1 2018-05-29 Martin Liska <mliska@suse.cz>
4 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
7 2018-05-29 Jakub Jelinek <jakub@redhat.com>
9 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
11 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
15 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
16 VEC_PACK_FLOAT_EXPR): New tree codes.
17 * tree-pretty-print.c (op_code_prio): Handle
18 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
19 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
20 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
21 * tree-inline.c (estimate_operator_cost): Likewise.
22 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
23 * fold-const.c (const_binop): Likewise.
24 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
25 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
26 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
27 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
28 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
29 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
30 * expr.c (expand_expr_real_2): Likewise.
31 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
32 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
33 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
35 * optabs.c (expand_widen_pattern_expr): For
36 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
37 sign from result type rather than operand's type.
38 (expand_binop_directly): For vec_packu_float_optab and
39 vec_packs_float_optab allow result type to be different from operand's
41 * optabs-tree.c (optab_for_tree_code): Handle
42 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
43 VEC_PACK_FLOAT_EXPR. Formatting fixes.
44 * tree-vect-generic.c (expand_vector_operations_1): Handle
45 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
47 * tree-vect-stmts.c (supportable_widening_operation): Handle
49 (supportable_narrowing_operation): Handle FLOAT_EXPR.
50 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
51 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
52 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
53 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
55 (vec_pack<floatprefix>_float_<mode>): New expander.
56 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
58 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
59 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
60 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
61 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
62 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
64 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
65 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
66 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
67 VEC_PACK_FLOAT_EXPR): Document.
69 2018-05-29 Richard Biener <rguenther@suse.de>
71 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
73 (stmt_vec_info_vec): Make pointer.
74 (init_stmt_vec_info_vec): Remove.
75 (free_stmt_vec_info_vec): Likewise.
76 (set_stmt_vec_info_vec): New function.
77 (free_stmt_vec_infos): Likewise.
78 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
79 (set_vinfo_for_stmt): Likewise.
80 (get_earlier_stmt): Likewise.
81 (get_later_stmt): Likewise.
82 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
83 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
84 (vec_info::~vec_info): Free stmt_vec_infos.
85 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
86 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
87 (pass_slp_vectorize::execute): Likewise.
88 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
89 (free_stmt_vec_info_vec): Likewise.
90 (set_stmt_vec_info_vec): New function.
91 (free_stmt_vec_infos): Likewise.
92 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
93 the global stmt_vec_info_vec.
94 * tree-parloops.c (gather_scalar_reductions): Use
95 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
98 2018-05-29 Richard Biener <rguenther@suse.de>
100 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
102 2018-05-29 Martin Liska <mliska@suse.cz>
103 David Malcolm <dmalcolm@redhat.com>
105 * vec.c (test_reverse): New.
106 (vec_c_tests): Add new test.
107 * vec.h (vl_ptr>::reverse): New function.
109 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
111 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
113 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
116 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
118 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
120 2018-05-28 Richard Biener <rguenther@suse.de>
122 PR tree-optimization/85933
123 * tree-vect-data-refs.c (vect_record_base_alignments): Only
124 look at stmts marked as vectorizable.
126 2018-05-28 Richard Biener <rguenther@suse.de>
128 PR tree-optimization/85934
129 * tree-vect-generic.c (expand_vector_operations_1): Hoist
130 vector boolean check before scalar optimization.
132 2018-05-28 Jakub Jelinek <jakub@redhat.com>
134 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
137 2018-05-28 Mark Wielaard <mark@klomp.org>
139 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
140 if it is an expression containing a minus sign.
142 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
144 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
146 2018-05-27 Paul Koning <ni1d@arrl.net>
148 * config/pdp11/pdp11.md (truncsihi2): Remove.
150 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
151 Chung-Ju Wu <jasonwucj@gmail.com>
153 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
155 (unaligned_store_dw): Ditto.
156 * config/nds32/nds32-memory-manipulation.c
157 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
158 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
159 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
160 (emit_setmem_word_loop): Rename to ...
161 (emit_setmem_doubleword_loop): ... this.
162 (nds32_gen_dup_4_byte_to_word_value): New function.
163 (nds32_gen_dup_8_byte_to_double_word_value): New function.
164 (nds32_expand_setmem_loop): Refine implementation.
165 (nds32_expand_setmem_loop_v3m): Ditto.
166 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
169 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
171 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
173 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
175 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
176 (nds32_init_machine_status): Initialize machine->attr_naked_p and
177 machine->attr_no_prologue_p.
178 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
179 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
180 (nds32_expand_epilogue): Consider attr_naked_p.
181 (nds32_expand_epilogue_v3pop): Likewise.
182 (nds32_can_use_return_insn): Likewise.
183 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
184 attr_no_prologue_p fields.
185 * config/nds32/nds32.opt (mret-in-naked-func): New option.
187 2018-05-27 Jakub Jelinek <jakub@redhat.com>
190 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
193 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
195 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
197 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
199 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
201 (*<floatsuffix>floatv2div2sf2): Rename to ...
202 (*float<floatunssuffix>v2div2sf2): ... this.
203 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
204 (float<floatunssuffix>v2div2sf2_mask): ... this.
205 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
206 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
207 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
209 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
211 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
213 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
215 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
217 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
219 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
220 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
221 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
222 gen_ufix_truncv8dfv8si2.
223 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
224 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
225 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
226 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
227 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
228 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
229 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
230 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
232 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
236 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
238 2018-05-25 Jim Wilson <jimw@sifive.com>
240 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
241 * config/riscv/riscv.c (struct machine_function): Add
242 interrupt_handler_p and attribute_checked_p fields.
243 (riscv_attribute_table): Add interrupt.
244 (riscv_interrupt_type_p): New.
245 (riscv_save_reg_p): Save extra regs for interrupt handler.
246 (riscv_use_save_libcall): Return false for interrupt handler.
247 (riscv_first_stack_step): Add forward declaration.
248 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
249 for interrupt handler with large frame. Use it for saved reg list.
250 (riscv_expand_prologue): Move flag_stack_usage_info support to
251 eliminate duplication.
252 (riscv_expand_epilogue): Generate mret for interrupt handler.
253 (riscv_epilogue_uses): New.
254 (riscv_can_use_return_insn): Return false for interrupt handler.
255 (riscv_function_ok_for_sibcall): Likewise.
256 (riscv_set_current_function): Add interrupt handler support.
257 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
258 * config/riscv/riscv.md (UNSPECV_MRET): New.
260 (riscv_frflags, riscv_fsflags): Use tab after opcode.
262 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
264 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
266 PR tree-optimization/85712
267 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
268 this candidate has already been replaced in-situ by a copy.
270 2018-05-25 Jason Merrill <jason@redhat.com>
272 PR c++/80485 - inline function non-zero address.
273 * symtab.c (nonzero_address): Check DECL_COMDAT.
275 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
278 * config/alpha/alpha.md (ashlsi3): New insn pattern.
279 (*ashlsi_se): Rename from *ashldi_se. Define as sign
280 extension of SImode operation. Use const123_operand predicate.
282 (*saddl_se_1): Ditto.
284 (*ssubl_se_1): Ditto.
285 * config/alpha/predicates.md (const123_operand): New predicate.
286 * config/alpha/constraints.md (P): Use IN_RANGE.
288 2018-05-25 Richard Biener <rguenther@suse.de>
290 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
292 (ref_maybe_used_by_stmt_p): Likewise.
293 (stmt_may_clobber_ref_p): Likewise.
294 (stmt_may_clobber_ref_p_1): Likewise.
295 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
297 (ref_maybe_used_by_stmt_p): Likewise.
298 (stmt_may_clobber_ref_p): Likewise.
299 (stmt_may_clobber_ref_p_1): Likewise.
300 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
301 the alias oracle to disambiguate DRs with stmts DR analysis
303 (vect_analyze_data_refs): Do not give up on not analyzable
304 DRs for BB vectorization. Remove code truncating the dataref
307 2018-05-25 Jakub Jelinek <jakub@redhat.com>
310 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
311 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
312 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
314 2018-05-25 Richard Biener <rguenther@suse.de>
316 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
317 function, combining stmt data ref gathering and fatal analysis
319 (vect_analyze_data_refs): Remove now redudnant code and simplify.
320 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
321 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
322 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
323 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
325 2018-05-25 Bin Cheng <bin.cheng@arm.com>
327 PR tree-optimization/85720
328 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
329 SCC if all partitions are builtins.
330 (version_loop_by_alias_check): New parameter. Generate cancelable
331 runtime alias check if all partitions are builtins.
332 (distribute_loop): Update call to above function.
334 2018-05-25 Bin Cheng <bin.cheng@arm.com>
336 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
337 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
338 (parm_default_def_partition_arg): Ditto.
339 (set_parm_default_def_partition): Ditto.
340 (get_parm_default_def_partitions): Ditto and make it static.
341 (get_undefined_value_partitions): Ditto and make it static.
342 (remove_ssa_form): Refactor call to init_var_map here.
343 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
344 computation for loop region.
345 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
346 (register_default_def): Delete.
347 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
348 (parm_default_def_partition_arg): Ditto.
349 (set_parm_default_def_partition): Ditto.
350 (get_parm_default_def_partitions): Ditto and make it static.
351 (get_undefined_value_partitions): Ditto and make it static.
352 (coalesce_with_default, coalesce_with_default): Update comment.
353 (create_coalesce_list_for_region): New func factored out from
354 create_outofssa_var_map.
355 (populate_coalesce_list_for_outofssa): New func factored out from
356 create_outofssa_var_map and coalesce_ssa_name.
357 (create_outofssa_var_map): Delete.
358 (coalesce_ssa_name): Refactor to support live range computation.
359 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
360 (get_parm_default_def_partitions): Delete.
361 (get_undefined_value_partitions): Ditto.
362 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
363 computation for loop region.
364 (new_tree_live_info, loe_visit_block): Ditto.
365 (live_worklist, set_var_live_on_entry): Ditto.
366 (calculate_live_on_exit, verify_live_on_entry): Ditto.
367 * tree-ssa-live.h (struct _var_map): New fields.
368 (init_var_map): Change decl.
369 (region_contains_p): New.
371 2018-05-25 Bin Cheng <bin.cheng@arm.com>
373 * tree-ssa-live.h (live_merge_and_clear): Delete.
375 2018-05-25 Richard Biener <rguenther@suse.de>
378 * tree-dump.c (dequeue_and_dump): Remove access to removed
379 operand 2 of a SWITCH_EXPR.
381 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
383 * doc/sourcebuild.texi (vect_double_cond_arith): Include
384 multiplication and division.
385 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
386 (cond_udiv@var{m}, cond_umod@var{m}): Document.
387 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
388 (cond_udiv_optab, cond_umod_optab): New optabs.
389 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
390 (IFN_COND_RDIV): New internal functions.
391 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
392 TRUNC_MOD_EXPR and RDIV_EXPR.
393 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
394 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
396 (SVE_INT_BINARY): Include mult.
397 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
398 (optab, sve_int_op): Handle mult.
399 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
401 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
402 for SVE_INT_BINARY_SD.
404 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
406 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
407 (optab, sve_int_op): Handle div and udiv.
408 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
409 for SVE_INT_BINARY_SD.
410 (*<optab><mode>3): New insn for the same.
412 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
414 * tree-vect-patterns.c: Include predict.h.
415 (vect_recog_divmod_pattern): Restrict check for division support
416 to when optimizing for size.
418 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
420 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
421 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
422 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
423 (gimple_match_op::set_op): Likewise.
424 (gimple_resimplify4): Declare.
425 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
426 (expr::gen_transform): Likewise.
427 (decision_tree::gen): Generate a simplification routine for 4 operands.
428 * gimple-match-head.c (gimple_simplify): Add an overload for
429 4 operands. In the top-level function, handle up to 4 call
430 arguments and call gimple_resimplify4.
431 (gimple_resimplify4): New function.
432 (build_call_internal): Pass a fourth operand.
433 (maybe_push_to_seq): Likewise.
434 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
435 Fold VEC_COND_EXPRs of an operation and a default value into
436 an IFN_COND_* function if possible.
437 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
439 (SVE_COND_FP_BINARY): Include them.
440 (optab, sve_fp_op): Handle them.
441 (SVE_INT_BINARY_REV): New code iterator.
442 (SVE_COND_FP_BINARY_REV): New int iterator.
443 (commutative): New int attribute.
444 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
446 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
448 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
449 (*cond_<optab><mode>): New patterns for reversed operands.
451 2018-05-25 Richard Biener <rguenther@suse.de>
453 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
454 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
455 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
456 (STMT_VINFO_GROUPED_ACCESS): Adjust.
457 * tree-vect-data-refs.c (everywhere): Adjust users.
458 * tree-vect-loop.c (everywhere): Likewise.
459 * tree-vect-slp.c (everywhere): Likewise.
460 * tree-vect-stmts.c (everywhere): Likewise.
461 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
463 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
465 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
467 (gcc_cv_as_section_exclude): ... this.
468 Try Solaris as #exclude syntax.
469 * configure: Regenerate.
470 * config.in: Regenerate.
471 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
473 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
474 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
476 * varasm.c (default_elf_asm_named_section): Don't check if
477 HAVE_GAS_SECTION_EXCLUDE is defined.
479 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
481 * doc/md.texi: Update the documentation of the cond_* optabs
482 to mention the new final operand. Fix GET_MODE_NUNITS call.
483 Describe the scalar case too.
484 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
485 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
487 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
488 (get_conditional_internal_fn): Update comment.
489 * tree-vect-loop.c (vectorizable_reduction): Pass the original
490 accumulator value as a final argument to conditional functions.
491 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
492 a define_expand and add an "else" operand. Assert for now that
493 the else operand is equal to operand 2. Use SVE_INT_BINARY and
494 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
495 (*cond_<optab><mode>): New patterns.
496 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
497 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
498 (UNSPEC_COND_EOR): Delete.
499 (optab): Remove associated mappings.
500 (SVE_INT_BINARY): New code iterator.
501 (sve_int_op): Remove int attribute and add "minus" to the code
503 (SVE_COND_INT_OP): Delete.
504 (SVE_COND_FP_OP): Rename to...
505 (SVE_COND_FP_BINARY): ...this.
507 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
509 * optabs.c (can_reuse_operands_p): New function.
510 (maybe_legitimize_operands): Try to reuse the results for
513 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
515 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
516 Add {q} suffix to insn mnemonic.
518 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
520 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
521 (msp430_warn_func_return): New.
523 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
525 * fold-const.c (tree_nonzero_bits): New function.
526 * fold-const.h (tree_nonzero_bits): Likewise.
527 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
528 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
530 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
534 * varasm.c (assemble_alias): Check ifunc_resolver only on
537 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
540 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
541 when memory input operand is handled.
543 2018-05-24 Luis Machado <luis.machado@linaro.org>
545 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
547 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
549 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
551 * match.pd: Delay FMA folds until after vectorization.
553 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
556 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
557 address check not strict.
559 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
561 * gimple-match.h (gimple_match_op): New class.
562 (mprts_hook): Replace parameters with a gimple_match_op *.
563 (maybe_build_generic_op): Likewise.
564 (gimple_simplified_result_is_gimple_val): Replace parameters with
565 a const gimple_match_op *.
566 (gimple_simplify): Replace code_helper * and tree * parameters with
567 a gimple_match_op * parameter.
568 (gimple_resimplify1): Replace code_helper *, tree and tree *
569 parameters with a gimple_match_op * parameter.
570 (gimple_resimplify2): Likewise.
571 (gimple_resimplify3): Likewise.
572 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
573 parameters with a gimple_match_op * parameter.
574 * gimple-match-head.c (gimple_simplify): Change prototypes of
575 auto-generated functions to take a gimple_match_op * instead of
576 separate code_helper * and tree * parameters. Make the same
577 change in the top-level overload and update calls to the
578 gimple_resimplify routines. Update calls to the auto-generated
579 functions and to maybe_push_res_to_seq in the publicly-facing
580 operation-specific gimple_simplify overloads.
581 (gimple_match_op::MAX_NUM_OPS): Define.
582 (gimple_resimplify1): Replace rcode and ops with a single res_op
583 parameter. Update call to gimple_simplify.
584 (gimple_resimplify2): Likewise.
585 (gimple_resimplify3): Likewise.
586 (mprts_hook): Replace parameters with a gimple_match_op *.
587 (maybe_build_generic_op): Likewise.
588 (build_call_internal): Replace type, nargs and ops with
590 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
591 with a single gimple_match_op *. Update calls to mprts_hook,
592 build_call_internal and gimple_simplified_result_is_gimple_val.
593 Factor out code that is common to the tree_code and combined_fn cases.
594 * genmatch.c (expr::gen_transform): Replace tem_code and
595 tem_ops with a gimple_match_op called tem_op. Update calls
596 to the gimple_resimplify functions and maybe_push_res_to_seq.
597 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
598 res_ops. Update call to the gimple_resimplify functions.
599 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
600 (decision_tree::gen): Make the functions take a gimple_match_op *
601 called res_op instead of separate res_code and res_ops parameters.
602 Update call accordingly.
603 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
604 and ops with a single res_op parameter. Update calls to
605 maybe_build_generic_op and maybe_push_res_to_seq.
606 (fold_stmt_1): Update calls to gimple_simplify and
607 replace_stmt_with_simplification.
608 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
609 and gimple_simplified_result_is_gimple_val.
610 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
612 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
613 with a gimple_match_op *.
614 (vn_nary_build_or_lookup): Likewise. Update call to
615 vn_nary_build_or_lookup_1.
616 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
617 gimple_match_op *. Update calls to the gimple_resimplify routines
618 and to gimple_simplified_result_is_gimple_val.
619 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
620 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
621 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
622 (visit_nary_op): Likewise.
623 (visit_reference_op_load): Likewise.
625 2018-05-23 Luis Machado <luis.machado@linaro.org>
627 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
628 modifier for printing the step amount.
630 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
633 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
636 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
638 * doc/sourcebuild.texi (Endianness): New subsubsection.
640 2018-05-23 Luis Machado <luis.machado@linaro.org>
642 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
643 <prefetch_dynamic_strides>: New const bool field.
644 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
645 prefetch_dynamic_strides.
646 (exynosm1_prefetch_tune): Likewise.
647 (thunderxt88_prefetch_tune): Likewise.
648 (thunderx_prefetch_tune): Likewise.
649 (thunderx2t99_prefetch_tune): Likewise.
650 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
652 (aarch64_override_options_internal): Update to set
653 PARAM_PREFETCH_DYNAMIC_STRIDES.
654 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
655 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
656 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
657 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
658 prefetch-dynamic-strides setting.
660 2018-05-23 Luis Machado <luis.machado@linaro.org>
662 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
663 <minimum_stride>: New const int field.
664 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
665 minimum_stride field defaulting to -1.
666 (exynosm1_prefetch_tune): Likewise.
667 (thunderxt88_prefetch_tune): Likewise.
668 (thunderx_prefetch_tune): Likewise.
669 (thunderx2t99_prefetch_tune): Likewise.
670 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
671 <default_opt_level>: Set to 3.
672 (aarch64_override_options_internal): Update to set
673 PARAM_PREFETCH_MINIMUM_STRIDE.
674 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
675 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
676 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
677 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
678 stride is constant and is below the minimum stride threshold.
680 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
682 * config/arm/arm-cpus.in (mode26): Delete.
683 (armv4): Delete mode26 reference.
684 * config/arm/arm.c (arm_configure_build_target): Delete use of
687 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
689 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
691 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
692 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
694 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
695 (floatunsdidf2): Ditto.
697 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
699 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
700 (fixuns_trunc<mode>si2_avx512f): Ditto.
701 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
702 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
703 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
705 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
707 PR rtl-optimization/79985
708 * df-scan.c (df_insn_refs_collect): Remove special case for
709 global registers and asm statements.
711 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
713 * extend.texi (Global Register Variables): Rewrite the bullet list.
714 Note that the register is available for allocation. Note that access
715 via inline asm must use constraints. Add note about async-signal
716 handlers. Remove paragraph about automagic register selection.
718 2018-05-23 Richard Biener <rguenther@suse.de>
720 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
721 of fixed offset from memset VN.
723 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
725 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
727 (alloc_cand_and_find_basis): Initialize first_interp field.
728 (slsr_process_mul): Modify first_interp field.
729 (slsr_process_add): Likewise.
730 (slsr_process_cast): Modify first_interp field for each new
732 (slsr_process_copy): Likewise.
733 (dump_candidate): Dump first_interp field.
734 (replace_mult_candidate): Process all interpretations, not just
736 (replace_rhs_if_not_dup): Likewise.
737 (replace_one_candidate): Likewise.
739 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
741 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
743 (aarch64_needs_frame_chain): New function.
744 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
746 2018-05-23 Sudakshina Das <sudi.das@arm.com>
749 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
750 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
751 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
752 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
753 as true for strict-align.
754 (aarch64_can_inline_p): Perform checks even when callee has no
755 attributes to check for strict alignment.
756 * doc/extend.texi (AArch64 Function Attributes): Document
758 * doc/invoke.texi: (AArch64 Options): Likewise.
760 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
762 PR tree-optimization/85853
763 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
764 the handling of the root of the node to...
765 (vect_slp_analyze_node_operations_1): ...this new function,
766 and run the whole thing with the child nodes' def types
767 set according to their SLP node's def type.
769 2018-05-23 Richard Biener <rguenther@suse.de>
772 * tree-data-ref.c (create_runtime_alias_checks): Defer
773 and ignore overflow warnings.
775 2018-05-23 Yury Gribov <tetra2005@gmail.com>
777 PR tree-optimization/85822
778 * tree-vrp.c (is_masked_range_test): Fix handling of negative
781 2018-05-23 Richard Biener <rguenther@suse.de>
783 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
784 memset constants via native_interpret_expr.
786 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
789 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
791 (cgraph_node::create_alias): Likewise.
792 (cgraph_node::get_availability): Check ifunc_resolver instead
793 of looking up ifunc attribute.
794 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
795 * varasm.c (do_assemble_alias): Likewise.
796 (assemble_alias): Likewise.
797 (default_binds_local_p_3): Likewise.
798 * cgraph.h (cgraph_node): Add ifunc_resolver.
799 (cgraph_node::only_called_directly_or_aliased_p): Return false
801 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
803 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
804 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
805 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
806 instead of looking up ifunc attribute.
808 2018-05-22 Luis Machado <luis.machado@linaro.org>
810 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
812 2018-05-22 Martin Sebor <msebor@redhat.com>
815 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
816 only when expasion succeeds.
817 (expand_builtin_strcmp): Same.
818 (expand_builtin_strncmp): Same.
820 2018-05-22 Martin Sebor <msebor@redhat.com>
822 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
824 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
825 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
827 * config/aarch64/aarch64-ldpstp.md: Replace uses of
828 aarch64_mem_pair_operand with memory_operand and delete operand swapping
830 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
831 Add check for legitimate_address.
832 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
833 (aarch64_swap_ldrstr_operands): New.
834 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
837 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
838 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
840 * config/aarch64/aarch64.md: New patterns to generate stp
842 (store_pair_sw, store_pair_dw): New patterns to generate stp for
843 single words and double words.
844 (load_pair_sw, load_pair_dw): Likewise.
845 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
847 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
849 * config/aarch64/aarch64-ldpstp.md: Modify peephole
850 for different mode ldpstp and add peephole for merged zero stores.
852 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
854 (aarch64_gen_store_pair): Rename calls to match new patterns.
855 (aarch64_gen_load_pair): Rename calls to match new patterns.
856 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
857 (load_pair<DREG:mode><DREG2:mode>): ... This.
858 (store_pair<mode>): Rename to...
859 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
860 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
862 (V_INT_EQUIV): Handle SImode.
863 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
866 2018-05-22 Martin Sebor <msebor@redhat.com>
869 * calls.c (maybe_warn_nonstring_arg): Use string length to set
870 or ajust the presumed bound on an operation to avoid unnecessary
873 2018-05-22 Martin Sebor <msebor@redhat.com>
875 PR tree-optimization/85826
876 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
877 assuming that a DECL necesarily has a constant size.
879 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
882 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
884 2018-05-22 Richard Biener <rguenther@suse.de>
886 PR tree-optimization/85834
887 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
888 non-constant and non-zero memset arguments.
890 2018-05-22 Martin Liska <mliska@suse.cz>
893 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
895 2018-05-22 Richard Biener <rguenther@suse.de>
897 PR tree-optimization/85863
898 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
899 comparisons when vectype is specified.
900 (vectorizable_condition): Do not specify vectype for
901 vect_is_simple_cond when SLP vectorizing.
903 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
906 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
907 define __ibm128 as long double.
908 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
909 as a distinct type when IEEE 128-bit support is enabled.
910 (init_float128_ieee): Fix up conversions between IFmode and IEEE
911 128-bit types to use the correct functions.
912 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
913 convert between 128-bit floating point types that have different
914 modes but the same representation, instead of using gen_lowpart to
916 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
918 (IFKF_reg): New attributes to give the register constraints for
920 (extend<mode>tf2_internal): New insns to mark an explicit
921 conversion between 128-bit floating point types that have a
922 different mode but share the same representation.
924 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
926 PR tree-optimization/85814
927 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
928 a null return from get_strinfo when unsharing the next
929 strinfo in the chain.
931 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
934 * varasm.c (weak_finish): Clean up weak_decls.
936 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
938 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
939 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
940 UNSPEC_UADALP values.
941 * config/aarch64/iterators.md (ABAL): New int iterator.
944 (sur): Add mappings for the above.
945 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
947 (aarch64_<sur>abal<mode>_4): Likewise.
948 (aarch64_<sur>adalp<mode>_3): Likewise.
949 (<sur>sadv16qi): New define_expand.
951 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
953 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
954 (*movdf_internal): Ditto.
955 (*rcpsf2_sse): Ditto.
956 (*rsqrtsf2_sse): Ditto.
957 (*sqrt<mode>2_sse): Ditto.
959 2018-05-21 Tamar Christina <tamar.christina@arm.com>
961 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
963 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
964 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
965 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
966 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
968 * config/aarch64/arm_neon.h: Likewise.
969 * config/aarch64/iterators.md (VQ_I): New.
971 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
973 * config.gcc: Add arc/t-multilib-linux to tmake_file for
975 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
978 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
980 * config/nds32/constraints.md (S): New constraint.
981 * config/nds32/nds32.md (call_internal): Use constraint S.
982 (call_value_internal): Likewise.
983 (sibcall_internal): Likewise.
984 (sibcall_value_internal): Likewise.
986 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
987 Chung-Ju Wu <jasonwucj@gmail.com>
989 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
992 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
993 Chung-Ju Wu <jasonwucj@gmail.com>
995 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
996 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
997 (nds32_rtx_costs_impl): Simplify.
998 (nds32_address_cost_impl): Simplify.
999 (nds32_init_rtx_costs): New function.
1000 (nds32_rtx_costs_speed_prefer): Likewise.
1001 (nds32_rtx_costs_size_prefer): Likewise.
1002 (nds32_address_cost_speed_prefer): Likewise.
1003 (nds32_address_cost_speed_fwprop): Likewise.
1004 (nds32_address_cost_size_prefer): Likewise.
1005 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
1006 * config/nds32/nds32.c (nds32_option_override): Use
1007 nds32_init_rtx_costs function.
1009 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1011 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
1012 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
1013 (TARGET_PIPELINE_N8): Likewise.
1014 (TARGET_PIPELINE_N10): Likewise.
1015 (TARGET_PIPELINE_N13): Likewise.
1016 (TARGET_PIPELINE_GRAYWOLF): Likewise.
1018 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
1020 * config/nds32/nds32-fpu.md: Update copyright year.
1022 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1024 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
1026 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1028 * config/nds32/nds32.c
1029 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
1030 * config/nds32/nds32.opt (minline-asm-r15): New option.
1032 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1034 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
1036 * config/nds32/nds32.md (abssi2): New pattern.
1038 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
1040 * config/i386/i386.md (rex64namesuffix): New mode attribute.
1041 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
1042 Merge insn pattern from sse_cvtsi2ss<round_name> and
1043 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
1044 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
1045 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
1046 using SWI48 mode iterator.
1047 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
1048 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
1049 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
1050 pattern from sse_cvttss2si<round_saeonly_name>
1051 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
1052 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
1053 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
1054 using SWI48 mode iterator.
1055 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
1056 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
1057 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
1058 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
1059 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
1060 using SWI48 mode iterator.
1061 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
1062 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
1063 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
1064 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
1065 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
1066 SWI48 mode iterator.
1067 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
1068 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
1069 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
1070 pattern from sse_cvttsd2si<round_saeonly_name>
1071 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
1073 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1075 * config/nds32/nds32-md-auxiliary.c
1076 (nds32_valid_smw_lwm_base_p): Refine.
1077 (nds32_output_smw_single_word): Refine.
1078 (nds32_output_smw_double_word): New.
1079 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
1081 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1083 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
1084 (nds32_output_stack_pop): Refine.
1085 (nds32_expand_unaligned_load): Refine.
1086 (nds32_expand_unaligned_store): Refine.
1088 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1089 Chung-Ju Wu <jasonwucj@gmail.com>
1091 * config/nds32/constants.md: Add TP_REGNUM constant.
1092 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
1093 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
1095 * config/nds32/nds32-doubleword.md: Consider flag_pic.
1096 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
1097 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
1098 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
1099 and PIC code generation.
1100 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
1102 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
1104 * config/nds32/nds32.md: Support TLS and PIC.
1105 * config/nds32/nds32.c: Support TLS and PIC.
1106 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
1107 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
1110 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1112 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
1113 mode with E_ prefix.
1115 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1116 Chung-Ju Wu <jasonwucj@gmail.com>
1118 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
1119 * config/nds32/nds32-md-auxiliary.c
1120 (symbolic_reference_mentioned_p): New.
1121 (nds32_legitimize_ict_address): New.
1122 (nds32_expand_ict_move): New.
1123 (nds32_indirect_call_referenced_p): New.
1124 (nds32_symbol_binds_local_p): Delete.
1125 (nds32_long_call_p): Modify.
1126 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
1127 * config/nds32/nds32-protos.h
1128 (symbolic_reference_mentioned_p): Declare.
1129 (nds32_legitimize_ict_address): Declare.
1130 (nds32_expand_ict_move): Declare.
1131 (nds32_indirect_call_referenced_p): Declare.
1132 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
1133 (nds32_relax_group): Use nds32_ict_const_p as condition.
1134 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
1135 (nds32_asm_file_start): Output ict_model directive in asm code.
1136 (nds32_legitimate_address_p): Consider indirect call.
1137 (nds32_print_operand): Consider indirect call.
1138 (nds32_print_operand_address): Consider indirect call.
1139 (nds32_insert_attributes): Handle "indirect_call" attribute.
1140 (TARGET_LEGITIMATE_ADDRESS_P): Define.
1141 (TARGET_LEGITIMATE_CONSTANT_P): Define.
1142 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
1143 (TARGET_DELEGITIMIZE_ADDRESS): Define.
1144 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
1145 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
1146 (TARGET_ICT_MODEL_SMALL): Define.
1147 (TARGET_ICT_MODEL_LARGE): Define.
1148 * config/nds32/nds32.md (movsi): Consider ict model.
1149 (call, call_value): Consider ict model.
1150 (sibcall, sibcall_value): Consider ict model.
1151 * config/nds32/nds32.opt (mict-model): New option.
1152 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
1155 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
1156 Monk Chiang <sh.chiang04@gmail.com>
1157 Jim Wilson <jimw@sifive.com>
1159 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
1160 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
1161 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
1162 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
1163 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
1164 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
1165 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
1166 compute save_libcall_adjustment properly.
1167 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
1168 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
1169 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
1170 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
1171 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
1172 (ABI_SPEC): Handle mabi=ilp32e.
1173 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
1174 (RVE): Add RVE mask.
1175 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
1176 <-march>: Add rv32e as an example.
1178 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
1181 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
1182 (intra_create_variable_infos): Handle C++ constructors.
1184 2018-05-18 Martin Liska <mliska@suse.cz>
1186 * passes.def: Remove a redundant pass.
1188 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
1191 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
1193 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1195 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
1197 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
1199 (armv2, armv2a, armv3, armv3m): Delete architectures.
1200 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
1201 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
1202 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
1204 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
1205 (*mulsidi3adddi): Likewise.
1206 (mulsidi3): Likewise.
1207 (*mulsidi3_nov6): Likewise.
1208 (umulsidi3): Likewise.
1209 (umulsidi3_nov6): Likewise.
1210 (umaddsidi4): Likewise.
1211 (*umulsidi3adddi): Likewise.
1212 (smulsi3_highpart): Likewise.
1213 (*smulsi3_highpart_nov6): Likewise.
1214 (umulsi3_highpart): Likewise.
1215 (*umulsi3_highpart_nov6): Likewise.
1216 * config/arm/arm.h (arm_arch3m): Delete.
1217 * config/arm/arm.c (arm_arch3m): Delete.
1218 (arm_option_override_internal): Update armv3-related comment.
1219 (arm_configure_build_target): Delete use of isa_bit_mode32.
1220 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
1221 (arm_rtx_costs_internal): Delete check of arm_arch3m.
1222 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
1224 (mulusa3): Likewise.
1225 * config/arm/arm-protos.h (arm_arch3m): Delete.
1226 * config/arm/arm-tables.opt: Regenerate.
1227 * config/arm/arm-tune.md: Likewise.
1228 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
1229 deleted architectures.
1231 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1233 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
1234 (armv5t, armv5te): New features.
1235 (ARMv5, ARMv5e): Delete fgroups.
1236 (ARMv5t, ARMv5te): Adjust for above changes.
1238 (armv5, armv5e): Delete arches.
1239 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
1241 (*call_reg_arm): Likewise.
1242 (*call_value_reg_armv5): Likewise.
1243 (*call_value_reg_arm): Likewise.
1244 (*call_symbol): Likewise.
1245 (*call_value_symbol): Likewise.
1246 (*sibcall_insn): Likewise.
1247 (*sibcall_value_insn): Likewise.
1249 (prefetch): Likewise.
