600 (mask_gather_load<mode>): New insns.
10602 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10603 Alan Hayward <alan.hayward@arm.com>
10604 David Sherwood <david.sherwood@arm.com>
10606 * optabs.def (fold_left_plus_optab): New optab.
10607 * doc/md.texi (fold_left_plus_@var{m}): Document.
10608 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
10609 * internal-fn.c (fold_left_direct): Define.
10610 (expand_fold_left_optab_fn): Likewise.
10611 (direct_fold_left_optab_supported_p): Likewise.
10612 * fold-const-call.c (fold_const_fold_left): New function.
10613 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
10614 * tree-parloops.c (valid_reduction_p): New function.
10615 (gather_scalar_reductions): Use it.
10616 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
10617 (vect_finish_replace_stmt): Declare.
10618 * tree-vect-loop.c (fold_left_reduction_fn): New function.
10619 (needs_fold_left_reduction_p): New function, split out from...
10620 (vect_is_simple_reduction): ...here. Accept reductions that
10621 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
10622 (vect_force_simple_reduction): Also store the reduction type in
10623 the assignment's STMT_VINFO_REDUC_TYPE.
10624 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
10625 (merge_with_identity): New function.
10626 (vect_expand_fold_left): Likewise.
10627 (vectorize_fold_left_reduction): Likewise.
10628 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
10629 scalar phi in place for it. Check for target support and reject
10630 cases that would reassociate the operation. Defer the transform
10631 phase to vectorize_fold_left_reduction.
10632 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
10633 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
10634 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
10636 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10638 * tree-if-conv.c (predicate_mem_writes): Remove redundant
10639 call to ifc_temp_var.
10641 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10642 Alan Hayward <alan.hayward@arm.com>
10643 David Sherwood <david.sherwood@arm.com>
10645 * target.def (legitimize_address_displacement): Take the original
10646 offset as a poly_int.
10647 * targhooks.h (default_legitimize_address_displacement): Update
10649 * targhooks.c (default_legitimize_address_displacement): Likewise.
10650 * doc/tm.texi: Regenerate.
10651 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
10652 as an argument, moving assert of ad->disp == ad->disp_term to...
10653 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
10654 Try calling targetm.legitimize_address_displacement before expanding
10655 the address rather than afterwards, and adjust for the new interface.
10656 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
10657 Match the new hook interface. Handle SVE addresses.
10658 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
10659 new hook interface.
10661 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10663 * Makefile.in (OBJS): Add early-remat.o.
10664 * target.def (select_early_remat_modes): New hook.
10665 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
10666 * doc/tm.texi: Regenerate.
10667 * targhooks.h (default_select_early_remat_modes): Declare.
10668 * targhooks.c (default_select_early_remat_modes): New function.
10669 * timevar.def (TV_EARLY_REMAT): New timevar.
10670 * passes.def (pass_early_remat): New pass.
10671 * tree-pass.h (make_pass_early_remat): Declare.
10672 * early-remat.c: New file.
10673 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
10675 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
10677 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10678 Alan Hayward <alan.hayward@arm.com>
10679 David Sherwood <david.sherwood@arm.com>
10681 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
10682 vfm1 with a bound_epilog parameter.
10683 (vect_do_peeling): Update calls accordingly, and move the prologue
10684 call earlier in the function. Treat the base bound_epilog as 0 for
10685 fully-masked loops and retain vf - 1 for other loops. Add 1 to
10686 this base when peeling for gaps.
10687 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
10688 with fully-masked loops.
10689 (vect_estimate_min_profitable_iters): Handle the single peeled
10690 iteration in that case.
10692 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10693 Alan Hayward <alan.hayward@arm.com>
10694 David Sherwood <david.sherwood@arm.com>
10696 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
10697 single-element interleaving even if the size is not a power of 2.
10698 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
10699 accesses for single-element interleaving if the group size is
10702 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10703 Alan Hayward <alan.hayward@arm.com>
10704 David Sherwood <david.sherwood@arm.com>
10706 * doc/md.texi (fold_extract_last_@var{m}): Document.
10707 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
10708 * optabs.def (fold_extract_last_optab): New optab.
10709 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
10710 * internal-fn.c (fold_extract_direct): New macro.
10711 (expand_fold_extract_optab_fn): Likewise.
10712 (direct_fold_extract_optab_supported_p): Likewise.
10713 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
10714 * tree-vect-loop.c (vect_model_reduction_cost): Handle
10715 EXTRACT_LAST_REDUCTION.
10716 (get_initial_def_for_reduction): Do not create an initial vector
10717 for EXTRACT_LAST_REDUCTION reductions.
10718 (vectorizable_reduction): Leave the scalar phi in place for
10719 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
10720 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
10721 epilogue code for EXTRACT_LAST_REDUCTION and defer the
10722 transform phase to vectorizable_condition.
10723 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
10725 (vect_finish_stmt_generation): ...here.
10726 (vect_finish_replace_stmt): New function.
10727 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
10728 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
10730 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
10732 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10733 Alan Hayward <alan.hayward@arm.com>
10734 David Sherwood <david.sherwood@arm.com>
10736 * doc/md.texi (extract_last_@var{m}): Document.
10737 * optabs.def (extract_last_optab): New optab.
10738 * internal-fn.def (EXTRACT_LAST): New internal function.
10739 * internal-fn.c (cond_unary_direct): New macro.
10740 (expand_cond_unary_optab_fn): Likewise.
10741 (direct_cond_unary_optab_supported_p): Likewise.
10742 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
10743 loops using EXTRACT_LAST.
10744 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
10745 (extract_last_<mode>): ...this optab.
10746 (vec_extract<mode><Vel>): Update accordingly.
10748 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10749 Alan Hayward <alan.hayward@arm.com>
10750 David Sherwood <david.sherwood@arm.com>
10752 * target.def (empty_mask_is_expensive): New hook.
10753 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
10754 * doc/tm.texi: Regenerate.
10755 * targhooks.h (default_empty_mask_is_expensive): Declare.
10756 * targhooks.c (default_empty_mask_is_expensive): New function.
10757 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
10758 if the target says that empty masks are expensive.
10759 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
10761 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
10763 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10764 Alan Hayward <alan.hayward@arm.com>
10765 David Sherwood <david.sherwood@arm.com>
10767 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
10768 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
10769 (vect_use_loop_mask_for_alignment_p): New function.
10770 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
10771 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
10772 niters_skip argument. Make sure that the first niters_skip elements
10773 of the first iteration are inactive.
10774 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
10775 Update call to vect_set_loop_masks_directly.
10776 (get_misalign_in_elems): New function, split out from...
10777 (vect_gen_prolog_loop_niters): ...here.
10778 (vect_update_init_of_dr): Take a code argument that specifies whether
10779 the adjustment should be added or subtracted.
10780 (vect_update_init_of_drs): Likewise.
10781 (vect_prepare_for_masked_peels): New function.
10782 (vect_do_peeling): Skip prologue peeling if we're using a mask
10783 instead. Update call to vect_update_inits_of_drs.
10784 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10786 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
10787 alignment. Do not include the number of peeled iterations in
10788 the minimum threshold in that case.
10789 (vectorizable_induction): Adjust the start value down by
10790 LOOP_VINFO_MASK_SKIP_NITERS iterations.
10791 (vect_transform_loop): Call vect_prepare_for_masked_peels.
10792 Take the number of skipped iterations into account when calculating
10794 * tree-vect-stmts.c (vect_gen_while_not): New function.
10796 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10797 Alan Hayward <alan.hayward@arm.com>
10798 David Sherwood <david.sherwood@arm.com>
10800 * doc/sourcebuild.texi (vect_fully_masked): Document.
10801 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
10802 default value to 0.
10803 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
10805 (vect_analyze_loop_2): ...here. Don't check the vectorization
10806 factor against the number of loop iterations if the loop is
10809 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10810 Alan Hayward <alan.hayward@arm.com>
10811 David Sherwood <david.sherwood@arm.com>
10813 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
10814 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
10815 (dump_groups): Update accordingly.
10816 (iv_use::mem_type): New member variable.
10817 (address_p): New function.
10818 (record_use): Add a mem_type argument and initialize the new
10820 (record_group_use): Add a mem_type argument. Use address_p.
10821 Remove obsolete null checks of base_object. Update call to record_use.
10822 (find_interesting_uses_op): Update call to record_group_use.
10823 (find_interesting_uses_cond): Likewise.
10824 (find_interesting_uses_address): Likewise.
10825 (get_mem_type_for_internal_fn): New function.
10826 (find_address_like_use): Likewise.
10827 (find_interesting_uses_stmt): Try find_address_like_use before
10828 calling find_interesting_uses_op.
10829 (addr_offset_valid_p): Use the iv mem_type field as the type
10830 of the addressed memory.
10831 (add_autoinc_candidates): Likewise.
10832 (get_address_cost): Likewise.
10833 (split_small_address_groups_p): Use address_p.
10834 (split_address_groups): Likewise.
10835 (add_iv_candidate_for_use): Likewise.
10836 (autoinc_possible_for_pair): Likewise.
10837 (rewrite_groups): Likewise.
10838 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
10839 (determine_group_iv_cost): Update after split of USE_ADDRESS.
10840 (get_alias_ptr_type_for_ptr_address): New function.
10841 (rewrite_use_address): Rewrite address uses in calls that were
10842 identified by find_address_like_use.
10844 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10845 Alan Hayward <alan.hayward@arm.com>
10846 David Sherwood <david.sherwood@arm.com>
10848 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
10850 * gimple-expr.h (is_gimple_addressable: Likewise.
10851 * gimple-expr.c (is_gimple_address): Likewise.
10852 * internal-fn.c (expand_call_mem_ref): New function.
10853 (expand_mask_load_optab_fn): Use it.
10854 (expand_mask_store_optab_fn): Likewise.
10856 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10857 Alan Hayward <alan.hayward@arm.com>
10858 David Sherwood <david.sherwood@arm.com>
10860 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
10861 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
10862 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
10863 (cond_umax@var{mode}): Document.
10864 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
10865 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
10866 (cond_umin_optab, cond_umax_optab): New optabs.
10867 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
10868 (COND_IOR, COND_XOR): New internal functions.
10869 * internal-fn.h (get_conditional_internal_fn): Declare.
10870 * internal-fn.c (cond_binary_direct): New macro.
10871 (expand_cond_binary_optab_fn): Likewise.
10872 (direct_cond_binary_optab_supported_p): Likewise.
10873 (get_conditional_internal_fn): New function.
10874 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
10875 Cope with reduction statements that are vectorized as calls rather
10877 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
10878 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
10879 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
10880 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
10881 (UNSPEC_COND_EOR): New unspecs.
10882 (optab): Add mappings for them.
10883 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
10884 (sve_int_op, sve_fp_op): New int attributes.
10886 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10887 Alan Hayward <alan.hayward@arm.com>
10888 David Sherwood <david.sherwood@arm.com>
10890 * optabs.def (while_ult_optab): New optab.
10891 * doc/md.texi (while_ult@var{m}@var{n}): Document.
10892 * internal-fn.def (WHILE_ULT): New internal function.
10893 * internal-fn.h (direct_internal_fn_supported_p): New override
10894 that takes two types as argument.
10895 * internal-fn.c (while_direct): New macro.
10896 (expand_while_optab_fn): New function.
10897 (convert_optab_supported_p): Likewise.
10898 (direct_while_optab_supported_p): New macro.
10899 * wide-int.h (wi::udiv_ceil): New function.
10900 * tree-vectorizer.h (rgroup_masks): New structure.
10901 (vec_loop_masks): New typedef.
10902 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
10903 and fully_masked_p.
10904 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
10905 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
10906 (vect_max_vf): New function.
10907 (slpeel_make_loop_iterate_ntimes): Delete.
10908 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
10909 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
10910 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
10911 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
10912 internal-fn.h, stor-layout.h and optabs-query.h.
10913 (vect_set_loop_mask): New function.
10914 (add_preheader_seq): Likewise.
10915 (add_header_seq): Likewise.
10916 (interleave_supported_p): Likewise.
10917 (vect_maybe_permute_loop_masks): Likewise.
10918 (vect_set_loop_masks_directly): Likewise.
10919 (vect_set_loop_condition_masked): Likewise.
10920 (vect_set_loop_condition_unmasked): New function, split out from
10921 slpeel_make_loop_iterate_ntimes.
10922 (slpeel_make_loop_iterate_ntimes): Rename to..
10923 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
10924 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
10925 (vect_do_peeling): Update call accordingly.
10926 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
10928 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10929 mask_compare_type, can_fully_mask_p and fully_masked_p.
10930 (release_vec_loop_masks): New function.
10931 (_loop_vec_info): Use it to free the loop masks.
10932 (can_produce_all_loop_masks_p): New function.
10933 (vect_get_max_nscalars_per_iter): Likewise.
10934 (vect_verify_full_masking): Likewise.
10935 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
10936 retries, and free the mask rgroups before retrying. Check loop-wide
10937 reasons for disallowing fully-masked loops. Make the final decision
10938 about whether use a fully-masked loop or not.
10939 (vect_estimate_min_profitable_iters): Do not assume that peeling
10940 for the number of iterations will be needed for fully-masked loops.
10941 (vectorizable_reduction): Disable fully-masked loops.
10942 (vectorizable_live_operation): Likewise.
10943 (vect_halve_mask_nunits): New function.
10944 (vect_double_mask_nunits): Likewise.
10945 (vect_record_loop_mask): Likewise.
10946 (vect_get_loop_mask): Likewise.
10947 (vect_transform_loop): Handle the case in which the final loop
10948 iteration might handle a partial vector. Call vect_set_loop_condition
10949 instead of slpeel_make_loop_iterate_ntimes.
10950 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
10951 (check_load_store_masking): New function.
10952 (prepare_load_store_mask): Likewise.
10953 (vectorizable_store): Handle fully-masked loops.
10954 (vectorizable_load): Likewise.
10955 (supportable_widening_operation): Use vect_halve_mask_nunits for
10957 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
10958 (vect_gen_while): New function.
10959 * config/aarch64/aarch64.md (umax<mode>3): New expander.
10960 (aarch64_uqdec<mode>): New insn.
10962 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10963 Alan Hayward <alan.hayward@arm.com>
10964 David Sherwood <david.sherwood@arm.com>
10966 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
10967 (reduc_xor_scal_optab): New optabs.
10968 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
10969 (reduc_xor_scal_@var{m}): Document.
10970 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
10971 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
10972 internal functions.
10973 * fold-const-call.c (fold_const_call): Handle them.
10974 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
10975 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
10976 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
10977 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
10978 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
10979 (UNSPEC_XORV): New unspecs.
10980 (optab): Add entries for them.
10981 (BITWISEV): New int iterator.
10982 (bit_reduc_op): New int attributes.
10984 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10985 Alan Hayward <alan.hayward@arm.com>
10986 David Sherwood <david.sherwood@arm.com>
10988 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
10989 * internal-fn.def (VEC_SHL_INSERT): New internal function.
10990 * optabs.def (vec_shl_insert_optab): New optab.
10991 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
10992 (duplicate_and_interleave): Likewise.
10993 * tree-vect-loop.c: Include internal-fn.h.
10994 (neutral_op_for_slp_reduction): New function, split out from
10995 get_initial_defs_for_reduction.
10996 (get_initial_def_for_reduction): Handle option 2 for variable-length
10997 vectors by loading the neutral value into a vector and then shifting
10998 the initial value into element 0.
10999 (get_initial_defs_for_reduction): Replace the code argument with
11000 the neutral value calculated by neutral_op_for_slp_reduction.
11001 Use gimple_build_vector for constant-length vectors.
11002 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
11003 but the first group_size elements have a neutral value.
11004 Use duplicate_and_interleave otherwise.
11005 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
11006 Update call to get_initial_defs_for_reduction. Handle SLP
11007 reductions for variable-length vectors by creating one vector
11008 result for each scalar result, with the elements associated
11009 with other scalar results stubbed out with the neutral value.