1250 (define_split and define_peephole2 dependent on arm_arch5):
1252 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
1254 (TARGET_ARM_QBIT): Likewise.
1255 (TARGET_DSP_MULTIPLY): Likewise.
1256 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
1257 (arm_arch5, arm_arch5e): Delete.
1258 (arm_arch5t, arm_arch5te): Declare.
1259 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
1260 (arm_arch5t): Declare.
1261 (arm_option_reconfigure_globals): Update for the above.
1262 (arm_options_perform_arch_sanity_checks): Update comment, replace
1263 use of arm_arch5 with arm_arch5t.
1264 (use_return_insn): Likewise.
1265 (arm_emit_call_insn): Likewise.
1266 (output_return_instruction): Likewise.
1267 (arm_final_prescan_insn): Likewise.
1268 (arm_coproc_builtin_available): Likewise.
1269 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
1270 arm_arch5e with arm_arch5t and arm_arch5te.
1271 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
1272 (arm_arch5t, arm_arch5te): Declare.
1273 * config/arm/arm-tables.opt: Regenerate.
1274 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
1275 * config/arm/t-multilib: Likewise.
1276 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
1277 instead of arm_arch5.
1278 (*call_reg_thumb1): Likewise.
1279 (*call_value_reg_thumb1_v5): Likewise.
1280 (*call_value_reg_thumb1): Likewise.
1281 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
1283 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
1285 2018-05-18 Martin Liska <mliska@suse.cz>
1287 PR gcov-profile/84846
1288 * doc/gcov.texi: Document -t option of gcov tool.
1290 2018-05-18 Martin Liska <mliska@suse.cz>
1292 PR gcov-profile/84846
1293 * gcov.c (print_usage): Add new -t option.
1294 (process_args): Handle the option.
1295 (generate_results): Use stdout as output when requested by
1298 2018-05-18 Martin Liska <mliska@suse.cz>
1300 PR gcov-profile/84846
1301 * coverage.c (coverage_init): Write PWD to .gcno file.
1302 * doc/gcov.texi: Document how working directory is printed.
1303 * gcov-dump.c (dump_gcov_file): Print PWD.
1304 * gcov.c (output_intermediate_file): Likewise.
1305 (read_graph_file): Read PWD string.
1306 (output_lines): Print PWD.
1308 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1311 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
1312 for retval and return false if all args to phi are zero.
1314 2018-05-18 Richard Biener <rguenther@suse.de>
1316 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
1318 (evrp_dom_walker::before_dom_children): Call it.
1320 2018-05-18 Richard Biener <rguenther@suse.de>
1322 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
1323 results when processing array refs with variable index.
1325 2018-05-18 Toon Moene <toon@moene.org>
1327 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
1328 directly after that of -floop-interchange. Indicate that both
1329 options are enabled by default when specifying -O3.
1331 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1333 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
1334 iterator. Delete separate integer-mode vec_set<mode> expander.
1335 (aarch64_simd_vec_setv2di): Delete.
1336 (vec_setv2di): Delete.
1337 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
1338 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
1339 the "w, r" alternative.
1341 2018-05-18 Martin Liska <mliska@suse.cz>
1343 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
1344 * tree-pass.h (make_pass_lower_switch_O0): New function.
1345 * tree-switch-conversion.c (node_has_low_bound): Remove.
1346 (node_has_high_bound): Likewise.
1347 (node_is_bounded): Likewise.
1348 (class pass_lower_switch): Make it a template type and create
1350 (pass_lower_switch::execute): Add template argument.
1351 (make_pass_lower_switch): New function.
1352 (make_pass_lower_switch_O0): New function.
1353 (do_jump_if_equal): Remove.
1354 (emit_case_nodes): Simplify to just handle all 3 cases and leave
1355 all the hard work to tree optimization passes.
1357 2018-05-18 Martin Liska <mliska@suse.cz>
1359 * dbgcnt.c (limit_low): Renamed from limit.
1360 (limit_high): New variable.
1361 (dbg_cnt_is_enabled): Check for upper limit.
1362 (dbg_cnt): Adjust dumping.
1363 (dbg_cnt_set_limit_by_index): Add new argument for high
1365 (dbg_cnt_set_limit_by_name): Likewise.
1366 (dbg_cnt_process_single_pair): Parse new format.
1367 (dbg_cnt_process_opt): Use strtok.
1368 (dbg_cnt_list_all_counters): Remove 'value' and add
1370 * doc/invoke.texi: Document changes.
1372 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
1374 * doc/sourcebuild.texi (scalar_all_fma): Document.
1375 * tree.def (FMA_EXPR): Delete.
1376 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
1377 * internal-fn.c (ternary_direct): New macro.
1378 (expand_ternary_optab_fn): Likewise.
1379 (direct_ternary_optab_supported_p): Likewise.
1380 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
1381 * builtins.c (fold_builtin_fma): Delete.
1382 (fold_builtin_3): Don't call it.
1383 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
1384 * expr.c (expand_expr_real_2): Likewise.
1385 * fold-const.c (operand_equal_p): Likewise.
1386 (fold_ternary_loc): Likewise.
1387 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
1388 * gimple.c (DEFTREECODE): Likewise.
1389 * gimplify.c (gimplify_expr): Likewise.
1390 * optabs-tree.c (optab_for_tree_code): Likewise.
1391 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
1392 * tree-eh.c (operation_could_trap_p): Likewise.
1393 (stmt_could_throw_1_p): Likewise.
1394 * tree-inline.c (estimate_operator_cost): Likewise.
1395 * tree-pretty-print.c (dump_generic_node): Likewise.
1396 (op_code_prio): Likewise.
1397 * tree-ssa-loop-im.c (stmt_cost): Likewise.
1398 * tree-ssa-operands.c (get_expr_operands): Likewise.
1399 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
1400 * fold-const-call.h (fold_fma): Delete.
1401 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
1402 CFN_FNMA and CFN_FNMS.
1404 * genmatch.c (combined_fn): New enum.
1405 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
1406 (commutative_op): New function.
1407 (commutate): Use it. Handle more than 2 operands.
1408 (dt_operand::gen_gimple_expr): Use commutative_op.
1409 (parser::parse_expr): Allow :c to be used with non-binary
1410 operators if the commutative operand is known.
1411 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
1412 CFN_FMS, CFN_FNMA and CFN_FNMS.
1413 (backprop::process_assign_use): Remove FMA_EXPR handling.
1414 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
1415 (gen_hsa_fma): New function.
1416 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
1417 IFN_FNMA and IFN_FNMS.
1418 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
1419 * gimple-fold.h (follow_all_ssa_edges): Declare.
1420 * gimple-fold.c (follow_all_ssa_edges): New function.
1421 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
1422 gimple_build interface and use follow_all_ssa_edges to fold the result.
1423 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
1424 instead of checking for optabs directly.
1425 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
1426 rather than FMA_EXPRs.
1427 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
1428 call to IFN_FMA instead of an FMA_EXPR.
1430 2018-05-17 Jim Wilson <jimw@sifive.com>
1432 * expr.c (do_tablejump): When converting index to Pmode, if we have a
1433 sign extended promoted subreg, and the range does not have the sign bit
1434 set, then do a sign extend.
1436 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
1437 test, check for sign extended subreg and/or constant operands, and
1438 do a sign extend in that case.
1440 2018-05-17 Steve Ellcey <sellcey@cavium.com>
1442 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
1443 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
1445 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
1446 Change logics_shift_reg to logics_shift_imm.
1447 (thunderx2t99_fp_loadpair_basic): Delete.
1448 (thunderx2t99_fp_storepair_basic): Delete.
1449 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
1450 (thunderx2t99_asimd_polynomial): Delete.
1451 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
1452 and neon_fp_mul_d_scalar_q.
1453 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
1454 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
1455 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
1456 (thunderx2t99_asimd_lut): Add missing tbl types.
1457 (thunderx2t99_asimd_ext): Delete.
1458 (thunderx2t99_asimd_load1_1_mult): Delete.
1459 (thunderx2t99_asimd_load1_2_mult): Delete.
1460 (thunderx2t99_asimd_load1_ldp): New.
1461 (thunderx2t99_asimd_load1): New.
1462 (thunderx2t99_asimd_load2): Add missing *load2* types.
1463 (thunderx2t99_asimd_load3): New.
1464 (thunderx2t99_asimd_load4): New.
1465 (thunderx2t99_asimd_store1_1_mult): Delete.
1466 (thunderx2t99_asimd_store1_2_mult): Delete.
1467 (thunderx2t99_asimd_store2_mult): Delete.
1468 (thunderx2t99_asimd_store2_onelane): Delete.
1469 (thunderx2t99_asimd_store_stp): New.
1470 (thunderx2t99_asimd_store1): New.
1471 (thunderx2t99_asimd_store2): New.
1472 (thunderx2t99_asimd_store3): New.
1473 (thunderx2t99_asimd_store4): New.
1475 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
1477 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
1478 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
1480 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
1481 Segher Boessenkool <segher@kernel.crashing.org>
1484 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
1487 2018-05-17 Richard Biener <rguenther@suse.de>
1489 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
1490 for pruning loop and prune defs feeding only already visited PHIs.
1492 2018-05-17 Richard Biener <rguenther@suse.de>
1494 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
1496 2018-05-17 Bin Cheng <bin.cheng@arm.com>
1497 Richard Biener <rguenther@suse.de>
1499 PR tree-optimization/85793
1500 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
1501 for VMAT_ELEMENTWISE.
1503 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1505 * internal-fn.h (lookup_internal_fn): Declare
1506 * internal-fn.c (lookup_internal_fn): New function.
1507 * gimple.c (gimple_build_call_from_tree): Handle calls to
1509 * gimple-pretty-print.c (dump_gimple_call): Print "." before
1510 internal function names.
1511 * tree-pretty-print.c (dump_generic_node): Likewise.
1512 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
1514 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1516 * gimple-fold.h (gimple_build): Make the function forms take
1517 combined_fn rather than built_in_function.
1518 (gimple_simplify): Likewise.
1519 * gimple-match-head.c (gimple_simplify): Likewise.
1520 * gimple-fold.c (gimple_build): Likewise.
1521 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
1522 rather than gimple_build_call_internal.
1523 (get_initial_defs_for_reduction): Likewise.
1524 (vect_create_epilog_for_reduction): Likewise.
1525 (vectorizable_live_operation): Likewise.
1527 2018-05-17 Martin Liska <mliska@suse.cz>
1529 * gimple-ssa-sprintf.c (format_directive): Do not use
1530 space in between 'G_' and '('.
1532 2018-05-17 Jakub Jelinek <jakub@redhat.com>
1535 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
1536 even if the mask is not all ones.
1539 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
1541 (ix86_gimple_fold_builtin): Likewise.
1544 * config/i386/i386.c: Include tree-vector-builder.h.
1545 (ix86_vector_shift_count): New function.
1546 (ix86_fold_builtin): Fold shift builtins by scalar count.
1547 (ix86_gimple_fold_builtin): Likewise.
1549 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
1550 _mm512_setzero): New intrinsics.
1552 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
1553 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1555 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
1556 code generation for cases where splatting a value is not useful.
1557 * simplify-rtx.c (simplify_ternary_operation): Simplify
1558 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
1559 mode to a vec_concat.
1561 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
1563 * config.gcc: Support "goldmont-plus".
1564 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1566 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1567 PROCESSOR_GOLDMONT_PLUS.
1568 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
1569 (processor_target_table): Add "goldmont-plus".
1570 (PTA_GOLDMONT_PLUS): Define.
1571 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
1572 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
1573 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
1574 (fold_builtin_cpu): Add "goldmont-plus".
1575 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
1576 (ix86_option_override_internal): Add "goldmont-plus".
1577 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
1578 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
1579 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
1580 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
1582 2018-05-17 Richard Biener <rguenther@suse.de>
1584 PR tree-optimization/85757
1585 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
1586 remove defs that only feed that PHI from further processing.
1588 2018-05-16 Jim Wilson <jimw@sifive.com>
1590 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
1592 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
1594 2018-05-16 Mark Wielaard <mark@klomp.org>
1596 * dwarf2out.c (count_index_strings): New function.
1597 (output_indirect_strings): Call count_index_strings and generate
1598 header for dwarf_version >= 5.
1600 2018-05-16 Mark Wielaard <mark@klomp.org>
1602 * dwarf2out.c (dwarf_FORM): New function.
1603 (set_indirect_string): Use dwarf_FORM.
1604 (reset_indirect_string): Likewise.
1605 (size_of_die): Likewise.
1606 (value_format): Likewise.
1607 (output_die): Likewise.
1608 (add_skeleton_AT_string): Likewise.
1609 (output_macinfo_op): Likewise.
1610 (index_string): Likewise.
1611 (output_index_string_offset): Likewise.
1612 (output_index_string): Likewise.
1613 (count_index_strings): Likewise.
1615 2018-05-16 Carl Love <cel@us.ibm.com>
1617 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1618 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
1620 2018-05-16 Martin Jambor <mjambor@suse.cz>
1622 * ipa-prop.c (ipa_free_all_edge_args): Remove.
1623 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
1625 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
1627 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
1628 (fnma<mode>4): Likewise.
1629 (fms<mode>4): Likewise.
1630 (fnms<mode>4): Likewise.
1631 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
1632 (aarch64_fnma<mode>4): Likewise.
1633 (aarch64_fms<mode>4): Likewise.
1634 (aarch64_fnms<mode>4): Likewise.
1635 (aarch64_fnmadd<mode>4): Likewise.
1637 2018-05-16 Jason Merrill <jason@redhat.com>
1639 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
1641 2018-05-16 Richard Biener <rguenther@suse.de>
1643 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
1644 (dump_stmt_cost): Declare.
1645 (add_stmt_cost): Dump cost we add.
1646 (add_stmt_costs): New function.
1647 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
1649 (vect_analyze_stmt): Adjust prototype.
1650 (vectorizable_condition): Likewise.
1651 (vectorizable_live_operation): Likewise.
1652 (vectorizable_reduction): Likewise.
1653 (vectorizable_induction): Likewise.
1654 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
1655 cost vector to pass to vectorizable_ and record afterwards.
1656 (vect_model_reduction_cost): Take cost vector argument and adjust.
1657 (vect_model_induction_cost): Likewise.
1658 (vectorizable_reduction): Likewise.
1659 (vectorizable_induction): Likewise.
1660 (vectorizable_live_operation): Likewise.
1661 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
1662 SLP_TREE_NUMBER_OF_VEC_STMTS.
1663 (vect_analyze_slp_cost_1): Remove.
1664 (vect_analyze_slp_cost): Likewise.
1665 (vect_slp_analyze_node_operations): Take visited args and
1666 a target cost vector. Avoid processing already visited stmt sets.
1667 (vect_slp_analyze_operations): Use a local cost vector to gather
1668 costs and register those of non-discarded instances.
1669 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
1670 (vect_schedule_slp_instance): Remove copying of
1671 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
1673 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
1674 adding cost. Record cost entry location.
1675 (vect_prologue_cost_for_slp_op): Function to compute cost of
1676 a constant or invariant generated for SLP vect in the prologue,
1677 split out from vect_analyze_slp_cost_1.
1678 (vect_model_simple_cost): Make static. Adjust for SLP costing.
1679 (vect_model_promotion_demotion_cost): Likewise.
1680 (vect_model_store_cost): Likewise, make static.
1681 (vect_model_load_cost): Likewise.
1682 (vectorizable_bswap): Add cost vector arg and adjust.
1683 (vectorizable_call): Likewise.
1684 (vectorizable_simd_clone_call): Likewise.
1685 (vectorizable_conversion): Likewise.
1686 (vectorizable_assignment): Likewise.
1687 (vectorizable_shift): Likewise.
1688 (vectorizable_operation): Likewise.
1689 (vectorizable_store): Likewise.
1690 (vectorizable_load): Likewise.
1691 (vectorizable_condition): Likewise.
1692 (vectorizable_comparison): Likewise.
1693 (can_vectorize_live_stmts): Likewise.
1694 (vect_analyze_stmt): Likewise.
1695 (vect_transform_stmt): Adjust calls to vectorizable_*.
1696 * tree-vectorizer.c: Include gimple-pretty-print.h.
1697 (dump_stmt_cost): New function.
1699 2018-05-16 Richard Biener <rguenther@suse.de>
1701 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
1702 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
1703 * tree-ssa-dse.c: Include tree-ssa-loop.h.
1704 (check_name): New callback.
1705 (dse_classify_store): Track cycles via a visited bitmap of PHI
1706 defs and simplify handling of in-loop and across loop dead stores
1707 and properly fail for loop-variant refs. Handle byte-tracking with
1708 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
1711 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
1713 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
1714 (vect_get_mask_type_for_stmt): Likewise.
1715 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
1717 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
1718 to determine the statement's vector type and the vector type that
1719 should be used for calculating nunits. Deal with cases in which
1720 the type has to be deferred.
1721 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
1722 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
1723 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
1724 (vect_determine_vf_for_stmt): New functions, split out from...
1725 (vect_determine_vectorization_factor): ...here.
1726 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
1727 (vect_get_mask_type_for_stmt): New functions, split out from
1728 vect_determine_vectorization_factor.
1730 2018-05-16 Richard Biener <rguenther@suse.de>
1732 * tree-cfg.c (verify_gimple_assign_ternary): Properly
1733 verify the [VEC_]COND_EXPR embedded comparison.
1735 2018-05-15 Martin Sebor <msebor@redhat.com>
1737 PR tree-optimization/85753
1738 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
1739 RECORD_TYPE in addition to ARRAY_TYPE.
1741 2018-05-15 Martin Sebor <msebor@redhat.com>
1744 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
1746 2018-05-15 Richard Biener <rguenther@suse.de>
1748 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
1749 add by_clobber_p one. Change algorithm to collect all defs
1750 representing uses we need to walk and try reducing them to
1751 a single one before failing.
1752 (dse_dom_walker::dse_optimize_stmt): Adjust.
1754 2018-05-13 Mark Wielaard <mark@klomp.org>
1756 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
1757 (size_of_loc_descr): Likewise.
1758 (output_loc_operands): Likewise.
1759 (output_loc_operands_raw): Likewise.
1760 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
1761 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
1762 (hash_loc_operands): Likewise.
1763 (compare_loc_operands): Likewise.
1765 2018-05-14 Mark Wielaard <mark@klomp.org>
1767 * dwarf2out.c (count_index_addrs): New function.
1768 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
1770 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1772 PR tree-optimization/83648
1773 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
1774 return value as malloc candidate.
1776 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1779 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
1780 as true in call to suggest_attribute.
1782 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
1784 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
1785 -mreadonly-in-sdata.
1787 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1789 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
1791 (aarch64_crypto_aesd_fused): Likewise.
1793 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
1795 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
1796 (movsi_aarch64): Likewise.
1797 (load_pairsi): Likewise.
1798 (load_pairdi): Likewise.
1799 (store_pairsi): Likewise.
1800 (store_pairdi): Likewise.
1801 (load_pairsf): Likewise.
1802 (load_pairdf): Likewise.
1803 (store_pairsf): Likewise.
1804 (store_pairdf): Likewise.
1805 (zero_extend): Likewise.
1806 (trunc): Swap alternatives.
1807 (fcvt_target): Add '?' to prefer w over r.
1809 2018-05-14 Jakub Jelinek <jakub@redhat.com>
1812 * config/i386/i386.md: Disallow non-commutative arithmetics in
1813 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
1814 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
1815 in the peephole2 before it.
1817 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
1819 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
1820 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
1821 (ix86_handle_option): Handle -mcldemote.
1822 * config.gcc: New header.
1823 * config/i386/cldemoteintrin.h: New file.
1824 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
1825 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1827 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1828 OPTION_MASK_ISA_CLDEMOTE.
1829 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
1830 (ix86_valid_target_attribute_inner_p): Ditto.
1831 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
1832 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
1833 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
1834 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
1835 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
1837 * config/i386/i386.opt: Add -mcldemote.
1838 * config/i386/x86intrin.h: New header.
1839 * doc/invoke.texi: Add -mcldemote.
1841 2018-05-14 Richard Biener <rguenther@suse.de>
1843 * doc/match-and-simplify.texi: Adjust :s documentation.
1845 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
1847 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
1848 intended memcpy size.
1849 (REORDER_45): Likewise.
1851 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
1853 * sort.cc: New file.
1854 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
1855 * vec.c (qsort_chk): Use gcc_qsort.
1856 * Makefile.in (OBJS-libcommon): Add sort.o.
1857 (build/sort.o): New target. Use it...
1858 (BUILD_RTL): ... here, and...
1859 (build/gencfn-macros): ... here, and...
1860 (build/genmatch): ... here.
1862 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1863 Chung-Ju Wu <jasonwucj@gmail.com>
1865 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
1866 * config/nds32/nds32-graywolf.md: New file.
1867 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
1868 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
1870 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
1871 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
1872 * config/nds32/nds32.md (pipeline_model): Add graywolf.
1873 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
1874 * config/nds32/pipelines.md: Include n15 settings.
1876 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1877 Chung-Ju Wu <jasonwucj@gmail.com>
1879 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
1880 * config/nds32/nds32-n13.md: New file.
1881 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
1882 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
1884 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
1885 * config/nds32/nds32.md (pipeline_model): Add n13.
1886 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
1887 * config/nds32/pipelines.md: Include n13 settings.
1889 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1890 Chung-Ju Wu <jasonwucj@gmail.com>
1892 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
1893 * config/nds32/nds32-n10.md: New file.
1894 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
1895 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
1897 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
1898 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
1899 * config/nds32/nds32.md (pipeline_model): Add n10.
1900 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
1901 * config/nds32/pipelines.md: Include n10 settings.
1903 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
1904 Kito Cheng <kito.cheng@gmail.com>
1905 Chung-Ju Wu <jasonwucj@gmail.com>
1907 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
1908 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1909 Add enum values for DSP extension instructions.
1910 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
1912 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
1913 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
1915 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
1916 * config/nds32/nds32-dspext.md: New file for DSP implementation.
1917 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
1918 * config/nds32/nds32-intrinsic.md: Likewise.
1919 * config/nds32/nds32_intrinsic.h: Likewise.
1920 * config/nds32/nds32-md-auxiliary.c: Likewise.
1921 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
1922 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
1923 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
1924 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
1925 * config/nds32/nds32-protos.h: New declarations for DSP extension.
1926 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
1927 TYPE_DMAC in switch statement.
1928 * config/nds32/nds32.c: New checking and implementation for DSP
1929 extension instructions.
1930 * config/nds32/nds32.h: Likewise.
1931 * config/nds32/nds32.md: Likewise.
1932 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
1933 * config/nds32/predicates.md: Implement new predicates for DSP
1936 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
1938 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
1939 Reformat alternatives and attributes so it is easier to identify
1940 which constraints/attributes go with which instruction.
1941 (mov<mode>_hardfloat32, FMOVE64): Likewise.
1942 (mov<mode>_softfloat32, FMOVE64): Likewise.
1943 (mov<mode>_hardfloat64, FMOVE64): Likewise.
1944 (mov<mode>_softfloat64, FMOVE64): Likewise.
1946 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
1948 * doc/extend.texi (PowerPC Built-in Functions): Rename this
1950 (Basic PowerPC Built-in Functions): The new name of the
1951 subsection previously known as "PowerPC Built-in Functions".
1952 (Basic PowerPC Built-in Functions Available on all Configurations):
1954 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
1955 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
1956 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
1957 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
1959 2018-05-11 Martin Jambor <mjambor@suse.cz>
1962 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
1965 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
1968 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
1970 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
1972 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
1973 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
1974 (ix86_handle_option): Handle -mwaitpkg.
1975 * config.gcc: New header.
1976 * config/i386/cpuid.h (bit_WAITPKG): New bit.
1977 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
1978 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
1980 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1981 OPTION_MASK_ISA_WAITPKG.
1982 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
1983 (ix86_option_override_internal): Add PTA_WAITPKG.
1984 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
1985 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
1986 IX86_BUILTIN_TPAUSE.
1987 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
1988 __builtin_ia32_umwait and __builtin_ia32_tpause.
1989 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
1990 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
1991 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
1992 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
1993 UNSPECV_TPAUSE): New.
1994 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
1995 * config/i386/i386.opt: Add -mwaitpkg.
1996 * config/i386/waitpkgintrin.h: New file.
1997 * config/i386/x86intrin.h: New header.
1998 * doc/invoke.texi: Add -mwaitpkg.
2000 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2003 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
2005 (cortex-m0): Use armv6s-m isa.
2006 (cortex-m0plus): Likewise.
2007 (cortex-m1): Likewise.
2008 (cortex-m0.small-multiply): Likewise.
2009 (cortex-m0plus.small-multiply): Likewise.
2010 (cortex-m1.small-multiply): Likewise.
2012 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
2013 Jakub Jelinek <jakub@redhat.com>
2015 PR tree-optimization/85692
2016 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
2017 source permute as well.
2019 2018-05-11 Martin Liska <mliska@suse.cz>
2022 * doc/extend.texi: Document LLVM style format for no_sanitize
2025 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
2027 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
2028 mode_supports_vsx_dform_quad to mode_supports_dq_form.
2029 (mode_supports_vsx_dform_quad): Likewise.
2030 (mode_supports_vmx_dform): Move these functions to be next to the
2031 other mode_supports functions.
2032 (mode_supports_dq_form): Likewise.
2033 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
2034 mode_supports_dq_form.
2035 (reg_offset_addressing_ok_p): Likewise.
2036 (offsettable_ok_by_alignment): Likewise.
2037 (rs6000_legitimate_offset_address_p): Likewise.
2038 (legitimate_lo_sum_address_p): Likewise.
2039 (rs6000_legitimize_address): Likewise.
2040 (rs6000_legitimize_reload_address): Likewise.
2041 (rs6000_secondary_reload_inner): Likewise.
2042 (rs6000_preferred_reload_class): Likewise.
2043 (rs6000_output_move_128bit): Likewise.
2045 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2047 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
2048 Generate SImode target register for null target.
2049 <case IX86_BUILTIN_XGETBV>: Ditto.
2050 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
2051 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
2053 2018-05-10 Carl Love <cel@us.ibm.com>
2055 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2056 dcbtt and dcbtstt if operands[2] is 0.
2058 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2061 * config/i386/sse.md (usadv64qi): New expander.
2063 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
2065 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
2066 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
2067 -maltivec=be support.
2068 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
2069 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
2070 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
2071 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
2072 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
2073 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
2074 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
2075 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
2076 altivec_vsumsws): Adjust.
2077 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
2078 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
2079 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
2081 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
2082 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
2083 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
2084 (altivec_lve<VI_char>x): Delete expand.
2085 (*altivec_lve<VI_char>x_internal): Rename to...
2086 (altivec_lve<VI_char>x): ... this.
2087 (altivec_lvxl_<mode>): Delete expand.
2088 (*altivec_lvxl_<mode>_internal): Rename to ...
2089 (altivec_lvxl_<mode>): ... this.
2090 (altivec_stvxl_<mode>): Delete expand.
2091 (*altivec_stvxl_<mode>_internal): Rename to ...
2092 (altivec_stvxl_<mode>): ... this.
2093 (altivec_stve<VI_char>x): Delete expand.
2094 (*altivec_stve<VI_char>x_internal): Rename to ...
2095 (altivec_stve<VI_char>x): ... this.
2096 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
2097 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
2098 reduc_plus_scal_<mode>): Adjust.
2099 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
2101 (rs6000_cpu_cpp_builtins): Adjust.
2102 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
2103 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
2104 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
2105 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
2106 -maltivec=be support.
2107 (rs6000_split_vec_extract_var): Adjust.
2108 (rs6000_split_v4si_init): Adjust.
2109 (swap_selector_for_mode): Delete.
2110 (altivec_expand_lvx_be, altivec_expand_stvx_be,
2111 altivec_expand_stvex_be): Delete.
2112 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
2113 -maltivec=be support.
2114 (rs6000_gimple_fold_builtin): Ditto.
2115 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
2117 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
2118 (TARGET_DIRECT_MOVE_64BIT): Adjust.
2119 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
2120 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
2121 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
2122 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
2123 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
2124 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
2125 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
2126 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
2127 anonymous split): Adjust.
2128 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
2129 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
2131 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
2133 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
2134 when --with-gxx-include-dir is also specified.
2135 * configure: Regenerate.
2137 2018-05-09 Jim Wilson <jimw@sifive.com>
2140 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
2141 * config/riscv/t-withmultilib: New.
2142 * config/riscv/withmultilib.h: New.
2143 * doc/install.texi: Document RISC-V --with-multilib-list support.
2145 2018-05-09 Richard Biener <rguenther@suse.de>
2147 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
2149 (vect_bb_vectorization_profitable_p): Adjust. Compute
2150 actual scalar cost using the cost vector and the add_stmt_cost
2153 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2155 PR rtl-optimization/85645
2156 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
2157 in the REG_CFA_REGISTER note for LR, don't leave it empty.
2159 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2161 PR rtl-optimization/85645
2162 * shrink-wrap.c (spread_components): Return a boolean saying if
2163 anything was changed.
2164 (try_shrink_wrapping_separate): Iterate spread_components until
2165 nothing changes anymore.
2167 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2169 PR rtl-optimization/85645
2170 * regrename.c (build_def_use): Also kill the chains that include the
2171 destination of a REG_CFA_REGISTER note.
2173 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2175 PR rtl-optimization/85645
2176 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
2177 insn that has a REG_CFA_REGISTER note.
2179 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
2181 * cfgexpand.c (expand_clobber): New function.
2182 (expand_gimple_stmt_1): Use it.
2183 * tree-vect-stmts.c (vect_clobber_variable): New function,
2185 (vectorizable_simd_clone_call): ...here.
2186 (vectorizable_store): Emit a clobber either side of an
2187 IFN_STORE_LANES sequence.
2188 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
2190 2018-05-09 Tom de Vries <tom@codesourcery.com>
2193 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
2194 (define_insn "trap_if_false"): Add exit after trap.
2196 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
2198 PR rtl-optimization/85638
2199 * bb-reorder.c: Include common/common-target.h.
2200 (create_forwarder_block): New function extracted from...
2201 (fix_up_crossing_landing_pad): ...here. Rename into...
2202 (dw2_fix_up_crossing_landing_pad): ...this.
2203 (sjlj_fix_up_crossing_landing_pad): New function.
2204 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
2205 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
2206 from both partitions and exit the loop after one iteration.
2208 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2211 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2213 (Basic PowerPC Built-in Functions): The new name of the
2214 subsection previously known as "PowerPC Built-in Functions".
2215 (Basic PowerPC Built-in Functions Available on all Configurations):
2217 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2219 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2221 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2223 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2226 2018-05-08 Jim Wilson <jimw@sifive.com>
2228 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
2229 (LD_EMUL_SUFFIX): New.
2230 (LINK_SPEC): Use it.
2232 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2234 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2236 (Basic PowerPC Built-in Functions): The new name of the
2237 subsection previously known as "PowerPC Built-in Functions".
2238 (Basic PowerPC Built-in Functions Available on all Configurations):
2240 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2242 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2244 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2246 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2249 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2252 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
2253 after cmpelim optimization.
2255 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
2257 * config.gcc: Support "goldmont".
2258 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
2259 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2261 * config/i386/i386.c (m_GOLDMONT): Define.
2262 (processor_target_table): Add "goldmont".
2263 (PTA_GOLDMONT): Define.
2264 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
2265 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
2266 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
2267 (fold_builtin_cpu): Add "goldmont".
2268 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
2269 (ix86_option_override_internal): Add "goldmont".
2270 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
2271 (processor_type): Add PROCESSOR_GOLDMONT.
2272 * config/i386/i386.md: Add CPU "glm".
2273 * config/i386/glm.md: New file.
2274 * config/i386/x86-tune.def: Add m_GOLDMONT.
2275 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
2277 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2280 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
2282 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
2283 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
2284 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
2287 * config/i386/i386.c (ix86_fold_builtin): Handle
2288 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
2291 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
2292 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
2294 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
2297 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
2298 (check_arch): Likewise.
2299 (check_fpu): Return the result rather than printing it.
2300 (end arch): Fix operator precedence.
2301 (end cpu): Likewise.
2302 (END): Print the result from check_fpu.
2304 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2305 Alan Hayward <alan.hayward@arm.com>
2306 David Sherwood <david.sherwood@arm.com>
2308 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
2309 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
2310 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
2311 (*fcmuo<mode>_and): New patterns.
2313 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2315 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
2316 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
2317 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
2318 (cmp_op, sve_imm_con): New code attributes.
2319 (SVE_COND_INT_CMP, imm_con): Delete.
2320 (cmp_op): Remove above unspecs from int attribute.
2321 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
2323 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
2324 comparison-specific unspecs.
2325 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
2326 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
2327 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
2328 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
2329 (*vec_fcm<cmp_op><mode>): Rename to...
2330 (*fcm<cmp_op><mode>): ...this and adjust likewise.
2331 (*vec_fcmuo<mode>): Rename to...
2332 (*fcmuo<mode>): ...this and adjust likewise.
2333 (*pred_fcm<cmp_op><mode>): New pattern.
2334 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
2335 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
2337 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
2339 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
2340 (aarch64_emit_sve_predicated_cond): New function.
2341 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
2342 (aarch64_emit_unspec_cond_or): Replace with...
2343 (aarch64_emit_sve_or_conds): ...this new function. Use
2344 aarch64_emit_sve_ptrue_op for the individual comparisons and
2345 aarch64_emit_binop to OR them together.
2346 (aarch64_emit_inverted_unspec_cond): Replace with...
2347 (aarch64_emit_sve_inverted_cond): ...this new function. Use
2348 aarch64_emit_sve_ptrue_op for the comparison and
2349 aarch64_emit_unop to invert the result.
2350 (aarch64_expand_sve_vec_cmp_float): Update after the above
2351 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
2353 2018-05-07 Nathan Sidwell <nathan@acm.org>
2355 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
2356 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
2357 (Backwards Compatibility): Likewise.
2359 2018-05-07 Luis Machado <luis.machado@linaro.org>
2363 2018-05-07 Luis Machado <luis.machado@linaro.org>
2365 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2366 <prefetch_dynamic_strides>: New const bool field.
2367 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2368 prefetch_dynamic_strides.
2369 (exynosm1_prefetch_tune): Likewise.
2370 (thunderxt88_prefetch_tune): Likewise.
2371 (thunderx_prefetch_tune): Likewise.
2372 (thunderx2t99_prefetch_tune): Likewise.
2373 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2374 (aarch64_override_options_internal): Update to set
2375 PARAM_PREFETCH_DYNAMIC_STRIDES.
2376 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2377 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2378 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2379 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2380 prefetch-dynamic-strides setting.
2382 2018-05-07 Luis Machado <luis.machado@linaro.org>
2384 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2385 <minimum_stride>: New const int field.
2386 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2387 minimum_stride field.
2388 (exynosm1_prefetch_tune): Likewise.
2389 (thunderxt88_prefetch_tune): Likewise.
2390 (thunderx_prefetch_tune): Likewise.
2391 (thunderx2t99_prefetch_tune): Likewise.
2392 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2393 (aarch64_override_options_internal): Update to set
2394 PARAM_PREFETCH_MINIMUM_STRIDE.
2395 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2396 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2397 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2398 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2399 stride is constant and is below the minimum stride threshold.
2401 2018-05-07 Luis Machado <luis.machado@linaro.org>
2403 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
2406 2018-05-07 Luis Machado <luis.machado@linaro.org>
2408 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2409 <prefetch_dynamic_strides>: New const bool field.
2410 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2411 prefetch_dynamic_strides.
2412 (exynosm1_prefetch_tune): Likewise.
2413 (thunderxt88_prefetch_tune): Likewise.
2414 (thunderx_prefetch_tune): Likewise.
2415 (thunderx2t99_prefetch_tune): Likewise.
2416 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2417 (aarch64_override_options_internal): Update to set
2418 PARAM_PREFETCH_DYNAMIC_STRIDES.
2419 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2420 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2421 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2422 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2423 prefetch-dynamic-strides setting.
2425 2018-05-07 Luis Machado <luis.machado@linaro.org>
2427 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2428 <minimum_stride>: New const int field.
2429 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2430 minimum_stride field.
2431 (exynosm1_prefetch_tune): Likewise.
2432 (thunderxt88_prefetch_tune): Likewise.
2433 (thunderx_prefetch_tune): Likewise.
2434 (thunderx2t99_prefetch_tune): Likewise.
2435 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2436 (aarch64_override_options_internal): Update to set
2437 PARAM_PREFETCH_MINIMUM_STRIDE.
2438 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2439 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2440 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2441 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2442 stride is constant and is below the minimum stride threshold.
2444 2018-05-06 Jakub Jelinek <jakub@redhat.com>
2447 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
2448 the type is addressable. Don't force op into register if it has
2451 2018-05-05 Roland McGrath <mcgrathr@google.com>
2454 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
2455 any section for which we don't know a specific type it should have,
2456 regardless of name. Previously this was done only for the exact
2457 names ".init_array", ".fini_array", and ".preinit_array".
2458 (default_elf_asm_named_section): Add comment about
2459 relationship with default_section_type_flags and SECTION_NOTYPE.
2460 (get_section): Don't consider it a type conflict if one side has
2461 SECTION_NOTYPE and the other doesn't, as long as neither has the
2462 SECTION_BSS et al used in the default_section_type_flags logic.
2464 2018-05-05 Tom de Vries <tom@codesourcery.com>
2467 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
2468 (workaround_barsyncs): New function.
2469 (nvptx_reorg): Use workaround_barsyncs.
2470 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2471 (define_expand "nvptx_membar_cta"): New define_expand.
2472 (define_insn "*nvptx_membar_cta"): New insn.
2474 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
2476 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
2477 To improve optimization opportunities.
2478 * builtin-types.def: The new needed builtin types for the above.
2480 2018-05-04 Richard Biener <rguenther@suse.de>
2482 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
2483 * gimple-ssa-store-merging.c
2484 (imm_store_chain_info::output_merged_store): Remove redundant create,
2485 release split_store vector contents on failure.
2486 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
2487 scalar stmt vector on cache hit.
2489 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
2491 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
2493 * config.gcc (powerpc-xilinx-eabi*): Remove.
2494 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
2496 (fusion_addis_mem_combo_load): Ditto.
2497 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
2499 (rs6000_cpu_cpp_builtins): Ditto.
2500 * config/rs6000/rs6000-linux.c
2501 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
2502 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
2503 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
2505 (rs6000_setup_reg_addr_masks): Ditto.
2506 (rs6000_init_hard_regno_mode_ok): Ditto.
2507 (rs6000_option_override_internal): Ditto.
2508 (legitimate_lo_sum_address_p): Ditto.
2509 (rs6000_legitimize_address): Ditto.
2510 (rs6000_legitimize_reload_address): Ditto.
2511 (rs6000_legitimate_address_p): Ditto.
2512 (abi_v4_pass_in_fpr): Ditto.
2513 (setup_incoming_varargs): Ditto.
2514 (rs6000_gimplify_va_arg): Ditto.
2515 (rs6000_split_multireg_move): Ditto.
2516 (rs6000_savres_strategy): Ditto.
2517 (rs6000_emit_prologue_components): Ditto.
2518 (rs6000_emit_epilogue_components): Ditto.
2519 (rs6000_emit_prologue): Ditto.
2520 (rs6000_emit_epilogue): Ditto.
2521 (rs6000_elf_file_end): Ditto.
2522 (rs6000_function_value): Ditto.
2523 (rs6000_libcall_value): Ditto.
2524 * config/rs6000/rs6000.h: Ditto.
2525 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
2526 (TARGET_MINMAX): ... this. New.
2527 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
2528 * config/rs6000/rs6000.md: Remove Xilinx FP support.
2529 (*movsi_internal1_single): Delete.
2530 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
2531 mfpu=, mxilinx-fpu): Delete.
2532 * config/rs6000/singlefp.h: Delete.
2533 * config/rs6000/sysv4.h: Remove Xilinx FP support.
2534 * config/rs6000/t-rs6000: Ditto.
2535 * config/rs6000/t-xilinx: Delete.
2536 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
2537 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
2538 (VStype_simple): Delete.
2539 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
2540 * config/rs6000/xfpu.h: Delete.
2541 * config/rs6000/xfpu.md: Delete.
2542 * config/rs6000/xilinx.h: Delete.
2543 * config/rs6000/xilinx.opt: Delete.
2544 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
2545 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
2547 2018-05-04 Tom de Vries <tom@codesourcery.com>
2550 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
2553 2018-05-04 Richard Biener <rguenther@suse.de>
2556 * tree-complex.c (update_complex_assignment): We are always in SSA form.
2557 (expand_complex_div_wide): Likewise.
2558 (expand_complex_operations_1): Likewise.
2559 (expand_complex_libcall): Preserve EH info of the original stmt.
2560 (tree_lower_complex): Handle removed blocks.
2561 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
2562 on complex multiplication and division libcall builtins.
2564 2018-05-04 Richard Biener <rguenther@suse.de>
2567 * fold-const.c (negate_expr_p): Restrict negation of operand
2568 zero of a division to when we know that can happen without
2570 (fold_negate_expr_1): Likewise.
2572 2018-05-04 Jakub Jelinek <jakub@redhat.com>
2575 * real.h (real_nextafter): Declare.
2576 * real.c (real_nextafter): New function.
2577 * fold-const-call.c (fold_const_nextafter): New function.
2578 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
2579 CASE_CFN_NEXTTOWARD.
2580 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
2581 even when arg1_mode is different from arg0_mode.
2583 2018-05-03 Nathan Sidwell <nathan@acm.org>
2585 * doc/extend.texi (Deprecated Features): Remove
2587 (Backwards Compatibility): Likewise.
2588 * doc/invoke.texi (C++ Language Options): Likewise.
2589 (C++ Dialect Options): Likewise.
2591 2018-05-03 Jakub Jelinek <jakub@redhat.com>
2594 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
2595 _mm512_mask_mullox_epi64): New intrinsics.
2597 2018-05-03 Tom de Vries <tom@codesourcery.com>
2600 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2601 dump files): Add offload-tree.
2603 2018-05-03 Richard Biener <rguenther@suse.de>
2605 PR tree-optimization/85615
2606 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
2607 to loops not nested in BBs loop father to avoid creating multi-entry
2610 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2612 PR tree-optimization/70291
2613 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
2614 arguments. Change return type to tree. Emit libcall as a new
2615 statement rather than replacing existing one when inplace_p is true.
2616 (expand_complex_multiplication_components): New function.
2617 (expand_complex_multiplication): Expand floating-point complex
2618 multiplication using the above.
2619 (expand_complex_division): Rename inner_type parameter to type.
2620 Update expand_complex_libcall call-site.
2621 (expand_complex_operations_1): Update expand_complex_multiplication
2622 and expand_complex_division call-sites.
2624 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2627 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2628 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2629 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
2630 the highest significant bit of the shift count mask is clear. In
2631 check whether and[sq]i3 is needed verify that all significant bits
2632 of the shift count other than the highest are set.
2634 2018-05-02 Tom de Vries <tom@codesourcery.com>
2637 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
2638 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
2639 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
2640 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
2641 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
2642 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
2643 __builtin_goacc_parlevel_size.
2645 2018-05-02 Richard Biener <rguenther@suse.de>
2647 PR tree-optimization/85597
2648 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
2649 do not use split vect_get_vec_defs call but call vect_get_slp_defs
2652 2018-05-02 Tom de Vries <tom@codesourcery.com>
2655 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2656 dump files): Add ltrans-tree.
2658 2018-05-02 Tom de Vries <tom@codesourcery.com>
2661 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2662 dump files): Add wpa-ipa.
2664 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
2666 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
2667 powerpc*-*-linux*paired* target.
2668 * config/rs6000/750cl.h: Delete.
2669 * config/rs6000/paired.h: Delete.
2670 * config/rs6000/paired.md: Delete.
2671 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
2673 * config/rs6000/rs6000-builtin.def: Remove paired float support.
2674 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
2675 comment. Remove paired float support.
2676 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
2677 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
2679 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
2680 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
2682 * config/rs6000/rs6000.c: Remove paired float support.
2683 (paired_expand_vector_init, paired_expand_vector_move,
2684 paired_emit_vector_compare, paired_emit_vector_cond_expr,
2685 (paired_expand_lv_builtin, paired_expand_stv_builtin,
2686 paired_expand_builtin, paired_expand_predicate_builtin,
2687 paired_init_builtins): Delete.
2688 * config/rs6000/rs6000.h: Remove paired float support.
2689 * config/rs6000/rs6000.md: Remove paired float support.
2690 (move_from_CR_ov_bit): Delete.
2691 * config/rs6000/rs6000.opt (mpaired): Delete.
2692 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
2693 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
2695 2018-05-02 Richard Biener <rguenther@suse.de>
2698 * gimplify.c (gimplify_save_expr): When in SSA form allow
2699 SAVE_EXPRs to compute to SSA vars.
2701 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2704 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2705 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2706 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
2707 clobber operands[2], instead use a new pseudo. Formatting fixes.
2709 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
2711 PR tree-optimization/85586
2712 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
2713 exit early for statements in the same group if the accesses are
2716 2018-05-02 Tom de Vries <tom@codesourcery.com>
2719 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
2722 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
2724 PR tree-optimization/85143
2725 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
2727 2018-05-01 Tom de Vries <tom@codesourcery.com>
2730 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
2731 not found" error message.
2733 2018-05-01 Tom de Vries <tom@codesourcery.com>
2736 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
2737 * vec.c (test_ordered_remove_if): New function.
2738 (vec_c_tests): Call test_ordered_remove_if.
2739 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
2740 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
2741 * tree-vect-patterns.c (vect_pattern_recog_1): Use
2742 VEC_ORDERED_REMOVE_IF.
2744 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2746 PR tree-optimization/82665
2747 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
2748 pointer subtraction where arguments come from a memchr call.
2750 2018-05-01 Jakub Jelinek <jakub@redhat.com>
2752 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
2753 --push-state --as-needed and --pop-state instead of --as-needed and
2754 --no-as-needed if ld supports it.
2755 * configure: Regenerated.
2758 * doc/install.texi2html: Replace _002d with - and _002a with * in
2759 generated html files using sed.
2761 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2764 * gcc-rich-location.c (blank_line_before_p): New function.
2765 (use_new_line): New function.
2766 (gcc_rich_location::add_fixit_insert_formatted): New function.
2767 * gcc-rich-location.h
2768 (gcc_rich_location::add_fixit_insert_formatted): New function.
2770 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2772 * selftest.c (assert_streq): Rename "expected" and "actual" to
2773 "val1" and "val2". Extend NULL-handling to cover both inputs
2774 symmetrically, while still requiring both to be non-NULL for a pass.
2775 * selftest.h (assert_streq): Rename "expected" and "actual" to
2777 (ASSERT_EQ): Likewise.
2778 (ASSERT_EQ_AT): Likewise.
2779 (ASSERT_KNOWN_EQ): Likewise.
2780 (ASSERT_KNOWN_EQ_AT): Likewise.
2781 (ASSERT_NE): Likewise.
2782 (ASSERT_MAYBE_NE): Likewise.
2783 (ASSERT_MAYBE_NE_AT): Likewise.
2784 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
2785 the assertion to pass.
2786 (ASSERT_STREQ_AT): Likewise.
2788 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2790 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
2791 interaction with -pie.
2793 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2795 * selftest.h: Fix alphabetization of per-source-file selftest
2798 2018-04-30 Jason Merrill <jason@redhat.com>
2800 PR c++/61982 - dead stores to destroyed objects.
2801 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
2804 2018-04-30 Jason Merrill <jason@redhat.com>
2806 * tree.c (build_clobber): New.
2807 * tree.h: Declare it.
2808 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
2810 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2812 * diagnostic-show-locus.c (layout::layout): Update for
2813 location_get_source_line returning a char_span.
2814 (struct char_span): Move to input.h.
2815 (struct correction): Update for fields in char_span becoming
2817 (struct source_line): Update for location_get_source_line
2818 returning a char_span.
2819 (layout::print_line): Likewise.
2820 * edit-context.c (edited_file::print_content): Likewise.
2821 (edited_file::print_diff_hunk): Likewise.
2822 (edited_file::print_run_of_changed_lines): Likewise.
2823 (edited_file::get_num_lines): Likewise.
2824 (edited_line::edited_line): Likewise.
2825 * final.c (asm_show_source): Likewise.
2826 * input.c (location_get_source_line): Convert return type
2827 from const char * to char_span, losing the final "line_len"
2829 (dump_location_info): Update for the above.
2830 (get_substring_ranges_for_loc): Likewise. Use a char_span
2831 when handling the literal within the line.
2832 (test_reading_source_line): Update for location_get_source_line
2833 returning a char_span.
2834 * input.h (class char_span): Move here from
2835 diagnostic-show-locus.c, converting from a struct to a class.
2836 Make data members private.
2837 (char_span::operator bool): New.
2838 (char_span::length): New.
2839 (char_span::get_buffer): New.
2840 (char_span::operator[]): New.
2841 (char_span::subspan): Make const.
2842 (char_span::xstrdup): New.
2843 (location_get_source_line): Convert return type from const char *
2844 to char_span, losing the final "line_size" param.
2846 2018-04-30 Jan Hubicka <jh@suse.cz>
2848 * lto-wrapper.c (ltrans_priorities): New static var.
2849 (cmp_priority): New.
2850 (run_gcc): Read priorities and if doing parallel build order
2851 the Makefile by them.
2853 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2855 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
2857 2018-04-30 Richard Biener <rguenther@suse.de>
2859 * tree-cfg.c (verify_address): Remove base argument, add
2860 flag whether to check TREE_ADDRESSABLE and do that.
2861 (verify_expr): Remove.
2862 (verify_types_in_gimple_reference): Add pieces from verify_expr.
2863 (verify_gimple_assign_single): Likewise.
2864 (verify_gimple_switch): Likewise.
2865 (verify_expr_location_1): Dereference tp once. Add (disabled)
2866 piece from verify_expr.
2867 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
2869 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2871 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
2873 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2875 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
2876 (small_data_pattern): Likewise.
2877 (arc_rewrite_small_data): Likewise.
2878 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
2879 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
2880 (get_symbol_alignment): New function.
2881 (legitimate_small_data_address_p): Likewise.
2882 (legitimate_scaled_address): Update, call
2883 legitimate_small_data_address_p.
2884 (output_sdata): New static variable.
2885 (arc_print_operand): Update how we handle small data operands.
2886 (arc_print_operand_address): Likewise.
2887 (arc_legitimate_address_p): Update, use
2888 legitimate_small_data_address_p.
2889 (arc_rewrite_small_data_p): Remove.
2890 (arc_rewrite_small_data_1): Likewise.
2891 (arc_rewrite_small_data): Likewise.
2892 (small_data_pattern): Likewise.
2893 (compact_sda_memory_operand): Update to use
2894 legitimate_small_data_address_p and get_symbol_alignment.
2895 (prepare_move_operands): Don't rewite sdata pattern.
2896 (prepare_extend_operands): Remove.
2897 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
2899 (zero_extendqisi2): Likewise.
2900 (zero_extendhisi2): Likewise.
2901 (extendqihi2): Likewise.
2902 (extendqisi2): Likewise.
2903 (extendhisi2): Likewise.
2907 * config/arc/constraints.md (Usd): Change it to memory constraint.
2909 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2911 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
2912 as source of std instructions.
2913 * config/arc/arc.md (movsi_insn): Update pattern predicate to
2914 allow 6-bit constants as source for store instructions.
2915 (movdi_insn): Update instruction pattern to allow 6-bit constants
2916 as source for store instructions.
2918 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2920 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
2922 2018-04-30 Nathan Sidwell <nathan@acm.org>
2923 Sandra Loosemore <sandra@codesourcery.com>
2925 * dumpfile.c (dump_open): Allow '-' for stdout.
2926 * doc/invoke.texi (Developer Options): Document dump filename
2927 determination early. Document stdin/stdout selection.
2929 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
2931 Microblaze Target: PIC data text relative
2933 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
2934 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
2936 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
2937 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
2938 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
2939 New addressing mode for data-text relative position indepenedent code.
2940 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
2941 'ADDRESS_SYMBOLIC_TXT_REL'.
2942 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
2943 (microblaze_legitimate_pic_operand): Exclude function calls from
2944 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
2945 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
2947 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2948 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
2949 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
2950 for 'address + offset'.
2951 (microblaze_expand_prologue): Add new function prologue call for
2953 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
2954 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
2955 table in case of TARGET_PIC_DATA_TEXT_REL.
2956 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
2957 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
2958 Add new macros 'UNSPEC_TEXT',
2959 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
2960 + exclude function calls from 'UNSPEC_PLT' in case of data text
2962 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2963 new target hook for generating address diff vector tables in case of
2965 * doc/tm.texi : Regenerate.
2966 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
2967 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
2968 of addr diff vector generation.
2969 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
2970 target hook definition.
2971 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
2972 Add default function for generate_pic_addr_diff_vec -> flag_pic.
2973 * doc/invoke.texi (Add new pic option): Add new microblaze pic
2974 option for data text relative.
2976 2018-04-30 Richard Biener <rguenther@suse.de>
2978 * tree-chrec.h (evolution_function_is_constant_p): Remove
2980 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
2982 2018-04-30 Richard Biener <rguenther@suse.de>
2985 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
2987 2018-04-30 Richard Biener <rguenther@suse.de>
2989 PR tree-optimization/28364
2990 PR tree-optimization/85275
2991 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
2992 copying first exit test.
2994 2018-04-28 Mark Wielaard <mark@klomp.org>
2996 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
2998 (dwarf_AT): Handle DW_AT_addr_base.
2999 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
3001 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
3004 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
3005 (*ashl<dwi>3_doubleword_mask_1): Ditto.
3006 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
3007 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
3009 2018-04-28 Richard Biener <rguenther@suse.de>
3011 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
3012 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
3013 to reflect use. Only add interesting stmts.
3015 2018-04-27 Martin Jambor <mjambor@suse.cz>
3018 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
3019 the jump function allows for passing through aggregate values.
3021 2018-04-27 David Malcolm <dmalcolm@redhat.com>
3023 * input.h (in_system_header_at): Convert from macro to inline
3025 (from_macro_expansion_at): Likewise.
3026 (from_macro_definition_at): Likewise.
3028 2018-04-27 Jeff Law <law@redhat.com>
3030 * config.gcc: Mark tile* targets as deprecated/obsolete.
3032 2018-04-27 Richard Biener <rguenther@suse.de>
3034 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
3037 2018-04-27 Richard Biener <rguenther@suse.de>
3039 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
3041 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
3043 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
3044 with Yd constraint. Set "preferred_for_speed" attribute from
3045 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
3047 (*movdi_internal): Ditto.
3048 (movti_interunit splitters): Remove
3049 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
3050 (movdi_interunit splitters): Ditto.
3051 * config/i386/constraints.md (Ye): Remove.
3052 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
3054 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3057 * config/aarch64/constraints.md (Usg): Limit to 31.
3060 2018-04-27 Jakub Jelinek <jakub@redhat.com>
3062 PR tree-optimization/85529
3063 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
3064 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
3065 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
3066 zero extension or masking of the MSB bit.
3067 (optimize_range_tests): Add FIRST_BB argument, pass it through
3068 to optimize_range_tests_var_bound.
3069 (maybe_optimize_range_tests, reassociate_bb): Adjust
3070 optimize_range_tests callers.
3072 2018-04-26 Richard Biener <rguenther@suse.de>
3073 Jakub Jelinek <jakub@redhat.com>
3075 * cgraph.h (symbol_table): Just declare debug method here.
3076 * symtab.c (symbol_table::debug): Define.
3078 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
3080 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
3082 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
3084 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
3085 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
3086 (*movdi_internal): Substitute Yi and Yj constraint with x
3087 and Ym and Yn constraint with y constraint. Update "isa"
3088 attribute and set "preferred_for_speed" attribute from
3089 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
3090 (*movsi_internal): Ditto.
3091 (*movdf_internal): Ditto.
3092 (*movsf_internal): Ditto.
3093 (*zero_extendsidi2): Ditto.
3094 * config/i386/sse.md (vec_set<mode>_0): Ditto.
3095 (sse2_loadld): Ditto.
3096 (*vec_extract<ssevecmodelower>_0): Ditto.
3097 (*vec_extractv4si_0_zext_sse4): Ditto.
3098 (vec_concatv2di): Ditto.
3099 (*vec_dup<mode>): Ditto.
3100 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
3101 * config/i386/constraints.md (Yi): Remove.
3106 2018-04-26 Nathan Sidwell <nathan@acm.org>
3108 * dumpfile.c (dump_open): New.
3109 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
3110 (dump_finish): Detect stdio/stderr by value not name.
3112 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
3114 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
3116 2018-04-26 Tom de Vries <tom@codesourcery.com>
3119 * config/nvptx/nvptx.c (verify_neutering_jumps)
3120 (verify_neutering_labels): New function
3121 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
3123 2018-04-26 Tom de Vries <tom@codesourcery.com>
3126 * config/nvptx/nvptx.c (needs_neutering_p): New function.
3127 (nvptx_single): Use needs_neutering_p to skip over insns that do not
3130 2018-04-26 Richard Biener <rguenther@suse.de>
3131 Tom de Vries <tom@codesourcery.com>
3134 * lto-streamer-out.c (output_function): Fixup loops if required to match
3135 discovery done in the reader.
3137 2018-04-26 Richard Biener <rguenther@suse.de>
3139 PR tree-optimization/85116
3140 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
3141 have a loop exit from the single latch predecessor. Remove
3142 case of header with just condition.
3143 (ch_base::copy_headers): Exclude infinite loops from any
3145 (pass_ch::execute): Record exits.
3147 2018-04-26 Richard Biener <rguenther@suse.de>
3149 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
3150 prologue cost vector and pass it to vect_get_load_cost.
3151 (vect_get_peeling_costs_all_drs): Likewise.
3152 (vect_peeling_hash_get_lowest_cost): Likewise.
3153 (vect_enhance_data_refs_alignment): Likewise.
3155 2018-04-26 Richard Biener <rguenther@suse.de>
3158 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
3159 checking of integer<->pointer conversions.
3160 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
3161 sign-/zero-extending pointer types.
3162 (expand_omp_for_static_chunk): Likewise.
3164 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
3165 Jean Lee <xiaoyur347@gmail.com>
3167 * config/mips/mips.c (mips_asan_shadow_offset): New function.
3168 (TARGET_ASAN_SHADOW_OFFSET): Define.
3169 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
3170 true for -fsanitize=address.
3172 2018-04-25 Mark Wielaard <mark@klomp.org>
3174 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
3177 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3179 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
3180 than "alu", remove explicit "memory" and "imm_disp" attributes.
3181 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
3184 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
3185 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
3186 gen_lowpart_no_emit.
3188 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
3191 * config/i386/i386.c (ix86_expand_builtin): Change memory
3192 operand to XI, extend p0 to Pmode.
3193 * config/i386/i386.md: Change unspec volatile and operand
3194 1 mode to XI, change operand 0 mode to P.
3196 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3198 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
3199 GET_MODE_MASK before any checking.
3200 (nds32_can_use_bset_p): Likewise.
3201 (nds32_can_use_btgl_p): Likewise.
3203 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3205 * config/nds32/nds32-doubleword.md: New define_split pattern for
3206 illegal register number.
3208 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3210 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
3212 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3214 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
3216 2018-04-25 Richard Biener <rguenther@suse.de>
3218 * lto-streamer.h (LTO_major_version): Bump to 8.
3220 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3222 * BASE-VER: Set to 9.0.0.
3224 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
3226 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
3227 in __abskf2 and __powikf2.
3229 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3232 * config/aarch64/constraints.md (Usg, Usj): New constraints.
3233 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
3234 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
3235 Use the above on operand 2. Reindent.
3236 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
3238 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3241 * common/config/i386/i386-common.c (ix86_handle_option): Don't
3243 * config/i386/i386.opt (mcet): Removed.
3244 * doc/install.texi: Remove -mcet documentation.
3245 * doc/invoke.texi: Likewise.
3247 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3250 * doc/install.texi: Remove -mcet from bootstrap-cet.
3252 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3255 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
3256 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
3260 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
3261 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
3262 containing a CONST_VECTOR.
3264 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
3266 * doc/install.texi: Update newlib dependency for nvptx.
3268 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3271 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
3272 instead of INTVAL when shifting x left.
3274 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
3276 PR tree-optimization/85478
3277 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
3278 vect_grouped_store_supported for single element vectors.
3280 2018-04-24 Richard Biener <rguenther@suse.de>
3283 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
3284 load cost increase to the case of non-constant step.
3286 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3289 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
3290 destination if any_malformed_asm.
3292 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
3295 * expr.c (store_field): In the bitfield case, if the value comes from
3296 a function call and is returned in registers by means of a PARALLEL,
3297 do not change the mode of the temporary unless BLKmode and VOIDmode.
3299 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
3301 PR rtl-optimization/85423
3302 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
3303 dependencies to debug insns when the previous insn is non-debug.
3305 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3307 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
3308 enums into a single definition.
3309 (fls): Fix predicates and printing.
3312 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3314 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
3315 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
3316 and short u6 immediate.
3317 (check_if_valid_sleep_operand): Remove.
3318 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
3320 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3322 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
3323 flag_always_save_lp condition.
3324 * config/nds32/nds32.opt (malways-save-lp): New option.
3326 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3328 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
3329 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
3330 * config/nds32/nds32.h
3331 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
3332 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
3334 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3336 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
3337 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
3339 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3340 Chung-Ju Wu <jasonwucj@gmail.com>
3342 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
3344 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
3345 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
3347 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3349 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
3351 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3353 * config/nds32/nds32-protos.h (nds32_data_alignment,
3354 nds32_local_alignment): Declare.
3355 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
3356 nds32_local_alignment): New functions.
3357 (TARGET_CONSTANT_ALIGNMENT): Define.
3358 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
3360 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3362 * config/nds32/nds32.c
3363 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
3364 (TARGET_MODES_TIEABLE_P): Likewise.
3366 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3368 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
3371 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
3372 Chung-Ju Wu <jasonwucj@gmail.com>
3374 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3375 for unaligned access.
3376 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3378 * config/nds32/nds32-intrinsic.md: Likewise.
3379 * config/nds32/nds32_intrinsic.h: Likewise.
3380 * config/nds32/nds32.h (nds32_builtins): Likewise.
3381 * config/nds32/nds32.opt (munaligned-access): New option.
3382 * config/nds32/nds32.c (nds32_asm_file_start): Display
3383 flag_unaligned_access status.
3385 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3387 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
3388 -mno-relax is present.
3389 * config/riscv/linux.h (LINK_SPEC): Ditto.
3391 2018-04-20 Martin Sebor <msebor@redhat.com>
3394 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
3396 (gimple_fold_builtin_stxcpy_chk): Same.
3397 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
3399 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
3402 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
3403 __powikf2 when long double is IEEE 128-bit.
3405 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3407 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
3408 step to make sure stack always aligned.
3410 2018-04-20 Carl Love <cel@us.ibm.com>
3413 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
3414 size check for arg0.
3416 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
3417 Tom de Vries <tom@codesourcery.com>
3420 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
3421 Emit insns for calls too.
3422 (nvptx_find_par): Always look for worker-level predecessor insn.
3423 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
3425 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
3426 (nvptx_process_pars): Propagate frames for calls.
3428 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
3431 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
3433 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
3434 (ix86_handle_option): Don't handle OPT_mibt.
3435 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
3437 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
3439 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
3440 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
3441 (ix86_target_macros): Define __CET__ with flag_cf_protection
3442 for -fcf-protection.
3443 * config/i386/i386.c (isa2_opts): Remove -mibt.
3444 * config/i386/i386.h (TARGET_IBT): Removed.
3445 (TARGET_IBT_P): Likewise.
3446 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
3447 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
3448 * config/i386/i386.opt (mcet): Update help message.
3451 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
3452 -mcet as an alias for -mshstk.
3454 2018-04-20 Richard Biener <rguenther@suse.de>
3457 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
3458 complexity by forcing a single use of the multiply operand.
3460 2018-04-20 Martin Jambor <mjambor@suse.cz>
3463 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
3464 recursion dependency to only apply to non-clones.
3466 2018-04-20 Martin Jambor <mjambor@suse.cz>
3469 * ipa-cp.c (create_specialized_node): Check that clones of
3470 self-recursive edges exist during IPA-CP.
3472 2018-04-19 Toon Moene <toon@moene.org>
3474 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
3477 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3479 PR tree-optimization/85467
3480 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
3481 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
3482 VECTOR_CST element to type.
3484 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3487 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
3488 * config/i386/i386.md (builtin_setjmp_setup): Removed.
3489 (builtin_longjmp): Likewise.
3490 (save_stack_nonlocal): New pattern.
3491 (restore_stack_nonlocal): Likewise.
3493 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3496 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3497 Replace ASM_OUTPUT_LABEL with fprintf.
3499 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3502 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3503 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
3504 * config/i386/i386-c.c (ix86_target_macros_internal): Also
3505 define __IBT__ and __SHSTK__ for -fcf-protection.
3506 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
3508 (ix86_trampoline_init): Likewise.
3509 (x86_output_mi_thunk): Likewise.
3510 (ix86_notrack_prefixed_insn_p): Likewise.
3511 (ix86_option_override_internal): Don't disallow -fcf-protection.
3512 * config/i386/i386.md (rdssp<mode>): Also enable for
3514 (incssp<mode>): Likewise.
3515 (nop_endbr): Likewise.
3516 * config/i386/i386.opt (mcet): Change help message to built-in
3520 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
3521 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
3522 enable CET built-in functions.
3524 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
3526 * common/config/i386/i386-common.c
3527 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
3528 OPTION_MASK_ISA_MOVDIRI_UNSET,
3529 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
3530 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
3531 * config.gcc (movdirintrin.h): New header.
3532 * config/i386/cpuid.h (bit_MOVDIRI,
3533 bit_MOVDIR64B): New bits.
3534 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
3536 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
3537 (VOID, PVOID, PCVOID)): New function types.
3538 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
3539 __builtin_ia32_directstoreu_u64,
3540 __builtin_ia32_movdir64b): New builtins.
3541 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
3542 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
3544 (ix86_valid_target_attribute_inner_p): Ditto.
3545 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
3546 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
3547 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
3548 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
3549 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
3550 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
3551 (movdiri<mode>, movdir64b_<mode>): New.
3552 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
3553 * config/i386/immintrin.h: Include movdirintrin.h.
3554 * config/i386/movdirintrin.h: New file.
3555 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
3557 2018-04-19 Richard Biener <rguenther@suse.de>
3560 * cfg.c (clear_bb_flags): When loop state says we have
3561 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
3563 2018-04-19 Richard Biener <rguenther@suse.de>
3565 PR tree-optimization/84737
3566 * tree-vect-data-refs.c (vect_copy_ref_info): New function
3567 copying restrict info.
3568 (vect_setup_realignment): Use it.
3569 * tree-vectorizer.h (vect_copy_ref_info): Declare.
3570 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
3571 the first DR to all generated stores.
3572 (vectorizable_load): Likewise for loads.
3574 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3576 PR tree-optimization/85446
3577 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
3578 the integral and pointer types to have the same precision.
3580 * doc/install.texi: Document --disable-cet being the default and
3583 2018-04-18 Martin Liska <mliska@suse.cz>
3585 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
3588 2018-04-18 Martin Liska <mliska@suse.cz>
3591 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3594 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3595 arguments if they are comparable.
3597 2018-04-18 Martin Liska <mliska@suse.cz>
3600 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3603 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3606 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
3609 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
3610 ENDBR after calling __morestack.
3612 2018-04-18 David Malcolm <dmalcolm@redhat.com>
3615 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
3616 by using gcc_base_ver to generate a gcc_driver_version, and use
3617 it when generating GCC_DRIVER_NAME.
3618 * configure: Regenerate.
3620 2018-04-18 Jakub Jelinek <jakub@redhat.com>
3623 * config.gcc: Obsolete powerpc*-*-*spe*.