11010 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
11011 Require IFN_VEC_SHL_INSERT for double reductions on
11012 variable-length vectors, or SLP reductions that have
11013 a neutral value. Require can_duplicate_and_interleave_p
11014 support for variable-length unchained SLP reductions if there
11015 is no neutral value, such as for MIN/MAX reductions. Also require
11016 the number of vector elements to be a multiple of the number of
11017 SLP statements when doing variable-length unchained SLP reductions.
11018 Update call to vect_create_epilog_for_reduction.
11019 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
11020 and remove initial values.
11021 (duplicate_and_interleave): Make public.
11022 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
11023 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
11025 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11026 Alan Hayward <alan.hayward@arm.com>
11027 David Sherwood <david.sherwood@arm.com>
11029 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
11030 (can_duplicate_and_interleave_p): New function.
11031 (vect_get_and_check_slp_defs): Take the vector of statements
11032 rather than just the current one. Remove excess parentheses.
11033 Restriction rejectinon of vect_constant_def and vect_external_def
11034 for variable-length vectors to boolean types, or types for which
11035 can_duplicate_and_interleave_p is false.
11036 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
11037 (duplicate_and_interleave): New function.
11038 (vect_get_constant_vectors): Use gimple_build_vector for
11039 constant-length vectors and suitable variable-length constant
11040 vectors. Use duplicate_and_interleave for other variable-length
11041 vectors. Don't defer the update when inserting new statements.
11043 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11044 Alan Hayward <alan.hayward@arm.com>
11045 David Sherwood <david.sherwood@arm.com>
11047 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
11048 min_profitable_iters doesn't go negative.
11050 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11051 Alan Hayward <alan.hayward@arm.com>
11052 David Sherwood <david.sherwood@arm.com>
11054 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
11055 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
11056 * optabs.def (vec_mask_load_lanes_optab): New optab.
11057 (vec_mask_store_lanes_optab): Likewise.
11058 * internal-fn.def (MASK_LOAD_LANES): New internal function.
11059 (MASK_STORE_LANES): Likewise.
11060 * internal-fn.c (mask_load_lanes_direct): New macro.
11061 (mask_store_lanes_direct): Likewise.
11062 (expand_mask_load_optab_fn): Handle masked operations.
11063 (expand_mask_load_lanes_optab_fn): New macro.
11064 (expand_mask_store_optab_fn): Handle masked operations.
11065 (expand_mask_store_lanes_optab_fn): New macro.
11066 (direct_mask_load_lanes_optab_supported_p): Likewise.
11067 (direct_mask_store_lanes_optab_supported_p): Likewise.
11068 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
11070 (vect_load_lanes_supported): Likewise.
11071 * tree-vect-data-refs.c (strip_conversion): New function.
11072 (can_group_stmts_p): Likewise.
11073 (vect_analyze_data_ref_accesses): Use it instead of checking
11074 for a pair of assignments.
11075 (vect_store_lanes_supported): Take a masked_p parameter.
11076 (vect_load_lanes_supported): Likewise.
11077 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
11078 vect_store_lanes_supported and vect_load_lanes_supported.
11079 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11080 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
11081 parameter. Don't allow gaps for masked accesses.
11082 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
11083 and vect_load_lanes_supported.
11084 (get_load_store_type): Take a masked_p parameter and update
11085 call to get_group_load_store_type.
11086 (vectorizable_store): Update call to get_load_store_type.
11087 Handle IFN_MASK_STORE_LANES.
11088 (vectorizable_load): Update call to get_load_store_type.
11089 Handle IFN_MASK_LOAD_LANES.
11091 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11092 Alan Hayward <alan.hayward@arm.com>
11093 David Sherwood <david.sherwood@arm.com>
11095 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
11097 * config/aarch64/aarch64-protos.h
11098 (aarch64_sve_struct_memory_operand_p): Declare.
11099 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
11100 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
11101 (VPRED, vpred): Handle SVE structure modes.
11102 * config/aarch64/constraints.md (Utx): New constraint.
11103 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
11104 (aarch64_sve_struct_nonimmediate_operand): New predicates.
11105 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
11106 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
11107 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
11108 structure modes. Split into pieces after RA.
11109 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
11110 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
11112 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
11113 SVE structure modes.
11114 (aarch64_classify_address): Likewise.
11115 (sizetochar): Move earlier in file.
11116 (aarch64_print_operand): Handle SVE register lists.
11117 (aarch64_array_mode): New function.
11118 (aarch64_sve_struct_memory_operand_p): Likewise.
11119 (TARGET_ARRAY_MODE): Redefine.
11121 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11122 Alan Hayward <alan.hayward@arm.com>
11123 David Sherwood <david.sherwood@arm.com>
11125 * target.def (array_mode): New target hook.
11126 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
11127 * doc/tm.texi: Regenerate.
11128 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
11129 * hooks.c (hook_optmode_mode_uhwi_none): New function.
11130 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
11131 targetm.array_mode.
11132 * stor-layout.c (mode_for_array): Likewise. Support polynomial
11135 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11136 Alan Hayward <alan.hayward@arm.com>
11137 David Sherwood <david.sherwood@arm.com>
11139 * fold-const.c (fold_binary_loc): Check the argument types
11140 rather than the result type when testing for a vector operation.
11142 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11144 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
11145 * doc/tm.texi: Regenerate.
11147 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11148 Alan Hayward <alan.hayward@arm.com>
11149 David Sherwood <david.sherwood@arm.com>
11151 * doc/invoke.texi (-msve-vector-bits=): Document new option.
11152 (sve): Document new AArch64 extension.
11153 * doc/md.texi (w): Extend the description of the AArch64
11154 constraint to include SVE vectors.
11155 (Upl, Upa): Document new AArch64 predicate constraints.
11156 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
11158 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
11159 (msve-vector-bits=): New option.
11160 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
11161 SVE when these are disabled.
11162 (sve): New extension.
11163 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
11164 modes. Adjust their number of units based on aarch64_sve_vg.
11165 (MAX_BITSIZE_MODE_ANY_MODE): Define.
11166 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
11167 aarch64_addr_query_type.
11168 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
11169 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
11170 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
11171 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
11172 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
11173 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
11174 (aarch64_simd_imm_zero_p): Delete.
11175 (aarch64_check_zero_based_sve_index_immediate): Declare.
11176 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11177 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11178 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11179 (aarch64_sve_float_mul_immediate_p): Likewise.
11180 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11181 rather than an rtx.
11182 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
11183 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
11184 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
11185 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
11186 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
11187 (aarch64_regmode_natural_size): Likewise.
11188 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
11189 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
11191 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
11192 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
11193 for VG and the SVE predicate registers.
11194 (V_ALIASES): Add a "z"-prefixed alias.
11195 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
11196 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
11197 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
11198 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
11199 (REG_CLASS_NAMES): Add entries for them.
11200 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
11201 and the predicate registers.
11202 (aarch64_sve_vg): Declare.
11203 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
11204 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
11205 (REGMODE_NATURAL_SIZE): Define.
11206 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
11208 * config/aarch64/aarch64.c: Include cfgrtl.h.
11209 (simd_immediate_info): Add a constructor for series vectors,
11210 and an associated step field.
11211 (aarch64_sve_vg): New variable.
11212 (aarch64_dbx_register_number): Handle VG and the predicate registers.
11213 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
11214 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
11215 (VEC_ANY_DATA, VEC_STRUCT): New constants.
11216 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
11217 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
11218 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
11219 (aarch64_get_mask_mode): New functions.
11220 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
11221 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
11222 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
11223 predicate modes and predicate registers. Explicitly restrict
11224 GPRs to modes of 16 bytes or smaller. Only allow FP registers
11225 to store a vector mode if it is recognized by
11226 aarch64_classify_vector_mode.
11227 (aarch64_regmode_natural_size): New function.
11228 (aarch64_hard_regno_caller_save_mode): Return the original mode
11230 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
11231 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
11232 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
11233 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
11235 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
11236 does not overlap dest if the function is frame-related. Handle
11238 (aarch64_split_add_offset): New function.
11239 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
11240 them aarch64_add_offset.
11241 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
11242 and update call to aarch64_sub_sp.
11243 (aarch64_add_cfa_expression): New function.
11244 (aarch64_expand_prologue): Pass extra temporary registers to the
11245 functions above. Handle the case in which we need to emit new
11246 DW_CFA_expressions for registers that were originally saved
11247 relative to the stack pointer, but now have to be expressed
11248 relative to the frame pointer.
11249 (aarch64_output_mi_thunk): Pass extra temporary registers to the
11251 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
11252 IP0 and IP1 values for SVE frames.
11253 (aarch64_expand_vec_series): New function.
11254 (aarch64_expand_sve_widened_duplicate): Likewise.
11255 (aarch64_expand_sve_const_vector): Likewise.
11256 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
11257 Handle SVE constants. Use emit_move_insn to move a force_const_mem
11258 into the register, rather than emitting a SET directly.
11259 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
11260 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
11261 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
11262 (offset_9bit_signed_scaled_p): New functions.
11263 (aarch64_replicate_bitmask_imm): New function.
11264 (aarch64_bitmask_imm): Use it.
11265 (aarch64_cannot_force_const_mem): Reject expressions involving
11266 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
11267 (aarch64_classify_index): Handle SVE indices, by requiring
11268 a plain register index with a scale that matches the element size.
11269 (aarch64_classify_address): Handle SVE addresses. Assert that
11270 the mode of the address is VOIDmode or an integer mode.
11271 Update call to aarch64_classify_symbol.
11272 (aarch64_classify_symbolic_expression): Update call to
11273 aarch64_classify_symbol.
11274 (aarch64_const_vec_all_in_range_p): New function.
11275 (aarch64_print_vector_float_operand): Likewise.
11276 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
11277 "vN" for FP registers with SVE modes. Handle (const ...) vectors
11278 and the FP immediates 1.0 and 0.5.
11279 (aarch64_print_address_internal): Handle SVE addresses.
11280 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
11281 (aarch64_regno_regclass): Handle predicate registers.
11282 (aarch64_secondary_reload): Handle big-endian reloads of SVE
11284 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
11285 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
11286 (aarch64_convert_sve_vector_bits): New function.
11287 (aarch64_override_options): Use it to handle -msve-vector-bits=.
11288 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11289 rather than an rtx.
11290 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
11291 Handle SVE vector and predicate modes. Accept VL-based constants
11292 that need only one temporary register, and VL offsets that require
11293 no temporary registers.
11294 (aarch64_conditional_register_usage): Mark the predicate registers
11295 as fixed if SVE isn't available.
11296 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
11297 Return true for SVE vector and predicate modes.
11298 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
11299 rather than an unsigned int. Handle SVE modes.
11300 (aarch64_preferred_simd_mode): Update call accordingly. Handle
11302 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
11304 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11305 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11306 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11307 (aarch64_sve_float_mul_immediate_p): New functions.
11308 (aarch64_sve_valid_immediate): New function.
11309 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
11310 Explicitly reject structure modes. Check for INDEX constants.
11311 Handle PTRUE and PFALSE constants.
11312 (aarch64_check_zero_based_sve_index_immediate): New function.
11313 (aarch64_simd_imm_zero_p): Delete.
11314 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
11315 vector modes. Accept constants in the range of CNT[BHWD].
11316 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
11317 ask for an Advanced SIMD mode.
11318 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
11319 (aarch64_simd_vector_alignment): Handle SVE predicates.
11320 (aarch64_vectorize_preferred_vector_alignment): New function.
11321 (aarch64_simd_vector_alignment_reachable): Use it instead of
11323 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
11324 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
11326 (MAX_VECT_LEN): Delete.
11327 (expand_vec_perm_d): Add a vec_flags field.
11328 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
11329 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
11330 (aarch64_evpc_ext): Don't apply a big-endian lane correction
11332 (aarch64_evpc_rev): Rename to...
11333 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
11334 (aarch64_evpc_rev_global): New function.
11335 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
11336 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
11338 (aarch64_evpc_sve_tbl): New function.
11339 (aarch64_expand_vec_perm_const_1): Update after rename of
11340 aarch64_evpc_rev. Handle SVE permutes too, trying
11341 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
11342 than aarch64_evpc_tbl.
11343 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
11344 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
11345 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
11346 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
11347 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
11348 (aarch64_expand_sve_vcond): New functions.
11349 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
11350 of aarch64_vector_mode_p.
11351 (aarch64_dwarf_poly_indeterminate_value): New function.
11352 (aarch64_compute_pressure_classes): Likewise.
11353 (aarch64_can_change_mode_class): Likewise.
11354 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
11355 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
11356 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
11357 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
11358 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
11359 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
11360 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
11361 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
11363 (Dn, Dl, Dr): Accept const as well as const_vector.
11364 (Dz): Likewise. Compare against CONST0_RTX.
11365 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
11366 of "vector" where appropriate.
11367 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
11368 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
11369 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
11370 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
11371 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
11372 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
11373 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
11374 (v_int_equiv): Extend to SVE modes.
11375 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
11377 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
11378 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
11379 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
11380 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
11381 (SVE_COND_FP_CMP): New int iterators.
11382 (perm_hilo): Handle the new unpack unspecs.
11383 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
11385 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
11386 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
11387 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
11388 (aarch64_equality_operator, aarch64_constant_vector_operand)
11389 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
11390 (aarch64_sve_nonimmediate_operand): Likewise.
11391 (aarch64_sve_general_operand): Likewise.
11392 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
11393 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
11394 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
11395 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
11396 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
11397 (aarch64_sve_float_arith_immediate): Likewise.
11398 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
11399 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
11400 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
11401 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
11402 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
11403 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
11404 (aarch64_sve_float_arith_operand): Likewise.
11405 (aarch64_sve_float_arith_with_sub_operand): Likewise.
11406 (aarch64_sve_float_mul_operand): Likewise.
11407 (aarch64_sve_vec_perm_operand): Likewise.
11408 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
11409 (aarch64_mov_operand): Accept const_poly_int and const_vector.
11410 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
11411 as well as const_vector.
11412 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
11413 in file. Use CONST0_RTX and CONSTM1_RTX.
11414 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
11415 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
11416 Use aarch64_simd_imm_zero.
11417 * config/aarch64/aarch64-sve.md: New file.
11418 * config/aarch64/aarch64.md: Include it.
11419 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
11420 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
11421 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
11422 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
11423 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
11424 (sve): New attribute.
11425 (enabled): Disable instructions with the sve attribute unless
11427 (movqi, movhi): Pass CONST_POLY_INT operaneds through
11428 aarch64_expand_mov_immediate.
11429 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
11430 CNT[BHSD] immediates.
11431 (movti): Split CONST_POLY_INT moves into two halves.
11432 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
11433 Split additions that need a temporary here if the destination
11434 is the stack pointer.
11435 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
11436 (*add<mode>3_poly_1): New instruction.
11437 (set_clobber_cc): New expander.
11439 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11441 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
11442 parameter and use it instead of GET_MODE_SIZE (innermode). Use
11443 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
11444 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
11445 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
11446 Change innermode from fixed_mode_size to machine_mode.
11447 (simplify_subreg): Update call accordingly. Handle a constant-sized
11448 subreg of a variable-length CONST_VECTOR.
11450 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11451 Alan Hayward <alan.hayward@arm.com>
11452 David Sherwood <david.sherwood@arm.com>
11454 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
11455 (add_offset_to_base): New function, split out from...
11456 (create_mem_ref): ...here. When handling a scale other than 1,
11457 check first whether the address is valid without the offset.
11458 Add it into the base if so, leaving the index and scale as-is.
11460 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11463 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
11464 fold_for_warn before checking if arg2 is INTEGER_CST.
11466 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
11468 * config/rs6000/predicates.md (load_multiple_operation): Delete.
11469 (store_multiple_operation): Delete.
11470 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
11471 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
11472 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
11473 guarded by TARGET_STRING.
11474 (rs6000_output_load_multiple): Delete.
11475 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
11476 OPTION_MASK_STRING / TARGET_STRING handling.
11477 (print_operand) <'N', 'O'>: Add comment that these are unused now.
11478 (const rs6000_opt_masks) <"string">: Change mask to 0.
11479 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
11480 (MASK_STRING): Delete.
11481 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
11483 (load_multiple): Delete.
11490 (store_multiple): Delete.
11497 (movmemsi_8reg): Delete.
11498 (corresponding unnamed define_insn): Delete.
11499 (movmemsi_6reg): Delete.
11500 (corresponding unnamed define_insn): Delete.