3625 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3628 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
3629 (stabstr_D): Change type of unum from unsigned int to
3630 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
3633 2018-04-17 Jim Wilson <jimw@sifive.com>
3636 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
3637 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
3638 Set arg_pointer_offset after using pretend_args_size.
3640 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3642 PR rtl-optimization/85431
3643 * dse.c (record_store): Ignore zero width stores.
3646 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
3647 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
3648 __builtin_stack_restore rather than after it.
3649 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
3650 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
3651 argument instead of virtual_dynamic_stack_rtx.
3653 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
3655 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
3657 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
3658 Add note to error message to explain internal mapping of overloaded
3659 built-in function name to non-overloaded built-in function name.
3660 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
3663 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3666 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
3667 where the inputs overlap with the output.
3669 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3671 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
3672 (=v, v) alternative and explicit "memory" attribute.
3673 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
3674 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3676 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
3677 "sselog1" type instead of "sselog".
3678 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
3679 "sselog". Remove explicit "memory" attribute.
3680 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
3681 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3683 (vec_extract_hi_v32hi): Merge all alternatives into one, use
3684 "sselog1" type instead of "sselog". Remove explicit "memory"
3686 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
3687 use "sselog1" type instead of "sselog". Remove explicit "memory"
3689 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
3690 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3692 (vec_extract_hi_v64qi): Merge all alternatives into one, use
3693 "sselog1" type instead of "sselog". Remove explicit "memory"
3695 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
3696 use "sselog1" type instead of "sselog". Remove explicit "memory"
3700 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
3703 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
3706 2018-04-17 Martin Jambor <mjambor@suse.cz>
3709 * ipa-cp.c (create_specialized_node): Call
3710 expand_all_artificial_thunks if necessary.
3712 2018-04-17 Martin Liska <mliska@suse.cz>
3715 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
3716 in message, remote space in between '_G' and '('.
3718 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3721 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
3722 avx512f_vmcmp<mode>3<round_saeonly_name>,
3723 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
3724 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
3725 avx512f_rndscale<mode><round_saeonly_name>,
3726 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
3727 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
3728 Use %<iptr>2 instead of %2 for -masm=intel.
3729 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
3730 avx512f_vcvttss2usi<round_saeonly_name>,
3731 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
3733 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
3734 avx512f_vcvttsd2usi<round_saeonly_name>,
3735 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
3736 Use %q1 instead of %1 for -masm=intel.
3737 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
3738 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
3739 of %3 for -masm=intel.
3740 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
3742 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
3744 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
3746 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
3747 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
3749 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
3751 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
3752 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
3753 %g1 and one with %0 and %1.
3754 (avx512er_vmrcp28<mode><round_saeonly_name>,
3755 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
3757 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
3758 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
3759 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
3760 of %0 and %{%4%} for -masm=intel.
3761 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
3762 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
3763 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
3764 order of %0 and %{%5%}%{z%} for -masm=intel.
3766 2018-04-17 Jan Hubicka <jh@suse.cz>
3769 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
3771 2018-04-17 Martin Liska <mliska@suse.cz>
3774 * multiple_target.c (create_dispatcher_calls): Set apostrophes
3775 for target_clone error message. Make default implementation
3776 clone to be a local declaration.
3777 (separate_attrs): Add new argument and check for an empty
3779 (expand_target_clones): Handle it.
3780 (ipa_target_clone): Make redirection just for target_clones
3783 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
3784 Tom de Vries <tom@codesourcery.com>
3787 * omp-expand.c (expand_oacc_for): Add dummy false branch for
3788 tiled basic blocks without omp continue statements.
3790 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
3793 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
3794 vec_extract expression as having side effects to make sure it gets
3797 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
3800 * config/i386/i386.c (get_builtin_code_for_version): Check
3803 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
3806 * gcc/config.gcc: Support "skylake".
3807 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
3809 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
3810 (processor_target_table): Add "skylake".
3811 (ix86_option_override_internal): Add "skylake".
3812 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
3813 PROCESSOR_CANNONLAKE.
3814 (get_builtin_code_for_version): Fix priority for
3815 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
3816 PROCESSOR_SKYLAKE-AVX512.
3817 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
3818 (processor_type): Add PROCESSOR_SKYLAKE.
3820 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
3821 Jason Merrill <jason@redhat.com>
3824 * convert.c (convert_to_integer_1): Use direct recursion for
3825 enumeral types and types with a precision less than the number
3826 of bits in their mode.
3828 2018-04-16 Julia Koval <julia.koval@intel.com>
3831 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
3832 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
3834 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
3837 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
3838 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
3839 and -mno-direct-move.
3841 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
3844 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
3845 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
3846 Ensure negative shifts result in {0}.
3848 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
3850 PR rtl-optimization/79916
3851 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
3852 regs (if any) to define how to gnerate SD moves when LRA is in
3855 2018-04-13 Jakub Jelinek <jakub@redhat.com>
3857 PR rtl-optimization/85393
3858 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
3859 * except.c (expand_dw2_landing_pad_for_region): Make static.
3860 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
3861 a label and unconditional jump to old_bb, rather than
3862 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
3865 PR rtl-optimization/85376
3866 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
3867 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
3868 instead of a specific value.
3870 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3871 Bin Cheng <bin.cheng@arm.com>
3873 PR tree-optimization/82965
3874 PR tree-optimization/83991
3875 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
3876 by_profile_only parameter.
3877 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
3878 information if the loop was predicted to iterate too many times.
3879 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
3881 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3884 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
3887 2018-04-13 Martin Liska <mliska@suse.cz>
3888 Jakub Jelinek <jakub@redhat.com>
3891 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
3892 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
3893 * builtins.c (expand_builtin_memory_copy_args): Use
3894 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
3895 handle dest_addr == pc_rtx.
3897 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
3900 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
3901 asked to not generate direct moves.
3902 (fix_trunc<mode>si2_stfiwx): Similar.
3903 (fix_trunc<mode>si2_internal): Similar.
3905 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3908 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
3909 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
3910 lookup if dest in some wider mode is known to be const0_rtx and
3911 if so, record permanent equivalence for it to be ZERO_EXTEND of
3912 the narrower mode destination.
3914 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3916 * lto-streamer-out.c (output_function): Revert 259346.
3917 * omp-expand.c (expand_oacc_for): Likewise.
3919 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
3921 PR rtl-optimization/85354
3922 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
3923 * sel-sched.c (sel_global_init): ... here.
3925 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
3928 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
3929 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
3930 mode for PE-COFF targets.
3931 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
3932 (i386_pe_asm_lto_end): Likewise.
3933 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
3934 (TARGET_ASM_LTO_END): Likewise.
3935 * config/i386/winnt.c (saved_debug_info_level): New static variable.
3936 (i386_pe_asm_lto_start): New function.
3937 (i386_pe_asm_lto_end): Likewise.
3939 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3940 Richard Biener <rguenther@suse.de>
3943 * lto-streamer-out.c (output_function): Fix CFG loop state before
3945 * omp-expand.c (expand_oacc_for): Handle calls to internal
3946 functions like regular functions.
3948 2018-04-12 Richard Biener <rguenther@suse.de>
3951 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
3952 for the early LTO debug to properly generate references to it
3953 during DIE emission. Do not re-use that for the skeleton for
3955 (dwarf2out_early_finish): Likewise.
3957 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3960 * config/i386/sse.md
3961 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
3962 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
3963 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
3964 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
3965 and output is a reg, avoid creating invalid lowpart subreg, but
3966 instead split into a 512-bit move. Don't split if not AVX512VL,
3967 input is xmm16+ reg and output is a mem.
3968 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
3969 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
3970 xmm16+ reg and output is a mem.
3972 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3974 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
3975 also for flag_dwarf2_cfi_asm.
3977 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3979 PR rtl-optimization/85342
3980 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
3981 a bool scalar var inside of the loop instead. Don't try to update
3982 recog_data.operand after failed apply_change_group.
3984 2018-04-12 Tom de Vries <tom@codesourcery.com>
3987 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
3988 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
3989 array with flexible array member as array without given dimension.
3990 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
3991 argument for undefined param to true.
3993 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
3996 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
3997 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
3998 from PowerPC section.
3999 * config/rs6000/sysv4.opt (mcall-): Improve help text.
4000 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
4001 help text that is too long.
4002 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
4003 help text that is too long.
4004 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
4005 help text that is too long.
4007 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
4009 * config/alpha/alpha.md (stack_probe_internal): Rename
4010 from "probe_stack". Update all callers.
4012 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4014 PR rtl-optimization/84566
4015 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
4016 sched_macro_fuse_insns.
4018 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4021 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
4022 (compute_block_dependences): ... from here.
4024 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4026 PR tree-optimization/85331
4027 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
4028 from int to HOST_WIDE_INT.
4030 2018-04-11 Martin Jambor <mjambor@suse.cz>
4033 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
4034 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
4035 not the same as the source val.
4036 (cgraph_edge_brings_value_p): New parameter.
4037 (gather_edges_for_value): Pass destination value to
4038 cgraph_edge_brings_value_p.
4039 (perhaps_add_new_callers): Likewise.
4040 (get_info_about_necessary_edges): Likewise and exclude values brought
4041 only by self-recursive edges.
4042 (create_specialized_node): Redirect only clones of self-calling edges.
4043 (+self_recursive_pass_through_p): New function.
4044 (find_more_scalar_values_for_callers_subset): Use it.
4045 (find_aggregate_values_for_callers_subset): Likewise.
4046 (known_aggs_to_agg_replacement_list): Removed.
4047 (decide_whether_version_node): Re-calculate known constants for all
4048 remaining context clones.
4050 2018-04-11 Richard Biener <rguenther@suse.de>
4053 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
4054 from early DWARF output.
4055 (dwarf2out_early_finish): Output line info unconditionally into
4056 early DWARF and add reference to it.
4058 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4061 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
4062 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
4063 other than V2DFmode using iptr mode attribute.
4064 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
4066 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4068 PR rtl-optimization/84659
4069 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
4071 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4074 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
4076 (output_loc_list): Pass address of a dummy size variable even in the
4077 locview handling loop.
4078 (index_location_lists): Add comment on why skip_loc_list_entry can't
4081 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
4084 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
4087 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
4090 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4091 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
4092 and -mstring-compare-inline-limit.
4094 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4097 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
4098 for stack clash protection in a register whenever we need it to be in
4101 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4103 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
4104 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
4106 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4109 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
4111 (mlong-double-): Ditto.
4112 * config/rs6000/sysv4.opt (msdata=): Ditto.
4113 (mtls-size=): Ditto.
4115 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4117 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4118 erroneous entries for
4119 "vector int vec_ldl (int, long int *)", and
4120 "vector unsigned int vec_ldl (int, unsigned long int *)".
4121 Add comments and entries for
4122 "vector bool char vec_ldl (int, bool char *)",
4123 "vector bool short vec_ldl (int, bool short *)",
4124 "vector bool int vec_ldl (int, bool int *)",
4125 "vector bool long long vec_ldl (int, bool long long *)",
4126 "vector pixel vec_ldl (int, pixel *)",
4127 "vector long long vec_ldl (int, long long *)",
4128 "vector unsigned long long vec_ldl (int, unsigned long long *)".
4129 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
4130 type tree bool_long_long_type_node and correct definition of
4131 bool_V2DI_type_node to make reference to this new type tree.
4132 (rs6000_mangle_type): Replace erroneous reference to
4133 bool_long_type_node with bool_long_long_type_node.
4134 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
4135 comments to emphasize sign distinctions for char and int types and
4136 replace RS6000_BTI_bool_long constant with
4137 RS6000_BTI_bool_long_long constant. Also add comment to restrict
4138 use of RS6000_BTI_pixel.
4139 (bool_long_type_node): Remove this macro definition.
4140 (bool_long_long_type_node): New macro definition
4142 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4144 PR rtl-optimization/85300
4145 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
4146 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
4147 simplify_unary_operation fails.
4149 2018-04-10 Martin Liska <mliska@suse.cz>
4151 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
4152 cgraph_edge and ipa_ref.
4154 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4158 * config/i386/sse.md
4159 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
4160 computation of the VEC_MERGE selector from mask.
4161 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
4162 Fix decoding of the VEC_MERGE selector into mask.
4164 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4166 PR tree-optimization/85286
4167 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
4169 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4171 * final.c (final_1): Set insn_last_address as well as
4172 insn_current_address.
4174 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4177 * explow.c (emit_stack_probe): Call validize_mem on memory location
4178 before passing it to gen_probe_stack. Create address operand and
4179 legitimize it for the probe_stack_address case.
4181 2018-04-09 Jan Hubicka <jh@suse.cz>
4184 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
4185 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
4186 * tree.c (free_lang_data_in_type): Fix handling of binfos;
4188 (free_lang_data): Rebuild type inheritance graph.
4190 2018-04-09 Martin Sebor <msebor@redhat.com>
4192 * invoke.texi (-finline-small-functions): Mention other optimization
4194 (-findirect-inlining, -fpartial-inlining): Same.
4195 (-finline-functions-called-once): Same.
4196 (-freorder-blocks-and-partition): Same.
4198 2018-04-09 Jan Hubicka <jh@suse.cz>
4201 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
4202 jumps; choose last target that matches the criteria (i.e.
4203 no partition changes for non-crossing jumps).
4204 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
4205 support for redirecting crossing jumps to non-crossing.
4207 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
4209 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
4210 also for naked functions.
4212 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
4214 * config/arc/arc.md (add_shift): New pattern.
4215 (add_shift2): Likewise.
4216 (sub_shift): Likewise.
4217 (sub_shift_cmp0_noout): Likewise.
4218 (compare_si_ashiftsi): Likewise.
4219 (xbfu_cmp0_noout): New combine pattern.
4220 (xbfu_cmp0"): Likewise.
4221 (movsi_set_cc_insn): Place the predicable variant first.
4222 (commutative_binary_cmp0_noout): Remove clobber.
4223 (commutative_binary_cmp0): New pattern.
4224 (noncommutative_binary_cmp0): Likewise.
4225 (noncommutative_binary_cmp0_noout): Likewise.
4226 (noncommutative_binary_comparison_result_used): Removed.
4227 (rsub_cmp0): New pattern.
4228 (rsub_cmp0_noout): Likewise.
4229 (extzvsi): Changed, keep only meaningful variants.
4230 (SQH, SEZ): New iterators.
4231 (SQH_postfix): New mode attribute.
4232 (SEZ_prefix): New code attribute.
4233 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
4234 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
4235 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
4237 (noncommutative_operator): Check the availability of barrel
4240 2018-04-09 Richard Biener <rguenther@suse.de>
4242 PR tree-optimization/85284
4243 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
4244 Only use the niter constraining form of simple_iv when the exit
4247 2018-04-09 Tom de Vries <tom@codesourcery.com>
4250 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4251 (define_expand "*memory_barrier"): New define_expand.
4252 (define_insn "memory_barrier"): New insn.
4254 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4256 PR rtl-optimization/80463
4257 PR rtl-optimization/83972
4258 PR rtl-optimization/83480
4260 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
4261 correct producer for the insn.
4262 (tidy_control_flow): Fixup seqnos in case of debug insns.
4264 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4266 PR rtl-optimization/83913
4268 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
4269 different sched-times when merging exprs.
4271 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4273 PR rtl-optimization/83962
4275 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
4276 tidy_fallthru_edge and tidy_control_flow.
4278 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4280 PR rtl-optimization/83530
4282 * sel-sched.c (force_next_insn): New global variable.
4283 (remove_insn_for_debug): When force_next_insn is true, also leave only
4284 next insn in the ready list.
4285 (sel_sched_region): When the region wasn't scheduled, make another pass
4286 over it with force_next_insn set to 1.
4288 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4290 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
4292 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4293 for interrupt control.
4294 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4295 functions for interrupt control.
4296 * config/nds32/nds32-intrinsic.md: Likewise.
4297 * config/nds32/nds32_intrinsic.h: Likewise.
4298 * config/nds32/nds32.h (nds32_builtins): Likewise.
4300 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4302 * config/nds32/nds32.c (nds32_init_machine_status,
4303 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
4304 strict_aligned_p field.
4305 (nds32_expand_to_rtl_hook): New function.
4306 (TARGET_EXPAND_TO_RTL_HOOK): Define.
4307 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
4309 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4310 Chung-Ju Wu <jasonwucj@gmail.com>
4312 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
4313 * config/nds32/nds32-n7.md: New file.
4314 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
4315 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
4317 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
4318 * config/nds32/nds32.md (pipeline_model): Add n7.
4319 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
4320 * config/nds32/pipelines.md: Include n7 settings.
4322 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4323 Chung-Ju Wu <jasonwucj@gmail.com>
4325 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
4326 * config/nds32/nds32-e8.md: New file.
4327 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
4328 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
4330 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
4331 * config/nds32/nds32.md (pipeline_model): Add e8.
4332 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
4333 * config/nds32/pipelines.md: Include e8 settings.
4335 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4336 Chung-Ju Wu <jasonwucj@gmail.com>
4338 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
4339 * config/nds32/nds32-n8.md: New file.
4340 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
4341 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
4343 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
4344 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
4345 * config/nds32/nds32.md (pipeline_model): Add n8.
4346 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
4347 * config/nds32/pipelines.md: Include n8 settings.
4349 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4350 Chung-Ju Wu <jasonwucj@gmail.com>
4352 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
4353 * config/nds32/nds32-n9-2r1w.md: New file.
4354 * config/nds32/nds32-n9-3r2w.md: New file.
4355 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
4356 nds32_register_ports): New or modify for cpu n9.
4357 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
4359 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
4360 * config/nds32/nds32-utils.c: New file.
4361 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
4362 TARGET_MUL_SLOW): Define.
4363 * config/nds32/nds32.md (pipeline_model): New attribute.
4364 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
4365 New options that support cpu n9.
4366 * config/nds32/pipelines.md: Include n9 settings.
4367 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
4369 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4371 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
4372 information if necessary.
4373 (output_cond_branch_compare_zero): Likewise.
4374 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
4375 (nds32_target_alignment): Refine for alignment.
4376 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
4377 (FUNCTION_BOUNDARY): Modify.
4378 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
4380 * config/nds32/nds32.opt (malways-align, malign-functions): New.
4382 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4384 * config/nds32/constants.md (unspec_volatile_element): Add values for
4385 TLB operation and data prefetch.
4386 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4387 functions for TLB operation and data prefetch.
4388 * config/nds32/nds32-intrinsic.md: Likewise.
4389 * config/nds32/nds32_intrinsic.h: Likewise.
4390 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
4391 (nds32_print_operand): Likewise.
4392 * config/nds32/nds32.h (nds32_builtins): Likewise.
4394 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
4395 Andrew Pinski <pinsika@gcc.gnu.org>
4398 * match.pd: Use constant_boolean_node of correct type instead of
4399 boolean_true_node or boolean_false_node for simplifying
4400 pointer comparisons to zero.
4402 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4404 PR tree-optimization/80021
4405 * tree.c (verify_type_variant): Make error call in verify_variant_match
4406 translatable and remove final full stop.
4408 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4410 * config/nds32/constants.md (unspec_volatile_element): Add
4411 UNSPEC_VOLATILE_EH_RETURN.
4412 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
4413 nds32_output_stack_pop): Support dwarf exception handling process.
4414 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
4415 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
4416 exception handling process.
4417 (nds32_compute_stack_frame): Likewise.
4418 (nds32_return_addr_rtx): Likewise.
4419 (nds32_initial_elimination_offset): Likewise.
4420 (nds32_expand_prologue): Likewise.
4421 (nds32_expand_epilogue): Likewise.
4422 (nds32_dynamic_chain_address): New function.
4423 * config/nds32/nds32.h (machine_function): Add fields for dwarf
4425 (DYNAMIC_CHAIN_ADDRESS): Define.
4426 (EH_RETURN_DATA_REGNO): Define.
4427 (EH_RETURN_STACKADJ_RTX): Define.
4428 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
4429 patterns for dwarf exception handling.
4431 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4433 * config/nds32/nds32.h: Clean up obsolete macros.
4435 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4437 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4438 Add enum values for particular instructions.
4439 * config/nds32/nds32-intrinsic.c: Implementation of expanding
4440 particular intrinsic functions.
4441 * config/nds32/nds32-intrinsic.md: Likewise.
4442 * config/nds32/nds32_intrinsic.h: Likewise.
4443 * config/nds32/nds32.h (nds32_builtins): Likewise.
4444 * config/nds32/nds32.md (type): Add pbsad and pbsada.
4445 (btst, ave): New patterns for particular instructions.
4447 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4449 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4450 Add enum values for atomic load/store and memory sync.
4451 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
4453 * config/nds32/nds32-intrinsic.md: Likewise.
4454 * config/nds32/nds32_intrinsic.h: Likewise.
4455 * config/nds32/nds32.h (nds32_builtins): Likewise.
4457 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4459 PR tree-optimization/85257
4460 * fold-const.c (native_encode_vector): If not all elts could fit
4461 and off is -1, return 0 rather than offset.
4462 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
4463 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
4464 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
4465 adjust buffer in native_interpret_expr call.
4467 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4469 * config/nds32/constants.md (unspec_volatile_element): Add cache
4470 control enum values.
4471 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
4472 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
4473 * config/nds32/nds32.c (nds32_cctl_names): New.
4474 (nds32_print_operand): Handle cache control register names.
4475 * config/nds32/nds32.h (nds32_builtins): New enum values.
4476 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
4478 * config/nds32/nds32.md (type): Add mmu.
4479 * config/nds32/pipelines.md (simple_insn): Add mmu.
4481 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4483 * config/nds32/nds32.md (type): Remove call.
4484 * config/nds32/pipelines.md (simple_insn): Likewise.
4486 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4488 * config/nds32/constants.md (unspec_volatile_element): Add
4489 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
4490 UNSPEC_VOLATILE_FMFCFG.
4491 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
4492 description for fmfcfg and fmfcsr.
4493 (bdesc_1arg): Add fmtcsr.
4494 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
4495 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
4496 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
4497 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
4498 unspec_fmfcfg): New patterns.
4499 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
4500 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
4501 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
4502 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
4503 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
4504 __nds32__fmfcfg): Define.
4506 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4508 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
4509 intrinsic register names.
4510 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
4511 intrinsic register enum values and macros.
4513 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4515 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
4516 for load/store addressing form.
4517 (nds32_print_operand_address): Likewise.
4519 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
4522 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
4523 based on LABEL_REF. Remove useless assertion.
4524 (pic_address_needs_scratch): Fix formatting.
4525 (sparc_legitimize_pic_address): Minor tweaks.
4526 (sparc_delegitimize_address): Adjust assertion accordingly.
4527 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
4528 into symbolic_operand.
4529 (movsi_high_pic_label_ref): Likewise.
4530 (movsi_lo_sum_pic_label_ref): Likewise.
4531 (movdi_pic_label_ref): Likewise.
4532 (movdi_high_pic_label_ref): Likewise.
4533 (movdi_lo_sum_pic_label_ref): Likewise.
4535 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
4537 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
4538 custom LIB_SPEC setup.
4540 2018-04-06 Ruslan Bukin <br@bsdpad.com>
4541 Kito Cheng <kito.cheng@gmail.com>
4543 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
4544 * config/riscv/freebsd.h: New.
4546 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4548 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
4549 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
4552 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4553 Kito Cheng <kito.cheng@gmail.com>
4555 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
4556 nds32_output_call, nds32_symbol_binds_local_p): New functions.
4557 * config/nds32/nds32-protos.h (nds32_output_call,
4558 nds32_output_return): Declare.
4559 * config/nds32/nds32.md: Refine all the call and return patterns.
4561 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4564 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
4565 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
4567 PR rtl-optimization/84872
4568 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
4569 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
4572 2018-04-06 Tamar Christina <tamar.christina@arm.com>
4574 * expr.c (copy_blkmode_to_reg): Revert 254862.
4575 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
4577 2018-04-06 Richard Biener <rguenther@suse.de>
4580 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
4581 after seeing a component reference with an adjacent field. Treat
4582 refs to arrays at struct end of external decls similar to
4583 refs to unconstrained commons.
4585 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4588 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
4589 look through SAVE_EXPRs with non-side-effects argument. Adjust
4591 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
4594 2018-04-06 Richard Biener <rguenther@suse.de>
4597 * alias.c (find_base_term): New wrapper around find_base_term
4598 unwinding CSELIB_VAL_PTR changes.
4599 (find_base_term): Do not restore CSELIB_VAL_PTR during the
4602 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4604 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
4606 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
4607 constant definitions.
4608 ("nop"): lr 0,0 -> nopr r0
4609 ("nop_lr0", "nop_lr1"): New insn definitions.
4611 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4613 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
4614 NDS32_V3PUSH_AVAILABLE_P macro.
4616 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
4617 Chung-Ju Wu <jasonwucj@gmail.com>
4619 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
4620 (nds32*-*-*): Add float and fpu_config into supported_defaults.
4621 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
4622 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
4623 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
4624 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
4625 * config/nds32/constraints.md: New constraints and checking for hard
4626 float configuration.
4627 * config/nds32/iterators.md: New mode iterator and attribute for hard
4628 float configuration.
4629 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
4631 * config/nds32/nds32-fpu.md: New file.
4632 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
4633 deal with hard float code generation.
4634 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
4636 (abi_type, float_reg_number): New enum type.
4637 * config/nds32/nds32-predicates.c: New predicates for hard float.
4638 * config/nds32/nds32-protos.h: Declare functions for hard float.
4639 * config/nds32/nds32.c: Implementation for hard float configuration.
4640 * config/nds32/nds32.h: Definitions for hard float configuration.
4641 * config/nds32/nds32.md: Include hard float machine description and
4642 modify patterns for hard float configuration.
4643 * config/nds32/nds32.opt: New options for hard float configuration.
4644 * config/nds32/predicates.md: New predicates for hard float
4647 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4649 * common/config/nds32/nds32-common.c
4650 (nds32_option_optimization_table): Enable -mreleax-hint by default.
4652 2018-04-05 Jakub Jelinek <jakub@redhat.com>
4655 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
4656 CONSTRUCTOR_ELT (ctor, ...)->value.
4658 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
4661 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
4663 2018-04-05 Tom de Vries <tom@codesourcery.com>
4666 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
4669 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4670 Kito Cheng <kito.cheng@gmail.com>
4672 * config/nds32/constraints.md (U33): Fine-tune checking condition.
4673 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
4674 * config/nds32/nds32.h (nds32_16bit_address_type): Add
4675 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
4677 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4678 Kito Cheng <kito.cheng@gmail.com>
4680 * config/nds32/constraints.md (Ufe): New memory constraint.
4681 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
4682 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
4683 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
4685 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
4686 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
4688 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4690 * config/nds32/nds32.md: Use optimize_size in the condition for
4691 alu-shift instructions.
4693 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4695 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
4697 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4699 * config/nds32/nds32.md (negsi2): Refine pattern.
4701 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4702 Chung-Ju Wu <jasonwucj@gmail.com>
4704 * config/nds32/iterators.md (shift_rotate): New code iterator.
4705 (shift): New code attribute.
4706 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
4707 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
4708 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
4709 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
4710 bit-wise operations.
4711 (andsi3, *andsi3): Ditto.
4712 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
4713 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
4714 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
4715 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
4716 nds32_ior_operand, nds32_xor_operand): New predicates.
4718 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4720 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
4721 (addsi3, subsi3): ... this.
4723 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4725 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
4727 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4729 * config/nds32/nds32.md: Adjust indention.
4731 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4733 * config/nds32/nds32.md (feature): New attribute.
4735 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4737 * config/nds32/nds32.md (subtype): New attribute.
4739 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
4742 * config/arm/arm-builtins.c (arm_expand_builtin): Change
4743 expansion to perform a bitwise AND of the argument followed by a
4744 boolean negation of the result.
4746 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
4748 PR rtl-optimization/84878
4749 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
4750 the basic block. Assert the use reference is not artificial and that
4751 it has an associated insn.
4753 2018-04-04 Michael Matz <matz@suse.de>
4755 * builtins.c (compute_objsize): Pass correct operand
4756 to array_at_struct_end_p.
4758 2018-04-04 Richard Biener <rguenther@suse.de>
4761 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
4762 from contexts for DINFO_LEVEL_TERSE and below.
4764 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4766 * config/nds32/nds32-doubleword.md (move_<mode>): Require
4767 resiter_operand condition.
4768 * config/nds32/nds32.md (*move<mode>): Ditto.
4770 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4771 Monk Chiang <sh.chiang04@gmail.com>
4773 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
4775 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4777 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
4779 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4780 Kito Cheng <kito.cheng@gmail.com>
4782 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
4783 nds32_cond_code_str, output_cond_branch,
4784 output_cond_branch_compare_zero, nds32_expand_cbranch,
4785 nds32_expand_cstore, nds32_expand_movcc,
4786 nds32_output_cbranchsi4_equality_zero,
4787 nds32_output_cbranchsi4_equality_reg,
4788 nds32_output_cbranchsi4_equality_reg_or_const_int,
4789 nds32_output_cbranchsi4_greater_less_zero: New functions.
4790 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
4791 nds32_expand_cstore, nds32_expand_movcc,
4792 nds32_output_cbranchsi4_equality_zero,
4793 nds32_output_cbranchsi4_equality_reg,
4794 nds32_output_cbranchsi4_equality_reg_or_const_int,
4795 nds32_output_cbranchsi4_greater_less_zero): Declare.
4796 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
4797 nds32_rimm11s_operand): New predicates.
4798 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
4799 * config/nds32/nds32.md: Rewrite all the branch and conditional move
4802 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4804 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
4805 * config/nds32/nds32.md: Ditto.
4806 * config/nds32/pipelines.md: Ditto.
4808 2018-04-04 Richard Biener <rguenther@suse.de>
4810 PR tree-optimization/85168
4811 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
4812 propagating abnormals.
4814 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4816 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
4818 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4819 Kito Cheng <kito.cheng@gmail.com>
4821 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
4822 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
4823 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
4824 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
4825 * config/nds32/nds32.md (sibcall_internal): New.
4826 (sibcall_register): Remove.
4827 (sibcall_immediate): Remove.
4828 (sibcall_value_internal): New.
4829 (sibcall_value_register): Remove.
4830 (sibcall_value_immediate): Remove.
4831 * config/nds32/predicates.md (nds32_general_register_operand): New.
4832 (nds32_call_address_operand): New.
4834 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4836 PR rtl-optimization/85167
4837 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
4838 bb_defs if *split_p, instead preinitialize it to NULL.
4840 PR tree-optimization/85156
4841 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
4842 evaluating the argument multiple times.
4844 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
4846 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
4848 (_mm_cvtpd_ps): Likewise.
4849 (_mm_cvttpd_epi32): Likewise.
4850 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
4851 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
4852 vector, pixel, and bool following altivec.h include.
4854 2018-04-03 Martin Sebor <msebor@redhat.com>
4856 * doc/extend.texi (Common Function Attributes): Clarify.
4857 (const attribute): Likewise.
4858 (pure attribute): Likewise.
4860 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4863 * config/i386/i386.c (ix86_expand_vector_set): Use
4864 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
4866 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
4868 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
4869 instructions when changing rounding bits to preserve precision bits
4870 in the x87 control word.
4872 2018-04-03 Martin Liska <mliska@suse.cz>
4874 PR tree-optimization/82491
4875 * rtl.h (strip_offset_and_add): Replace += suboffset with
4876 poly_uint64 () + suboffset.
4878 2018-03-29 Martin Liska <mliska@suse.cz>
4879 Martin Jambor <mjambor@suse.cz>
4882 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
4883 param_type is not an integral or pointer type.
4885 2018-04-03 Richard Biener <rguenther@suse.de>
4887 * sese.h (recompute_all_dominators): Remove.
4889 2018-04-02 Martin Sebor <msebor@redhat.com>
4891 * doc/invoke.texi (-Wrestrict): Fix typos.
4893 2018-04-02 Jim Wilson <jimw@sifive.com>
4895 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
4896 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
4897 (<optab>di3, <optab>si3_extend): Likewise.
4898 (<optab>si3_mask, <optab>si3_mask_1): New.
4899 (<optab>di3_mask, <optab>di3_mask_1): New.
4900 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
4901 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
4902 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
4904 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
4906 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
4909 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
4911 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
4912 (nds32_canonicalize_comparison): New function.
4914 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4915 Kito Cheng <kito.cheng@gmail.com>
4916 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4918 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
4919 * config/nds32/constants.md (unspec_volatile_element): Add
4920 UNSPEC_VOLATILE_RELAX_GROUP.
4921 * config/nds32/nds32-relax-opt.c: New file.
4922 * config/nds32/nds32-predicates.c
4923 (nds32_symbol_load_store_p): New function.
4924 * config/nds32/nds32-protos.h
4925 (nds32_symbol_load_store_p): Declare function.
4926 (make_pass_nds32_relax_opt): Declare new rtl pass function.
4927 * config/nds32/nds32.c
4928 (nds32_register_pass): New function to register pass.
4929 (nds32_register_passes): New function to register passes.
4930 * config/nds32/nds32.md (relax_group): New pattern.
4931 * config/nds32/nds32.opt (mrelax-hint): New option.
4932 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
4934 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
4936 * config/nds32/t-nds32: Modify files dependency.
4938 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4940 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
4941 (PROFILE_HOOK): Define its implementation.
4943 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4945 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
4946 type and 32-bit size.
4948 2018-04-01 Jakub Jelinek <jakub@redhat.com>
4951 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
4952 (V_128_256): New mode iterator.
4953 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
4954 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
4955 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
4957 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
4958 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
4960 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
4963 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
4964 NaN inputs correctly.
4966 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
4969 * config/rs6000/vsx.md (??r): New mode attribute.
4970 (*vsx_mov<mode>_64bit): Use it.
4971 (*vsx_mov<mode>_32bit): Likewise.
4973 2018-03-30 Martin Sebor <msebor@redhat.com>
4975 PR tree-optimization/84818
4976 * builtins.c (check_access): Use warning_n.
4978 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4981 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
4983 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
4986 2018-03-30 Julia Koval <julia.koval@intel.com>
4989 * x86-tune.def (movx, partial_reg_dependency): Enable for
4992 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
4995 * lra-constraints.c (process_alt_operands): Move setting
4996 this_alternative_matches below.
4998 2018-03-29 Martin Liska <mliska@suse.cz>
5001 * doc/invoke.texi: Document how LTO works with debug info.
5002 Describe auto-load support of binutils. Mention 'x86-64'
5003 as valid option value of -march option.
5005 2018-03-29 Jakub Jelinek <jakub@redhat.com>
5007 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
5010 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
5011 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
5012 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
5015 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5018 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
5019 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
5020 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
5021 for RS6000_BTM_POWERPC64.
5022 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
5023 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
5024 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
5029 2018-03-28 Carl Love <cel@us.ibm.com>
5032 2017-09-27 Carl Love <cel@us.ibm.com>
5034 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
5035 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
5036 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
5039 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5041 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
5042 instead of __vector bool.
5043 (_mm_max_pu8): Likewise.
5044 (_mm_min_pi16): Likewise.
5046 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5049 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
5050 (DIVWEUO): Likewise.
5052 (DIVDEUO): Likewise.
5053 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
5054 DIVWEUO and DIVDEUO.
5055 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
5056 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
5057 (div_extend): Likewise.
5058 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
5060 (__builtin_divweuo): Likewise.
5061 (__builtin_divdeo): Likewise.
5062 (__builtin_divdeuo): Likewise.
5064 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5067 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
5068 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
5070 PR tree-optimization/82004
5071 * gimple-match-head.c (optimize_pow_to_exp): New function.
5072 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
5073 Don't fold to exp if optimize_pow_to_exp is false.
5075 2018-03-28 Martin Liska <mliska@suse.cz>
5078 * calls.c (initialize_argument_information): Fix trailing space.
5079 * common.opt: Fix typo and provide better explanation for
5080 -fsanitize-coverage option.
5081 * config/i386/i386.opt: Fix typo.
5083 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5084 Martin Liska <mliska@suse.cz>
5087 * gimplify.c (asan_poison_variable): Don't do the check for
5088 gimplify_omp_ctxp here.
5089 (gimplify_decl_expr): Do it here.
5090 (gimplify_target_expr): Likewise.
5092 2018-03-28 Martin Liska <mliska@suse.cz>
5095 * config/i386/i386.c (ix86_function_arg_advance): Do not call
5096 chkp_type_bounds_count if MPX is not enabled.
5098 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
5100 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
5102 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
5105 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
5106 function to create the function decl for complex long double
5107 multiply and divide for -mabi=ieeelongdouble.
5108 (init_float128_ieee): Call it.
5110 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
5113 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
5114 -fcf-protection=branch -mibt.
5115 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
5117 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5120 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
5122 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
5125 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
5126 extern array declarations.
5128 2018-03-27 Richard Biener <rguenther@suse.de>
5131 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
5132 explicit single_use checks.
5134 2018-03-27 Richard Biener <rguenther@suse.de>
5136 PR tree-optimization/85082
5137 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
5140 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5142 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
5143 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
5144 Turn on fasynchronous-unwind-tables and funwind-tables.
5146 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
5149 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
5150 (*bmi_blsr_<mode>_ccz): Ditto.
5152 2018-03-26 Tom de Vries <tom@codesourcery.com>
5154 PR tree-optimization/85063
5155 * omp-general.c (offloading_function_p): New function. Factor out
5157 * omp-offload.c (pass_omp_target_link::gate): ... here.
5158 * omp-general.h (offloading_function_p): Declare.
5159 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
5160 with attribute omp declare target for offloading functions.
5162 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
5164 PR tree-optimization/84005
5165 * tree-data-ref.h (get_base_for_alignment): Declare.
5166 * tree-data-ref.c (get_base_for_alignment_1): New function.
5167 (get_base_for_alignment): Likewise.
5168 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
5169 get_base_for_alignment to find a suitable base object, instead
5170 of always using drb->base_address.
5172 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5175 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
5176 known size by default.
5178 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
5181 * lra-constraints.c (process_alt_operands): Don't match BLKmode
5182 and non BLKmode operands.
5184 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5187 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
5188 Clean up attributes.
5190 2018-03-23 Richard Biener <rguenther@suse.de>
5193 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
5194 we are going to emit early debug for LTO.
5196 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5199 * function.c (match_asm_constraints_1): Don't optimize if input
5200 doesn't satisfy general_operand predicate for output's mode.
5203 * alias.c (write_dependence_p): Don't require for x_canonicalized
5204 non-VOIDmode if x has VOIDmode.
5207 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
5208 just don't try to optimize it rather than assert it never happens.
5210 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5212 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
5213 macro expansions for definition of ST_INTERNAL_<mode> and
5214 LD_INTERNAL_<mode> builtins.
5215 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
5217 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
5219 (altivec_expand_st_builtin): Likewise.
5220 (altivec_expand_builtin): Remove calls to deleted functions.
5221 (rs6000_address_for_altivec): Delete this function.
5222 * config/rs6000/vector.md: Remove expands for
5223 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
5225 2018-03-22 Sudakshina Das <sudi.das@arm.com>
5228 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
5229 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
5230 re-computing once computed.
5231 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
5232 (arm_init_machine_status): Initialize
5233 machine->static_chain_stack_bytes.
5235 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5238 * doc/extend.texi: Add four new prototypes for vec_ld.
5239 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
5240 definitions for more logical presentation.
5241 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
5242 entries for V1TI variants of __builtin_altivec_ld builtin.
5243 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
5244 handling of V1TI variant of LVX icode pattern.
5245 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
5246 (rs6000_gimple_fold_builtin): Likewise.
5247 (altivec_init_builtins): Add code to define
5248 __builtin_altivec_lvx_v1ti function.
5250 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5253 * function.c (match_asm_constraints_1): Don't do the optimization
5254 if input isn't a REG, SUBREG, MEM or constant.
5256 2018-03-22 Tom de Vries <tom@codesourcery.com>
5258 PR tree-optimization/84956
5259 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
5260 bb_has_abnormal_pred.
5262 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5265 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
5266 DECL_INITIAL (decl) to decl at the end.
5267 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
5270 2018-03-21 Joseph Myers <joseph@codesourcery.com>
5272 * doc/extend.texi (__builtin_tgmath): Document when complex
5273 integer types are treated as _Complex _Float64.
5275 2018-03-21 Tom de Vries <tom@codesourcery.com>
5277 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
5279 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5281 PR tree-optimization/84960
5282 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
5283 if it is ENTRY block, move them into single succ of ENTRY in that case.
5285 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
5287 PR tree-optimization/84811
5288 * poly-int.h (poly_span_traits): Remove the T3 parameter and
5289 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
5290 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
5291 (known_subrange_p): Update accordingly. Cast each value involved
5292 in the size comparison, rather than casting the result of the
5295 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5297 PR tree-optimization/84982
5298 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
5299 by flipping the least significant bit rather than all bits from
5300 bitpos to bitpos + bitsize - 1.
5302 2018-03-21 Nathan Sidwell <nathan@acm.org>
5304 * doc/extend.texi (Deprecated Features): Remove mention of
5305 long-deleted deprecations.
5307 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5310 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
5311 * configure: Regenerate.
5313 2018-03-21 Tom de Vries <tom@codesourcery.com>
5315 PR tree-optimization/83126
5316 * tree-parloops.c (num_phis): New function.
5317 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
5319 2018-03-21 Nathan Sidwell <nathan@acm.org>
5321 * doc/extend.texi (Deprecated Features): Update deprecated flags,
5322 mention anon-struct/union members and trailing attributes.
5324 2018-03-21 Bin Cheng <bin.cheng@arm.com>
5326 PR tree-optimization/84969
5327 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
5328 builtin memset partitions if they set different rhs values.
5330 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5332 PR rtl-optimization/84989
5333 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
5334 VEC_DUPLICATE with scalar result mode.
5336 2018-03-21 Martin Liska <mliska@suse.cz>
5339 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
5340 not intended return statement.
5342 2018-03-21 Martin Liska <mliska@suse.cz>
5345 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
5346 (chkp_find_bound_slots_1): Limit number of iterations.
5348 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
5351 * Minor grammar fixes for x86 options.
5353 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5356 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
5357 holding REG_CFA_RESTORE notes, instead turn them into a USE.
5359 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
5362 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
5363 (altivec_lvx_<mode>_1op): Likewise.
5364 (altivec_stvx_<mode>_2op): Likewise.
5365 (altivec_stvx_<mode>_1op): Likewise.
5366 (altivec_lvx_<VM2:mode>): New define_expand.
5367 (altivec_stvx_<VM2:mode>): Likewise.
5368 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
5369 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5370 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
5371 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5372 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
5373 (rs6000_gen_lvx): Likewise.
5374 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
5375 (altivec_expand_stv_builtin): Likewise.
5376 (altivec_expand_builtin): Likewise.
5377 * config/rs6000/vector.md: Likewise.
5379 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5382 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
5385 2018-03-20 Richard Biener <rguenther@suse.de>
5388 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
5389 sign-conversions as zero, fall back to standard scalar_stmt
5392 2018-03-20 Martin Liska <mliska@suse.cz>
5395 * predict.c (rebuild_frequencies): Handle case when we have
5396 PROFILE_ABSENT, but flag_guess_branch_prob is false.
5398 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5401 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
5402 flag_section_anchors.
5403 * varasm.c (use_blocks_for_decl_p): Remove hack for
5404 dw2_force_const_mem.
5407 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
5409 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
5410 be created, use lowpart_subreg of operands[0] rather than operands[0]
5412 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
5413 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
5414 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
5415 and n constraint instead of aarch64_shift_imm_di and Usd.
5416 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
5417 (*aarch64_<optab>_reg_minus<mode>3): ... this.
5419 2018-03-20 Sudakshina Das <sudi.das@arm.com>
5422 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
5423 to favor GPR over NEON registers.
5424 (<shift>di3_neon): Likewise.
5426 2018-03-20 Tom de Vries <tom@codesourcery.com>
5429 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
5430 (nvptx_process_pars): Emit bar.sync asap and alap.
5432 2018-03-20 Tom de Vries <tom@codesourcery.com>
5435 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
5436 seen_label if seen_label is already set.
5438 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5441 * config/i386/i386.c (fold_builtin_cpu): For features above 31
5442 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
5443 Use 1U instead of 1. Formatting fixes.
5446 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
5447 instead of TREE_TYPE (s1) for the return value.
5449 2018-03-19 Jakub Jelinek <jakub@redhat.com>
5451 PR tree-optimization/84946
5452 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
5453 bitsize + bitsize in poly_uint64 rather than poly_int64.
5456 * dwarf2asm.c: Include fold-const.c.
5457 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
5458 of decl rather than decl itself.
5460 PR rtl-optimization/84643
5461 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
5463 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
5466 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
5467 calling assemble_variable.
5469 2018-03-19 Sudakshina Das <sudi.das@arm.com>
5472 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
5473 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
5475 2018-03-19 Jim Wilson <jimw@sifive.com>
5478 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
5479 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
5480 (riscv_first_stack_step): Likewise.
5481 (riscv_option_override): Use STACK_BOUNDARY instead of
5483 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
5485 (BIGGEST_ALIGNMENT): Set to 128.
5486 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
5487 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
5490 2018-03-19 Richard Biener <rguenther@suse.de>
5492 PR tree-optimization/84933
5493 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
5494 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
5496 2018-03-19 Richard Biener <rguenther@suse.de>
5498 PR tree-optimization/84859
5499 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
5500 (cond_if_else_store_replacement): Perform sinking operation on
5501 single-store BBs regardless of MAX_STORES_TO_SINK setting.
5502 Generalize what a BB with a single eligible store is.
5504 2018-03-19 Richard Biener <rguenther@suse.de>
5506 PR tree-optimization/84929
5507 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
5508 chrec_is_positive against non-chrec arg.
5510 2018-03-19 Tamar Christina <tamar.christina@arm.com>
5513 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
5515 2018-03-18 Martin Liska <mliska@suse.cz>
5517 PR rtl-optimization/84635
5518 * regrename.c (build_def_use): Use matches_mode only when
5521 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
5523 PR tree-optimization/84913
5524 * tree-vect-loop.c (vectorizable_reduction): Don't try to
5525 vectorize chains of COND_EXPRs.
5527 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5529 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
5531 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5533 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
5535 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5537 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
5539 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
5540 Kito Cheng <kito.cheng@gmail.com>
5542 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
5543 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
5544 (nds32_adjust_reg_alloc_order): New function.
5545 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
5547 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
5549 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
5550 nds32_print_operand, nds32_print_operand_address): Use
5551 HOST_WIDE_INT_PRINT_DEC instead.
5553 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
5555 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
5557 2018-03-17 Jakub Jelinek <jakub@redhat.com>
5560 * config/i386/i386.c (initial_ix86_tune_features,
5561 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
5563 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
5564 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
5565 rather than 1u << ix86_tune. Formatting fix.
5566 (ix86_option_override_internal): Change ix86_arch_mask from
5567 unsigned int to unsigned HOST_WIDE_INT, initialize to
5568 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
5569 (ix86_function_specific_restore): Likewise.
5571 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5574 * postreload.c (reload_combine_recognize_pattern): Perform
5575 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
5576 truncate_int_for_mode the result for the destination's mode.
5579 * hsa-gen.c (mem_type_for_type): Fix comment typo.
5580 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
5582 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5585 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
5588 * lra-assigns.c (lra_split_hard_reg_for): Don't use
5589 regno_allocno_class_array and sorted_pseudos.
5590 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
5591 insns where regno is used.
5593 2018-03-16 Martin Liska <mliska@suse.cz>
5596 * multiple_target.c (create_dispatcher_calls): Redirect
5597 reference in the symbol table.
5599 2018-03-16 Martin Liska <mliska@suse.cz>
5602 * multiple_target.c (create_dispatcher_calls): Redirect also
5605 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5609 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
5610 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
5611 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
5613 2018-03-16 Julia Koval <julia.koval@intel.com>
5615 * doc/invoke.texi (Skylake Server): Add CLWB.
5616 Cannonlake): Remove CLWB.
5618 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5620 PR tree-optimization/84841
5621 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
5623 (FLOAT_ONE_CONST_TYPE): Define.
5624 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
5625 (sort_by_operand_rank): Put entries with higher constant_type last
5626 rather than first to match comments.
5628 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
5630 * config/nios2/nios2.md (movsi_internal): Fix thinko in
5633 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5636 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
5637 check and use address of target always.
5639 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
5642 * config/i386/i386.c (indirect_thunk_needed): Update comments.
5643 (indirect_thunk_bnd_needed): Likewise.
5644 (indirect_thunks_used): Likewise.
5645 (indirect_thunks_bnd_used): Likewise.
5646 (indirect_return_needed): New.
5647 (indirect_return_bnd_needed): Likewise.
5648 (output_indirect_thunk_function): Add a bool argument for
5650 (output_indirect_thunk_function): Don't generate alias for
5651 function return thunk.
5652 (ix86_code_end): Call output_indirect_thunk_function to generate
5653 function return thunks.
5654 (ix86_output_function_return): Set indirect_return_bnd_needed
5655 and indirect_return_needed instead of indirect_thunk_bnd_needed
5656 and indirect_thunk_needed.
5658 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
5660 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
5661 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
5662 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
5664 2018-03-15 David Malcolm <dmalcolm@redhat.com>
5665 Paul Hua <paul.hua.gm@gmail.com>
5668 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
5670 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
5672 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
5673 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
5676 2018-03-15 Tamar Christina <tamar.christina@arm.com>
5679 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
5680 instead of GET_MODE_SIZE when comparing Units.
5682 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
5685 * varasm.c (hash_section): Return an unchangeble hash value
5686 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
5687 Return !aarch64_can_use_per_function_literal_pools_p ().
5689 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5692 * optabs.c (emit_conditional_move): Pass address of cmode's copy
5693 rather than address of cmode as last argument to prepare_cmp_insn.
5695 2018-03-15 Julia Koval <julia.koval@intel.com>
5697 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
5698 F_AVX512VNNI, F_AVX512BITALG): New.
5700 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
5703 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
5704 insn for floating-point loads and stores.
5706 2018-03-14 Carl Love <cel@us.ibm.com>
5708 * config/rs6000/rs6000-c.c: Add macro definitions for
5709 ALTIVEC_BUILTIN_VEC_PERMXOR.
5710 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
5711 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
5712 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
5713 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
5715 * config/doc/extend.texi: Add prototypes for vec_permxor.
5717 2018-03-14 David Malcolm <dmalcolm@redhat.com>
5720 * diagnostic-show-locus.c (class layout_point): Convert m_line
5721 from int to linenum_type.
5722 (line_span::comparator): Use linenum "compare" function when
5723 comparing line numbers.
5724 (test_line_span): New function.
5725 (layout_range::contains_point): Convert param "row" from int to
5727 (layout_range::intersects_line_p): Likewise.
5728 (layout::will_show_line_p): Likewise.
5729 (layout::print_source_line): Likewise.
5730 (layout::should_print_annotation_line_p): Likewise.
5731 (layout::print_annotation_line): Likewise.
5732 (layout::print_leading_fixits): Likewise.
5733 (layout::annotation_line_showed_range_p): Likewise.
5734 (struct line_corrections): Likewise for field m_row.
5735 (line_corrections::line_corrections): Likewise for param "row".
5736 (layout::print_trailing_fixits): Likewise.
5737 (layout::get_state_at_point): Likewise.
5738 (layout::get_x_bound_for_row): Likewise.
5739 (layout::print_line): Likewise.
5740 (diagnostic_show_locus): Likewise for locals "last_line" and
5742 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
5743 * input.c (selftest::test_linenum_comparisons): New function.
5744 (selftest::input_c_tests): Call it.
5745 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
5746 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
5747 * selftest.h (ASSERT_GT): New macro.
5748 (ASSERT_GT_AT): New macro.
5749 (ASSERT_LT): New macro.
5750 (ASSERT_LT_AT): New macro.
5752 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
5754 PR rtl-optimization/84780
5755 * combine.c (distribute_links): Don't make a link based on pc_rtx.
5757 2018-03-14 Martin Liska <mliska@suse.cz>
5759 * tree.c (record_node_allocation_statistics): Use
5760 get_stats_node_kind.
5761 (get_stats_node_kind): New function extracted from
5762 record_node_allocation_statistics.
5763 (free_node): Use get_stats_node_kind.
5765 2018-03-14 Richard Biener <rguenther@suse.de>
5767 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
5768 that the value-set of ANTIC_IN doesn't grow.
5771 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5773 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5774 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5776 2018-03-14 Julia Koval <julia.koval@intel.com>
5778 * config.gcc (icelake-client, icelake-server): New.
5780 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
5781 (initial_ix86_arch_features): Ditto.
5782 (PTA_SKYLAKE): Add SGX.
5783 (PTA_ICELAKE): Remove.
5784 (PTA_ICELAKE_CLIENT): New.
5785 (PTA_ICELAKE_SERVER): New.
5786 (ix86_option_override_internal): Split up icelake on icelake client and
5788 (get_builtin_code_for_version): Ditto.
5789 (fold_builtin_cpu): Ditto.
5790 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
5791 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
5792 * config/i386/i386.h (processor_type): Ditto.
5793 * doc/invoke.texi: Ditto.
5795 2018-03-14 Jakub Jelinek <jakub@redhat.com>
5798 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
5799 INTEGER_CST offset, add it together with bitpos / 8 and
5800 sign extend based on POINTER_SIZE.
5804 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
5807 * config/i386/constraints.md (Yc): New register constraint.
5808 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
5809 Use Yc constraint for alternative 2 of operand 0. Remove
5810 preferred_for_speed attribute.
5812 2018-03-14 Richard Biener <rguenther@suse.de>
5814 PR tree-optimization/84830
5815 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
5816 with the old one to avoid oscillations.
5818 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
5821 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
5823 (assign_by_spills): Return a flag of reload assignment failure.
5824 Do not process the reload assignment failures. Do not spill other
5825 reload pseudos if they has the same reg class. Update n if
5827 (lra_assign): Add a return arg. Set up from the result of
5828 assign_by_spills call.
5829 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5830 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5831 usage_insns if it is not NULL.
5832 (spill_hard_reg_in_range): New function.
5833 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5834 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5835 function prototypes.
5836 (lra_assign): Change prototype.
5837 * lra.c (lra): Add code to deal with fails by splitting hard reg
5840 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
5842 * config/riscv/riscv.opt (mrelax): New option.
5843 * config/riscv/riscv.c (riscv_file_start): Emit ".option
5844 "norelax" when riscv_mrelax is disabled.
5845 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
5847 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5850 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
5851 reassociation for int modes.
5853 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5855 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
5856 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
5858 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
5859 * config/aarch64/aarch64-sve.md
5860 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
5861 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
5862 (*extend<mode><Vwide>2): Rename to...
5863 (aarch64_sve_extend<mode><Vwide>2): ...this.
5864 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
5865 renaming the old pattern to...
5866 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
5868 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
5869 define_expand, renaming the old pattern to...
5870 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
5871 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
5872 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
5873 account when deciding which SVE instruction the optab should use.
5874 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
5876 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5878 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
5879 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
5880 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
5881 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
5882 (tlsdesc_small_<mode>): Turn a define_expand and use
5883 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
5884 (tlsdesc_small_advsimd_<mode>): ...this.
5885 (tlsdesc_small_sve_<mode>): New pattern.
5887 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5889 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
5890 (UNSPEC_UMUL_HIGHPART): New constants.
5891 (MUL_HIGHPART): New int iteraor.
5892 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
5893 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
5895 (*<su>mul<mode>3_highpart): New define_insn.
5897 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
5900 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
5903 2018-03-13 Martin Liska <mliska@suse.cz>
5906 * (sem_item_optimizer::sem_item_optimizer): Initialize new
5908 (sem_item_optimizer::~sem_item_optimizer): Release it.
5909 (sem_item_optimizer::merge_classes): Register variable aliases.
5910 (sem_item_optimizer::fixup_pt_set): New function.
5911 (sem_item_optimizer::fixup_points_to_sets): Likewise.
5912 * ipa-icf.h: Declare new variables and functions.
5914 2018-03-13 Jakub Jelinek <jakub@redhat.com>
5917 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
5918 integer_pow2p@2 and test integer_pow2p in condition.
5919 (A < 0 ? C : 0): Similarly for @1.
5922 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
5923 characters starting at p contain '\0' character, don't look beyond
5927 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
5928 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
5931 * reg-stack.c (change_stack): Change update_end var from int to
5932 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
5933 also call set_block_for_insn on the newly added insns and rescan.
5936 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
5937 on the last operand.
5940 * tree.c (stabilize_reference_1): Return save_expr (e) for
5941 STATEMENT_LIST even if it doesn't have side-effects.
5943 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
5945 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
5947 2018-03-12 Renlin Li <renlin.li@arm.com>
5949 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
5950 aarch64_output_scalar_simd_mov_immediate.
5952 2018-03-12 Martin Sebor <msebor@redhat.com>
5954 PR tree-optimization/83456
5955 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
5956 for perfectly overlapping calls to memcpy.
5957 (gimple_fold_builtin_memory_chk): Same.
5958 (gimple_fold_builtin_strcpy): Handle no-warning.
5959 (gimple_fold_builtin_stxcpy_chk): Same.
5960 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
5962 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5964 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
5965 parameter. Use it for SFmode.
5966 (rs6000_function_arg_advance_1): Adjust.
5967 (rs6000_function_arg): Adjust.
5968 (rs6000_gimplify_va_arg): Pass false for that new parameter.
5970 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
5972 PR rtl-optimization/84169
5973 PR rtl-optimization/84780
5974 * combine.c (can_combine_p): Check for a 2-insn combination whether
5975 the destination register is used between the two insns, too.
5977 2018-03-12 Richard Biener <rguenther@suse.de>
5979 PR tree-optimization/84803
5980 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
5981 for refs DR analysis didn't process.
5983 2018-03-12 Richard Biener <rguenther@suse.de>
5985 PR tree-optimization/84777
5986 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
5987 force-vectorize loops ignore whether we are optimizing for size.
5989 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
5991 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
5992 (TARGET_MD_ASM_ADJUST): Define.
5994 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
5995 Kito Cheng <kito.cheng@gmail.com>
5996 Chung-Ju Wu <jasonwucj@gmail.com>
5998 * config/nds32/nds32.c (nds32_compute_stack_frame,
5999 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
6000 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
6001 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
6002 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
6003 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
6004 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
6005 * config/nds32/nds32.md (prologue, epilogue): Use macro
6006 NDS32_V3PUSH_AVAILABLE_P to do checking.
6008 2018-03-11 Jakub Jelinek <jakub@redhat.com>
6011 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
6012 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
6013 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
6014 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
6015 addition of most attributes on !orig_type_die or the attribute not
6016 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
6018 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6019 Chung-Ju Wu <jasonwucj@gmail.com>
6021 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
6023 * config/nds32/nds32.opt (mvh): New option.
6025 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6026 Chung-Ju Wu <jasonwucj@gmail.com>
6028 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
6030 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
6031 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
6034 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6035 Chung-Ju Wu <jasonwucj@gmail.com>
6037 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
6039 * config/nds32/nds32-multiple.md (strlensi): New pattern.
6040 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
6042 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
6043 Kito Cheng <kito.cheng@gmail.com>
6044 Chung-Ju Wu <jasonwucj@gmail.com>
6046 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
6047 UNSPEC_FFMISM and UNSPEC_FLMISM.
6048 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
6049 for ffb, ffmism and flmism.
6050 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
6051 (unspec_ffmism): Ditto.
6052 (unspec_flmism): Ditto.
6053 (nds32_expand_builtin_impl): Check if string extension is available.
6054 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
6055 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
6057 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
6060 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6063 * lra-assigns.c (assign_by_spills): Return a flag of reload
6064 assignment failure. Do not process the reload assignment
6065 failures. Do not spill other reload pseudos if they has the same
6067 (lra_assign): Add a return arg. Set up from the result of
6068 assign_by_spills call.
6069 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6070 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6071 usage_insns if it is not NULL.
6072 (spill_hard_reg_in_range): New function.
6073 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6074 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6075 function prototypes.
6076 (lra_assign): Change prototype.
6077 * lra.c (lra): Add code to deal with fails by splitting hard reg
6080 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
6083 * config/i386/i386.opt: Replace Enforcment with Enforcement.
6085 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
6088 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
6089 (dw_val_node): Add val_symbolic_view.
6090 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
6091 (symview_upper_bound): New.
6092 (new_line_info_table): Initialize symviews_since_reset.
6093 (dwarf2out_source_line): Count symviews_since_reset and set
6094 symview_upper_bound.
6095 (dw_val_equal_p): Handle symview.
6096 (add_AT_symview): New.
6097 (print_dw_val): Handle symview.
6098 (attr_checksum, attr_checksum_ordered): Likewise.
6099 (same_dw_val_p, size_of_die): Likewise.
6100 (value_format, output_die): Likewise.
6101 (add_high_low_attributes): Use add_AT_symview for entry_view.
6102 (dwarf2out_finish): Reset symview_upper_bound, clear
6105 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
6108 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
6109 Add strict argument and use it.
6110 (rs6000_split_multireg_move): Update for new strict argument.
6111 (mem_operand_gpr): Disallow all non-offsettable addresses.
6112 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
6114 2018-03-09 Jakub Jelinek <jakub@redhat.com>
6117 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
6118 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
6119 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
6122 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
6123 decl, use remap_type if we want to use the type.
6125 2018-03-09 Martin Sebor <msebor@redhat.com>
6127 PR tree-optimization/84526
6128 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6130 (builtin_access::generic_overlap): Be prepared to handle non-array
6133 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
6135 PR rtl-optimization/84682
6136 * lra-constraints.c (process_address_1): Check is_address flag
6137 for address constraints.
6138 (process_alt_operands): Likewise.
6139 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
6140 preprocess_constraints.
6141 * recog.h (preprocess_constraints): Add oploc parameter.
6143 * recog.c (preprocess_constraints): Test address_operand for
6144 CT_ADDRESS constraints.
6146 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6149 * lra-assigns.c (assign_by_spills): Return a flag of reload
6150 assignment failure. Do not process the reload assignment
6151 failures. Do not spill other reload pseudos if they has the same
6153 (lra_assign): Add a return arg. Set up from the result of
6154 assign_by_spills call.
6155 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6156 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6157 usage_insns if it is not NULL.
6158 (spill_hard_reg_in_range): New function.
6159 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6160 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6161 function prototypes.
6162 (lra_assign): Change prototype.
6163 * lra.c (lra): Add code to deal with fails by splitting hard reg
6166 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6169 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
6170 Accept complain bool parameter. Only emit errors if it is true.
6171 (arm_parse_cpu_option_name): Likewise.
6172 (arm_target_thumb_only): Adjust callers of the above.
6173 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
6174 prototype to take a default true bool parameter.
6175 (arm_parse_arch_option_name): Likewise.
6177 2018-03-09 David Malcolm <dmalcolm@redhat.com>
6178 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
6182 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
6183 * configure: Regenerate.
6184 * configure.ac ("linker --version-script option"): New.
6185 ("linker soname option"): New.
6187 2018-03-09 Richard Biener <rguenther@suse.de>
6189 PR tree-optimization/84775
6190 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
6191 immediate uses of predicate stmts and mark them modified.
6194 PR tree-optimization/84178
6195 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6197 (version_loop_for_if_conversion): Delay update_ssa call.
6198 (tree_if_conversion): Delay update_ssa until after predicate
6201 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
6204 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
6205 when the function accesses prior frames.
6207 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6210 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
6211 gen_llsym, otherwise call maybe_gen_llsym.
6214 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
6215 has ',' character inside of it.
6217 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6220 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
6221 as clobbering CC_REGNUM.
6223 2018-03-08 Richard Biener <rguenther@suse.de>
6226 * tree-scalar-evolution.c: Include tree-into-ssa.h.
6227 (follow_copies_to_constant): Do not follow SSA names registered
6230 2018-03-08 Richard Biener <rguenther@suse.de>
6232 PR tree-optimization/84178
6233 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6235 (version_loop_for_if_conversion): Delay update_ssa call.
6236 (tree_if_conversion): Delay update_ssa until after predicate
6239 2018-03-08 David Malcolm <dmalcolm@redhat.com>
6241 PR tree-optimization/84178
6242 * tree-if-conv.c (release_bb_predicate): Remove the
6243 the assertion that the stmts have NULL use_ops.
6244 Discard the statements, asserting that they haven't
6245 yet been added to a BB.
6247 2018-03-08 Richard Biener <rguenther@suse.de>
6249 PR tree-optimization/84746
6250 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
6251 (phi_translate): Pass in destination ANTIC_OUT set.
6252 (phi_translate_1): Likewise. For a simplified result lookup
6253 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
6254 (phi_translate_set): Adjust.
6255 (do_pre_regular_insertion): Likewise.
6256 (do_pre_partial_partial_insertion): Likewise.
6258 2018-03-08 Martin Liska <mliska@suse.cz>
6260 PR gcov-profile/84735
6261 * doc/gcov.texi: Document usage of profile files.
6262 * gcov-io.h: Document changes in the format.
6264 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
6268 * dwarf2out.c (struct dw_line_info_table): Update comments for
6270 (FORCE_RESET_NEXT_VIEW): New.
6271 (FORCE_RESETTING_VIEW_P): New.
6272 (RESETTING_VIEW_P): Check for -1 too.
6273 (ZERO_VIEW_P): Likewise.
6274 (new_line_info_table): Force-reset next view.
6275 (dwarf2out_begin_function): Likewise.
6276 (dwarf2out_source_line): Simplify zero_view_p initialization.
6277 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
6278 view directly. Omit view when omitting .loc at line 0.
6280 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6282 PR tree-optimization/84740
6283 * tree-switch-conversion.c (process_switch): Call build_constructors
6284 only if info.phi_count is non-zero.
6286 PR tree-optimization/84739
6287 * tree-tailcall.c (find_tail_calls): Check call arguments against
6288 DECL_ARGUMENTS (current_function_decl) rather than
6289 DECL_ARGUMENTS (func) when checking for tail recursion.
6291 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6293 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
6294 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
6295 Volker Reichelt's entry and add entries for people that perform
6296 GCC fuzzy testing and report numerous bugs.
6298 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
6301 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
6302 readonly data in sdata, if that is disabled.
6303 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
6304 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
6305 -mreadonly-in-sdata option.
6307 2018-03-07 Martin Sebor <msebor@redhat.com>
6309 PR tree-optimization/84468
6310 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
6311 basic block when looking for nul assignment.
6313 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
6316 * except.h (output_function_exception_table): Adjust prototype.
6317 * except.c (output_function_exception_table): Remove FNNAME parameter
6318 and add SECTION parameter. Ouput one part of the table at a time.
6319 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
6320 the first part of the exception table and emit unwind directives.
6321 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
6322 (i386_pe_seh_cold_init): Likewise.
6323 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
6324 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
6325 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
6326 (ix86_output_call_insn): Emit a nop in one more case for SEH.
6327 * config/i386/winnt.c: Include except.h.
6328 (struct seh_frame_state): Add reg_offset, after_prologue and
6329 in_cold_section fields.
6330 (i386_pe_seh_end_prologue): Set seh->after_prologue.
6331 (i386_pe_seh_cold_init): New function.
6332 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
6333 to seh->in_cold_section.
6334 (seh_emit_push): Record the offset of the push.
6335 (seh_emit_save): Record the offet of the save.
6336 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
6337 Test seh->after_prologue to disregard the epilogue.
6338 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
6339 (i386_pe_end_cold_function): New function.
6341 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6344 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
6345 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
6348 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
6350 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
6351 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
6354 * multiple_target.c: Include tree-inline.h and intl.h.
6355 (expand_target_clones): Diagnose and fail if node->definition and
6356 !tree_versionable_function_p (node->decl).
6358 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
6360 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
6362 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
6363 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
6364 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
6366 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6369 * combine.c (try_combine): Use reg_or_subregno instead of handling
6370 just paradoxical SUBREGs and REGs.