11501 (movmemsi_4reg): Delete.
11502 (corresponding unnamed define_insn): Delete.
11503 (movmemsi_2reg): Delete.
11504 (corresponding unnamed define_insn): Delete.
11505 (movmemsi_1reg): Delete.
11506 (corresponding unnamed define_insn): Delete.
11507 * config/rs6000/rs6000.opt (mno-string): New.
11508 (mstring): Replace by deprecation warning stub.
11509 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
11511 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11513 * regrename.c (regrename_do_replace): If replacing the same
11514 reg multiple times, try to reuse last created gen_raw_REG.
11517 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
11518 main to workaround a bug in GDB.
11520 2018-01-12 Tom de Vries <tom@codesourcery.com>
11523 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
11525 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
11527 PR rtl-optimization/80481
11528 * ira-color.c (get_cap_member): New function.
11529 (allocnos_conflict_by_live_ranges_p): Use it.
11530 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
11531 (setup_slot_coalesced_allocno_live_ranges): Ditto.
11533 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
11536 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
11537 (*saddl_se_1): Ditto.
11538 (*ssubsi_1): Ditto.
11539 (*ssubl_se_1): Ditto.
11541 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11543 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
11544 rather than wi::to_widest for DR_INITs.
11545 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
11546 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
11547 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
11549 (vect_analyze_group_access_1): Note that here.
11551 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11553 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
11554 polynomial type sizes.
11556 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11558 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
11559 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
11560 (gimple_add_tmp_var): Likewise.
11562 2018-01-12 Martin Liska <mliska@suse.cz>
11564 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
11565 (gimple_alloc_sizes): Likewise.
11566 (dump_gimple_statistics): Use PRIu64 in printf format.
11567 * gimple.h: Change uint64_t to int.
11569 2018-01-12 Martin Liska <mliska@suse.cz>
11571 * tree-core.h: Use uint64_t instead of int.
11572 * tree.c (tree_node_counts): Likewise.
11573 (tree_node_sizes): Likewise.
11574 (dump_tree_statistics): Use PRIu64 in printf format.
11576 2018-01-12 Martin Liska <mliska@suse.cz>
11578 * Makefile.in: As qsort_chk is implemented in vec.c, add
11579 vec.o to linkage of gencfn-macros.
11580 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
11581 passing the info to record_node_allocation_statistics.
11582 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
11584 * ggc-common.c (struct ggc_usage): Add operator== and use
11585 it in operator< and compare function.
11586 * mem-stats.h (struct mem_usage): Likewise.
11587 * vec.c (struct vec_usage): Remove operator< and compare
11588 function. Can be simply inherited.
11590 2018-01-12 Martin Jambor <mjambor@suse.cz>
11593 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
11594 * tree-ssa-math-opts.c: Include domwalk.h.
11595 (convert_mult_to_fma_1): New function.
11596 (fma_transformation_info): New type.
11597 (fma_deferring_state): Likewise.
11598 (cancel_fma_deferring): New function.
11599 (result_of_phi): Likewise.
11600 (last_fma_candidate_feeds_initial_phi): Likewise.
11601 (convert_mult_to_fma): Added deferring logic, split actual
11602 transformation to convert_mult_to_fma_1.
11603 (math_opts_dom_walker): New type.
11604 (math_opts_dom_walker::after_dom_children): New method, body moved
11605 here from pass_optimize_widening_mul::execute, added deferring logic
11607 (pass_optimize_widening_mul::execute): Moved most of code to
11608 math_opts_dom_walker::after_dom_children.
11609 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
11610 * config/i386/i386.c (ix86_option_override_internal): Added
11611 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
11613 2018-01-12 Richard Biener <rguenther@suse.de>
11616 * dwarf2out.c (gen_variable_die): Do not reset old_die for
11617 inline instance vars.
11619 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
11622 * config/rx/rx.c (rx_is_restricted_memory_address):
11623 Handle SUBREG case.
11625 2018-01-12 Richard Biener <rguenther@suse.de>
11627 PR tree-optimization/80846
11628 * target.def (split_reduction): New target hook.
11629 * targhooks.c (default_split_reduction): New function.
11630 * targhooks.h (default_split_reduction): Declare.
11631 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
11632 target requests first reduce vectors by combining low and high
11634 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
11635 (get_vectype_for_scalar_type_and_size): Export.
11636 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
11637 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
11638 * doc/tm.texi: Regenerate.
11639 * config/i386/i386.c (ix86_split_reduction): Implement
11640 TARGET_VECTORIZE_SPLIT_REDUCTION.
11642 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11645 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
11646 in PIC mode except for TARGET_VXWORKS_RTP.
11647 * config/sparc/sparc.c: Include cfgrtl.h.
11648 (TARGET_INIT_PIC_REG): Define.
11649 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
11650 (sparc_pic_register_p): New predicate.
11651 (sparc_legitimate_address_p): Use it.
11652 (sparc_legitimize_pic_address): Likewise.
11653 (sparc_delegitimize_address): Likewise.
11654 (sparc_mode_dependent_address_p): Likewise.
11655 (gen_load_pcrel_sym): Remove 4th parameter.
11656 (load_got_register): Adjust call to above. Remove obsolete stuff.
11657 (sparc_expand_prologue): Do not call load_got_register here.
11658 (sparc_flat_expand_prologue): Likewise.
11659 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
11660 (sparc_use_pseudo_pic_reg): New function.
11661 (sparc_init_pic_reg): Likewise.
11662 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11663 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
11665 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
11667 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
11668 Add item for branch_cost.
11670 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11672 PR rtl-optimization/83565
11673 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
11674 not extend the result to a larger mode for rotate operations.
11675 (num_sign_bit_copies1): Likewise.
11677 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11680 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
11682 Use values-Xc.o for -pedantic.
11683 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
11685 2018-01-12 Martin Liska <mliska@suse.cz>
11688 * ipa-devirt.c (final_warning_record::grow_type_warnings):
11690 (possible_polymorphic_call_targets): Use it.
11691 (ipa_devirt): Likewise.
11693 2018-01-12 Martin Liska <mliska@suse.cz>
11695 * profile-count.h (enum profile_quality): Use 0 as invalid
11696 enum value of profile_quality.
11698 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
11700 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
11701 -mext-string options.
11703 2018-01-12 Richard Biener <rguenther@suse.de>
11705 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
11706 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
11707 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
11709 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
11711 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
11713 * configure.ac (--with-long-double-format): Add support for the
11714 configuration option to change the default long double format on
11716 * config.gcc (powerpc*-linux*-*): Likewise.
11717 * configure: Regenerate.
11718 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
11719 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
11720 used without modification.
11722 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11724 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
11725 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
11726 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
11727 MISC_BUILTIN_SPEC_BARRIER.
11728 (rs6000_init_builtins): Likewise.
11729 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
11731 (speculation_barrier): New define_insn.
11732 * doc/extend.texi: Document __builtin_speculation_barrier.
11734 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11737 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
11738 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
11739 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
11741 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
11742 integral modes instead of "ss" and "sd".
11743 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
11744 vectors with 32-bit and 64-bit elements.
11745 (vecdupssescalarmodesuffix): New mode attribute.
11746 (vec_dup<mode>): Use it.
11748 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
11751 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
11752 frame if argument is passed on stack.
11754 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11757 * ree.c (combine_reaching_defs): Optimize also
11758 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
11759 reg2=any_extend(exp); reg1=reg2;, formatting fix.
11761 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11763 PR middle-end/83189
11764 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
11766 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11768 PR middle-end/83718
11769 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
11770 after they are computed.
11772 2018-01-11 Bin Cheng <bin.cheng@arm.com>
11774 PR tree-optimization/83695
11775 * gimple-loop-linterchange.cc
11776 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
11777 reset cached scev information after interchange.
11778 (pass_linterchange::execute): Remove call to scev_reset_htab.
11780 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11782 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
11783 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
11784 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
11785 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
11786 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
11787 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
11788 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
11789 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
11790 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
11791 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
11792 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
11793 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
11794 (V_lane_reg): Likewise.
11795 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
11797 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
11798 (vfmal_lane_low<mode>_intrinsic,
11799 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
11800 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
11801 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
11802 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
11803 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
11804 vfmsl_lane_high<mode>_intrinsic): New define_insns.
11806 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11808 * config/arm/arm-cpus.in (fp16fml): New feature.
11809 (ALL_SIMD): Add fp16fml.
11810 (armv8.2-a): Add fp16fml as an option.
11811 (armv8.3-a): Likewise.
11812 (armv8.4-a): Add fp16fml as part of fp16.
11813 * config/arm/arm.h (TARGET_FP16FML): Define.
11814 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
11816 * config/arm/arm-modes.def (V2HF): Define.
11817 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
11818 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
11819 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
11820 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
11821 vfmsl_low, vfmsl_high): New set of builtins.
11822 * config/arm/iterators.md (PLUSMINUS): New code iterator.
11823 (vfml_op): New code attribute.
11824 (VFMLHALVES): New int iterator.
11825 (VFML, VFMLSEL): New mode attributes.
11826 (V_reg): Define mapping for V2HF.
11827 (V_hi, V_lo): New mode attributes.
11828 (VF_constraint): Likewise.
11829 (vfml_half, vfml_half_selector): New int attributes.
11830 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
11832 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
11833 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
11835 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
11836 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
11837 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
11838 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
11840 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
11841 Document new effective target and option set.
11843 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11845 * config/arm/arm-cpus.in (armv8_4): New feature.
11846 (ARMv8_4a): New fgroup.
11847 (armv8.4-a): New arch.
11848 * config/arm/arm-tables.opt: Regenerate.
11849 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
11850 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
11851 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
11852 Add matching rules for -march=armv8.4-a and extensions.
11853 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
11855 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
11858 * config/rx/rx.md (BW): New mode attribute.
11859 (sync_lock_test_and_setsi): Add mode suffix to insn output.
11861 2018-01-11 Richard Biener <rguenther@suse.de>
11863 PR tree-optimization/83435
11864 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
11865 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
11866 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
11868 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11869 Alan Hayward <alan.hayward@arm.com>
11870 David Sherwood <david.sherwood@arm.com>
11872 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
11874 (aarch64_classify_address): Initialize it. Track polynomial offsets.
11875 (aarch64_print_address_internal): Use it to check for a zero offset.
11877 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11878 Alan Hayward <alan.hayward@arm.com>
11879 David Sherwood <david.sherwood@arm.com>
11881 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
11882 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
11883 Return a poly_int64 rather than a HOST_WIDE_INT.
11884 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
11885 rather than a HOST_WIDE_INT.
11886 * config/aarch64/aarch64.h (aarch64_frame): Protect with
11887 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
11888 hard_fp_offset, frame_size, initial_adjust, callee_offset and
11889 final_offset from HOST_WIDE_INT to poly_int64.
11890 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
11891 to_constant when getting the number of units in an Advanced SIMD
11893 (aarch64_builtin_vectorized_function): Check for a constant number
11895 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
11897 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
11898 attribute instead of GET_MODE_NUNITS.
11899 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
11900 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
11901 GET_MODE_SIZE for fixed-size registers.
11902 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
11903 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
11904 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
11905 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
11906 (aarch64_print_operand, aarch64_print_address_internal)
11907 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
11908 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
11909 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
11910 Handle polynomial GET_MODE_SIZE.
11911 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
11912 wider than SImode without modification.
11913 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
11914 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
11915 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
11916 passing and returning SVE modes.
11917 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
11918 rather than GEN_INT.
11919 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
11920 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
11921 (aarch64_allocate_and_probe_stack_space): Likewise.
11922 (aarch64_layout_frame): Cope with polynomial offsets.
11923 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
11924 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
11925 polynomial offsets.
11926 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
11927 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
11928 poly_int64 rather than a HOST_WIDE_INT.
11929 (aarch64_get_separate_components, aarch64_process_components)
11930 (aarch64_expand_prologue, aarch64_expand_epilogue)
11931 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
11932 (aarch64_anchor_offset): New function, split out from...
11933 (aarch64_legitimize_address): ...here.
11934 (aarch64_builtin_vectorization_cost): Handle polynomial
11935 TYPE_VECTOR_SUBPARTS.
11936 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
11938 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
11939 number of elements from the PARALLEL rather than the mode.
11940 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
11941 rather than GET_MODE_BITSIZE.
11942 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
11943 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
11944 (aarch64_expand_vec_perm_const_1): Handle polynomial
11945 d->perm.length () and d->perm elements.
11946 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
11947 Apply to_constant to d->perm elements.
11948 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
11949 polynomial CONST_VECTOR_NUNITS.
11950 (aarch64_move_pointer): Take amount as a poly_int64 rather
11952 (aarch64_progress_pointer): Avoid temporary variable.
11953 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
11954 the mode attribute instead of GET_MODE.
11956 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11957 Alan Hayward <alan.hayward@arm.com>
11958 David Sherwood <david.sherwood@arm.com>
11960 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
11961 x exists before using it.
11962 (aarch64_add_constant_internal): Rename to...
11963 (aarch64_add_offset_1): ...this. Replace regnum with separate
11964 src and dest rtxes. Handle the case in which they're different,
11965 including when the offset is zero. Replace scratchreg with an rtx.
11966 Use 2 additions if there is no spare register into which we can
11967 move a 16-bit constant.
11968 (aarch64_add_constant): Delete.
11969 (aarch64_add_offset): Replace reg with separate src and dest
11970 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
11971 Use aarch64_add_offset_1.
11972 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
11973 an rtx rather than an int. Take the delta as a poly_int64
11974 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
11975 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
11976 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
11977 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
11978 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
11979 and aarch64_add_sp.
11980 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
11981 aarch64_add_constant.
11983 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11985 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
11986 Use scalar_float_mode.
11988 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11990 * config/aarch64/aarch64-simd.md
11991 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
11992 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
11993 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
11994 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
11995 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
11996 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
11997 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
11998 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
11999 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
12000 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
12002 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12005 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
12006 targ_options->x_arm_arch_string is non NULL.
12008 2018-01-11 Tamar Christina <tamar.christina@arm.com>
12010 * config/aarch64/aarch64.h
12011 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
12013 2018-01-11 Sudakshina Das <sudi.das@arm.com>
12016 * expmed.c (emit_store_flag_force): Swap if const op0
12017 and change VOIDmode to mode of op0.
12019 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12021 PR rtl-optimization/83761
12022 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
12023 than bytes to mode_for_size.
12025 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12027 PR middle-end/83189
12028 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
12029 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
12032 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12034 PR middle-end/83575
12035 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
12036 when in layout mode.
12037 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
12038 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
12041 2018-01-10 Michael Collison <michael.collison@arm.com>
12043 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
12044 * config/aarch64/aarch64-option-extension.def: Add
12045 AARCH64_OPT_EXTENSION of 'fp16fml'.
12046 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12047 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
12048 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
12049 * config/aarch64/constraints.md (Ui7): New constraint.
12050 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
12051 (VFMLA_SEL_W): Ditto.
12054 (VFMLA16_LOW): New int iterator.
12055 (VFMLA16_HIGH): Ditto.
12056 (UNSPEC_FMLAL): New unspec.
12057 (UNSPEC_FMLSL): Ditto.
12058 (UNSPEC_FMLAL2): Ditto.
12059 (UNSPEC_FMLSL2): Ditto.
12060 (f16mac): New code attribute.
12061 * config/aarch64/aarch64-simd-builtins.def
12062 (aarch64_fmlal_lowv2sf): Ditto.
12063 (aarch64_fmlsl_lowv2sf): Ditto.
12064 (aarch64_fmlalq_lowv4sf): Ditto.
12065 (aarch64_fmlslq_lowv4sf): Ditto.
12066 (aarch64_fmlal_highv2sf): Ditto.
12067 (aarch64_fmlsl_highv2sf): Ditto.
12068 (aarch64_fmlalq_highv4sf): Ditto.
12069 (aarch64_fmlslq_highv4sf): Ditto.
12070 (aarch64_fmlal_lane_lowv2sf): Ditto.
12071 (aarch64_fmlsl_lane_lowv2sf): Ditto.
12072 (aarch64_fmlal_laneq_lowv2sf): Ditto.
12073 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
12074 (aarch64_fmlalq_lane_lowv4sf): Ditto.