6372 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
6374 * config/arc/arc.c (arc_finalize_pic): Remove function.
6375 (arc_must_save_register): We use single base PIC register, remove
6376 checks to save/restore the PIC register.
6377 (arc_expand_prologue): Likewise.
6378 * config/arc/arc-protos.h (arc_set_default_type_attributes):
6380 (arc_verify_short): Likewise.
6381 (arc_attr_type): Likewise.
6382 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
6383 (walk_stores): Likewise.
6384 (arc_address_cost): Make it static.
6385 (arc_verify_short): Likewise.
6386 (branch_dest): Likewise.
6387 (arc_attr_type): Likewise.
6388 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
6389 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
6390 (arc_final_prescan_insn): Remove inserting the nops due to
6391 hardware hazards. It is done in reorg step.
6392 (insn_length_variant_t): Remove.
6393 (insn_length_parameters_t): Likewise.
6394 (arc_insn_length_parameters): Likewise.
6395 (arc_get_insn_variants): Likewise.
6396 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
6398 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6401 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
6404 PR tree-optimization/84687
6405 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
6407 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
6409 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6411 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
6412 Rename to ppc_speculation_barrier.
6413 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
6414 __builtin_ppc_speculation_barrier.
6416 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6419 * combine.c (combine_simplify_rtx): Don't try to simplify if
6420 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
6423 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
6425 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
6426 to 32 bytes when compiling for POWER9.
6428 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6431 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
6432 regparm >= 3 with no arg reg available also for calls with
6433 flag_force_indirect_call. Pass decl to ix86_function_regparm.
6436 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
6438 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
6440 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
6443 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
6445 2018-03-05 Richard Biener <rguenther@suse.de>
6447 PR tree-optimization/84486
6448 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
6449 When inserting a __builtin_assume_aligned call set the LHS
6450 SSA name alignment info accordingly.
6452 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
6454 PR tree-optimization/84114
6455 * config/aarch64/aarch64.c (aarch64_reassociation_width)
6456 Avoid reassociation of FLOAT_MODE addition.
6458 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
6460 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
6461 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
6462 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
6463 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
6464 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
6465 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
6466 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
6468 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
6469 __builtin_ia32_wbinvd): New builtins.
6470 (SPECIAL_ARGS2): New.
6471 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
6472 (SPECIAL_ARGS2): New.
6473 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
6474 (ix86_valid_target_attribute_inner_p): Ditto.
6475 (ix86_init_mmx_sse_builtins): Add special_args2.
6476 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
6477 TARGET_WBNOINVD_P): New.
6478 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
6479 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
6480 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
6481 * config/i386/immintrin.h (_wbinvd): New intrinsic.
6482 * config/i386/pconfigintrin.h: New file.
6483 * config/i386/wbnoinvdintrin.h: Ditto.
6484 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
6485 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
6487 2018-03-05 Richard Biener <rguenther@suse.de>
6489 PR tree-optimization/84670
6490 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6492 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6493 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6494 (compute_antic_aux): Only assert the number of values in ANTIC_IN
6495 doesn't grow if all successors (recursively) were visited at least
6498 2018-03-05 Richard Biener <rguenther@suse.de>
6500 PR tree-optimization/84650
6501 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
6502 if executed in the loop pipeline.
6504 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
6506 * doc/configfiles.texi (Configuration Files): Move info about
6507 conditionalizing $target-protos.h to...
6508 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
6509 differs from $target-protos.h.
6511 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
6512 Chung-Ju Wu <jasonwucj@gmail.com>
6514 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
6515 * config/nds32/nds32-multiple.md (setmemsi): Define.
6516 * config/nds32/nds32-memory-manipulation.c
6517 (nds32_gen_dup_4_byte_to_word_value): New.
6518 (emit_setmem_word_loop): New.
6519 (emit_setmem_byte_loop): New.
6520 (nds32_expand_setmem_loop): New.
6521 (nds32_expand_setmem_loop_v3m): New.
6522 (nds32_expand_setmem_unroll): New.
6523 (nds32_expand_setmem): New.
6525 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6526 Chung-Ju Wu <jasonwucj@gmail.com>
6528 * config/nds32/nds32-memory-manipulation.c
6529 (nds32_emit_load_store): New.
6530 (nds32_emit_post_inc_load_store): New.
6531 (nds32_emit_mem_move): New.
6532 (nds32_emit_mem_move_block): New.
6533 (nds32_expand_movmemsi_loop_unknown_size): New.
6534 (nds32_expand_movmemsi_loop_known_size): New.
6535 (nds32_expand_movmemsi_loop): New.
6536 (nds32_expand_movmemsi_unroll): New.
6537 (nds32_expand_movmemqi): Rename ...
6538 (nds32_expand_movmemsi): ... to this.
6539 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
6540 (movmemsi): ... to this.
6541 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
6542 (nds32_expand_movmemsi): ... to this.
6544 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6545 Monk Chiang <sh.chiang04@gmail.com>
6546 Chung-Ju Wu <jasonwucj@gmail.com>
6548 * config/nds32/nds32-protos.h
6549 (nds32_expand_load_multiple): New arguments.
6550 (nds32_expand_store_multiple): Ditto.
6551 (nds32_valid_multiple_load_store): Rename ...
6552 (nds32_valid_multiple_load_store_p): ... to this.
6553 * config/nds32/nds32-memory-manipulation.c
6554 (nds32_expand_load_multiple): Refine implementation.
6555 (nds32_expand_store_multiple): Ditto.
6556 * config/nds32/nds32-multiple.md
6557 (load_multiple): Update nds32_expand_load_multiple interface.
6558 (store_multiple): Update nds32_expand_store_multiple interface.
6559 * config/nds32/nds32-predicates.c
6560 (nds32_valid_multiple_load_store): Rename ...
6561 (nds32_valid_multiple_load_store_p): ... to this and refine
6563 * config/nds32/predicates.md
6564 (nds32_load_multiple_and_update_address_operation): New predicate.
6565 (nds32_store_multiple_and_update_address_operation): New predicate.
6567 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6568 Chung-Ju Wu <jasonwucj@gmail.com>
6570 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
6571 (combo): New attribute.
6572 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
6574 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
6576 * config/nds32/nds32.opt: Change -mcmodel= default value.
6578 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
6579 Monk Chiang <sh.chiang04@gmail.com>
6580 Chung-Ju Wu <jasonwucj@gmail.com>
6582 * config/nds32/constants.md (unspec_element): New enum.
6583 * config/nds32/constraints.md (Umw): New constraint.
6584 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
6585 * config/nds32/nds32-intrinsic.md: Likewise.
6586 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
6587 (nds32_valid_smw_lwm_base_p): New.
6588 (nds32_output_smw_single_word): New.
6589 (nds32_output_lmw_single_word): New.
6590 (nds32_expand_unaligned_load): New.
6591 (nds32_expand_unaligned_store): New.
6592 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
6593 (nds32_output_smw_single_word): Declare.
6594 (nds32_output_lmw_single_word): Declare.
6595 (nds32_expand_unaligned_load): Declare.
6596 (nds32_expand_unaligned_store): Declare.
6597 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
6598 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
6599 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
6600 NDS32_BUILTIN_UASTORE_DW.
6601 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
6604 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
6605 Kito Cheng <kito.cheng@gmail.com>
6606 Chung-Ju Wu <jasonwucj@gmail.com>
6608 * config/nds32/nds32-intrinsic.c
6609 (nds32_expand_builtin_null_ftype_reg): Delete.
6610 (nds32_expand_builtin_reg_ftype_imm): Ditto.
6611 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
6612 (nds32_read_argument): New.
6613 (nds32_legitimize_target): Ditto.
6614 (nds32_legitimize_argument): Ditto.
6615 (nds32_check_constant_argument): Ditto.
6616 (nds32_expand_unop_builtin): Ditto.
6617 (nds32_expand_unopimm_builtin): Ditto.
6618 (nds32_expand_binop_builtin): Ditto.
6619 (nds32_builtin_decl_impl): Ditto.
6620 (builtin_description): Ditto.
6621 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
6622 (nds32_init_builtins_impl): Ditto.
6623 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
6624 (nds32_builtin_decl): New.
6625 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
6626 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
6628 2018-03-02 Jeff Law <law@redhat.com>
6630 * reorg.c (stop_search_p): Handle DEBUG_INSN.
6631 (redundant_insn, fill_simple_delay_slots): Likewise.
6632 (fill_slots_from_thread): Likewise.
6633 * resource.c (mark_referenced_resources): Likewise.
6634 (mark_set_resources, find_dead_or_set_registers): Likewise.
6636 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6638 * substring-locations.h (format_warning_va): Formatting fix for
6640 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
6642 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
6643 * substring-locations.c: Include intl.h.
6644 (format_warning_va): Turned into small wrapper around
6645 format_warning_n_va, renamed to ...
6646 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
6647 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
6649 (format_warning_at_substring_n): New function.
6650 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
6651 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
6652 format_warning_at_substring with just a shorter name instead of
6653 const function pointer.
6654 (fmtwarn_n): New function.
6655 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
6656 appropriate, get rid of all the fmtstr temporaries, move conditionals
6657 with G_() wrapped string literals directly into fmtwarn arguments,
6658 cast dir.len to (int), formatting fixes.
6660 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
6662 * doc/invoke.texi: Remove "Cilk Plus" references.
6664 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6665 Richard Biener <rguenther@suse.de>
6668 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
6669 for error or warning attributes if CALL_FROM_THUNK_P is set.
6672 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6675 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
6676 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
6679 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
6680 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
6682 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
6683 instead of -1U in last predictors element's probability member.
6685 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
6688 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
6689 arguments if they are comparable.
6691 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6693 PR tree-optimization/84634
6694 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
6695 masks and masked_loop_p with a single loop_masks, making sure it's
6696 null for bb vectorization.
6698 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6700 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
6701 (vect_analyze_data_ref_access): Use loop->safe_len rather than
6702 loop->force_vectorize to check whether there is no alias.
6704 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6707 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
6709 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
6711 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
6712 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
6713 instead of a loop around prev_real_insn.
6714 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
6718 * config/i386/i386.c (ix86_print_operand): Use conditional
6719 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
6722 2018-03-02 Richard Biener <rguenther@suse.de>
6724 PR tree-optimization/84427
6725 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
6726 (bitmap_set_subtract_values): Rewrite to handle multiple
6729 (prune_clobbered_mems): Likewise.
6730 (phi_translate): Take edge instead of pred/phiblock.
6731 (phi_translate_1): Likewise.
6732 (phi_translate_set): Likewise. Insert all translated
6733 exprs for a value into the set, keeping possibly multiple
6734 expressions per value.
6735 (compute_antic_aux): Adjust for phi_translate changes.
6736 When intersecting union the expressions and prune those
6737 not in the final value set, keeping possibly multiple
6738 expressions per value. Do not use value-insertion
6739 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
6740 all expressions. Add verification that the value-sets
6741 only shrink during iteration.
6742 (compute_partial_antic_aux): Adjust for the phi_translate changes.
6743 (do_pre_regular_insertion): Likewise.
6744 (do_pre_partial_partial_insertion): Likewise.
6746 2018-03-02 Richard Biener <rguenther@suse.de>
6749 * config/darwin.c (saved_debug_info_level): New static global.
6750 (darwin_asm_lto_start): Disable debug info generation for LTO out.
6751 (darwin_asm_lto_end): Restore debug info generation settings.
6753 2018-03-01 Martin Liska <mliska@suse.cz>
6756 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
6759 2018-03-01 Richard Biener <rguenther@suse.de>
6762 * dwarf2out.c (gen_variable_die): Properly handle late VLA
6763 type annotation with LTO when debug was disabled at compile-time.
6765 2018-03-01 Matthew Fortune <mfortune@gmail.com>
6767 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
6769 (mips_final_postscan_insn): Likewise.
6771 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
6773 PR rtl-optimization/84528
6774 * alias.c (init_alias_target): Add commentary.
6775 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
6776 a unique base value if the frame pointer is not eliminated
6777 to the stack pointer.
6779 2018-03-01 Tom de Vries <tom@codesourcery.com>
6781 PR rtl-optimization/83327
6782 * lra-int.h (hard_regs_spilled_into): Declare.
6783 * lra.c (hard_regs_spilled_into): Define.
6784 (init_reg_info): Init hard_regs_spilled_into.
6785 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
6786 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
6787 (process_bb_lives): Handle hard_regs_spilled_into.
6788 (lra_create_live_ranges_1): Before doing liveness propagation, clear
6789 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
6791 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
6793 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
6794 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
6795 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
6796 * config/rs6000/aix72.h: New file.
6798 2018-02-28 Jakub Jelinek <jakub@redhat.com>
6800 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
6801 instead of warning_at with conditional singular and plural messages
6805 * stor-layout.c (update_alignment_for_field): For
6806 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
6807 && !DECL_PACKED (field), do the alignment update, just use
6808 only desired_align instead of MAX (type_align, desired_align)
6810 (place_field): Don't do known_align < desired_align handling
6811 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
6812 is non-NULL, instead do it after rli->prev_field handling and
6813 only if not within a bitfield word. For DECL_PACKED (field)
6814 use type_align of BITS_PER_UNIT.
6816 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
6818 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
6819 superfluous parentheses and trailing spaces.
6821 2018-02-28 Richard Biener <rguenther@suse.de>
6823 PR tree-optimization/84584
6824 * graphite-scop-detection.c (scop_detection::add_scop): Discard
6825 SCoPs with fake exit edge.
6827 2018-02-28 Martin Liska <mliska@suse.cz>
6830 * timevar.c (timer::print): Fix format to properly print 100%
6833 2018-02-28 Richard Biener <rguenther@suse.de>
6836 * genmatch.c (capture_info::walk_match): Do not mark
6837 captured expressions without operands as expr_p given
6838 they act more like predicates and should be subject to
6839 "lost tail" side-effect preserving.
6841 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
6843 PR rtl-optimization/81611
6844 * auto-inc-dec.c (attempt_change): Move dead note from
6845 mem_insn if it's the next use of regno
6846 (find_address): Take address use of reg holding
6847 non-incremented value. Add parm to limit search to the named
6849 (merge_in_block): Attempt to use a mem insn that is the next
6850 use of the original regno.
6852 2018-02-27 Martin Sebor <msebor@redhat.com>
6855 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
6856 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
6858 2018-02-27 Martin Sebor <msebor@redhat.com>
6860 PR translation/84207
6861 * diagnostic-core.h (warning_n, error_n, inform_n): Change
6862 n argument to unsigned HOST_WIDE_INT.
6863 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
6864 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
6865 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
6866 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
6868 2018-02-27 Richard Biener <rguenther@suse.de>
6870 PR tree-optimization/84512
6871 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
6872 Do not use the estimate returned from record_stmt_cost for
6873 the scalar iteration cost but sum properly using add_stmt_cost.
6875 2018-02-27 Richard Biener <rguenther@suse.de>
6877 PR tree-optimization/84466
6878 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6879 Adjust last change to less strictly validate use operands.
6881 2018-02-27 Martin Liska <mliska@suse.cz>
6883 PR gcov-profile/84548
6884 * gcov.c (process_file): Allow partial overlap and consider it
6885 also as group functions.
6886 (output_lines): Properly calculate range of lines for a group.
6888 2018-02-27 Martin Liska <mliska@suse.cz>
6890 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
6891 'ggc' suffixes. Change first column width.
6892 (timer::print): Fix formatting of the column.
6894 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
6896 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
6897 preserve inline entry blocks for the sake of debug inline
6898 entry point markers alone.
6899 (remove_unused_locals): Suggest in comments a better place to
6900 force the preservation of inline entry blocks that are
6901 otherwise unused, but do not preserve them.
6903 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6905 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
6907 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6910 * config/i386/constraints.md (Bs): Replace
6911 ix86_indirect_branch_register with
6912 TARGET_INDIRECT_BRANCH_REGISTER.
6914 * config/i386/i386.md (indirect_jump): Likewise.
6915 (tablejump): Likewise.
6916 (*sibcall_memory): Likewise.
6917 (*sibcall_value_memory): Likewise.
6918 Peepholes of indirect call and jump via memory: Likewise.
6919 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
6920 (*sibcall_value_GOT_32): Likewise.
6921 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6922 (GOT_memory_operand): Likewise.
6923 (call_insn_operand): Likewise.
6924 (sibcall_insn_operand): Likewise.
6925 (GOT32_symbol_operand): Likewise.
6926 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
6928 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6930 PR rtl-optimization/83496
6931 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
6932 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
6933 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
6934 redundant insn, if any.
6935 (relax_delay_slots): Likewise.
6936 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
6938 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
6940 PR tree-optimization/83965
6941 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
6942 that grouped statements are part of a reduction chain. Return
6943 true if the statement is not marked as a reduction itself but
6945 (vect_recog_dot_prod_pattern): Don't check whether the statement
6946 is part of a group here.
6947 (vect_recog_sad_pattern): Likewise.
6948 (vect_recog_widen_sum_pattern): Likewise.
6950 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6953 * final.c (rest_of_clean_state): Also look for calls inside sequences.
6955 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6958 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
6960 (ix86_output_indirect_function_return): New prototype.
6961 (ix86_split_simple_return_pop_internal): Likewise.
6962 * config/i386/i386.c (indirect_return_via_cx): New.
6963 (indirect_return_via_cx_bnd): Likewise.
6964 (indirect_thunk_name): Handle return va CX_REG.
6965 (output_indirect_thunk_function): Create alias for
6966 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
6967 (ix86_output_indirect_jmp): Remove the bool argument.
6968 (ix86_output_indirect_function_return): New function.
6969 (ix86_split_simple_return_pop_internal): Likewise.
6970 * config/i386/i386.md (*indirect_jump): Don't pass false
6971 to ix86_output_indirect_jmp.
6972 (*tablejump_1): Likewise.
6973 (simple_return_pop_internal): Change it to define_insn_and_split.
6974 Call ix86_split_simple_return_pop_internal to split it for
6976 (simple_return_indirect_internal): Call
6977 ix86_output_indirect_function_return instead of
6978 ix86_output_indirect_jmp.
6980 2018-02-26 Jakub Jelinek <jakub@redhat.com>
6983 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
6984 memset and value initialization afterwards.
6986 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
6988 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
6990 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
6993 * common/config/aarch64/aarch64-common.c
6994 (aarch_option_optimization_table[]): Switch
6995 off fomit-frame-pointer
6997 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
6998 Chung-Ju Wu <jasonwucj@gmail.com>
7000 * config/nds32/nds32-multiple.md (load_multiple): Disallow
7002 (store_multiple): Ditto.
7004 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7006 * config.gcc: Add --with-cpu support for nds32 target.
7007 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
7008 * config/nds32/nds32.opt: Add -mcpu= option.
7010 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
7012 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
7013 isel=yes): Warn for these deprecated options.
7015 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
7017 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
7018 ISA_2_5_MASKS_EMBEDDED.
7020 2018-02-23 Jakub Jelinek <jakub@redhat.com>
7022 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
7023 p->max as pointers rather than using iterative_hash_expr.
7025 2018-02-23 Carl Love <cel@us.ibm.com>
7027 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
7028 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
7030 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
7031 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
7032 P8V_BUILTIN_VEC_VUNSIGNED2.
7034 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
7037 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
7038 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
7040 * lra-constraints.c (curr_insn_transform): Set up
7041 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
7043 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
7044 * lra-eliminations.c (spill_pseudos): Ditto.
7045 (process_insn_for_elimination): Ditto.
7046 * lra-lives.c (reg_early_clobber_p): Use the new macros.
7047 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
7048 LRA_NON_CLOBBERED_ALT.
7050 2018-02-22 Martin Sebor <msebor@redhat.com>
7052 PR tree-optimization/84480
7053 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
7054 to maybe_diag_stxncpy_trunc. Call it.
7055 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
7056 from gimple_fold_builtin_strcpy. Print inlining stack.
7057 (handle_builtin_stxncpy): Print inlining stack.
7058 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
7060 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
7063 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
7064 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
7065 and -fcheck-pointer-bounds are used together.
7066 (indirect_thunk_prefix): New enum.
7067 (indirect_thunk_need_prefix): New function.
7068 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
7069 "_nt" instead of "_bnd" for NOTRACK prefix.
7070 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
7071 (output_indirect_thunk_function): Likewise.
7073 (ix86_code_end): Update output_indirect_thunk_function calls.
7074 (ix86_output_indirect_branch_via_reg): Replace
7075 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
7076 (ix86_output_indirect_branch_via_push): Likewise.
7077 (ix86_output_function_return): Likewise.
7078 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
7079 incompatible with -fcf-protection=branch and
7080 -fcheck-pointer-bounds.
7082 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7085 * config/aarch64/aarch64.c (aarch64_print_address_internal):
7086 Change gcc_assert call to output_operand_lossage.
7088 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7090 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
7092 2018-02-22 DJ Delorie <dj@redhat.com>
7093 Sebastian Perta <sebastian.perta@renesas.com>
7094 Oleg Endo <olegendo@gcc.gnu.org>
7096 * config/rx/rx.c (rx_rtx_costs): New function.
7097 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
7099 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
7101 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
7103 2018-02-22 Martin Liska <mliska@suse.cz>
7106 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
7107 Add "native" as a possible value.
7109 2018-02-22 Martin Liska <mliska@suse.cz>
7112 * config/i386/i386.c (ix86_option_override_internal):
7113 Add "native" as a possible value for -march and -mtune.
7115 2018-02-22 Jakub Jelinek <jakub@redhat.com>
7118 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
7119 to all type variants.
7121 PR tree-optimization/84503
7122 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
7123 width as info->bitpos + info->bitsize - start.
7124 (merged_store_group::merge_overlapping): Simplify width computation.
7125 (check_no_overlap): New function.
7126 (imm_store_chain_info::try_coalesce_bswap): Compute expected
7127 start + width and last_order of the group, fail if check_no_overlap
7129 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
7130 to group if check_no_overlap fails.
7132 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7134 * config/rs6000/altivec.md: Delete contraint arguments to
7135 define_expand, define_split, and define_peephole2, and in
7136 define_insn_and_split if always unused.
7137 * config/rs6000/darwin.md: Ditto.
7138 * config/rs6000/dfp.md: Ditto.
7139 * config/rs6000/rs6000.md: Ditto.
7140 * config/rs6000/sync.md: Ditto.
7141 * config/rs6000/vector.md: Ditto.
7142 * config/rs6000/vsx.md: Ditto.
7144 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7146 * config/rs6000/altivec.md: Write output control strings as braced
7147 blocks instead of double-quoted strings.
7148 * config/rs6000/darwin.md: Ditto.
7149 * config/rs6000/rs6000.md: Ditto.
7150 * config/rs6000/vector.md: Ditto.
7151 * config/rs6000/vsx.md: Ditto.
7153 2018-02-21 Jason Merrill <jason@redhat.com>
7155 PR c++/84314 - ICE with templates and fastcall attribute.
7156 * attribs.c (build_type_attribute_qual_variant): Remove assert.
7158 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7160 * ipa-cp.c (determine_versionability): Fix comment typos.
7162 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7165 * ipa-cp.c (determine_versionability): Do not version functions caling
7168 2018-02-21 Martin Liska <mliska@suse.cz>
7171 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
7172 Add "native" as a possible value.
7173 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
7174 the macro when native cpu detection is available.
7176 2018-02-21 Martin Liska <mliska@suse.cz>
7179 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
7180 Add "native" as a possible value.
7181 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
7182 when native cpu detection is available.
7184 2018-02-21 Jakub Jelinek <jakub@redhat.com>
7185 Martin Sebor <msebor@redhat.com>
7187 PR tree-optimization/84478
7188 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
7190 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
7191 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
7192 support which is conservatively correct, for 2 only stay conservative
7193 for maxlen. Formatting and comment capitalization fixes. Add STRICT
7194 argument to the 2 argument get_range_strlen, adjust 6 arg
7195 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
7197 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
7198 (gimple_fold_builtin_strlen): Pass true as last argument to
7201 2018-02-20 Martin Sebor <msebor@redhat.com>
7204 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
7205 (builtin_memref::set_base_and_offset): Same. Handle inner references.
7206 (builtin_memref::builtin_memref): Factor out parts into
7207 set_base_and_offset and call it.
7209 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7212 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
7213 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
7214 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
7215 search at the associated MODE_INT.
7217 2018-02-20 Jeff Law <law@redhat.com>
7220 PR tree-optimization/81592
7222 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
7223 for range data rather than using global data.
7224 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
7225 range data rather than using global data.
7226 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
7227 pass it to children as needed.
7228 (struct directive::fmtresult): Similarly.
7229 (struct directive::set_width): Similarly.
7230 (struct directive::set_precision): Similarly.
7231 (format_integer, format_directive, parse_directive): Similarly.
7232 (format_none): Accept unnamed vr_values parameter.
7233 (format_percent, format_floating, format_character): Similarly.
7234 (format_string, format_plain): Similarly.
7235 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
7236 the EVRP range analyzer for range data rather than using global data.
7237 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
7238 gimple-ssa-evrp-analyze.h
7239 (class sprintf_dom_walker): Add after_dom_children member function.
7240 Add evrp_range_analyzer member.
7241 (sprintf_dom_walker::before_dom_children): Call into the EVRP
7242 range analyzer as needed.
7243 (sprintf_dom_walker::after_dom_children): New member function.
7244 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
7246 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
7247 (evrp_range_analyzer::pop_to_marker): Likewise.
7249 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7251 PR tree-optimization/84419
7252 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
7253 with the required type if its current type is compatible but
7256 2018-02-20 Jakub Jelinek <jakub@redhat.com>
7259 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
7260 after vectorization.
7262 2018-02-20 Martin Liska <mliska@suse.cz>
7265 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
7266 possible values if we don't have a hint.
7268 2018-02-20 Martin Liska <mliska@suse.cz>
7272 * final.c (shorten_branches): Build align_tab array with one
7274 * opts.c (finish_options): Add alignment option limit check.
7275 (MAX_CODE_ALIGN): Likewise.
7276 (MAX_CODE_ALIGN_VALUE): Likewise.
7277 * doc/invoke.texi: Document maximum allowed option value for
7278 all -falign-* options.
7280 2018-02-19 Jakub Jelinek <jakub@redhat.com>
7283 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
7284 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
7285 * var-tracking.c (emit_note_insn_var_location): Remove all references
7286 to NOTE_INSN_CALL_ARG_LOCATION.
7287 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
7288 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
7289 Use copy_rtx_if_shared.
7290 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
7291 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
7292 (dwarf2out_var_location): Remove handling of
7293 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
7295 * final.c (final_scan_insn): Remove all references to
7296 NOTE_INSN_CALL_ARG_LOCATION.
7297 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
7298 before dumping final insns.
7299 * except.c (emit_note_eh_region_end): Remove all references to
7300 NOTE_INSN_CALL_ARG_LOCATION.
7301 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
7302 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
7303 * config/arc/arc.c (hwloop_optimize): Likewise.
7304 * config/arm/arm.c (create_fix_barrier): Likewise.
7305 * config/s390/s390.c (s390_chunkify_start): Likewise.
7306 * config/sh/sh.c (find_barrier): Likewise.
7307 * config/i386/i386.c (rest_of_insert_endbranch,
7308 ix86_seh_fixup_eh_fallthru): Likewise.
7309 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
7310 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
7311 * config/frv/frv.c (frv_function_prologue): Likewise.
7312 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
7314 (note_outside_basic_block_p): Remove all references to
7315 NOTE_INSN_CALL_ARG_LOCATION.
7316 * gengtype.c (adjust_field_rtx_def): Likewise.
7317 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
7319 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
7320 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
7323 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
7326 PR tree-optimization/84452
7327 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
7328 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
7331 2018-02-19 Martin Liska <mliska@suse.cz>
7334 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
7336 2018-02-19 Martin Liska <mliska@suse.cz>
7337 Richard Sandiford <richard.sandiford@linaro.org>
7339 PR tree-optimization/82491
7340 * gimple-fold.c (get_base_constructor): Make earlier bail out
7343 2018-02-19 Carl Love <cel@us.ibm.com>
7345 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
7346 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
7348 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
7349 P8V_BUILTIN_VEC_NEG.
7351 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
7353 * config/rl78/rl78.md (movdf): New define expand.
7355 2018-02-19 Martin Liska <mliska@suse.cz>
7358 * doc/invoke.texi: Fix typo.
7359 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
7361 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
7363 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
7364 handle rs6000_single_float and rs6000_double_float specially for
7367 2018-02-16 Jeff Law <law@redhat.com>
7369 * config/rx/rx.c (add_pop_cfi_notes): New function.;
7372 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7375 * ipa-inline.c (inline_small_functions): Fix a typo.
7377 2018-02-16 Nathan Sidwell <nathan@acm.org>
7379 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
7381 2018-02-16 Carl Love <cel@us.ibm.com>
7383 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
7384 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
7385 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
7386 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
7387 expansion to P8V_BUILTIN_VEC_FLOAT2.
7389 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
7391 PR rtl-optimization/70023
7392 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
7393 src_regno into account.
7395 2018-02-16 Carl Love <cel@us.ibm.com>
7397 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
7398 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
7399 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
7400 * config/rs6000/rs6000.c: Remove case statements for
7401 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
7402 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
7403 and P9V_BUILTIN_VEC_VINSERT4B.
7404 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
7405 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
7406 * config/rs6000/vsx.md:
7407 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
7410 2018-02-16 Carl Love <cel@us.ibm.com>
7412 * config/rs6000/altivec.h: Add builtin names vec_extract4b
7414 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
7416 * config/rs6000/rs6000-c.c: Add the definitions for
7417 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
7418 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
7419 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
7420 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
7421 definition for insert4b and define insn *insert3b_internal.
7422 * doc/extend.texi: Add documentation for vec_extract4b.
7424 2018-02-16 Nathan Sidwell <nathan@acm.org>
7426 * doc/extend.texi (Backwards Compatibility): Mention friend
7427 injection. Note for-scope is deprecated.
7428 * doc/invoke.texi (-ffriend-injection): Deprecate.
7430 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
7432 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
7433 that moved to I2, also allow destinations that are a paradoxical
7434 subreg (instead of a normal reg).
7436 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
7439 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
7442 2018-02-16 Richard Biener <rguenther@suse.de>
7444 PR tree-optimization/84037
7445 PR tree-optimization/84016
7447 * config/i386/i386.c (ix86_builtin_vectorization_cost):
7448 Adjust vec_construct for the fact we need additional higher latency
7449 128bit inserts for AVX256 and AVX512 vector builds.
7450 (ix86_add_stmt_cost): Scale vector construction cost for
7453 2018-02-16 Richard Biener <rguenther@suse.de>
7455 PR tree-optimization/84417
7456 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
7457 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
7458 (non_rewritable_lvalue_p): Likewise, use poly-ints.
7460 2018-02-16 Martin Liska <mliska@suse.cz>
7463 * internal-fn.def (ASAN_CHECK): Set proper flags.
7464 (ASAN_MARK): Likewise.
7466 2018-02-16 Julia Koval <julia.koval@intel.com>
7468 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
7469 from PTA_CANNONLAKE.
7471 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7474 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
7475 Use ++iter rather than iter++ for std::list iterators.
7476 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
7477 defer deleting them until all nodes in the forest are processed. Do
7478 free even leaf nodes. Change to_process into auto_vec.
7481 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
7482 * vec.h (vec_default_construct): Use memset instead of placement new
7483 if BROKEN_VALUE_INITIALIZATION is defined.
7484 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
7485 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
7488 PR rtl-optimization/83723
7489 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
7490 * lra.c (lra_substitute_pseudo): Likewise. If true, use
7491 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
7493 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
7495 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
7497 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
7499 PR rtl-optimization/81443
7500 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
7501 from inner REGs to paradoxical SUBREGs.
7503 2018-02-16 Richard Biener <rguenther@suse.de>
7505 PR tree-optimization/84399
7506 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7507 For operands we can analyze at their definition make sure we can
7508 analyze them at each use as well.
7510 2018-02-16 Richard Biener <rguenther@suse.de>
7512 PR tree-optimization/84190
7513 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
7514 volatile accesses if the decl isn't volatile.
7516 2018-02-15 Jason Merrill <jason@redhat.com>
7518 PR c++/84314 - ICE with templates and fastcall attribute.
7519 * attribs.c (build_type_attribute_qual_variant): Don't clobber
7520 TYPE_CANONICAL on an existing type.
7522 2018-02-15 Jakub Jelinek <jakub@redhat.com>
7524 PR tree-optimization/84383
7525 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
7526 dstoff nor call operand_equal_p if dstbase is NULL.
7528 PR tree-optimization/84334
7529 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
7530 also a CONSTANT_CLASS_P, punt.
7532 2018-02-14 Jim Wilson <jimw@sifive.com>
7534 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
7535 first SMALL_OPERAND check. New local min_second_step. Move assert
7536 to where locals are set. Add TARGET_RVC support.
7537 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
7539 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
7541 * doc/invoke.texi: Correct -Wformat-overflow code sample.
7543 2018-02-14 Martin Sebor <msebor@redhat.com>
7545 PR tree-optimization/83698
7546 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
7547 arrays constrain the offset range to their bounds.
7548 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
7549 (builtin_access::overlap): Avoid setting the size of overlap if it's
7551 (maybe_diag_overlap): Also consider arrays when deciding what values
7552 of offsets to include in diagnostics.
7554 2018-02-14 Martin Sebor <msebor@redhat.com>
7557 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
7558 that correspond to the kind of a declaration.
7560 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
7563 * config/pa/pa.md: Load address of PIC label using the linkage table
7564 if the label is nonlocal.
7566 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
7568 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
7569 warning message if user requests -maltivec=be.
7570 * doc/invoke.texi: Document deprecation of -maltivec=be.
7572 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
7575 * config/rs6000/rs6000-c.c: Update definitions for
7576 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
7577 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
7579 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7582 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
7583 add _get_ssp intrinsics. Remove argument from
7584 __builtin_ia32_rdssp[d|q].