12075 (aarch64_fmlsl_lane_lowv4sf): Ditto.
12076 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
12077 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
12078 (aarch64_fmlal_lane_highv2sf): Ditto.
12079 (aarch64_fmlsl_lane_highv2sf): Ditto.
12080 (aarch64_fmlal_laneq_highv2sf): Ditto.
12081 (aarch64_fmlsl_laneq_highv2sf): Ditto.
12082 (aarch64_fmlalq_lane_highv4sf): Ditto.
12083 (aarch64_fmlsl_lane_highv4sf): Ditto.
12084 (aarch64_fmlalq_laneq_highv4sf): Ditto.
12085 (aarch64_fmlsl_laneq_highv4sf): Ditto.
12086 * config/aarch64/aarch64-simd.md:
12087 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
12088 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12089 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
12090 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12091 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
12092 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
12093 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
12094 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
12095 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
12096 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
12097 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
12098 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
12099 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
12100 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
12101 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
12102 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
12103 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
12104 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
12105 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
12106 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
12107 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
12108 (vfmlsl_low_u32): Ditto.
12109 (vfmlalq_low_u32): Ditto.
12110 (vfmlslq_low_u32): Ditto.
12111 (vfmlal_high_u32): Ditto.
12112 (vfmlsl_high_u32): Ditto.
12113 (vfmlalq_high_u32): Ditto.
12114 (vfmlslq_high_u32): Ditto.
12115 (vfmlal_lane_low_u32): Ditto.
12116 (vfmlsl_lane_low_u32): Ditto.
12117 (vfmlal_laneq_low_u32): Ditto.
12118 (vfmlsl_laneq_low_u32): Ditto.
12119 (vfmlalq_lane_low_u32): Ditto.
12120 (vfmlslq_lane_low_u32): Ditto.
12121 (vfmlalq_laneq_low_u32): Ditto.
12122 (vfmlslq_laneq_low_u32): Ditto.
12123 (vfmlal_lane_high_u32): Ditto.
12124 (vfmlsl_lane_high_u32): Ditto.
12125 (vfmlal_laneq_high_u32): Ditto.
12126 (vfmlsl_laneq_high_u32): Ditto.
12127 (vfmlalq_lane_high_u32): Ditto.
12128 (vfmlslq_lane_high_u32): Ditto.
12129 (vfmlalq_laneq_high_u32): Ditto.
12130 (vfmlslq_laneq_high_u32): Ditto.
12131 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
12132 (AARCH64_FL_FOR_ARCH8_4): New.
12133 (AARCH64_ISA_F16FML): New ISA flag.
12134 (TARGET_F16FML): New feature flag for fp16fml.
12135 (doc/invoke.texi): Document new fp16fml option.
12137 2018-01-10 Michael Collison <michael.collison@arm.com>
12139 * config/aarch64/aarch64-builtins.c:
12140 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
12141 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12142 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
12143 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
12144 (AARCH64_ISA_SHA3): New ISA flag.
12145 (TARGET_SHA3): New feature flag for sha3.
12146 * config/aarch64/iterators.md (sha512_op): New int attribute.
12147 (CRYPTO_SHA512): New int iterator.
12148 (UNSPEC_SHA512H): New unspec.
12149 (UNSPEC_SHA512H2): Ditto.
12150 (UNSPEC_SHA512SU0): Ditto.
12151 (UNSPEC_SHA512SU1): Ditto.
12152 * config/aarch64/aarch64-simd-builtins.def
12153 (aarch64_crypto_sha512hqv2di): New builtin.
12154 (aarch64_crypto_sha512h2qv2di): Ditto.
12155 (aarch64_crypto_sha512su0qv2di): Ditto.
12156 (aarch64_crypto_sha512su1qv2di): Ditto.
12157 (aarch64_eor3qv8hi): Ditto.
12158 (aarch64_rax1qv2di): Ditto.
12159 (aarch64_xarqv2di): Ditto.
12160 (aarch64_bcaxqv8hi): Ditto.
12161 * config/aarch64/aarch64-simd.md:
12162 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
12163 (aarch64_crypto_sha512su0qv2di): Ditto.
12164 (aarch64_crypto_sha512su1qv2di): Ditto.
12165 (aarch64_eor3qv8hi): Ditto.
12166 (aarch64_rax1qv2di): Ditto.
12167 (aarch64_xarqv2di): Ditto.
12168 (aarch64_bcaxqv8hi): Ditto.
12169 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
12170 (vsha512h2q_u64): Ditto.
12171 (vsha512su0q_u64): Ditto.
12172 (vsha512su1q_u64): Ditto.
12173 (veor3q_u16): Ditto.
12174 (vrax1q_u64): Ditto.
12175 (vxarq_u64): Ditto.
12176 (vbcaxq_u16): Ditto.
12177 * config/arm/types.md (crypto_sha512): New type attribute.
12178 (crypto_sha3): Ditto.
12179 (doc/invoke.texi): Document new sha3 option.
12181 2018-01-10 Michael Collison <michael.collison@arm.com>
12183 * config/aarch64/aarch64-builtins.c:
12184 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
12185 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12186 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
12187 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
12188 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
12189 (AARCH64_ISA_SM4): New ISA flag.
12190 (TARGET_SM4): New feature flag for sm4.
12191 * config/aarch64/aarch64-simd-builtins.def
12192 (aarch64_sm3ss1qv4si): Ditto.
12193 (aarch64_sm3tt1aq4si): Ditto.
12194 (aarch64_sm3tt1bq4si): Ditto.
12195 (aarch64_sm3tt2aq4si): Ditto.
12196 (aarch64_sm3tt2bq4si): Ditto.
12197 (aarch64_sm3partw1qv4si): Ditto.
12198 (aarch64_sm3partw2qv4si): Ditto.
12199 (aarch64_sm4eqv4si): Ditto.
12200 (aarch64_sm4ekeyqv4si): Ditto.
12201 * config/aarch64/aarch64-simd.md:
12202 (aarch64_sm3ss1qv4si): Ditto.
12203 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
12204 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
12205 (aarch64_sm4eqv4si): Ditto.
12206 (aarch64_sm4ekeyqv4si): Ditto.
12207 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
12208 (sm3part_op): Ditto.
12209 (CRYPTO_SM3TT): Ditto.
12210 (CRYPTO_SM3PART): Ditto.
12211 (UNSPEC_SM3SS1): New unspec.
12212 (UNSPEC_SM3TT1A): Ditto.
12213 (UNSPEC_SM3TT1B): Ditto.
12214 (UNSPEC_SM3TT2A): Ditto.
12215 (UNSPEC_SM3TT2B): Ditto.
12216 (UNSPEC_SM3PARTW1): Ditto.
12217 (UNSPEC_SM3PARTW2): Ditto.
12218 (UNSPEC_SM4E): Ditto.
12219 (UNSPEC_SM4EKEY): Ditto.
12220 * config/aarch64/constraints.md (Ui2): New constraint.
12221 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
12222 * config/arm/types.md (crypto_sm3): New type attribute.
12223 (crypto_sm4): Ditto.
12224 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
12225 (vsm3tt1aq_u32): Ditto.
12226 (vsm3tt1bq_u32): Ditto.
12227 (vsm3tt2aq_u32): Ditto.
12228 (vsm3tt2bq_u32): Ditto.
12229 (vsm3partw1q_u32): Ditto.
12230 (vsm3partw2q_u32): Ditto.
12231 (vsm4eq_u32): Ditto.
12232 (vsm4ekeyq_u32): Ditto.
12233 (doc/invoke.texi): Document new sm4 option.
12235 2018-01-10 Michael Collison <michael.collison@arm.com>
12237 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
12238 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
12239 (AARCH64_FL_FOR_ARCH8_4): New.
12240 (AARCH64_FL_V8_4): New flag.
12241 (doc/invoke.texi): Document new armv8.4-a option.
12243 2018-01-10 Michael Collison <michael.collison@arm.com>
12245 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12246 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
12247 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
12248 * config/aarch64/aarch64-option-extension.def: Add
12249 AARCH64_OPT_EXTENSION of 'sha2'.
12250 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
12251 (crypto): Disable sha2 and aes if crypto disabled.
12252 (crypto): Enable aes and sha2 if enabled.
12253 (simd): Disable sha2 and aes if simd disabled.
12254 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
12256 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
12257 (TARGET_SHA2): New feature flag for sha2.
12258 (TARGET_AES): New feature flag for aes.
12259 * config/aarch64/aarch64-simd.md:
12260 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
12261 conditional on TARGET_AES.
12262 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
12263 (aarch64_crypto_sha1hsi): Make pattern conditional
12265 (aarch64_crypto_sha1hv4si): Ditto.
12266 (aarch64_be_crypto_sha1hv4si): Ditto.
12267 (aarch64_crypto_sha1su1v4si): Ditto.
12268 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
12269 (aarch64_crypto_sha1su0v4si): Ditto.
12270 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
12271 (aarch64_crypto_sha256su0v4si): Ditto.
12272 (aarch64_crypto_sha256su1v4si): Ditto.
12273 (doc/invoke.texi): Document new aes and sha2 options.
12275 2018-01-10 Martin Sebor <msebor@redhat.com>
12277 PR tree-optimization/83781
12278 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
12281 2018-01-11 Martin Sebor <msebor@gmail.com>
12282 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12284 PR tree-optimization/83501
12285 PR tree-optimization/81703
12287 * tree-ssa-strlen.c (get_string_cst): Rename...
12288 (get_string_len): ...to this. Handle global constants.
12289 (handle_char_store): Adjust.
12291 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
12292 Jim Wilson <jimw@sifive.com>
12294 * config/riscv/riscv-protos.h (riscv_output_return): New.
12295 * config/riscv/riscv.c (struct machine_function): New naked_p field.
12296 (riscv_attribute_table, riscv_output_return),
12297 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
12298 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
12299 (riscv_compute_frame_info): Only compute frame->mask if not a naked
12301 (riscv_expand_prologue): Add early return for naked function.
12302 (riscv_expand_epilogue): Likewise.
12303 (riscv_function_ok_for_sibcall): Return false for naked function.
12304 (riscv_set_current_function): New.
12305 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
12306 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
12307 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
12308 * doc/extend.texi (RISC-V Function Attributes): New.
12310 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
12312 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
12313 check for 128-bit long double before checking TCmode.
12314 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
12315 128-bit long doubles before checking TFmode or TCmode.
12316 (FLOAT128_IBM_P): Likewise.
12318 2018-01-10 Martin Sebor <msebor@redhat.com>
12320 PR tree-optimization/83671
12321 * builtins.c (c_strlen): Unconditionally return zero for the empty
12323 Use -Warray-bounds for warnings.
12324 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
12325 for non-constant array indices with COMPONENT_REF, arrays of
12326 arrays, and pointers to arrays.
12327 (gimple_fold_builtin_strlen): Determine and set length range for
12328 non-constant character arrays.
12330 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
12332 PR middle-end/81897
12333 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
12336 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
12338 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
12340 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12343 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
12344 VECTOR_MEM_ALTIVEC_OR_VSX_P.
12345 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
12346 indexed_or_indirect_operand predicate.
12347 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
12348 (*vsx_le_perm_load_v8hi): Likewise.
12349 (*vsx_le_perm_load_v16qi): Likewise.
12350 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
12351 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
12352 (*vsx_le_perm_store_v8hi): Likewise.
12353 (*vsx_le_perm_store_v16qi): Likewise.
12354 (eight unnamed splitters): Likewise.
12356 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12358 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
12359 * config/rs6000/emmintrin.h: Likewise.
12360 * config/rs6000/mmintrin.h: Likewise.
12361 * config/rs6000/xmmintrin.h: Likewise.
12363 2018-01-10 David Malcolm <dmalcolm@redhat.com>
12366 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
12368 * tree.c (tree_nop_conversion): Return true for location wrapper
12370 (maybe_wrap_with_location): New function.
12371 (selftest::check_strip_nops): New function.
12372 (selftest::test_location_wrappers): New function.
12373 (selftest::tree_c_tests): Call it.
12374 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
12375 (maybe_wrap_with_location): New decl.
12376 (EXPR_LOCATION_WRAPPER_P): New macro.
12377 (location_wrapper_p): New inline function.
12378 (tree_strip_any_location_wrapper): New inline function.
12380 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
12383 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
12384 stack_realign_offset for the largest alignment of stack slot
12386 (ix86_find_max_used_stack_alignment): New function.
12387 (ix86_finalize_stack_frame_flags): Use it. Set
12388 max_used_stack_alignment if we don't realign stack.
12389 * config/i386/i386.h (machine_function): Add
12390 max_used_stack_alignment.
12392 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
12394 * config/arm/arm.opt (-mbranch-cost): New option.
12395 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
12398 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
12401 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
12402 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
12404 2018-01-10 Richard Biener <rguenther@suse.de>
12407 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
12408 early out so it also covers the case where we have a non-NULL
12411 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12413 PR tree-optimization/83753
12414 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
12415 for non-strided grouped accesses if the number of elements is 1.
12417 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12420 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
12421 * i386.h (TARGET_USE_GATHER): Define.
12422 * x86-tune.def (X86_TUNE_USE_GATHER): New.
12424 2018-01-10 Martin Liska <mliska@suse.cz>
12427 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
12428 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
12430 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
12431 CLEANUP_NO_PARTITIONING is not set.
12433 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12435 * doc/rtl.texi: Remove documentation of (const ...) wrappers
12436 for vectors, as a partial revert of r254296.
12437 * rtl.h (const_vec_p): Delete.
12438 (const_vec_duplicate_p): Don't test for vector CONSTs.
12439 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
12440 * expmed.c (make_tree): Likewise.
12443 * common.md (E, F): Use CONSTANT_P instead of checking for
12445 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
12446 checking for CONST_VECTOR.
12448 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12450 PR middle-end/83575
12451 * predict.c (force_edge_cold): Handle in more sane way edges
12452 with no prediction.
12454 2018-01-09 Carl Love <cel@us.ibm.com>
12456 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
12458 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
12459 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
12460 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
12461 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
12462 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
12463 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
12464 * config/rs6000/rs6000-protos.h: Add extern defition for
12465 rs6000_generate_float2_double_code.
12466 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
12468 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
12469 (float2_v2df): Add define_expand.
12471 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
12474 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
12475 op_mode in the force_to_mode call.
12477 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12479 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
12480 instead of checking each element individually.
12481 (aarch64_evpc_uzp): Likewise.
12482 (aarch64_evpc_zip): Likewise.
12483 (aarch64_evpc_ext): Likewise.
12484 (aarch64_evpc_rev): Likewise.
12485 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
12486 instead of checking each element individually. Return true without
12488 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
12489 whether all selected elements come from the same input, instead of
12490 checking each element individually. Remove calls to gen_rtx_REG,
12491 start_sequence and end_sequence and instead assert that no rtl is
12494 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12496 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
12497 order of HIGH and CONST checks.
12499 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12501 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
12502 if the destination isn't an SSA_NAME.
12504 2018-01-09 Richard Biener <rguenther@suse.de>
12506 PR tree-optimization/83668
12507 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
12509 (canonicalize_loop_form): ... here, renamed from ...
12510 (canonicalize_loop_closed_ssa_form): ... this and amended to
12511 swap successor edges for loop exit blocks to make us use
12512 the RPO order we need for initial schedule generation.
12514 2018-01-09 Joseph Myers <joseph@codesourcery.com>
12516 PR tree-optimization/64811
12517 * match.pd: When optimizing comparisons with Inf, avoid
12518 introducing or losing exceptions from comparisons with NaN.
12520 2018-01-09 Martin Liska <mliska@suse.cz>
12523 * asan.c (shadow_mem_size): Add gcc_assert.
12525 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
12527 Don't save registers in main().
12530 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
12531 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
12532 * config/avr/avr.c (avr_set_current_function): Don't error if
12533 naked, OS_task or OS_main are specified at the same time.
12534 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
12536 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
12538 * common/config/avr/avr-common.c (avr_option_optimization_table):
12539 Switch on -mmain-is-OS_task for optimizing compilations.
12541 2018-01-09 Richard Biener <rguenther@suse.de>
12543 PR tree-optimization/83572
12544 * graphite.c: Include cfganal.h.
12545 (graphite_transform_loops): Connect infinite loops to exit
12546 and remove fake edges at the end.
12548 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12550 * ipa-inline.c (edge_badness): Revert accidental checkin.
12552 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12555 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
12556 symbols; not inline clones.
12558 2018-01-09 Jakub Jelinek <jakub@redhat.com>
12561 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
12562 hard registers. Formatting fixes.
12564 PR preprocessor/83722
12565 * gcc.c (try_generate_repro): Pass
12566 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
12567 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
12570 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
12571 Kito Cheng <kito.cheng@gmail.com>
12573 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
12574 (riscv_leaf_function_p): Delete.
12575 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
12577 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12579 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
12581 (do_ifelse): New function.
12582 (do_isel): New function.
12583 (do_sub3): New function.
12584 (do_add3): New function.
12585 (do_load_mask_compare): New function.
12586 (do_overlap_load_compare): New function.
12587 (expand_compare_loop): New function.
12588 (expand_block_compare): Call expand_compare_loop() when appropriate.
12589 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
12590 option description.
12591 (-mblock-compare-inline-loop-limit): New option.
12593 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12596 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
12597 Reverse order of second and third operands in first alternative.
12598 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
12599 of first and second elements in UNSPEC_VPERMR vector.
12600 (altivec_expand_vec_perm_le): Likewise.
12602 2018-01-08 Jeff Law <law@redhat.com>
12604 PR rtl-optimizatin/81308
12605 * tree-switch-conversion.c (cfg_altered): New file scoped static.
12606 (process_switch): If group_case_labels makes a change, then set
12608 (pass_convert_switch::execute): If a switch is converted, then
12609 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
12611 PR rtl-optimization/81308
12612 * recog.c (split_all_insns): Conditionally cleanup the CFG after
12615 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
12617 PR target/83663 - Revert r255946
12618 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
12619 generation for cases where splatting a value is not useful.
12620 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
12621 across a vec_duplicate and a paradoxical subreg forming a vector
12622 mode to a vec_concat.
12624 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12626 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
12627 -march=armv8.3-a variants.
12628 * config/arm/t-multilib: Likewise.
12629 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
12631 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12633 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
12635 (cceq_ior_compare_complement): Give it a name so I can use it, and
12636 change boolean_or_operator predicate to boolean_operator so it can
12637 be used to generate a crand.
12638 (eqne): New code iterator.
12639 (bd/bd_neg): New code_attrs.
12640 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
12641 a single define_insn.
12642 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
12643 decrement (bdnzt/bdnzf/bdzt/bdzf).
12644 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
12645 with the new names of the branch decrement patterns, and added the
12646 names of the branch decrement conditional patterns.
12648 2018-01-08 Richard Biener <rguenther@suse.de>
12650 PR tree-optimization/83563
12651 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
12654 2018-01-08 Richard Biener <rguenther@suse.de>
12656 PR middle-end/83713
12657 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
12659 2018-01-08 Richard Biener <rguenther@suse.de>
12661 PR tree-optimization/83685
12662 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
12663 references to abnormals.
12665 2018-01-08 Richard Biener <rguenther@suse.de>
12668 * dwarf2out.c (output_indirect_strings): Handle empty
12669 skeleton_debug_str_hash.
12670 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
12672 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12674 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
12675 (emit_store_direct): Likewise.
12676 (arc_trampoline_adjust_address): Likewise.
12677 (arc_asm_trampoline_template): New function.
12678 (arc_initialize_trampoline): Use asm_trampoline_template.
12679 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
12680 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
12681 * config/arc/arc.md (flush_icache): Delete pattern.
12683 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12685 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
12686 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
12689 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12692 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
12693 by not USED_FOR_TARGET.
12694 (make_pass_resolve_sw_modes): Likewise.
12696 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12698 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
12701 2018-01-08 Richard Biener <rguenther@suse.de>
12703 PR middle-end/83580
12704 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
12706 2018-01-08 Richard Biener <rguenther@suse.de>
12708 PR middle-end/83517
12709 * match.pd ((t * 2) / 2) -> t): Add missing :c.
12711 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
12713 PR middle-end/81897
12714 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
12715 basic blocks with a small number of successors.
12716 (convert_control_dep_chain_into_preds): Improve handling of
12718 (dump_predicates): Split apart into...
12719 (dump_pred_chain): ...here...
12720 (dump_pred_info): ...and here.
12721 (can_one_predicate_be_invalidated_p): Add debugging printfs.
12722 (can_chain_union_be_invalidated_p): Improve check for invalidation
12724 (uninit_uses_cannot_happen): Avoid unnecessary if
12725 convert_control_dep_chain_into_preds yielded nothing.
12727 2018-01-06 Martin Sebor <msebor@redhat.com>
12729 PR tree-optimization/83640
12730 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
12731 subtracting negative offset from size.
12732 (builtin_access::overlap): Adjust offset bounds of the access to fall
12733 within the size of the object if possible.
12735 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
12737 PR rtl-optimization/83699
12738 * expmed.c (extract_bit_field_1): Restrict the vector usage of
12739 extract_bit_field_as_subreg to cases in which the extracted
12740 value is also a vector.
12742 * lra-constraints.c (process_alt_operands): Test for the equivalence
12743 substitutions when detecting a possible reload cycle.
12745 2018-01-06 Jakub Jelinek <jakub@redhat.com>
12748 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
12749 by default if flag_selective_schedling{,2}. Formatting fixes.
12751 PR rtl-optimization/83682
12752 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
12753 if it has non-VECTOR_MODE element mode.
12754 (vec_duplicate_p): Likewise.
12756 PR middle-end/83694
12757 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
12758 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
12760 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12763 * config/i386/i386-builtin.def
12764 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
12765 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
12766 Require also OPTION_MASK_ISA_AVX512F in addition to
12767 OPTION_MASK_ISA_GFNI.
12768 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
12769 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
12770 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
12771 to OPTION_MASK_ISA_GFNI.
12772 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
12773 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
12774 OPTION_MASK_ISA_AVX512BW.
12775 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
12776 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
12777 addition to OPTION_MASK_ISA_GFNI.
12778 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
12779 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
12780 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
12781 to OPTION_MASK_ISA_GFNI.
12782 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
12783 a requirement for all ISAs rather than any of them with a few
12785 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
12787 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
12788 bitmasks to be enabled with 3 exceptions, instead of requiring any
12789 enabled ISA with lots of exceptions.
12790 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
12791 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
12792 Change avx512bw in isa attribute to avx512f.
12793 * config/i386/sgxintrin.h: Add license boilerplate.
12794 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
12795 to __AVX512F__ and __AVX512VL to __AVX512VL__.
12796 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
12797 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
12799 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
12800 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
12801 temporarily sse2 rather than sse if not enabled already.
12804 * config/i386/sse.md (VI248_VLBW): Rename to ...
12805 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
12806 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
12807 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
12808 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
12809 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
12810 mode iterator instead of VI248_VLBW.
12812 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
12814 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
12815 (record_modified): Skip clobbers; add debug output.
12816 (param_change_prob): Use sreal frequencies.
12818 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12820 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
12821 punt for user-aligned variables.
12823 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12825 * tree-chrec.c (chrec_contains_symbols): Return true for
12828 2018-01-05 Sudakshina Das <sudi.das@arm.com>
12831 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
12832 of (x|y) == x for BICS pattern.
12834 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12836 PR tree-optimization/83605
12837 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
12838 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
12841 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
12843 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
12844 * config/epiphany/rtems.h: New file.
12846 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12847 Uros Bizjak <ubizjak@gmail.com>
12850 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
12851 QIreg_operand instead of register_operand predicate.
12852 * config/i386/i386.c (ix86_rop_should_change_byte_p,
12853 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
12854 comments instead of -fmitigate[-_]rop.
12856 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12859 * cgraphunit.c (symbol_table::compile): Switch to text_section
12860 before calling assembly_start debug hook.
12861 * run-rtl-passes.c (run_rtl_passes): Likewise.
12864 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12866 * tree-vrp.c (extract_range_from_binary_expr_1): Check
12867 range_int_cst_p rather than !symbolic_range_p before calling
12868 extract_range_from_multiplicative_op_1.
12870 2018-01-04 Jeff Law <law@redhat.com>
12872 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
12873 redundant test in assertion.
12875 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12877 * doc/rtl.texi: Document machine_mode wrapper classes.
12879 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12881 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
12882 using tree_to_uhwi.
12884 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12886 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
12887 the VEC_PERM_EXPR fold to fail.
12889 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12892 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
12893 to switched_sections.
12895 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12898 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
12899 test for d.testing.
12901 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
12904 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
12905 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
12907 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12910 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
12911 is BLKmode and bitpos not zero or mode change is needed.
12913 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12916 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
12919 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
12922 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
12923 instead of MULT rtx. Update all corresponding splitters.
12924 (*saddl_se): Ditto.
12925 (*ssub<modesuffix>): Ditto.
12926 (*ssubl_se): Ditto.
12927 (*cmp_sadd_di): Update split patterns.
12928 (*cmp_sadd_si): Ditto.
12929 (*cmp_sadd_sidi): Ditto.
12930 (*cmp_ssub_di): Ditto.
12931 (*cmp_ssub_si): Ditto.
12932 (*cmp_ssub_sidi): Ditto.
12933 * config/alpha/predicates.md (const23_operand): New predicate.
12934 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
12935 Look for ASHIFT, not MULT inner operand.
12936 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
12938 2018-01-04 Martin Liska <mliska@suse.cz>
12940 PR gcov-profile/83669
12941 * gcov.c (output_intermediate_file): Add version to intermediate
12943 * doc/gcov.texi: Document new field 'version' in intermediate
12944 file format. Fix location of '-k' option of gcov command.
12946 2018-01-04 Martin Liska <mliska@suse.cz>
12949 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
12951 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12953 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
12955 2018-01-03 Martin Sebor <msebor@redhat.com>
12957 PR tree-optimization/83655
12958 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
12959 checking calls with invalid arguments.
12961 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12963 * tree-vect-stmts.c (vect_get_store_rhs): New function.
12964 (vectorizable_mask_load_store): Delete.
12965 (vectorizable_call): Return false for masked loads and stores.
12966 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
12967 instead of gimple_assign_rhs1.
12968 (vectorizable_load): Handle IFN_MASK_LOAD.
12969 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
12971 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12973 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
12975 (vectorizable_mask_load_store): ...here.
12976 (vectorizable_load): ...and here.
12978 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12980 * tree-vect-stmts.c (vect_build_all_ones_mask)
12981 (vect_build_zero_merge_argument): New functions, split out from...
12982 (vectorizable_load): ...here.
12984 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12986 * tree-vect-stmts.c (vect_check_store_rhs): New function,
12988 (vectorizable_mask_load_store): ...here.
12989 (vectorizable_store): ...and here.
12991 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12993 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
12995 (vectorizable_mask_load_store): ...here.
12997 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12999 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
13000 (vect_model_store_cost): Take a vec_load_store_type instead of a
13002 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
13003 (vect_model_store_cost): Take a vec_load_store_type instead of a
13005 (vectorizable_mask_load_store): Update accordingly.
13006 (vectorizable_store): Likewise.
13007 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
13009 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13011 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
13012 IFN_MASK_LOAD calls here rather than...
13013 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
13015 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13016 Alan Hayward <alan.hayward@arm.com>
13017 David Sherwood <david.sherwood@arm.com>
13019 * expmed.c (extract_bit_field_1): For vector extracts,
13020 fall back to extract_bit_field_as_subreg if vec_extract
13023 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13024 Alan Hayward <alan.hayward@arm.com>
13025 David Sherwood <david.sherwood@arm.com>
13027 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
13028 they are variable or constant sized.
13029 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
13030 slots for constant-sized data.
13032 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13033 Alan Hayward <alan.hayward@arm.com>
13034 David Sherwood <david.sherwood@arm.com>
13036 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
13037 handling COND_EXPRs with boolean comparisons, try to find a better
13038 basis for the mask type than the boolean itself.
13040 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13042 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
13043 is calculated and how it can be overridden.
13044 * genmodes.c (max_bitsize_mode_any_mode): New variable.
13045 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
13047 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
13050 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13051 Alan Hayward <alan.hayward@arm.com>
13052 David Sherwood <david.sherwood@arm.com>
13054 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
13055 Remove the mode argument.
13056 (aarch64_simd_valid_immediate): Remove the mode and inverse
13058 * config/aarch64/iterators.md (bitsize): New iterator.
13059 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
13060 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
13061 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
13062 aarch64_simd_valid_immediate.
13063 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
13064 (aarch64_reg_or_bic_imm): Likewise.
13065 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
13066 with an insn_type enum and msl with a modifier_type enum.
13067 Replace element_width with a scalar_mode. Change the shift
13068 to unsigned int. Add constructors for scalar_float_mode and
13069 scalar_int_mode elements.
13070 (aarch64_vect_float_const_representable_p): Delete.
13071 (aarch64_can_const_movi_rtx_p)
13072 (aarch64_simd_scalar_immediate_valid_for_move)
13073 (aarch64_simd_make_constant): Update call to
13074 aarch64_simd_valid_immediate.
13075 (aarch64_advsimd_valid_immediate_hs): New function.
13076 (aarch64_advsimd_valid_immediate): Likewise.
13077 (aarch64_simd_valid_immediate): Remove mode and inverse
13078 arguments. Rewrite to use the above. Use const_vec_duplicate_p
13079 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
13080 and aarch64_float_const_representable_p on the result.
13081 (aarch64_output_simd_mov_immediate): Remove mode argument.
13082 Update call to aarch64_simd_valid_immediate and use of
13083 simd_immediate_info.
13084 (aarch64_output_scalar_simd_mov_immediate): Update call
13087 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13088 Alan Hayward <alan.hayward@arm.com>
13089 David Sherwood <david.sherwood@arm.com>
13091 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
13092 (mode_nunits): Likewise CONST_MODE_NUNITS.
13093 * machmode.def (ADJUST_NUNITS): Document.
13094 * genmodes.c (mode_data::need_nunits_adj): New field.
13095 (blank_mode): Update accordingly.
13096 (adj_nunits): New variable.
13097 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
13099 (emit_mode_size_inline): Set need_bytesize_adj for all modes
13100 listed in adj_nunits.
13101 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
13102 listed in adj_nunits. Don't emit case statements for such modes.
13103 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
13104 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
13105 nothing if adj_nunits is nonnull.
13106 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
13107 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
13108 (emit_mode_fbit): Update use of print_maybe_const_decl.
13109 (emit_move_size): Likewise. Treat the array as non-const
13111 (emit_mode_adjustments): Handle adj_nunits.
13113 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13115 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
13116 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
13117 (VECTOR_MODES): Use it.
13118 (make_vector_modes): Take the prefix as an argument.
13120 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13121 Alan Hayward <alan.hayward@arm.com>
13122 David Sherwood <david.sherwood@arm.com>
13124 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
13125 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
13126 for MODE_VECTOR_BOOL.
13127 * machmode.def (VECTOR_BOOL_MODE): Document.
13128 * genmodes.c (VECTOR_BOOL_MODE): New macro.
13129 (make_vector_bool_mode): New function.
13130 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
13132 * lto-streamer-in.c (lto_input_mode_table): Likewise.
13133 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
13135 * stor-layout.c (int_mode_for_mode): Likewise.
13136 * tree.c (build_vector_type_for_mode): Likewise.
13137 * varasm.c (output_constant_pool_2): Likewise.
13138 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
13139 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
13140 for MODE_VECTOR_BOOL.
13141 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
13142 of mode class checks.
13143 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
13144 instead of a list of mode class checks.
13145 (expand_vector_scalar_condition): Likewise.
13146 (type_for_widest_vector_mode): Handle BImode as an inner mode.
13148 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13149 Alan Hayward <alan.hayward@arm.com>
13150 David Sherwood <david.sherwood@arm.com>
13152 * machmode.h (mode_size): Change from unsigned short to
13154 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
13155 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13156 or if measurement_type is not polynomial.
13157 (fixed_size_mode::includes_p): Check for constant-sized modes.
13158 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
13159 return a poly_uint16 rather than an unsigned short.
13160 (emit_mode_size): Change the type of mode_size from unsigned short
13161 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
13162 (emit_mode_adjustments): Cope with polynomial vector sizes.