7585 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
7586 * config/i386/i386-builtin.def: Remove argument from
7587 __builtin_ia32_rdssp[d|q].
7588 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
7589 ix86_expand_special_args_builtin for _rdssp[d|q].
7590 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
7591 Clear register before usage.
7592 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
7593 Add documentation for new _get_ssp and _inc_ssp intrinsics.
7595 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
7597 PR tree-optimization/84357
7598 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
7599 operand 1 of an ARRAY_REF too.
7601 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
7604 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
7605 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7607 (set_of_reg): New struct.
7608 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
7609 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
7610 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7612 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
7613 Split into bitclr, bitset, bitinvert patterns if appropriate.
7614 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
7615 use rx_fuse_in_memory_bitop.
7616 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
7617 to named insn, correct maximum insn length.
7619 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
7622 * machmode.def: Define a complex mode for PARTIAL_INT.
7623 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
7625 * doc/rtl.texi: Document CSPImode.
7626 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
7628 (msp430_hard_regno_nregs_with_padding): Likewise.
7630 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
7633 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
7635 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
7637 PR rtl-optimization/84169
7638 * combine.c (try_combine): New variable split_i2i3. Set it to true if
7639 we generated a parallel as new i3 and we split that to new i2 and i3
7640 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
7641 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
7642 those to i2, not i1. Partially rewrite this scan code.
7644 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7647 * stor-layout.c (place_field): For variable length fields, adjust
7648 offset_align afterwards not just based on the field's alignment,
7649 but also on the size.
7652 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
7653 of exps and logs in the use_exp2 case.
7655 2018-02-13 Jeff Law <law@redhat.com>
7657 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
7660 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
7663 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
7667 * common.opt (gas-loc-support, gas-locview-support): New.
7668 (ginline-points, ginternal-reset-location-views): New.
7669 * doc/invoke.texi: Document them. Use @itemx where intended.
7670 (gvariable-location-views): Adjust.
7671 * target.def (reset_location_view): New.
7672 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
7673 (TARGET_RESET_LOCATION_VIEW): New.
7674 * doc/tm.texi: Rebuilt.
7675 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
7676 (dwarf2out_default_as_locview_support): New.
7677 (output_asm_line_debug_info): Use option variables.
7678 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
7679 (output_loc_list): Likewise.
7680 (add_high_low_attributes): Check option variables.
7681 Don't output entry view attribute in strict mode.
7682 (gen_inlined_subroutine_die): Check option variables.
7683 (dwarf2out_inline_entry): Likewise.
7684 (init_sections_and_labels): Likewise.
7685 (dwarf2out_early_finish): Likewise.
7686 (maybe_reset_location_view): New, from...
7687 (dwarf2out_var_location): ... here. Call it.
7688 * debug.h (dwarf2out_default_as_loc_support): Declare.
7689 (dwarf2out_default_as_locview_support): Declare.
7690 * hooks.c (hook_int_rtx_insn_0): New.
7691 * hooks.h (hook_int_rtx_insn_0): Declare.
7692 * toplev.c (process_options): Take -gas-loc-support and
7693 -gas-locview-support from dwarf2out. Enable
7694 -gvariable-location-views by default only with locview
7695 assembler support. Enable -ginternal-reset-location-views by
7696 default only if the target defines the corresponding hook.
7697 Enable -ginline-points by default if location views are
7698 enabled; force it disabled if statement frontiers are
7700 * tree-inline.c (expand_call_inline): Check option variables.
7701 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
7703 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
7705 PR tree-optimization/84321
7706 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
7707 handling. Also check whether the anti-range contains any values
7708 that satisfy the mask; switch to a VR_RANGE if not.
7710 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
7713 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
7715 2018-02-13 Martin Jambor <mjambor@suse.cz>
7718 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
7719 of call statements, also set location of a load to a temporary.
7721 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
7723 * config/rl78/rl78.c (add_vector_labels): New function.
7724 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
7725 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
7726 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
7727 which checks that no arguments are passed.
7728 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
7729 * doc/extend.texi: Documentation for the new attribute.
7731 2018-02-13 Andreas Schwab <schwab@suse.de>
7733 * config/riscv/linux.h (CPP_SPEC): Define.
7735 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7738 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
7739 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
7740 OPTION_MASK_ISA_AES as first argument to def_builtin_const
7741 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
7742 instead of OPTION_MASK_ISA_PCLMUL as first argument to
7743 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
7744 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
7745 temporarily for AES and PCLMUL builtins.
7747 PR tree-optimization/84339
7748 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
7749 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
7753 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
7754 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
7755 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
7757 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
7759 * omp-simd-clone.h: New file.
7760 * omp-simd-clone.c: Include omp-simd-clone.h.
7761 (expand_simd_clones): No longer static.
7762 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
7763 cgraph.h and omp-simd-clone.h.
7764 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
7765 (vect_recog_widen_shift_pattern): Formatting fix.
7766 (vect_pattern_recog_1): Don't check optab for calls.
7769 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
7770 operands[2] into a REG before using gen_lowpart on it.
7772 2018-02-12 Jeff Law <law@redhat.com>
7775 * config/sh/sh.c (find_barrier): Consider a sibling call
7778 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
7779 successfully back substituting a reg.
7781 2018-02-12 Richard Biener <rguenther@suse.de>
7783 PR tree-optimization/84037
7784 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
7785 parameter, move visited init to caller.
7786 (vect_slp_analyze_operations): Separate cost from validity
7787 check, initialize visited once for all instances.
7788 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
7790 * tree-vect-stmts.c (vect_model_simple_cost): Make early
7792 (vect_model_promotion_demotion_cost): Likewise.
7793 (vectorizable_bswap): Guard cost modeling with !slp_node
7794 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
7796 (vectorizable_call): Likewise.
7797 (vectorizable_conversion): Likewise.
7798 (vectorizable_assignment): Likewise.
7799 (vectorizable_shift): Likewise.
7800 (vectorizable_operation): Likewise.
7801 (vectorizable_store): Likewise.
7802 (vectorizable_load): Likewise.
7803 (vectorizable_condition): Likewise.
7804 (vectorizable_comparison): Likewise.
7806 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
7809 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
7810 (ASAN_MARK): Fix fnspec to account for return value, change pointer
7811 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
7813 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7816 * params.def (inline-min-speedup): Increase from 8 to 15.
7817 (max-inline-insns-auto): Decrease from 40 to 30.
7818 * ipa-split.c (consider_split): Add some buffer for function to
7819 be considered inlining candidate.
7820 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
7823 2018-02-12 Richard Biener <rguenther@suse.de>
7825 PR tree-optimization/84037
7826 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
7827 matched stmts if we cannot swap the non-matched ones.
7829 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7831 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
7832 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
7833 _mm_maskz_scalef_round_ss): New intrinsics.
7834 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
7835 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
7836 __builtin_ia32_scalefss_round): Remove.
7837 (__builtin_ia32_scalefsd_mask_round,
7838 __builtin_ia32_scalefss_mask_round): New intrinsics.
7839 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
7840 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
7841 ((match_operand:VF_128 2 "<round_nimm_predicate>"
7842 "<round_constraint>")): Changed to ...
7843 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
7844 "<round_scalar_constraint>")): ... this.
7845 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
7846 %0, %1, %2<round_op3>}"): Changed to ...
7847 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
7848 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
7849 %2<round_scalar_mask_op3>}"): ... this.
7850 * config/i386/subst.md (round_scalar_nimm_predicate): New.
7852 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7854 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
7855 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
7856 (_mm_maskz_sqrt_round_ss): New intrinsics.
7857 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
7858 (__builtin_ia32_sqrtsd_mask_round)
7859 (__builtin_ia32_sqrtss_mask_round): New builtins.
7860 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
7861 (__builtin_ia32_sqrtss_round): Remove.
7862 (__builtin_ia32_sqrtsd_mask_round)
7863 (__builtin_ia32_sqrtss_mask_round): New builtins.
7864 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
7865 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
7866 ((match_operand:VF_128 1 "vector_operand"
7867 "xBm,<round_constraint>")): Changed to ...
7868 ((match_operand:VF_128 1 "vector_operand"
7869 "xBm,<round_scalar_constraint>")): ... this.
7870 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
7871 %0, %2, %<iptr>1<round_op3>}): Changed to ...
7872 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
7873 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
7874 %<iptr>1<round_scalar_mask_op3>}): ... this.
7875 ((set_attr "prefix" "<round_prefix>")): Changed to ...
7876 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
7878 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
7881 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
7882 Cast vec_cmpeq result to correct type.
7883 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
7884 Cast vec_cmpgt result to correct type.
7886 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
7888 * final.c (final_scan_insn_1): Renamed from...
7889 (final_scan_insn): ... this. New wrapper, to recover
7890 seen from the outermost call in recursive ones.
7891 * config/sparc/sparc.c (output_return): Drop seen from call.
7892 (output_sibcall): Likewise.
7893 * config/visium/visium.c (output_branch): Likewise.
7895 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
7897 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
7900 2018-02-10 Alan Modra <amodra@gmail.com>
7903 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
7904 Specify LR as an input.
7906 2018-02-10 Jakub Jelinek <jakub@redhat.com>
7909 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
7910 remove_member_access_dummy_vars): New functions.
7911 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
7912 lower_omp_1, execute_lower_omp): Use them.
7914 PR rtl-optimization/84308
7915 * shrink-wrap.c (spread_components): Release todo vector.
7917 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
7919 PR rtl-optimization/57193
7920 * ira-color.c (struct allocno_color_data): Add member
7921 conflict_allocno_hard_prefs.
7922 (update_conflict_allocno_hard_prefs): New.
7923 (bucket_allocno_compare_func): Add a preference based on
7924 conflict_allocno_hard_prefs.
7925 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
7926 (color_allocnos): Remove a dead code. Initiate
7927 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
7929 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7932 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
7933 constraint from =wa to wa. Avoid a subreg on the output operand,
7934 instead use a pseudo and subreg it in a move.
7935 (p9_xxbrd_<mode>): Changed to ...
7936 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
7937 (p9_xxbrd_v2df): New expander.
7938 (p9_xxbrw_<mode>): Changed to ...
7939 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
7940 (p9_xxbrw_v4sf): New expander.
7942 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7944 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
7946 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
7949 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
7950 multiply in 32-bit mode.
7951 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
7952 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
7955 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
7957 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
7958 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
7959 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
7960 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
7962 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
7965 * dwarf2out.c (is_trivial_indirect_ref): New function.
7966 (dwarf2out_late_global_decl): Do not generate a location
7967 attribute for variables that have a non-trivial DECL_VALUE_EXPR
7968 and that are not defined in the current unit.
7970 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7972 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
7973 instead of a libcall for UNORDERED.
7975 2018-02-09 Tamar Christina <tamar.christina@arm.com>
7978 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
7979 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
7981 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7984 * config/s390/s390.c (s390_set_current_function): Invoke
7985 s390_indirect_branch_settings also if fndecl didn't change.
7987 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
7989 * config/rs6000/rs6000.md (blockage): Set length to zero.
7991 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
7993 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
7995 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7998 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
7999 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
8003 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
8004 PARALLEL incoming that failed vt_get_decl_and_offset check.
8007 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
8008 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
8010 (get_variable_section): For decls in named .bss* sections pass true as
8011 second argument to bss_initializer_p.
8013 2018-02-09 Marek Polacek <polacek@redhat.com>
8014 Jakub Jelinek <jakub@redhat.com>
8017 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
8018 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
8019 Sync some changes from cxx_fold_indirect_ref.
8021 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8023 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
8025 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
8026 (BLOCK_INLINE_ENTRY_LABEL): New.
8027 (dwarf2out_var_location): Disregard inline entry markers.
8028 (inline_entry_data): New struct.
8029 (inline_entry_data_hasher): New hashtable type.
8030 (inline_entry_data_hasher::hash): New.
8031 (inline_entry_data_hasher::equal): New.
8032 (inline_entry_data_table): New variable.
8033 (add_high_low_attributes): Add DW_AT_entry_pc and
8034 DW_AT_GNU_entry_view attributes if a pending entry is found
8035 in inline_entry_data_table. Add old entry_pc attribute only
8036 if debug nonbinding markers are disabled.
8037 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
8038 markers are enabled.
8039 (block_within_block_p, dwarf2out_inline_entry): New.
8040 (dwarf2out_finish): Check that no entries remained in
8041 inline_entry_data_table.
8042 * final.c (reemit_insn_block_notes): Handle inline entry notes.
8043 (final_scan_insn, notice_source_line): Likewise.
8044 (rest_of_clean_state): Skip inline entry markers.
8045 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
8047 * gimple.c (gimple_build_debug_inline_entry): New.
8048 * gimple.h (enum gimple_debug_subcode): Add
8049 GIMPLE_DEBUG_INLINE_ENTRY.
8050 (gimple_build_debug_inline_entry): Declare.
8051 (gimple_debug_inline_entry_p): New.
8052 (gimple_debug_nonbind_marker_p): Adjust.
8053 * insn-notes.def (INLINE_ENTRY): New.
8054 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
8055 inline entry marker notes.
8056 (print_insn): Likewise.
8057 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
8058 (INSN_DEBUG_MARKER_KIND): Likewise.
8059 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
8060 * tree-inline.c (expand_call_inline): Build and insert
8061 debug_inline_entry stmt.
8062 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
8063 inline entry blocks early, if nonbind markers are enabled.
8064 (dump_scope_block): Dump fragment info.
8065 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
8066 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
8067 (gimple_build_debug_inline_entry): New.
8068 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
8069 Enable/disable inline entry points too.
8070 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
8071 (DEBUG_INSN): Describe inline entry markers.
8073 * common.opt (gvariable-location-views): New.
8074 (gvariable-location-views=incompat5): New.
8075 * config.in: Rebuilt.
8076 * configure: Rebuilt.
8077 * configure.ac: Test assembler for view support.
8078 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
8079 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
8080 * dwarf2out.c (var_loc_view): New typedef.
8081 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
8082 (dwarf2out_locviews_in_attribute): New.
8083 (dwarf2out_locviews_in_loclist): New.
8084 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
8085 (enum dw_line_info_opcode): Add LI_adv_address.
8086 (struct dw_line_info_table): Add view.
8087 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
8088 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
8089 (zero_view_p): New variable.
8090 (ZERO_VIEW_P): New macro.
8091 (output_asm_line_debug_info): New.
8092 (struct var_loc_node): Add view.
8093 (add_AT_view_list, AT_loc_list): New.
8094 (add_var_loc_to_decl): Add view param. Test it against last.
8095 (new_loc_list): Add view params. Record them.
8096 (AT_loc_list_ptr): Handle loc and view lists.
8097 (view_list_to_loc_list_val_node): New.
8098 (print_dw_val): Handle dw_val_class_view_list.
8099 (size_of_die): Likewise.
8100 (value_format): Likewise.
8101 (loc_list_has_views): New.
8102 (gen_llsym): Set vl_symbol too.
8103 (maybe_gen_llsym, skip_loc_list_entry): New.
8104 (dwarf2out_maybe_output_loclist_view_pair): New.
8105 (output_loc_list): Output view list or entries too.
8106 (output_view_list_offset): New.
8107 (output_die): Handle dw_val_class_view_list.
8108 (output_dwarf_version): New.
8109 (output_compilation_unit_header): Use it.
8110 (output_skeleton_debug_sections): Likewise.
8111 (output_rnglists, output_line_info): Likewise.
8112 (output_pubnames, output_aranges): Update version comments.
8113 (output_one_line_info_table): Output view numbers in asm comments.
8114 (dw_loc_list): Determine current endview, pass it to new_loc_list.
8115 Call maybe_gen_llsym.
8116 (loc_list_from_tree_1): Adjust.
8117 (add_AT_location_description): Create view list attribute if
8118 needed, check it's absent otherwise.
8119 (convert_cfa_to_fb_loc_list): Adjust.
8120 (maybe_emit_file): Call output_asm_line_debug_info for test.
8121 (dwarf2out_var_location): Reset views as needed. Precompute
8122 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
8123 attribute. Set view.
8124 (new_line_info_table): Reset next view.
8125 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
8126 (dwarf2out_source_line): Likewise. Output view resets and labels to
8127 the assembler, or select appropriate line info opcodes.
8128 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
8129 (optimize_string_length): Catch it. Adjust.
8130 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
8131 dw_val_class_view_list, and remove it if no longer needed.
8132 (hash_loc_list): Hash view numbers.
8133 (loc_list_hasher::equal): Compare them.
8134 (optimize_location_lists): Check whether a view list symbol is
8135 needed, and whether the locview attribute is present, and
8136 whether they match. Remove the locview attribute if no longer
8138 (index_location_lists): Call skip_loc_list_entry for test.
8139 (dwarf2out_finish): Call output_asm_line_debug_info for test.
8140 Use output_dwarf_version.
8141 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
8142 (struct dw_val_node): Add val_view_list.
8143 * final.c (SEEN_NEXT_VIEW): New.
8144 (set_next_view_needed): New.
8145 (clear_next_view_needed): New.
8146 (maybe_output_next_view): New.
8147 (final_start_function): Rename to...
8148 (final_start_function_1): ... this. Take pointer to FIRST,
8149 add SEEN parameter. Emit param bindings in the initial view.
8150 (final_start_function): Reintroduce SEEN-less interface.
8151 (final): Rename to...
8152 (final_1): ... this. Take SEEN parameter. Output final pending
8153 next view at the end.
8154 (final): Reintroduce seen-less interface.
8155 (final_scan_insn): Output pending next view before switching
8156 sections or ending a block. Mark the next view as needed when
8157 outputting variable locations. Notify debug backend of section
8158 changes, and of location view changes.
8159 (rest_of_handle_final): Adjust.
8160 * toplev.c (process_options): Autodetect value for debug variable
8161 location views option. Warn on incompat5 without -gdwarf-5.
8162 * doc/invoke.texi (gvariable-location-views): New.
8163 (gvariable-location-views=incompat5): New.
8164 (gno-variable-location-views): New.
8166 2018-02-08 David Malcolm <dmalcolm@redhat.com>
8168 PR tree-optimization/84136
8169 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
8170 that the result of find_edge is non-NULL.
8172 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8175 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
8176 storing integer register in SImode. Fix cost of 256 and 512
8177 byte aligned SSE register store.
8179 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8181 * config/i386/i386.c (ix86_multiplication_cost): Fix
8182 multiplication cost for TARGET_AVX512DQ.
8184 2018-02-08 Marek Polacek <polacek@redhat.com>
8186 PR tree-optimization/84238
8187 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
8190 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8192 PR tree-optimization/84265
8193 * tree-vect-stmts.c (vectorizable_store): Don't treat
8194 VMAT_CONTIGUOUS accesses as grouped.
8195 (vectorizable_load): Likewise.
8197 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8199 PR tree-optimization/81635
8200 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
8201 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
8202 (test_round_for_mask): New functions.
8203 (wide_int_cc_tests): Call test_round_for_mask.
8204 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
8205 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
8206 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
8207 range returned by get_range_info.
8209 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8212 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
8213 * symtab.c: Include builtins.h
8214 (symtab_node::output_to_lto_symbol_table_p): Move here
8215 from lto-streamer-out.c:output_symbol_p.
8216 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
8217 (output_symbol_p): Move all logic to symtab.c
8218 (produce_symtab): Update.
8220 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8222 * config/s390/s390-opts.h (enum indirect_branch): Define.
8223 * config/s390/s390-protos.h (s390_return_addr_from_memory)
8224 (s390_indirect_branch_via_thunk)
8225 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
8226 (enum s390_indirect_branch_type): Define.
8227 * config/s390/s390.c (struct s390_frame_layout, struct
8228 machine_function): Remove.
8229 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
8230 (indirect_branch_table_label_no, indirect_branch_table_name):
8232 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
8233 (enum s390_indirect_branch_option): Define.
8234 (s390_return_addr_from_memory): New function.
8235 (s390_handle_string_attribute): New function.
8236 (s390_attribute_table): Add new attribute handler.
8237 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
8238 (s390_indirect_branch_via_thunk): New function.
8239 (s390_indirect_branch_via_inline_thunk): New function.
8240 (s390_function_ok_for_sibcall): When jumping via thunk disallow
8241 sibling call optimization for non z10 compiles.
8242 (s390_emit_call): Force indirect branch target to be a single
8243 register. Add r1 clobber for non-z10 compiles.
8244 (s390_emit_epilogue): Emit return jump via return_use expander.
8245 (s390_reorg): Handle JUMP_INSNs as execute targets.
8246 (s390_option_override_internal): Perform validity checks for the
8247 new command line options.
8248 (s390_indirect_branch_attrvalue): New function.
8249 (s390_indirect_branch_settings): New function.
8250 (s390_set_current_function): Invoke s390_indirect_branch_settings.
8251 (s390_output_indirect_thunk_function): New function.
8252 (s390_code_end): Implement target hook.
8253 (s390_case_values_threshold): Implement target hook.
8254 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
8256 * config/s390/s390.h (struct s390_frame_layout)
8257 (struct machine_function): Move here from s390.c.
8258 (TARGET_INDIRECT_BRANCH_NOBP_RET)
8259 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
8260 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
8261 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
8262 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
8263 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
8264 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
8265 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
8266 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
8267 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
8268 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
8269 (mnemonic attribute): Add values which aren't recognized
8271 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
8272 pattern for branch conversion. Fix mnemonic attribute.
8273 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
8274 indirect branch via thunk if requested.
8275 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
8276 ("*indirect_jump"): Disable for branch conversion using out of
8278 ("indirect_jump_via_thunk<mode>_z10")
8279 ("indirect_jump_via_thunk<mode>")
8280 ("indirect_jump_via_inlinethunk<mode>_z10")
8281 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
8282 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
8283 ("casesi_jump_via_inlinethunk<mode>_z10")
8284 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
8285 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
8286 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
8287 ("*indirect2_jump"): Disable for branch conversion.
8288 ("casesi_jump"): Turn into expander and expand patterns for branch
8290 ("return_use"): New expander.
8291 ("*return"): Emit return via thunk and rename it to ...
8292 ("*return<mode>"): ... this one.
8293 * config/s390/s390.opt: Add new options and and enum for the
8296 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8298 * lra-constraints.c (match_reload): Unconditionally use
8299 gen_lowpart_SUBREG, rather than selecting between that
8300 and equivalent gen_rtx_SUBREG code.
8302 2018-02-08 Richard Biener <rguenther@suse.de>
8304 PR tree-optimization/84233
8305 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
8306 changed flag instead of boguously re-using phi_inserted.
8308 2018-02-08 Martin Jambor <mjambor@suse.cz>
8310 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
8311 static local variables.
8313 2018-02-08 Richard Biener <rguenther@suse.de>
8315 PR tree-optimization/84278
8316 * tree-vect-stmts.c (vectorizable_store): When looking for
8317 smaller vector types to perform grouped strided loads/stores
8318 make sure the mode is supported by the target.
8319 (vectorizable_load): Likewise.
8321 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8323 * config/aarch64/aarch64.c (aarch64_components_for_bb):
8324 Increase LDP/STP opportunities by adding adjacent callee-saves.
8326 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8328 PR rtl-optimization/84068
8329 PR rtl-optimization/83459
8330 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
8332 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
8334 PR tree-optimization/84224
8335 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
8336 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
8339 2018-02-07 Iain Sandoe <iain@codesourcery.com>
8342 * config/rs6000/altivec.md (*restore_world): Remove LR use.
8343 * config/rs6000/predicates.md (restore_world_operation): Adjust op
8344 count, remove one USE.
8346 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
8348 * doc/install.texi (Configuration): Document the
8349 --with-long-double-format={ibm,ieee} PowerPC configuration
8353 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
8354 Convert from define_expand to be define_insn_and_split. Rework
8355 float/double/_Float128 conversions to QI/HI/SImode to work with
8356 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
8357 conversions to QI/HImode types did a store and then a load to
8358 truncate the value. For conversions to VSX registers, don't split
8359 the insn, instead emit the code directly. Use the code iterator
8360 any_fix to combine signed and unsigned conversions.
8361 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
8362 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
8363 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8364 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
8365 (fix_<mode>di2_hw): Likewise.
8366 (fixuns_<mode>di2_hw): Likewise.
8367 (fix_<mode>si2_hw): Likewise.
8368 (fixuns_<mode>si2_hw): Likewise.
8369 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
8370 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8371 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
8372 fix<uns>_trunc<SFDF:mode>si2_p8.
8373 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
8375 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
8376 (fix<uns>_<mode>_mem): Likewise.
8377 (fctiw<u>z_<mode>_mem): Likewise.
8378 (fix<uns>_<mode>_mem): Likewise.
8379 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
8380 the register allocator from doing a direct move to the GPRs to do
8381 a store, and instead use the ISA 3.0 store byte/half-word from
8382 vector register instruction. For IEEE 128-bit floating point,
8383 also optimize stores of 32-bit ints.
8384 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
8386 2018-02-07 Alan Hayward <alan.hayward@arm.com>
8388 * genextract.c (push_pathstr_operand): New function to support
8390 (walk_rtx): Call push_pathstr_operand.
8391 (print_path): Support [a-zA-Z].
8393 2018-02-07 Richard Biener <rguenther@suse.de>
8395 PR tree-optimization/84037
8396 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
8397 (cse_and_gimplify_to_preheader): Declare.
8398 (vect_get_place_in_interleaving_chain): Likewise.
8399 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8401 (_loop_vec_info::~_loop_vec_info): Delete it.
8402 (cse_and_gimplify_to_preheader): New function.
8403 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
8404 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
8405 (vectorizable_load): Likewise. For grouped stores always base
8406 the IV on the first element.
8407 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
8408 condition before gimplifying.
8410 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8412 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
8413 *DIV_EXPR and *MOD_EXPR.
8415 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
8418 * config/i386/i386.c (ix86_option_override_internal): Mask out
8419 the CF_SET bit when checking -fcf-protection.
8421 2018-02-07 Tom de Vries <tom@codesourcery.com>
8424 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
8427 2018-02-07 Richard Biener <rguenther@suse.de>
8429 PR tree-optimization/84204
8430 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
8433 PR tree-optimization/84205
8434 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
8435 special-case isl_ast_op_zdiv_r.
8437 PR tree-optimization/84223
8438 * graphite-scop-detection.c (gather_bbs::before_dom_children):
8439 Only add conditions from within the region.
8440 (gather_bbs::after_dom_children): Adjust.
8442 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
8445 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
8446 * config/avr/avr.md: Only post-reload split REG-REG moves if
8447 either register is GENERAL_REG_P.
8449 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8451 PR tree-optimization/84235
8452 * tree-ssa-scopedtables.c
8453 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
8454 if the subtraction is performed in floating point type where NaNs are
8455 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
8456 build 1. Formatting fix.
8458 2018-02-06 Jakub Jelinek <jakub@redhat.com>
8461 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
8462 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
8463 and skip it regardless of bb boundaries. Use CALL_P macro,
8464 don't test INSN_P (insn) together with CALL_P or JUMP_P check
8465 unnecessarily, formatting fix.
8467 2018-02-06 Michael Collison <michael.collison@arm.com>
8469 * config/arm/thumb2.md:
8470 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
8471 (*thumb_mov_notscc): Ditto.
8473 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
8476 * config/rs6000/rs6000.md (su code attribute): Use "u" for
8477 unsigned_fix, not "s".
8479 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8481 * configure.ac (gcc_fn_eh_frame_ro): New function.
8482 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
8483 correct .eh_frame permissions.
8484 * configure: Regenerate.
8486 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
8488 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
8491 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8493 * config/rs6000/rs6000.c (rs6000_option_override_internal):
8494 Display warning message for -mno-speculate-indirect-jumps.
8496 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
8498 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
8500 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
8502 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
8504 PR tree-optimization/84225
8505 * tree-eh.c (find_trapping_overflow): Only call
8506 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
8508 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8511 * config/i386/i386.c: Reimplement the check of possible options
8512 -mibt/-mshstk conbination. Change error messages.
8513 * doc/invoke.texi: Fix a typo: remove extra '='.
8515 2018-02-06 Marek Polacek <polacek@redhat.com>
8517 PR tree-optimization/84228
8518 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
8520 2018-02-06 Tamar Christina <tamar.christina@arm.com>
8523 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
8524 emitted arch directives.
8525 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
8526 __ARM_FEATURE_COPROC before changing architectures.
8528 2018-02-06 Richard Biener <rguenther@suse.de>
8530 * config/i386/i386.c (print_reg): Fix typo.
8531 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
8533 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
8535 * configure: Regenerate.
8537 2018-02-05 Martin Sebor <msebor@redhat.com>
8539 PR tree-optimization/83369
8540 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
8543 2018-02-05 Martin Liska <mliska@suse.cz>
8545 * doc/invoke.texi: Cherry-pick upstream r323995.
8547 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
8549 * ira.c (ira_init_register_move_cost): Adjust comment.
8551 2018-02-05 Martin Liska <mliska@suse.cz>
8553 PR gcov-profile/84137
8554 * doc/gcov.texi: Fix typo in documentation.
8556 2018-02-05 Martin Liska <mliska@suse.cz>
8558 PR gcov-profile/83879
8559 * doc/gcov.texi: Document necessity of --dynamic-list-data when
8560 using dlopen functionality.
8562 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
8564 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
8565 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
8566 _mm_maskz_range_ss, _mm_mask_range_round_ss,
8567 _mm_maskz_range_round_ss): New intrinsics.
8568 (__builtin_ia32_rangesd128_round)
8569 (__builtin_ia32_rangess128_round): Remove.
8570 (__builtin_ia32_rangesd128_mask_round,
8571 __builtin_ia32_rangess128_mask_round): New builtins.
8572 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
8573 __builtin_ia32_rangess128_round): Remove.
8574 (__builtin_ia32_rangesd128_mask_round,
8575 __builtin_ia32_rangess128_mask_round): New builtins.
8576 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
8577 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
8578 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
8579 "<round_saeonly_constraint>")): Changed to ...
8580 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
8581 "<round_saeonly_scalar_constraint>")): ... this.
8582 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
8583 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
8584 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
8585 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
8586 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
8588 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
8590 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
8592 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
8593 Remove all values except native, 8540 and 8548.
8595 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
8597 * config/i386/i386.c (ix86_output_function_return): Pass
8598 INVALID_REGNUM, instead of -1, as invalid register number to
8599 indirect_thunk_name and output_indirect_thunk.
8601 2018-02-02 Julia Koval <julia.koval@intel.com>
8603 * config.gcc: Add -march=icelake.
8604 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
8605 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
8606 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
8607 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
8608 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
8609 (processor_target_table): Add icelake.
8610 (ix86_option_override_internal): Handle new PTAs.
8611 (get_builtin_code_for_version): Handle icelake.
8612 (M_INTEL_COREI7_ICELAKE): New.
8613 (fold_builtin_cpu): Handle icelake.
8614 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
8615 * doc/invoke.texi: Add -march=icelake.
8617 2018-02-02 Julia Koval <julia.koval@intel.com>
8619 * config/i386/i386.c (ix86_option_override_internal): Change flags type
8620 to wide_int_bitmask.
8621 * wide-int-bitmask.h: New.
8623 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8626 * config/i386/i386.md: Replace Pmode with word_mode in
8627 builtin_setjmp_setup and builtin_longjmp to support x32.
8629 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
8633 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
8635 (rs6000_supported_cpu_names): New static variable.
8636 (linux_cpu_translation_table): Likewise.
8637 (elf_platform) <cpu>: Define new static variable and use it.
8638 Translate kernel AT_PLATFORM name to canonical name if needed.
8639 Error if platform name is unknown.
8641 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
8644 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
8646 2018-02-01 Jeff Law <law@redhat.com>
8649 * config/i386/i386.c (release_scratch_register_on_entry): Add new
8650 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
8651 the scratch if RELEASE_VIA_POP is false.
8652 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
8653 If we have to save a temporary register, decrement SIZE appropriately.
8654 Pass new arguments to release_scratch_register_on_entry.
8655 (ix86_adjust_stack_and_probe): Likewise.
8656 (ix86_emit_probe_stack_range): Pass new arguments to
8657 release_scratch_register_on_entry.
8659 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
8661 PR rtl-optimization/84157
8662 * combine.c (change_zero_ext): Use REG_P predicate in
8663 front of HARD_REGISTER_P predicate.
8665 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
8667 * config/avr/avr.c (avr_option_override): Move disabling of
8668 -fdelete-null-pointer-checks to...
8669 * common/config/avr/avr-common.c (avr_option_optimization_table):
8672 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8674 PR tree-optimization/81635
8675 * tree-data-ref.c (split_constant_offset_1): For types that
8676 wrap on overflow, try to use range info to prove that wrapping
8679 2018-02-01 Renlin Li <renlin.li@arm.com>
8682 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
8684 (aarch64_register_move_cost): Likewise.
8685 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
8687 (REG_CLASS_NAMES): Likewise.
8688 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
8689 TAILCALL_ADDR_REGS. Remove IP registers.
8690 * config/aarch64/aarch64.md (Ucs): Update register constraint.
8692 2018-02-01 Richard Biener <rguenther@suse.de>
8694 * domwalk.h (dom_walker::dom_walker): Add additional constructor
8695 for specifying RPO order and allow NULL for that.
8696 * domwalk.c (dom_walker::dom_walker): Likewise.
8697 (dom_walker::walk): Handle NULL RPO order.
8698 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
8700 (rewrite_update_dom_walker): Likewise.
8701 (mark_def_dom_walker): Likewise.
8703 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8705 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
8706 (aarch64_maybe_expand_sve_subreg_move): Declare.
8707 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
8708 * config/aarch64/predicates.md (aarch64_any_register_operand): New
8710 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
8711 that are semantically a reverse operation.
8712 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
8713 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
8714 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
8716 (aarch64_can_change_mode_class): For big-endian, forbid changes
8717 between two SVE modes if they have different element sizes.
8719 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8721 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
8722 the TImode handling for big-endian targets.
8724 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8726 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
8727 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
8729 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
8730 Remove BSWAP handing for big-endian targets and use the form of
8731 LD1RQ appropariate for the mode.
8733 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8735 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
8736 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
8739 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8742 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
8743 check for operands that need to go through aarch64_sve_reload_be.