13163 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13165 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13167 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
13168 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
13169 * caller-save.c (setup_save_areas): Likewise.
13170 (replace_reg_with_saved_mem): Likewise.
13171 * calls.c (emit_library_call_value_1): Likewise.
13172 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
13173 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
13174 (gen_lowpart_for_combine): Likewise.
13175 * convert.c (convert_to_integer_1): Likewise.
13176 * cse.c (equiv_constant, cse_insn): Likewise.
13177 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
13178 (cselib_subst_to_values): Likewise.
13179 * dce.c (word_dce_process_block): Likewise.
13180 * df-problems.c (df_word_lr_mark_ref): Likewise.
13181 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
13182 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
13183 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
13184 (rtl_for_decl_location): Likewise.
13185 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
13186 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
13187 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
13188 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
13189 (expand_expr_real_1): Likewise.
13190 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
13191 (pad_below): Likewise.
13192 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13193 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
13194 * ira.c (get_subreg_tracking_sizes): Likewise.
13195 * ira-build.c (ira_create_allocno_objects): Likewise.
13196 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
13197 (ira_sort_regnos_for_alter_reg): Likewise.
13198 * ira-costs.c (record_operand_costs): Likewise.
13199 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
13200 (resolve_simple_move): Likewise.
13201 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
13202 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
13203 (lra_constraints): Likewise.
13204 (CONST_POOL_OK_P): Reject variable-sized modes.
13205 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
13206 (add_pseudo_to_slot, lra_spill): Likewise.
13207 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13208 * optabs-query.c (get_best_extraction_insn): Likewise.
13209 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13210 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
13211 (expand_mult_highpart, valid_multiword_target_p): Likewise.
13212 * recog.c (offsettable_address_addr_space_p): Likewise.
13213 * regcprop.c (maybe_mode_change): Likewise.
13214 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
13215 * regrename.c (build_def_use): Likewise.
13216 * regstat.c (dump_reg_info): Likewise.
13217 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
13218 (find_reloads, find_reloads_subreg_address): Likewise.
13219 * reload1.c (eliminate_regs_1): Likewise.
13220 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
13221 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
13222 (simplify_binary_operation_1, simplify_subreg): Likewise.
13223 * targhooks.c (default_function_arg_padding): Likewise.
13224 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
13225 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
13226 (verify_gimple_assign_ternary): Likewise.
13227 * tree-inline.c (estimate_move_cost): Likewise.
13228 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13229 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
13230 (get_address_cost_ainc): Likewise.
13231 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
13232 (vect_supportable_dr_alignment): Likewise.
13233 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13234 (vectorizable_reduction): Likewise.
13235 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
13236 (vectorizable_operation, vectorizable_load): Likewise.
13237 * tree.c (build_same_sized_truth_vector_type): Likewise.
13238 * valtrack.c (cleanup_auto_inc_dec): Likewise.
13239 * var-tracking.c (emit_note_insn_var_location): Likewise.
13240 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
13241 (ADDR_VEC_ALIGN): Likewise.
13243 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13244 Alan Hayward <alan.hayward@arm.com>
13245 David Sherwood <david.sherwood@arm.com>
13247 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
13249 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13250 or if measurement_type is polynomial.
13251 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
13252 * combine.c (make_extraction): Likewise.
13253 * dse.c (find_shift_sequence): Likewise.
13254 * dwarf2out.c (mem_loc_descriptor): Likewise.
13255 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
13256 (extract_bit_field, extract_low_bits): Likewise.
13257 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
13258 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
13259 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
13260 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
13261 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13262 * reload.c (find_reloads): Likewise.
13263 * reload1.c (alter_reg): Likewise.
13264 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
13265 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
13266 * tree-if-conv.c (predicate_mem_writes): Likewise.
13267 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
13268 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
13269 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
13270 * valtrack.c (dead_debug_insert_temp): Likewise.
13271 * varasm.c (mergeable_constant_section): Likewise.
13272 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
13274 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13275 Alan Hayward <alan.hayward@arm.com>
13276 David Sherwood <david.sherwood@arm.com>
13278 * expr.c (expand_assignment): Cope with polynomial mode sizes
13279 when assigning to a CONCAT.
13281 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13282 Alan Hayward <alan.hayward@arm.com>
13283 David Sherwood <david.sherwood@arm.com>
13285 * machmode.h (mode_precision): Change from unsigned short to
13287 (mode_to_precision): Return a poly_uint16 rather than an unsigned
13289 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
13290 or if measurement_type is not polynomial.
13291 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
13292 in which the mode is already known to be a scalar_int_mode.
13293 * genmodes.c (emit_mode_precision): Change the type of mode_precision
13294 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
13296 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13297 for GET_MODE_PRECISION.
13298 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13299 for GET_MODE_PRECISION.
13300 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
13302 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
13303 (expand_field_assignment, make_extraction): Likewise.
13304 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
13305 (get_last_value): Likewise.
13306 * convert.c (convert_to_integer_1): Likewise.
13307 * cse.c (cse_insn): Likewise.
13308 * expr.c (expand_expr_real_1): Likewise.
13309 * lra-constraints.c (simplify_operand_subreg): Likewise.
13310 * optabs-query.c (can_atomic_load_p): Likewise.
13311 * optabs.c (expand_atomic_load): Likewise.
13312 (expand_atomic_store): Likewise.
13313 * ree.c (combine_reaching_defs): Likewise.
13314 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
13315 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
13316 * tree.h (type_has_mode_precision_p): Likewise.
13317 * ubsan.c (instrument_si_overflow): Likewise.
13319 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13320 Alan Hayward <alan.hayward@arm.com>
13321 David Sherwood <david.sherwood@arm.com>
13323 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
13324 polynomial numbers of units.
13325 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
13326 (valid_vector_subparts_p): New function.
13327 (build_vector_type): Remove temporary shim and take the number
13328 of units as a poly_uint64 rather than an int.
13329 (build_opaque_vector_type): Take the number of units as a
13330 poly_uint64 rather than an int.
13331 * tree.c (build_vector_from_ctor): Handle polynomial
13332 TYPE_VECTOR_SUBPARTS.
13333 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
13334 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
13335 (build_vector_from_val): If the number of units is variable,
13336 use build_vec_duplicate_cst for constant operands and
13337 VEC_DUPLICATE_EXPR otherwise.
13338 (make_vector_type): Remove temporary is_constant ().
13339 (build_vector_type, build_opaque_vector_type): Take the number of
13340 units as a poly_uint64 rather than an int.
13341 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
13343 * cfgexpand.c (expand_debug_expr): Likewise.
13344 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
13345 (store_constructor, expand_expr_real_1): Likewise.
13346 (const_scalar_mask_from_tree): Likewise.
13347 * fold-const-call.c (fold_const_reduction): Likewise.
13348 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
13349 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
13350 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
13351 (fold_relational_const): Likewise.
13352 (native_interpret_vector): Likewise. Change the size from an
13353 int to an unsigned int.
13354 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
13355 TYPE_VECTOR_SUBPARTS.
13356 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
13357 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
13358 duplicating a non-constant operand into a variable-length vector.
13359 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
13360 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
13361 * ipa-icf.c (sem_variable::equals): Likewise.
13362 * match.pd: Likewise.
13363 * omp-simd-clone.c (simd_clone_subparts): Likewise.
13364 * print-tree.c (print_node): Likewise.
13365 * stor-layout.c (layout_type): Likewise.
13366 * targhooks.c (default_builtin_vectorization_cost): Likewise.
13367 * tree-cfg.c (verify_gimple_comparison): Likewise.
13368 (verify_gimple_assign_binary): Likewise.
13369 (verify_gimple_assign_ternary): Likewise.
13370 (verify_gimple_assign_single): Likewise.
13371 * tree-pretty-print.c (dump_generic_node): Likewise.
13372 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13373 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
13374 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
13375 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
13376 (vect_shift_permute_load_chain): Likewise.
13377 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
13378 (expand_vector_condition, optimize_vector_constructor): Likewise.
13379 (lower_vec_perm, get_compute_type): Likewise.
13380 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13381 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
13382 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
13383 (vect_recog_mask_conversion_pattern): Likewise.
13384 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
13385 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
13386 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13387 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
13388 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
13389 (vectorizable_shift, vectorizable_operation, vectorizable_store)
13390 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
13391 (supportable_widening_operation): Likewise.
13392 (supportable_narrowing_operation): Likewise.
13393 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
13395 * varasm.c (output_constant): Likewise.
13397 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13398 Alan Hayward <alan.hayward@arm.com>
13399 David Sherwood <david.sherwood@arm.com>
13401 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
13402 so that both the length == 3 and length != 3 cases set up their
13403 own permute vectors. Add comments explaining why we know the
13404 number of elements is constant.
13405 (vect_permute_load_chain): Likewise.
13407 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13408 Alan Hayward <alan.hayward@arm.com>
13409 David Sherwood <david.sherwood@arm.com>
13411 * machmode.h (mode_nunits): Change from unsigned char to
13413 (ONLY_FIXED_SIZE_MODES): New macro.
13414 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
13415 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
13416 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
13418 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
13419 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
13420 or if measurement_type is not polynomial.
13421 * genmodes.c (ZERO_COEFFS): New macro.
13422 (emit_mode_nunits_inline): Make mode_nunits_inline return a
13424 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
13425 Use ZERO_COEFFS when emitting initializers.
13426 * data-streamer.h (bp_pack_poly_value): New function.
13427 (bp_unpack_poly_value): Likewise.
13428 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13429 for GET_MODE_NUNITS.
13430 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13431 for GET_MODE_NUNITS.
13432 * tree.c (make_vector_type): Remove temporary shim and make
13433 the real function take the number of units as a poly_uint64
13434 rather than an int.
13435 (build_vector_type_for_mode): Handle polynomial nunits.
13436 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
13437 * emit-rtl.c (const_vec_series_p_1): Likewise.
13438 (gen_rtx_CONST_VECTOR): Likewise.
13439 * fold-const.c (test_vec_duplicate_folding): Likewise.
13440 * genrecog.c (validate_pattern): Likewise.
13441 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
13442 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13443 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
13444 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
13445 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
13446 * rtlanal.c (subreg_get_info): Likewise.
13447 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13448 (vect_grouped_load_supported): Likewise.
13449 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
13450 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
13451 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
13452 (simplify_const_unary_operation, simplify_binary_operation_1)
13453 (simplify_const_binary_operation, simplify_ternary_operation)
13454 (test_vector_ops_duplicate, test_vector_ops): Likewise.
13455 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
13456 instead of CONST_VECTOR_NUNITS.
13457 * varasm.c (output_constant_pool_2): Likewise.
13458 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
13459 explicit-encoded elements in the XVEC for variable-length vectors.
13461 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13463 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
13465 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13466 Alan Hayward <alan.hayward@arm.com>
13467 David Sherwood <david.sherwood@arm.com>
13469 * coretypes.h (fixed_size_mode): Declare.
13470 (fixed_size_mode_pod): New typedef.
13471 * builtins.h (target_builtins::x_apply_args_mode)
13472 (target_builtins::x_apply_result_mode): Change type to
13473 fixed_size_mode_pod.
13474 * builtins.c (apply_args_size, apply_result_size, result_vector)
13475 (expand_builtin_apply_args_1, expand_builtin_apply)
13476 (expand_builtin_return): Update accordingly.
13478 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13480 * cse.c (hash_rtx_cb): Hash only the encoded elements.
13481 * cselib.c (cselib_hash_rtx): Likewise.
13482 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
13483 CONST_VECTOR encoding.
13485 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13486 Jeff Law <law@redhat.com>
13489 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
13490 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
13491 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
13492 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
13495 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
13496 explicitly probe *sp in a noreturn function if there were any callee
13497 register saves or frame pointer is needed.
13499 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13502 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
13503 BLKmode for ternary, binary or unary expressions.
13506 * var-tracking.c (delete_vta_debug_insn): New inline function.
13507 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
13508 insns from get_insns () to NULL instead of each bb separately.
13509 Use delete_vta_debug_insn. No longer static.
13510 (vt_debug_insns_local, variable_tracking_main_1): Adjust
13511 delete_vta_debug_insns callers.
13512 * rtl.h (delete_vta_debug_insns): Declare.
13513 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
13514 instead of variable_tracking_main.
13516 2018-01-03 Martin Sebor <msebor@redhat.com>
13518 PR tree-optimization/83603
13519 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
13520 arguments past the endof the argument list in functions declared
13521 without a prototype.
13522 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
13523 Avoid checking when arguments are null.
13525 2018-01-03 Martin Sebor <msebor@redhat.com>
13528 * doc/extend.texi (attribute const): Fix a typo.
13529 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
13530 issuing -Wsuggest-attribute for void functions.
13532 2018-01-03 Martin Sebor <msebor@redhat.com>
13534 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
13535 offset_int::from instead of wide_int::to_shwi.
13536 (maybe_diag_overlap): Remove assertion.
13537 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
13538 * gimple-ssa-sprintf.c (format_directive): Same.
13539 (parse_directive): Same.
13540 (sprintf_dom_walker::compute_format_length): Same.
13541 (try_substitute_return_value): Same.
13543 2018-01-03 Jeff Law <law@redhat.com>
13545 PR middle-end/83654
13546 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
13547 non-constant residual for zero at runtime and avoid probing in
13548 that case. Reorganize code for trailing problem to mirror handling
13551 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13553 PR tree-optimization/83501
13554 * tree-ssa-strlen.c (get_string_cst): New.
13555 (handle_char_store): Call get_string_cst.
13557 2018-01-03 Martin Liska <mliska@suse.cz>
13559 PR tree-optimization/83593
13560 * tree-ssa-strlen.c: Include tree-cfg.h.
13561 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
13562 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
13563 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
13565 (strlen_dom_walker::before_dom_children): Call
13566 gimple_purge_dead_eh_edges. Dump tranformation with details
13568 (strlen_dom_walker::before_dom_children): Update call by adding
13569 new argument cleanup_eh.
13570 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
13572 2018-01-03 Martin Liska <mliska@suse.cz>
13575 * cif-code.def (VARIADIC_THUNK): New enum value.
13576 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
13579 2018-01-03 Jan Beulich <jbeulich@suse.com>
13581 * sse.md (mov<mode>_internal): Tighten condition for when to use
13582 vmovdqu<ssescalarsize> for TI and OI modes.
13584 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13586 Update copyright years.
13588 2018-01-03 Martin Liska <mliska@suse.cz>
13591 * ipa-visibility.c (function_and_variable_visibility): Skip
13592 functions with noipa attribure.
13594 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13596 * gcc.c (process_command): Update copyright notice dates.
13597 * gcov-dump.c (print_version): Ditto.
13598 * gcov.c (print_version): Ditto.
13599 * gcov-tool.c (print_version): Ditto.
13600 * gengtype.c (create_file): Ditto.
13601 * doc/cpp.texi: Bump @copying's copyright year.
13602 * doc/cppinternals.texi: Ditto.
13603 * doc/gcc.texi: Ditto.
13604 * doc/gccint.texi: Ditto.
13605 * doc/gcov.texi: Ditto.
13606 * doc/install.texi: Ditto.
13607 * doc/invoke.texi: Ditto.
13609 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13611 * vector-builder.h (vector_builder::m_full_nelts): Change from
13612 unsigned int to poly_uint64.
13613 (vector_builder::full_nelts): Update prototype accordingly.
13614 (vector_builder::new_vector): Likewise.
13615 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
13616 (vector_builder::operator ==): Likewise.
13617 (vector_builder::finalize): Likewise.
13618 * int-vector-builder.h (int_vector_builder::int_vector_builder):
13619 Take the number of elements as a poly_uint64 rather than an
13621 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
13622 from unsigned int to poly_uint64.
13623 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
13624 (vec_perm_indices::new_vector): Likewise.
13625 (vec_perm_indices::length): Likewise.
13626 (vec_perm_indices::nelts_per_input): Likewise.
13627 (vec_perm_indices::input_nelts): Likewise.
13628 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
13629 number of elements per input as a poly_uint64 rather than an
13630 unsigned int. Use the original encoding for variable-length
13631 vectors, rather than clamping each individual element.
13632 For the second and subsequent elements in each pattern,
13633 clamp the step and base before clamping their sum.
13634 (vec_perm_indices::series_p): Handle polynomial element counts.
13635 (vec_perm_indices::all_in_range_p): Likewise.
13636 (vec_perm_indices_to_tree): Likewise.
13637 (vec_perm_indices_to_rtx): Likewise.
13638 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
13639 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
13640 (tree_vector_builder::new_binary_operation): Handle polynomial
13641 element counts. Return false if we need to know the number
13642 of elements at compile time.
13643 * fold-const.c (fold_vec_perm): Punt if the number of elements
13644 isn't known at compile time.
13646 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13648 * vec-perm-indices.h (vec_perm_builder): Change element type
13649 from HOST_WIDE_INT to poly_int64.
13650 (vec_perm_indices::element_type): Update accordingly.
13651 (vec_perm_indices::clamp): Handle polynomial element_types.
13652 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13653 (vec_perm_indices::all_in_range_p): Likewise.
13654 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
13656 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
13657 polynomial vec_perm_indices element types.
13658 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
13659 * fold-const.c (fold_vec_perm): Likewise.
13660 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
13661 * tree-vect-generic.c (lower_vec_perm): Likewise.
13662 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13663 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
13664 element type to HOST_WIDE_INT.
13666 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13667 Alan Hayward <alan.hayward@arm.com>
13668 David Sherwood <david.sherwood@arm.com>
13670 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
13671 rather than an int. Use plus_constant.
13672 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
13673 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
13675 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13676 Alan Hayward <alan.hayward@arm.com>
13677 David Sherwood <david.sherwood@arm.com>
13679 * calls.c (emit_call_1, expand_call): Change struct_value_size from
13680 a HOST_WIDE_INT to a poly_int64.
13682 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13683 Alan Hayward <alan.hayward@arm.com>
13684 David Sherwood <david.sherwood@arm.com>
13686 * calls.c (load_register_parameters): Cope with polynomial
13687 mode sizes. Require a constant size for BLKmode parameters
13688 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
13689 forces a parameter to be padded at the lsb end in order to
13690 fill a complete number of words, require the parameter size
13691 to be ordered wrt UNITS_PER_WORD.
13693 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13694 Alan Hayward <alan.hayward@arm.com>
13695 David Sherwood <david.sherwood@arm.com>
13697 * reload1.c (spill_stack_slot_width): Change element type
13698 from unsigned int to poly_uint64_pod.
13699 (alter_reg): Treat mode sizes as polynomial.
13701 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13702 Alan Hayward <alan.hayward@arm.com>
13703 David Sherwood <david.sherwood@arm.com>
13705 * reload.c (complex_word_subreg_p): New function.
13706 (reload_inner_reg_of_subreg, push_reload): Use it.
13708 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13709 Alan Hayward <alan.hayward@arm.com>
13710 David Sherwood <david.sherwood@arm.com>
13712 * lra-constraints.c (process_alt_operands): Reject matched
13713 operands whose sizes aren't ordered.
13714 (match_reload): Refer to this check here.
13716 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13717 Alan Hayward <alan.hayward@arm.com>
13718 David Sherwood <david.sherwood@arm.com>
13720 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
13721 that the mode size is in the set {1, 2, 4, 8, 16}.
13723 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13724 Alan Hayward <alan.hayward@arm.com>
13725 David Sherwood <david.sherwood@arm.com>
13727 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
13728 Use plus_constant instead of gen_rtx_PLUS.
13730 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13731 Alan Hayward <alan.hayward@arm.com>
13732 David Sherwood <david.sherwood@arm.com>
13734 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
13735 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
13736 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
13737 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
13738 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
13739 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
13740 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
13741 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
13742 * config/i386/i386.c (ix86_push_rounding): ...this new function.
13743 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
13745 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
13746 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
13747 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
13748 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
13749 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
13750 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
13751 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
13752 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
13753 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
13754 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
13756 * expr.c (emit_move_resolve_push): Treat the input and result
13757 of PUSH_ROUNDING as a poly_int64.
13758 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
13759 (emit_push_insn): Likewise.
13760 * lra-eliminations.c (mark_not_eliminable): Likewise.
13761 * recog.c (push_operand): Likewise.
13762 * reload1.c (elimination_effects): Likewise.
13763 * rtlanal.c (nonzero_bits1): Likewise.
13764 * calls.c (store_one_arg): Likewise. Require the padding to be
13765 known at compile time.
13767 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13768 Alan Hayward <alan.hayward@arm.com>
13769 David Sherwood <david.sherwood@arm.com>
13771 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
13772 Use plus_constant instead of gen_rtx_PLUS.
13774 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13775 Alan Hayward <alan.hayward@arm.com>
13776 David Sherwood <david.sherwood@arm.com>
13778 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
13779 rather than an int.
13781 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13782 Alan Hayward <alan.hayward@arm.com>
13783 David Sherwood <david.sherwood@arm.com>
13785 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
13786 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
13787 via stack temporaries. Treat the mode size as polynomial too.
13789 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13790 Alan Hayward <alan.hayward@arm.com>
13791 David Sherwood <david.sherwood@arm.com>
13793 * expr.c (expand_expr_real_2): When handling conversions involving
13794 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
13795 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
13796 as a poly_uint64 too.
13798 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13799 Alan Hayward <alan.hayward@arm.com>
13800 David Sherwood <david.sherwood@arm.com>
13802 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
13804 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13805 Alan Hayward <alan.hayward@arm.com>
13806 David Sherwood <david.sherwood@arm.com>
13808 * combine.c (can_change_dest_mode): Handle polynomial
13809 REGMODE_NATURAL_SIZE.
13810 * expmed.c (store_bit_field_1): Likewise.
13811 * expr.c (store_constructor): Likewise.
13812 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
13813 and polynomial REGMODE_NATURAL_SIZE.
13814 (gen_lowpart_common): Likewise.
13815 * reginfo.c (record_subregs_of_mode): Likewise.
13816 * rtlanal.c (read_modify_subreg_p): Likewise.
13818 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13819 Alan Hayward <alan.hayward@arm.com>
13820 David Sherwood <david.sherwood@arm.com>
13822 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
13823 numbers of elements.
13825 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13826 Alan Hayward <alan.hayward@arm.com>
13827 David Sherwood <david.sherwood@arm.com>
13829 * match.pd: Cope with polynomial numbers of vector elements.
13831 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13832 Alan Hayward <alan.hayward@arm.com>
13833 David Sherwood <david.sherwood@arm.com>
13835 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
13836 in a POINTER_PLUS_EXPR.
13838 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13839 Alan Hayward <alan.hayward@arm.com>
13840 David Sherwood <david.sherwood@arm.com>
13842 * omp-simd-clone.c (simd_clone_subparts): New function.
13843 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
13844 (ipa_simd_modify_function_body): Likewise.
13846 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13847 Alan Hayward <alan.hayward@arm.com>
13848 David Sherwood <david.sherwood@arm.com>
13850 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
13851 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
13852 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
13853 (expand_vector_condition, vector_element): Likewise.
13854 (subparts_gt): New function.
13855 (get_compute_type): Use subparts_gt.
13856 (count_type_subparts): Delete.
13857 (expand_vector_operations_1): Use subparts_gt instead of
13858 count_type_subparts.
13860 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13861 Alan Hayward <alan.hayward@arm.com>
13862 David Sherwood <david.sherwood@arm.com>
13864 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
13865 (vect_compile_time_alias): ...this new function. Do the calculation
13866 on poly_ints rather than trees.
13867 (vect_prune_runtime_alias_test_list): Update call accordingly.
13869 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13870 Alan Hayward <alan.hayward@arm.com>
13871 David Sherwood <david.sherwood@arm.com>
13873 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
13875 (vect_schedule_slp_instance): Likewise.
13877 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13878 Alan Hayward <alan.hayward@arm.com>
13879 David Sherwood <david.sherwood@arm.com>
13881 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
13882 constant and extern definitions for variable-length vectors.
13883 (vect_get_constant_vectors): Note that the number of units
13884 is known to be constant.
13886 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13887 Alan Hayward <alan.hayward@arm.com>
13888 David Sherwood <david.sherwood@arm.com>
13890 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
13891 of units as polynomial. Choose between WIDE and NARROW based
13894 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13895 Alan Hayward <alan.hayward@arm.com>
13896 David Sherwood <david.sherwood@arm.com>
13898 * tree-vect-stmts.c (simd_clone_subparts): New function.
13899 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
13901 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13902 Alan Hayward <alan.hayward@arm.com>
13903 David Sherwood <david.sherwood@arm.com>
13905 * tree-vect-stmts.c (vectorizable_call): Treat the number of
13906 vectors as polynomial. Use build_index_vector for
13907 IFN_GOMP_SIMD_LANE.
13909 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13910 Alan Hayward <alan.hayward@arm.com>
13911 David Sherwood <david.sherwood@arm.com>
13913 * tree-vect-stmts.c (get_load_store_type): Treat the number of
13914 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
13915 for variable-length vectors.
13916 (vectorizable_mask_load_store): Treat the number of units as
13917 polynomial, asserting that it is constant if the condition has
13918 already been enforced.
13919 (vectorizable_store, vectorizable_load): Likewise.
13921 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13922 Alan Hayward <alan.hayward@arm.com>
13923 David Sherwood <david.sherwood@arm.com>
13925 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
13926 of units as polynomial. Punt if we can't tell at compile time
13927 which vector contains the final result.
13929 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13930 Alan Hayward <alan.hayward@arm.com>
13931 David Sherwood <david.sherwood@arm.com>
13933 * tree-vect-loop.c (vectorizable_induction): Treat the number
13934 of units as polynomial. Punt on SLP inductions. Use an integer
13935 VEC_SERIES_EXPR for variable-length integer reductions. Use a
13936 cast of such a series for variable-length floating-point
13939 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13940 Alan Hayward <alan.hayward@arm.com>
13941 David Sherwood <david.sherwood@arm.com>
13943 * tree.h (build_index_vector): Declare.
13944 * tree.c (build_index_vector): New function.
13945 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
13946 of units as polynomial, forcibly converting it to a constant if
13947 vectorizable_reduction has already enforced the condition.
13948 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
13949 to create a {1,2,3,...} vector.
13950 (vectorizable_reduction): Treat the number of units as polynomial.
13951 Choose vectype_in based on the largest scalar element size rather
13952 than the smallest number of units. Enforce the restrictions
13955 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13956 Alan Hayward <alan.hayward@arm.com>
13957 David Sherwood <david.sherwood@arm.com>
13959 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
13960 number of units as polynomial.
13962 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13963 Alan Hayward <alan.hayward@arm.com>
13964 David Sherwood <david.sherwood@arm.com>
13966 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
13967 * target.def (autovectorize_vector_sizes): Return the vector sizes
13968 by pointer, using vector_sizes rather than a bitmask.
13969 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
13970 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
13971 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
13973 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
13974 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
13975 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
13976 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
13977 * omp-general.c (omp_max_vf): Likewise.
13978 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13979 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
13980 * tree-vect-loop.c (vect_analyze_loop): Likewise.
13981 * tree-vect-slp.c (vect_slp_bb): Likewise.
13982 * doc/tm.texi: Regenerate.
13983 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
13985 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
13986 the vector size as a poly_uint64 rather than an unsigned int.
13987 (current_vector_size): Change from an unsigned int to a poly_uint64.
13988 (get_vectype_for_scalar_type): Update accordingly.
13989 * tree.h (build_truth_vector_type): Take the size and number of
13990 units as a poly_uint64 rather than an unsigned int.
13991 (build_vector_type): Add a temporary overload that takes
13992 the number of units as a poly_uint64 rather than an unsigned int.
13993 * tree.c (make_vector_type): Likewise.
13994 (build_truth_vector_type): Take the number of units as a poly_uint64
13995 rather than an unsigned int.
13997 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13998 Alan Hayward <alan.hayward@arm.com>
13999 David Sherwood <david.sherwood@arm.com>
14001 * target.def (get_mask_mode): Take the number of units and length
14002 as poly_uint64s rather than unsigned ints.
14003 * targhooks.h (default_get_mask_mode): Update accordingly.
14004 * targhooks.c (default_get_mask_mode): Likewise.
14005 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
14006 * doc/tm.texi: Regenerate.
14008 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14009 Alan Hayward <alan.hayward@arm.com>
14010 David Sherwood <david.sherwood@arm.com>
14012 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
14013 * omp-general.c (omp_max_vf): Likewise.
14014 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
14015 (expand_omp_simd): Handle polynomial safelen.
14016 * omp-low.c (omplow_simd_context): Add a default constructor.
14017 (omplow_simd_context::max_vf): Change from int to poly_uint64.
14018 (lower_rec_simd_input_clauses): Update accordingly.
14019 (lower_rec_input_clauses): Likewise.
14021 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14022 Alan Hayward <alan.hayward@arm.com>
14023 David Sherwood <david.sherwood@arm.com>
14025 * tree-vectorizer.h (vect_nunits_for_cost): New function.
14026 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
14027 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
14028 (vect_analyze_slp_cost): Likewise.
14029 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
14030 (vect_model_load_cost): Likewise.
14032 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14033 Alan Hayward <alan.hayward@arm.com>
14034 David Sherwood <david.sherwood@arm.com>
14036 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
14037 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
14038 from an unsigned int * to a poly_uint64_pod *.
14039 (calculate_unrolling_factor): New function.
14040 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
14042 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14043 Alan Hayward <alan.hayward@arm.com>
14044 David Sherwood <david.sherwood@arm.com>
14046 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
14047 from an unsigned int to a poly_uint64.
14048 (_loop_vec_info::slp_unrolling_factor): Likewise.
14049 (_loop_vec_info::vectorization_factor): Change from an int
14051 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
14052 (vect_get_num_vectors): New function.
14053 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
14054 (vect_get_num_copies): Use vect_get_num_vectors.
14055 (vect_analyze_data_ref_dependences): Change max_vf from an int *
14056 to an unsigned int *.
14057 (vect_analyze_data_refs): Change min_vf from an int * to a
14059 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14060 than an unsigned HOST_WIDE_INT.
14061 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
14062 (vect_analyze_data_ref_dependence): Change max_vf from an int *
14063 to an unsigned int *.
14064 (vect_analyze_data_ref_dependences): Likewise.
14065 (vect_compute_data_ref_alignment): Handle polynomial vf.
14066 (vect_enhance_data_refs_alignment): Likewise.
14067 (vect_prune_runtime_alias_test_list): Likewise.
14068 (vect_shift_permute_load_chain): Likewise.
14069 (vect_supportable_dr_alignment): Likewise.
14070 (dependence_distance_ge_vf): Take the vectorization factor as a
14071 poly_uint64 rather than an unsigned HOST_WIDE_INT.
14072 (vect_analyze_data_refs): Change min_vf from an int * to a
14074 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
14075 vfm1 as a poly_uint64 rather than an int. Make the same change
14076 for the returned bound_scalar.
14077 (vect_gen_vector_loop_niters): Handle polynomial vf.
14078 (vect_do_peeling): Likewise. Update call to
14079 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
14080 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
14082 * tree-vect-loop.c (vect_determine_vectorization_factor)
14083 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
14084 (vect_get_known_peeling_cost): Likewise.
14085 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
14086 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
14087 (vect_transform_loop): Likewise. Use the lowest possible VF when
14088 updating the upper bounds of the loop.
14089 (vect_min_worthwhile_factor): Make static. Return an unsigned int
14090 rather than an int.
14091 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
14092 polynomial unroll factors.
14093 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
14094 (vect_make_slp_decision): Likewise.
14095 (vect_supported_load_permutation_p): Likewise, and polynomial
14097 (vect_analyze_slp_cost): Handle polynomial vf.
14098 (vect_slp_analyze_node_operations): Likewise.
14099 (vect_slp_analyze_bb_1): Likewise.
14100 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14101 than an unsigned HOST_WIDE_INT.
14102 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
14103 (vectorizable_load): Handle polynomial vf.
14104 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
14106 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
14108 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14109 Alan Hayward <alan.hayward@arm.com>
14110 David Sherwood <david.sherwood@arm.com>
14112 * match.pd: Handle bit operations involving three constants
14113 and try to fold one pair.
14115 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14117 * tree-vect-loop-manip.c: Include gimple-fold.h.
14118 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
14119 niters_maybe_zero parameters. Handle other cases besides a step of 1.
14120 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
14121 Add a path that uses a step of VF instead of 1, but disable it
14123 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
14124 and niters_no_overflow parameters. Update calls to
14125 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
14126 Create a new SSA name if the latter choses to use a ste other
14127 than zero, and return it via niters_vector_mult_vf_var.
14128 * tree-vect-loop.c (vect_transform_loop): Update calls to
14129 vect_do_peeling, vect_gen_vector_loop_niters and
14130 slpeel_make_loop_iterate_ntimes.
14131 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
14132 (vect_gen_vector_loop_niters): Update declarations after above changes.
14134 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
14136 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
14137 128-bit round to integer instructions.
14138 (ceil<mode>2): Likewise.
14139 (btrunc<mode>2): Likewise.
14140 (round<mode>2): Likewise.
14142 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14144 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
14145 unaligned VSX load/store on P8/P9.
14146 (expand_block_clear): Allow the use of unaligned VSX
14147 load/store on P8/P9.
14149 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14151 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
14153 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
14154 swap associated with both a load and a store.
14156 2018-01-02 Andrew Waterman <andrew@sifive.com>
14158 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
14159 * config/riscv/riscv.md (clear_cache): Use it.
14161 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
14163 * web.c: Remove out-of-date comment.
14165 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14167 * expr.c (fixup_args_size_notes): Check that any existing
14168 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
14169 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
14170 (emit_single_push_insn): ...here.
14172 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14174 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
14175 (const_vector_encoded_nelts): New function.
14176 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
14177 (const_vector_int_elt, const_vector_elt): Declare.
14178 * emit-rtl.c (const_vector_int_elt_1): New function.
14179 (const_vector_elt): Likewise.
14180 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
14181 of CONST_VECTOR_ELT.
14183 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14185 * expr.c: Include rtx-vector-builder.h.
14186 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
14187 directly on the tree encoding.
14188 (const_vector_from_tree): Likewise.
14189 * optabs.c: Include rtx-vector-builder.h.
14190 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
14191 sequence of "u" values.
14192 * vec-perm-indices.c: Include rtx-vector-builder.h.
14193 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
14194 directly on the vec_perm_indices encoding.
14196 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14198 * doc/rtl.texi (const_vector): Describe new encoding scheme.
14199 * Makefile.in (OBJS): Add rtx-vector-builder.o.
14200 * rtx-vector-builder.h: New file.
14201 * rtx-vector-builder.c: Likewise.
14202 * rtl.h (rtx_def::u2): Add a const_vector field.
14203 (CONST_VECTOR_NPATTERNS): New macro.
14204 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
14205 (CONST_VECTOR_DUPLICATE_P): Likewise.
14206 (CONST_VECTOR_STEPPED_P): Likewise.
14207 (CONST_VECTOR_ENCODED_ELT): Likewise.
14208 (const_vec_duplicate_p): Check for a duplicated vector encoding.
14209 (unwrap_const_vec_duplicate): Likewise.
14210 (const_vec_series_p): Check for a non-duplicated vector encoding.
14211 Say that the function only returns true for integer vectors.
14212 * emit-rtl.c: Include rtx-vector-builder.h.
14213 (gen_const_vec_duplicate_1): Delete.
14214 (gen_const_vector): Call gen_const_vec_duplicate instead of
14215 gen_const_vec_duplicate_1.
14216 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
14217 (gen_const_vec_duplicate): Use rtx_vector_builder.
14218 (gen_const_vec_series): Likewise.
14219 (gen_rtx_CONST_VECTOR): Likewise.
14220 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
14221 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14222 Build a new vector rather than modifying a CONST_VECTOR in-place.
14223 (handle_special_swappables): Update call accordingly.
14224 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
14225 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14226 Build a new vector rather than modifying a CONST_VECTOR in-place.
14227 (handle_special_swappables): Update call accordingly.
14229 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14231 * simplify-rtx.c (simplify_const_binary_operation): Use
14232 CONST_VECTOR_ELT instead of XVECEXP.
14234 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14236 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
14237 the selector elements to be different from the data elements
14238 if the selector is a VECTOR_CST.
14239 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
14240 ssizetype for the selector.
14242 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14244 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
14245 before testing each element individually.
14246 * tree-vect-generic.c (lower_vec_perm): Likewise.
14248 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14250 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
14251 * selftest-run-tests.c (selftest::run_tests): Call it.
14252 * vector-builder.h (vector_builder::operator ==): New function.
14253 (vector_builder::operator !=): Likewise.
14254 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
14255 (vec_perm_indices::all_from_input_p): New function.
14256 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14257 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
14258 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
14259 instead of reading the VECTOR_CST directly. Detect whether both
14260 vector inputs are the same before constructing the vec_perm_indices,
14261 and update the number of inputs argument accordingly. Use the
14262 utility functions added above. Only construct sel2 if we need to.
14264 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14266 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
14267 the broadcast of the low byte.
14268 (expand_mult_highpart): Use an explicit encoding for the permutes.
14269 * optabs-query.c (can_mult_highpart_p): Likewise.
14270 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
14271 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14272 (vectorizable_bswap): Likewise.
14273 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
14274 explicit encoding for the power-of-2 permutes.
14275 (vect_permute_store_chain): Likewise.
14276 (vect_grouped_load_supported): Likewise.
14277 (vect_permute_load_chain): Likewise.
14279 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14281 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
14282 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
14283 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
14284 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14285 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
14286 (vect_gen_perm_mask_any): Likewise.
14288 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14290 * int-vector-builder.h: New file.
14291 * vec-perm-indices.h: Include int-vector-builder.h.
14292 (vec_perm_indices): Redefine as an int_vector_builder.
14293 (auto_vec_perm_indices): Delete.
14294 (vec_perm_builder): Redefine as a stand-alone class.
14295 (vec_perm_indices::vec_perm_indices): New function.
14296 (vec_perm_indices::clamp): Likewise.
14297 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
14298 (vec_perm_indices::new_vector): New function.
14299 (vec_perm_indices::new_expanded_vector): Update for new
14300 vec_perm_indices class.
14301 (vec_perm_indices::rotate_inputs): New function.
14302 (vec_perm_indices::all_in_range_p): Operate directly on the
14303 encoded form, without computing elided elements.
14304 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
14305 encoding. Update for new vec_perm_indices class.
14306 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
14307 the given vec_perm_builder.
14308 (expand_vec_perm_var): Update vec_perm_builder constructor.
14309 (expand_mult_highpart): Use vec_perm_builder instead of
14310 auto_vec_perm_indices.
14311 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
14312 vec_perm_indices instead of auto_vec_perm_indices. Use a single
14313 or double series encoding as appropriate.
14314 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
14315 vec_perm_indices instead of auto_vec_perm_indices.
14316 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14317 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14318 (vect_permute_store_chain): Likewise.
14319 (vect_grouped_load_supported): Likewise.
14320 (vect_permute_load_chain): Likewise.
14321 (vect_shift_permute_load_chain): Likewise.
14322 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14323 (vect_transform_slp_perm_load): Likewise.
14324 (vect_schedule_slp_instance): Likewise.
14325 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14326 (vectorizable_mask_load_store): Likewise.
14327 (vectorizable_bswap): Likewise.
14328 (vectorizable_store): Likewise.
14329 (vectorizable_load): Likewise.
14330 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
14331 vec_perm_indices instead of auto_vec_perm_indices. Use
14332 tree_to_vec_perm_builder to read the vector from a tree.
14333 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
14334 vec_perm_builder instead of a vec_perm_indices.
14335 (have_whole_vector_shift): Use vec_perm_builder and
14336 vec_perm_indices instead of auto_vec_perm_indices. Leave the
14337 truncation to calc_vec_perm_mask_for_shift.
14338 (vect_create_epilog_for_reduction): Likewise.
14339 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
14340 from auto_vec_perm_indices to vec_perm_indices.
14341 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14342 instead of changing individual elements.
14343 (aarch64_vectorize_vec_perm_const): Use new_vector to install
14344 the vector in d.perm.
14345 * config/arm/arm.c (expand_vec_perm_d::perm): Change
14346 from auto_vec_perm_indices to vec_perm_indices.
14347 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14348 instead of changing individual elements.
14349 (arm_vectorize_vec_perm_const): Use new_vector to install
14350 the vector in d.perm.
14351 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
14352 Update vec_perm_builder constructor.
14353 (rs6000_expand_interleave): Likewise.
14354 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
14355 (rs6000_expand_interleave): Likewise.
14357 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14359 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
14360 to qimode could truncate the indices.
14361 * optabs.c (expand_vec_perm_var): Likewise.
14363 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14365 * Makefile.in (OBJS): Add vec-perm-indices.o.
14366 * vec-perm-indices.h: New file.
14367 * vec-perm-indices.c: Likewise.
14368 * target.h (vec_perm_indices): Replace with a forward class
14370 (auto_vec_perm_indices): Move to vec-perm-indices.h.
14371 * optabs.h: Include vec-perm-indices.h.
14372 (expand_vec_perm): Delete.
14373 (selector_fits_mode_p, expand_vec_perm_var): Declare.
14374 (expand_vec_perm_const): Declare.
14375 * target.def (vec_perm_const_ok): Replace with...
14376 (vec_perm_const): ...this new hook.
14377 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
14378 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
14379 * doc/tm.texi: Regenerate.
14380 * optabs.def (vec_perm_const): Delete.
14381 * doc/md.texi (vec_perm_const): Likewise.
14382 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
14383 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
14384 expand_vec_perm for constant permutation vectors. Assert that
14385 the mode of variable permutation vectors is the integer equivalent
14386 of the mode that is being permuted.
14387 * optabs-query.h (selector_fits_mode_p): Declare.
14388 * optabs-query.c: Include vec-perm-indices.h.
14389 (selector_fits_mode_p): New function.
14390 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
14391 is defined, instead of checking whether the vec_perm_const_optab
14392 exists. Use targetm.vectorize.vec_perm_const instead of
14393 targetm.vectorize.vec_perm_const_ok. Check whether the indices
14394 fit in the vector mode before using a variable permute.
14395 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
14396 vec_perm_indices instead of an rtx.
14397 (expand_vec_perm): Replace with...
14398 (expand_vec_perm_const): ...this new function. Take the selector
14399 as a vec_perm_indices rather than an rtx. Also take the mode of
14400 the selector. Update call to shift_amt_for_vec_perm_mask.
14401 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
14402 Use vec_perm_indices::new_expanded_vector to expand the original
14403 selector into bytes. Check whether the indices fit in the vector
14404 mode before using a variable permute.
14405 (expand_vec_perm_var): Make global.
14406 (expand_mult_highpart): Use expand_vec_perm_const.
14407 * fold-const.c: Includes vec-perm-indices.h.
14408 * tree-ssa-forwprop.c: Likewise.
14409 * tree-vect-data-refs.c: Likewise.
14410 * tree-vect-generic.c: Likewise.
14411 * tree-vect-loop.c: Likewise.
14412 * tree-vect-slp.c: Likewise.
14413 * tree-vect-stmts.c: Likewise.
14414 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
14416 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
14417 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
14418 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
14419 (aarch64_vectorize_vec_perm_const): ...this new function.
14420 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14421 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14422 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
14423 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
14424 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14425 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14426 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
14428 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
14429 check for NEON modes.
14430 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
14431 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
14432 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
14433 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
14435 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
14436 the old VEC_PERM_CONST conditions.
14437 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
14438 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
14439 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
14440 (ia64_vectorize_vec_perm_const_ok): Merge into...
14441 (ia64_vectorize_vec_perm_const): ...this new function.
14442 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
14443 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
14444 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
14445 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
14446 * config/mips/mips.c (mips_expand_vec_perm_const)
14447 (mips_vectorize_vec_perm_const_ok): Merge into...
14448 (mips_vectorize_vec_perm_const): ...this new function.
14449 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
14450 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
14451 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
14452 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
14453 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
14454 (rs6000_expand_vec_perm_const): Delete.
14455 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
14457 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14458 (altivec_expand_vec_perm_const_le): Take each operand individually.
14459 Operate on constant selectors rather than rtxes.
14460 (altivec_expand_vec_perm_const): Likewise. Update call to
14461 altivec_expand_vec_perm_const_le.
14462 (rs6000_expand_vec_perm_const): Delete.
14463 (rs6000_vectorize_vec_perm_const_ok): Delete.
14464 (rs6000_vectorize_vec_perm_const): New function.
14465 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14466 an element count and rtx array.
14467 (rs6000_expand_extract_even): Update call accordingly.
14468 (rs6000_expand_interleave): Likewise.
14469 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
14470 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
14471 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
14472 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
14473 (rs6000_expand_vec_perm_const): Delete.
14474 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14475 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14476 (altivec_expand_vec_perm_const_le): Take each operand individually.
14477 Operate on constant selectors rather than rtxes.
14478 (altivec_expand_vec_perm_const): Likewise. Update call to
14479 altivec_expand_vec_perm_const_le.
14480 (rs6000_expand_vec_perm_const): Delete.
14481 (rs6000_vectorize_vec_perm_const_ok): Delete.
14482 (rs6000_vectorize_vec_perm_const): New function. Remove stray
14483 reference to the SPE evmerge intructions.
14484 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14485 an element count and rtx array.
14486 (rs6000_expand_extract_even): Update call accordingly.
14487 (rs6000_expand_interleave): Likewise.
14488 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
14489 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
14491 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14493 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14495 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
14496 vector mode and that that mode matches the mode of the data
14498 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
14499 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
14500 directly using expand_vec_perm_1 when forcing selectors into
14502 (expand_vec_perm_var): New function, split out from expand_vec_perm.
14504 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14506 * optabs-query.h (can_vec_perm_p): Delete.
14507 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
14508 * optabs-query.c (can_vec_perm_p): Split into...
14509 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
14510 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
14511 particular selector is valid.
14512 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14513 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14514 (vect_grouped_load_supported): Likewise.
14515 (vect_shift_permute_load_chain): Likewise.
14516 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14517 (vect_transform_slp_perm_load): Likewise.
14518 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14519 (vectorizable_bswap): Likewise.
14520 (vect_gen_perm_mask_checked): Likewise.
14521 * fold-const.c (fold_ternary_loc): Likewise. Don't take
14522 implementations of variable permutation vectors into account
14523 when deciding which selector to use.
14524 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
14525 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
14526 with a false third argument.
14527 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
14528 to test whether the constant selector is valid and can_vec_perm_var_p
14529 to test whether a variable selector is valid.
14531 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14533 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
14534 * optabs-query.c (can_vec_perm_p): Likewise.
14535 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
14536 instead of vec_perm_indices.
14537 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
14538 (vect_gen_perm_mask_checked): Likewise,
14539 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
14540 (vect_gen_perm_mask_checked): Likewise,
14542 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14544 * optabs-query.h (qimode_for_vec_perm): Declare.
14545 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
14546 (qimode_for_vec_perm): ...this new function.
14547 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
14549 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14551 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
14552 does not have a conditional at the top.
14554 2018-01-02 Richard Biener <rguenther@suse.de>
14556 * ipa-inline.c (big_speedup_p): Fix expression.
14558 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14561 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
14564 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14568 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
14569 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
14570 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
14571 cond_taken_branch_cost 3->4.
14573 2018-01-01 Jakub Jelinek <jakub@redhat.com>
14575 PR tree-optimization/83581
14576 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
14577 TODO_cleanup_cfg if any changes have been made.
14579 PR middle-end/83608
14580 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
14581 convert_modes if target mode has the right side, but different mode
14584 PR middle-end/83609
14585 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
14586 last argument when extracting from CONCAT. If either from_real or
14587 from_imag is NULL, use expansion through memory. If result is not
14588 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
14589 the parts directly to inner mode, if even that fails, use expansion
14592 PR middle-end/83623
14593 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
14594 check for bswap in mode rather than HImode and use that in expand_unop
14597 Copyright (C) 2018 Free Software Foundation, Inc.
14599 Copying and distribution of this file, with or without modification,
14600 are permitted in any medium without royalty provided the copyright
14601 notice and this notice are preserved.