8745 2018-02-01 Jakub Jelinek <jakub@redhat.com>
8747 PR tree-optimization/81661
8748 PR tree-optimization/84117
8749 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
8750 * tree-eh.c: Include gimplify.h.
8751 (find_trapping_overflow, replace_trapping_overflow,
8752 rewrite_to_non_trapping_overflow): New functions.
8753 * tree-vect-loop.c: Include tree-eh.h.
8754 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
8755 * tree-data-ref.c: Include tree-eh.h.
8756 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
8758 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
8760 PR rtl-optimization/84123
8761 * combine.c (change_zero_ext): Check if hard register satisfies
8762 can_change_dest_mode before calling gen_lowpart_SUBREG.
8764 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
8767 * ira.c (ira_init_register_move_cost): Remove assert.
8769 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8771 PR rtl-optimization/84071
8772 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
8773 * doc/tm.texi: Regenerate.
8775 2018-01-31 Richard Biener <rguenther@suse.de>
8777 PR tree-optimization/84132
8778 * tree-data-ref.c (analyze_miv_subscript): Properly
8779 check whether evolution_function_is_affine_multivariate_p
8780 before calling gcd_of_steps_may_divide_p.
8782 2018-01-31 Julia Koval <julia.koval@intel.com>
8785 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
8786 * config/i386/i386.md (rdpid_rex64) New.
8787 (rdpid): Make 32bit only.
8789 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
8792 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
8793 an IDENTIFIER_NODE for FUNCTION_TYPE's.
8795 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8798 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8800 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8802 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8804 PR rtl-optimization/84071
8805 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
8806 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
8808 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8810 * config/arc/arc.c (arc_handle_aux_attribute): New function.
8811 (arc_attribute_table): Add 'aux' attribute.
8812 (arc_in_small_data_p): Consider aux like variables.
8813 (arc_is_aux_reg_p): New function.
8814 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
8815 (arc_get_aux_arg): New function.
8816 (prepare_move_operands): Handle aux-register access.
8817 (arc_handle_aux_attribute): New function.
8818 * doc/extend.texi (ARC Variable attributes): Add subsection.
8820 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8822 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
8823 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
8824 (arc_attribute_table): Add 'uncached' attribute.
8825 (arc_print_operand): Print '.di' flag for uncached memory
8827 (arc_in_small_data_p): Do not consider for small data the uncached
8829 (arc_is_uncached_mem_p): New function.
8830 * config/arc/predicates.md (compact_store_memory_operand): Check
8831 for uncached memory accesses.
8832 (nonvol_nonimm_operand): Likewise.
8833 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
8835 2018-01-31 Jakub Jelinek <jakub@redhat.com>
8838 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
8839 falign-loops=): Add Optimization flag.
8841 2018-01-30 Jeff Law <law@redhat.com>
8844 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
8845 INT_REGISTERS_SAVED. Check it prior to calling
8846 get_scratch_register_on_entry.
8847 (ix86_adjust_stack_and_probe): Similarly.
8848 (ix86_emit_probe_stack_range): Similarly.
8849 (ix86_expand_prologue): Corresponding changes.
8851 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8854 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
8855 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
8857 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
8860 * lra-constraints.c (curr_insn_transform): Process AND in the
8863 2018-01-30 Jakub Jelinek <jakub@redhat.com>
8865 PR rtl-optimization/83986
8866 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
8867 dependence against last_pending_memory_flush in addition to
8870 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
8872 PR tree-optimization/81611
8873 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
8876 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8879 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
8882 2018-01-30 Richard Biener <rguenther@suse.de>
8883 Jakub Jelinek <jakub@redhat.com>
8885 PR tree-optimization/84111
8886 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
8887 inner loops added during recursion, as they don't have up-to-date
8890 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8893 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
8894 (can_inline_edge_by_limits_p): ... here.
8895 (can_early_inline_edge_p, check_callers,
8896 update_caller_keys, update_callee_keys, recursive_inlining,
8897 add_new_edges_to_heap, speculation_useful_p,
8898 inline_small_functions,
8899 inline_small_functions, flatten_function,
8900 inline_to_all_callers_1): Update.
8902 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8904 * profile-count.c (profile_count::combine_with_ipa_count): Handle
8907 2018-01-30 Richard Biener <rguenther@suse.de>
8909 PR tree-optimization/83008
8910 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
8911 invariant and constant vector uses in stmts when they need
8914 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8917 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
8918 * configure: Regenerate.
8920 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8922 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
8924 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
8925 Use gen_rtx_REG rather than gen_lowpart.
8927 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8929 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
8930 rather than 0 when creating partial subregs.
8932 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8934 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
8937 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
8940 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
8941 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
8942 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
8943 flags. This restores the settings used before the 2017-07-24.
8944 Turning off pre increment/decrement/modify allows IVOPTS to
8945 optimize DF/SF loops where the index is an int.
8947 2018-01-29 Richard Biener <rguenther@suse.de>
8948 Kelvin Nilsen <kelvin@gcc.gnu.org>
8951 * tree-vect-stmts.c (vectorizable_call): Don't call
8952 targetm.vectorize_builtin_md_vectorized_function if callee is
8955 2018-01-22 Carl Love <cel@us.ibm.com>
8957 * doc/extend.tex: Fix typo in second arg in
8958 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
8960 2018-01-29 Richard Biener <rguenther@suse.de>
8962 PR tree-optimization/84086
8963 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
8964 (flush_ssaname_freelist): When SSA names were released reset
8965 the SCEV hash table.
8967 2018-01-29 Richard Biener <rguenther@suse.de>
8969 PR tree-optimization/84057
8970 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
8971 removed paths when removing edges.
8973 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
8975 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
8976 -mfunction-return=@var{choice}.
8978 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
8981 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
8983 (layout::print_source_line): Likewise.
8984 (test_get_line_width_without_trailing_whitespace): Add test cases.
8986 2018-01-27 Jakub Jelinek <jakub@redhat.com>
8989 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
8992 2018-01-26 Jim Wilson <jimw@sifive.com>
8994 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
8996 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
8999 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9001 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
9002 and CMP + SUB-immediate -> SUBS.
9004 2018-01-26 Martin Sebor <msebor@redhat.com>
9006 PR tree-optimization/83896
9007 * tree-ssa-strlen.c (get_string_len): Rename...
9008 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
9009 Avoid assuming length is constant.
9010 (handle_char_store): Use HOST_WIDE_INT for string length.
9012 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
9015 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
9016 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
9018 2018-01-26 Richard Biener <rguenther@suse.de>
9020 PR rtl-optimization/84003
9021 * dse.c (record_store): Only record redundant stores when
9022 the earlier store aliases at least all accesses the later one does.
9024 2018-01-26 Jakub Jelinek <jakub@redhat.com>
9026 PR rtl-optimization/83985
9027 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
9028 REG_CFA_RESTORE insns.
9029 (delete_unmarked_insns): Don't ignore separate shrink wrapping
9030 REG_CFA_RESTORE insns here.
9033 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
9034 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
9036 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9038 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
9039 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
9040 (arc_init): Likewise.
9041 (arc_override_options): Likewise.
9042 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
9044 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
9046 * config/arc/arc.h (TARGET_DBNZ): Define.
9047 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
9048 properly set the tune attribute.
9049 (dbnz): Use TARGET_DBNZ guard.
9050 * config/arc/arc.opt (mtune): Add core3 option.
9052 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9054 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
9055 recognize new pic like addresses.
9056 (arc_delegitimize_address): Clean up.
9058 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9060 * config/arc/arc-arches.def: Option mrf16 valid for all
9062 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
9063 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
9064 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
9065 * config/arc/arc-tables.opt: Regenerate.
9066 * config/arc/arc.c (arc_conditional_register_usage): Handle
9067 reduced register file case.
9068 (arc_file_start): Set must have build attributes.
9069 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
9071 * config/arc/arc.opt (mrf16): Add new option.
9072 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
9073 * config/arc/genmultilib.awk: Handle new mrf16 option.
9074 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
9075 * config/arc/t-multilib: Regenerate.
9076 * doc/invoke.texi (ARC Options): Document mrf16 option.
9078 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9080 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
9081 * config/arc/arc.c (arc_handle_secure_attribute): New function.
9082 (arc_attribute_table): Add 'secure_call' attribute.
9083 (arc_print_operand): Print secure call operand.
9084 (arc_function_ok_for_sibcall): Don't optimize tail calls when
9086 (arc_is_secure_call_p): New function. * config/arc/arc.md
9087 (call_i): Add support for sjli instruction.
9088 (call_value_i): Likewise.
9089 * config/arc/constraints.md (Csc): New constraint.
9091 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9092 John Eric Martin <John.Martin@emmicro-us.com>
9094 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
9095 * config/arc/arc.c (_arc_jli_section): New struct.
9096 (arc_jli_section): New type.
9097 (rc_jli_sections): New static variable.
9098 (arc_handle_jli_attribute): New function.
9099 (arc_attribute_table): Add jli_always and jli_fixed attribute.
9100 (arc_file_end): New function.
9101 (TARGET_ASM_FILE_END): Define.
9102 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
9103 (arc_add_jli_section): New function.
9104 (jli_call_scan): Likewise.
9105 (arc_reorg): Call jli_call_scan.
9106 (arc_output_addsi): Remove 'S' from printing asm operand.
9107 (arc_is_jli_call_p): New function.
9108 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
9110 (movhi_insn): Likewise.
9111 (movsi_insn): Likewise.
9112 (movsi_set_cc_insn): Likewise.
9113 (loadqi_update): Likewise.
9114 (load_zeroextendqisi_update): Likewise.
9115 (load_signextendqisi_update): Likewise.
9116 (loadhi_update): Likewise.
9117 (load_zeroextendhisi_update): Likewise.
9118 (load_signextendhisi_update): Likewise.
9119 (loadsi_update): Likewise.
9120 (loadsf_update): Likewise.
9121 (movsicc_insn): Likewise.
9122 (bset_insn): Likewise.
9123 (bxor_insn): Likewise.
9124 (bclr_insn): Likewise.
9125 (bmsk_insn): Likewise.
9126 (bicsi3_insn): Likewise.
9127 (cmpsi_cc_c_insn): Likewise.
9128 (movsi_ne): Likewise.
9129 (movsi_cond_exec): Likewise.
9130 (clrsbsi2): Likewise.
9141 (call_i): Remove 'S' asm letter, add jli instruction.
9142 (call_value_i): Likewise.
9143 * config/arc/arc.op (mjli-always): New option.
9144 * config/arc/constraints.md (Cji): New constraint.
9145 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
9147 (subsf3_fpx): Likewise.
9148 (mulsf3_fpx): Likewise.
9149 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
9151 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
9153 * doc/invoke.texi (ARC): Document mjli-always option.
9155 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
9157 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
9158 avoid addition with 0 and use incw and decw where possible.
9160 2018-01-26 Richard Biener <rguenther@suse.de>
9162 PR tree-optimization/81082
9163 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
9164 association if it requires casting to unsigned.
9165 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
9166 from fold_plusminus_mult_expr to catch important cases late when
9167 range info is available.
9169 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9171 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
9172 * configure.ac (hidden_linkonce): New test.
9173 * configure: Regenerate.
9174 * config.in: Regenerate.
9176 2018-01-26 Julia Koval <julia.koval@intel.com>
9178 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
9179 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
9180 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
9181 _mm_mask_bitshuffle_epi64_mask): Fix type.
9182 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
9183 USI_FTYPE_V4DI_V4DI_USI): Remove.
9184 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
9185 __builtin_ia32_vpshufbitqmb256_mask,
9186 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
9187 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
9188 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
9190 2018-01-26 Alan Modra <amodra@gmail.com>
9193 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
9194 UNSPEC_VBPERMQ. Sort other unspecs.
9196 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
9198 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
9200 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
9203 * predict.c (drop_profile): Do not push/pop cfun; update also
9205 (handle_missing_profiles): Fix logic looking for zero profiles.
9207 2018-01-25 Jakub Jelinek <jakub@redhat.com>
9210 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
9211 on functions with #pragma omp declare simd or functions with simd
9213 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
9214 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
9215 Remove trailing \n from warning_at calls.
9217 2018-01-25 Tom de Vries <tom@codesourcery.com>
9220 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9221 for neutered workers.
9223 2018-01-24 Joseph Myers <joseph@codesourcery.com>
9226 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
9227 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
9229 2018-01-24 Jeff Law <law@redhat.com>
9232 * i386.c (get_probe_interval): Move to earlier point.
9233 (ix86_compute_frame_layout): If -fstack-clash-protection and
9234 the frame is larger than the probe interval, then use pushes
9235 to save registers rather than reg->mem moves.
9236 (ix86_expand_prologue): Remove conditional for int_registers_saved
9239 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
9242 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
9243 min/max for never referenced object.
9245 2018-01-24 Jakub Jelinek <jakub@redhat.com>
9248 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
9250 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
9251 attributes from DECL_ATTRIBUTES (decl) without affecting
9252 DECL_ATTRIBUTES (current_function_decl).
9253 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
9254 functions with non-NULL DECL_ABSTRACT_ORIGIN.
9256 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
9258 PR tree-optimization/83979
9259 * fold-const.c (fold_comparison): Use constant_boolean_node
9260 instead of boolean_{true,false}_node.
9262 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
9264 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
9267 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9269 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9270 Simplify the clause that sets the length attribute.
9271 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9272 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
9273 clause that sets the length attribute.
9274 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9276 2018-01-24 Tom de Vries <tom@codesourcery.com>
9279 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
9280 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
9281 Add strict parameter.
9282 (prevent_branch_around_nothing): Insert dummy insn between branch to
9283 label and label with no ptx insn inbetween.
9284 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
9286 2018-01-24 Tom de Vries <tom@codesourcery.com>
9289 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9290 for neutered threads in warp.
9291 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
9293 2018-01-24 Richard Biener <rguenther@suse.de>
9295 PR tree-optimization/83176
9296 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
9299 2018-01-24 Richard Biener <rguenther@suse.de>
9301 PR tree-optimization/82819
9302 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
9303 code generating pluses that are no-ops in the target precision.
9305 2018-01-24 Richard Biener <rguenther@suse.de>
9308 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
9310 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9312 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
9313 to merge probabilities.
9314 * predict.c (probably_never_executed): Also mark as cold functions
9315 with global 0 profile and guessed local profile.
9316 * profile-count.c (profile_probability::combine_with_count): New
9318 * profile-count.h (profile_probability::operator*,
9319 profile_probability::operator*=, profile_probability::operator/,
9320 profile_probability::operator/=): Reduce precision to adjusted
9321 and set value to guessed on contradictory divisions.
9322 (profile_probability::combine_with_freq): Remove.
9323 (profile_probability::combine_wiht_count): Declare.
9324 (profile_count::force_nonzero):: Set to adjusted.
9325 (profile_count::probability_in):: Set quality to adjusted.
9326 * tree-ssa-tail-merge.c (replace_block_by): Use
9329 2018-01-23 Andrew Waterman <andrew@sifive.com>
9330 Jim Wilson <jimw@sifive.com>
9332 * config/riscv/riscv.c (riscv_stack_boundary): New.
9333 (riscv_option_override): Set riscv_stack_boundary. Handle
9334 riscv_preferred_stack_boundary_arg.
9335 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
9336 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
9337 (STACK_BOUNDARY): Set to riscv_stack_boundary.
9338 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
9339 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
9340 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
9342 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
9345 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
9346 of struct ix86_frame.
9347 (ix86_expand_epilogue): Likewise. Add a local variable for
9348 the reg_save_offset field in struct ix86_frame.
9350 2018-01-23 Bin Cheng <bin.cheng@arm.com>
9352 PR tree-optimization/82604
9353 * tree-loop-distribution.c (enum partition_kind): New enum item
9354 PKIND_PARTIAL_MEMSET.
9355 (partition_builtin_p): Support above new enum item.
9356 (generate_code_for_partition): Ditto.
9357 (compute_access_range): Differentiate cases that equality can be
9358 proven at all loops, the innermost loops or no loops.
9359 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
9360 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
9361 (finalize_partitions, distribute_loop): Don't fuse partition of
9362 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
9363 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
9366 2018-01-23 Martin Liska <mliska@suse.cz>
9368 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
9369 order to ignore the predictor.
9370 (PRED_POLYMORPHIC_CALL): Likewise.
9371 (PRED_RECURSIVE_CALL): Likewise.
9373 2018-01-23 Martin Liska <mliska@suse.cz>
9375 * tree-profile.c (tree_profiling): Print function header to
9376 aware reader which function we are working on.
9377 * value-prof.c (gimple_find_values_to_profile): Do not print
9378 not interesting value histograms.
9380 2018-01-23 Martin Liska <mliska@suse.cz>
9382 * profile-count.h (enum profile_quality): Add
9383 profile_uninitialized as the first value. Do not number values
9384 as they are zero based.
9385 (profile_count::verify): Update sanity check.
9386 (profile_probability::verify): Likewise.
9388 2018-01-23 Nathan Sidwell <nathan@acm.org>
9390 * doc/invoke.texi (ffor-scope): Deprecate.
9392 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9394 PR tree-optimization/83510
9395 * domwalk.c (set_all_edges_as_executable): New function.
9396 (dom_walker::dom_walker): Convert bool param
9397 "skip_unreachable_blocks" to enum reachability. Move setup of
9398 edge flags to set_all_edges_as_executable and only do it when
9399 reachability is REACHABLE_BLOCKS.
9400 * domwalk.h (enum dom_walker::reachability): New enum.
9401 (dom_walker::dom_walker): Convert bool param
9402 "skip_unreachable_blocks" to enum reachability.
9403 (set_all_edges_as_executable): New decl.
9404 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
9405 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
9407 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
9408 but converting true to REACHABLE_BLOCKS.
9409 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
9411 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
9412 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
9413 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
9415 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
9416 if check_all_array_refs will be called.
9418 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9420 * tree.c (selftest::test_location_wrappers): Add more test
9423 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9425 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
9426 (selftest::test_bit_in_range): Likewise.
9428 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9431 * doc/sourcebuild.texi (vect_float): Say that the selector
9432 only describes the situation when -funsafe-math-optimizations is on.
9433 (vect_float_strict): Document.
9435 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9437 PR tree-optimization/83965
9438 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
9439 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
9440 instead of checking only for a reduction.
9441 (vect_recog_widen_sum_pattern): Likewise.
9443 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9445 * predict.c (probably_never_executed): Only use precise profile info.
9446 (compute_function_frequency): Skip after inlining hack since we now
9447 have quality checking.
9449 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9451 * profile-count.h (profile_probability::very_unlikely,
9452 profile_probability::unlikely, profile_probability::even): Set
9453 precision to guessed.
9455 2018-01-23 Richard Biener <rguenther@suse.de>
9457 PR tree-optimization/83963
9458 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
9459 Properly terminate dominator walk when crossing the exit edge not
9460 when visiting its source block.
9462 2018-01-23 Jakub Jelinek <jakub@redhat.com>
9465 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
9466 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
9468 2018-01-22 Jakub Jelinek <jakub@redhat.com>
9470 PR tree-optimization/83957
9471 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
9472 semicolon after for body surrounded by braces.
9474 PR tree-optimization/83081
9475 * profile-count.h (profile_probability::split): New method.
9476 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
9477 Use profile_probability::split.
9478 (do_compare_rtx_and_jump): Fix adjustment of probabilities
9479 when splitting a single conditional jump into 2.
9481 2018-01-22 David Malcolm <dmalcolm@redhat.com>
9483 PR tree-optimization/69452
9484 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
9487 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9489 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
9490 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
9491 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
9493 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9495 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
9496 * config/rl78/rl78.md (movdi): New define_expand.
9497 * config/rl78/rl78.c (rl78_split_movdi): New function.
9499 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
9502 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
9504 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
9505 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
9506 128-bit to produce an UNSPEC move to get the double word with the
9507 signbit and then a shift directly to do signbit.
9508 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
9509 implementation with a new version that just does either a direct
9510 move or a regular move. Move memory interface to separate insns.
9511 Move insns so they are next to the expander.
9512 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
9513 with signbit move. Split big and little endian case.
9514 (signbit<mode>2_dm_mem_le): Likewise.
9515 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
9516 (signbit<mode>2_dm2): Likewise.
9518 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9520 * config/rl78/rl78.md (anddi3): New define_expand.
9522 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9524 * config/rl78/rl78.md (umindi3): New define_expand.
9526 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9528 * config/rl78/rl78.md (smindi3): New define_expand.
9530 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9532 * config/rl78/rl78.md (smaxdi3): New define_expand.
9534 2018-01-22 Carl Love <cel@us.ibm.com>
9536 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
9537 LVX_V1TI): Add macro expansion.
9538 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
9539 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
9540 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
9541 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
9542 Change check to determine if the instruction is a byte reversing
9543 entry. Fix typo in comment.
9544 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
9545 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
9546 Add def_builtin calls for new builtins.
9547 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
9548 Add define_insn expansion.
9550 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9552 * config/rl78/rl78.md (umaxdi3): New define_expand.
9554 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9556 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
9557 for non-QImode registers.
9559 2018-01-22 Richard Biener <rguenther@suse.de>
9561 PR tree-optimization/83963
9562 * graphite-scop-detection.c (scop_detection::get_sese): Delay
9563 including the loop exit block.
9564 (scop_detection::merge_sese): Likewise.
9565 (scop_detection::add_scop): Do it here instead.
9567 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9569 * doc/sourcebuild.texi (arm_softfloat): Document.
9571 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
9574 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
9575 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
9576 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
9578 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9579 David Edelsohn <dje.gcc@gmail.com>
9582 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9583 Change "crset eq" to "crset 2".
9584 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9585 (*call_indirect_aix<mode>_nospec): Likewise.
9586 (*call_value_indirect_aix<mode>_nospec): Likewise.
9587 (*call_indirect_elfv2<mode>_nospec): Likewise.
9588 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
9589 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
9590 change assembly output from . to $.
9591 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9592 (indirect_jump<mode>_nospec): Change assembly output from . to $.
9593 (*tablejump<mode>_internal1_nospec): Likewise.
9595 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
9598 * config/sh/sh_optimize_sett_clrt.cc:
9599 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
9601 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
9603 PR tree-optimization/83940
9604 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
9605 offset_dt to vect_constant_def rather than vect_unknown_def_type.
9606 (vect_check_load_store_mask): Add a mask_dt_out parameter and
9607 use it to pass back the definition type.
9608 (vect_check_store_rhs): Likewise rhs_dt_out.
9609 (vect_build_gather_load_calls): Add a mask_dt argument and use
9610 it instead of a call to vect_is_simple_use.
9611 (vectorizable_store): Update calls to vect_check_load_store_mask
9612 and vect_check_store_rhs. Use the dt returned by the latter instead
9613 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
9614 instead of calls to vect_is_simple_use. Pass the scalar rather
9615 than the vector operand to vect_is_simple_use when handling
9616 second and subsequent copies of an rhs value.
9617 (vectorizable_load): Update calls to vect_check_load_store_mask
9618 and vect_build_gather_load_calls. Use the cached mask_dt and
9619 gs_info.offset_dt instead of calls to vect_is_simple_use.
9621 2018-01-20 Jakub Jelinek <jakub@redhat.com>
9624 * tree-emutls.c: Include gimplify.h.
9625 (lower_emutls_2): New function.
9626 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
9627 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
9628 it before further processing.
9631 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
9632 UINTVAL (trueop1) instead of INTVAL (op1).
9634 2018-01-19 Jakub Jelinek <jakub@redhat.com>
9638 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
9639 INCOMING_FRAME_SP_OFFSET if not defined.
9640 (scan_trace): Add ENTRY argument. If true and
9641 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
9642 emit a note to adjust the CFA offset.
9643 (create_cfi_notes): Adjust scan_trace callers.
9644 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
9645 INCOMING_FRAME_SP_OFFSET in the CIE.
9646 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
9647 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
9649 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
9650 * doc/tm.texi: Regenerated.
9652 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9654 PR rtl-optimization/83147
9655 * lra-constraints.c (remove_inheritance_pseudos): Use
9656 lra_substitute_pseudo_within_insn.
9658 2018-01-19 Tom de Vries <tom@codesourcery.com>
9659 Cesar Philippidis <cesar@codesourcery.com>
9662 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
9664 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
9667 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
9668 spaces for function labels.
9670 2018-01-19 Martin Liska <mliska@suse.cz>
9672 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
9673 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
9674 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
9675 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
9676 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
9677 (PRED_CONST_RETURN): Change from 69 to 65.
9678 (PRED_NULL_RETURN): Change from 91 to 71.
9679 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
9680 (PRED_LOOP_GUARD): Change from 66 to 73.
9682 2018-01-19 Martin Liska <mliska@suse.cz>
9684 * predict.c (predict_insn_def): Add new assert.
9685 (struct branch_predictor): Change type to signed integer.
9686 (test_prediction_value_range): Amend test to cover
9688 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
9689 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
9690 (PRED_LOOP_ITERATIONS_MAX): Likewise.
9691 (PRED_LOOP_IV_COMPARE): Likewise.
9692 * predict.h (PROB_UNINITIALIZED): Define new constant.
9694 2018-01-19 Martin Liska <mliska@suse.cz>
9696 * predict.c (dump_prediction): Add new format for
9697 analyze_brprob.py script which is enabled with -details
9699 * profile-count.h (precise_p): New function.
9701 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9703 PR tree-optimization/83922
9704 * tree-vect-loop.c (vect_verify_full_masking): Return false if
9705 there are no statements that need masking.
9706 (vect_active_double_reduction_p): New function.
9707 (vect_analyze_loop_operations): Use it when handling phis that
9708 are not in the loop header.
9710 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9712 PR tree-optimization/83914
9713 * tree-vect-loop.c (vectorizable_induction): Don't convert
9714 init_expr or apply the peeling adjustment for inductions
9715 that are nested within the vectorized loop.
9717 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9719 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
9722 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9726 * function.h (gimplify_parameters): Add gimple_seq * argument.
9727 * function.c: Include gimple.h and options.h.
9728 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
9729 for the added local temporaries if needed.
9730 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
9731 if there are any parameter cleanups, wrap whole body into a
9732 try/finally with the cleanups.
9734 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
9737 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
9738 Use GET_MODE_CLASS for scalar floating point.
9740 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9744 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
9745 Fix call of call_cgraph_insertion_hooks.
9747 2018-01-18 Martin Sebor <msebor@redhat.com>
9749 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
9751 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9754 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
9757 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
9760 * common.opt: (-ffile-prefix-map): New option.
9761 * opts.c (common_handle_option): Defer it.
9762 * opts-global.c (handle_common_deferred_options): Handle it.
9763 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
9764 * file-prefix-map.h: New file.
9765 (remap_debug_filename, add_debug_prefix_map): ...here.
9766 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
9767 * final.c (debug_prefix_map, add_debug_prefix_map
9768 remap_debug_filename): Move to...
9769 * file-prefix-map.c: New file.
9770 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
9771 generalize, get rid of alloca(), use strrchr() instead of strchr().
9772 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
9773 Implement in terms of add_prefix_map().
9774 (remap_macro_filename, remap_debug_filename): Implement in term of
9776 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
9777 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
9778 * dbxout.c: Include file-prefix-map.h.
9779 * varasm.c: Likewise.
9780 * vmsdbgout.c: Likewise.
9781 * xcoffout.c: Likewise.
9782 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
9783 * doc/cppopts.texi (-fmacro-prefix-map): Document.
9784 * doc/invoke.texi (-ffile-prefix-map): Document.
9785 (-fdebug-prefix-map): Update description.
9787 2018-01-18 Martin Liska <mliska@suse.cz>
9789 * config/i386/i386.c (indirect_thunk_name): Document that also
9791 (output_indirect_thunk): Document why both instructions
9792 (pause and lfence) are generated.
9794 2018-01-18 Richard Biener <rguenther@suse.de>
9796 PR tree-optimization/83887
9797 * graphite-scop-detection.c
9798 (scop_detection::get_nearest_dom_with_single_entry): Remove.
9799 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
9800 (scop_detection::merge_sese): Re-implement with a flood-fill
9801 algorithm that properly finds a SESE region if it exists.
9803 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9806 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
9807 pointer_diff optimizations use view_convert instead of convert.
9809 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9811 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9812 Generate different code for -mno-speculate-indirect-jumps.
9813 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9814 (*call_indirect_aix<mode>): Disable for
9815 -mno-speculate-indirect-jumps.
9816 (*call_indirect_aix<mode>_nospec): New define_insn.
9817 (*call_value_indirect_aix<mode>): Disable for
9818 -mno-speculate-indirect-jumps.
9819 (*call_value_indirect_aix<mode>_nospec): New define_insn.
9820 (*sibcall_nonlocal_sysv<mode>): Generate different code for
9821 -mno-speculate-indirect-jumps.
9822 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9824 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
9826 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
9827 long double type, set the flags for noting the default long double
9828 type, even if we don't pass or return a long double type.
9830 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
9833 * ipa-inline.c (flatten_function): Do not overwrite final inlining
9836 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
9838 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
9839 support for merge[hl].
9840 (fold_mergehl_helper): New helper function.
9841 (tree-vector-builder.h): New #include for tree_vector_builder usage.
9842 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
9843 (altivec_vmrglw_direct): Add xxmrglw insn.
9845 2018-01-17 Andrew Waterman <andrew@sifive.com>
9847 * config/riscv/riscv.c (riscv_conditional_register_usage): If
9848 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
9850 2018-01-17 David Malcolm <dmalcolm@redhat.com>
9853 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
9854 call the lto_location_cache before reading the
9855 DECL_SOURCE_LOCATION of the types.
9857 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
9858 Richard Sandiford <richard.sandiford@linaro.org>
9860 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
9861 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
9862 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
9863 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
9864 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
9866 * config/aarch64/constraints.md (aarch64_movti_operand):
9868 * config/aarch64/predicates.md (Uti): Add new constraint.
9870 2018-01-17 Carl Love <cel@us.ibm.com>
9872 * config/rs6000/vsx.md (define_expand xl_len_r,
9873 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
9874 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
9876 (define_expand, define_insn): Move the shift left from the
9877 define_insn to the define_expand for lxvl and stxvl instructions.
9878 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
9879 and XL_LEN_R definitions to PURE.
9881 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
9883 * config/i386/i386.c (indirect_thunk_name): Declare regno
9884 as unsigned int. Compare regno with INVALID_REGNUM.
9885 (output_indirect_thunk): Ditto.
9886 (output_indirect_thunk_function): Ditto.
9887 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
9888 in the call to output_indirect_thunk_function.
9890 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
9893 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
9894 rather than the size of inner_type to determine the stack slot size
9895 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
9897 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
9900 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
9903 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
9905 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
9906 endian Linux systems to optionally enable multilibs for selecting
9907 the long double type if the user configured an explicit type.
9908 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
9909 have no long double multilibs if not defined.
9910 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
9911 warn if the user used -mabi={ieee,ibm}longdouble and we built
9912 multilibs for long double.
9913 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
9914 appropriate multilib option.
9915 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
9917 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
9918 for building long double multilibs.
9919 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
9921 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
9923 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
9926 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
9928 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
9931 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
9934 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
9937 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9939 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
9940 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
9942 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
9944 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
9945 different rtl trees depending on TARGET_64BIT.
9946 (rs6000_gen_lvx): Likewise.
9948 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
9950 * config/visium/visium.md (nop): Tweak comment.
9951 (hazard_nop): Likewise.
9953 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9955 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
9956 -mspeculate-indirect-jumps.
9957 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
9958 for -mno-speculate-indirect-jumps.
9959 (*call_indirect_elfv2<mode>_nospec): New define_insn.
9960 (*call_value_indirect_elfv2<mode>): Disable for
9961 -mno-speculate-indirect-jumps.
9962 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
9963 (indirect_jump): Emit different RTL for
9964 -mno-speculate-indirect-jumps.
9965 (*indirect_jump<mode>): Disable for
9966 -mno-speculate-indirect-jumps.
9967 (*indirect_jump<mode>_nospec): New define_insn.
9968 (tablejump): Emit different RTL for
9969 -mno-speculate-indirect-jumps.
9970 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
9971 (tablejumpsi_nospec): New define_expand.
9972 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
9973 (tablejumpdi_nospec): New define_expand.
9974 (*tablejump<mode>_internal1): Disable for
9975 -mno-speculate-indirect-jumps.
9976 (*tablejump<mode>_internal1_nospec): New define_insn.
9977 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
9980 2018-01-16 Artyom Skrobov tyomitch@gmail.com
9982 * caller-save.c (insert_save): Drop unnecessary parameter. All
9985 2018-01-16 Jakub Jelinek <jakub@redhat.com>
9986 Richard Biener <rguenth@suse.de>
9989 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
9990 return early, inline manually is_gimple_sizepos. Make sure if we
9991 call gimplify_expr we don't end up with a gimple constant.
9992 * tree.c (variably_modified_type_p): Don't return true for
9993 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
9994 * gimplify.h (is_gimple_sizepos): Remove.
9996 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
9998 PR tree-optimization/83857
9999 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
10000 vectorizable_live_operation for pure SLP statements.
10001 (vectorizable_live_operation): Handle PHIs.
10003 2018-01-16 Richard Biener <rguenther@suse.de>
10005 PR tree-optimization/83867
10006 * tree-vect-stmts.c (vect_transform_stmt): Precompute
10007 nested_in_vect_loop_p since the scalar stmt may get invalidated.
10009 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10012 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
10013 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
10014 If off is not INTEGER_CST, issue a may not be aligned warning
10015 rather than isn't aligned. Use isn%'t rather than isn't.
10016 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
10018 <case MULT_EXPR>: Improve the case when bottom and one of the
10019 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
10020 operand, in that case check if the other operand is multiple of
10021 bottom divided by the INTEGER_CST operand.
10023 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10026 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
10027 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
10028 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
10029 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
10030 * config/pa/pa.c (pa_function_arg_advance): Likewise.
10031 (pa_function_arg, pa_arg_partial_bytes): Likewise.
10032 (pa_function_arg_size): New function.
10034 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10036 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
10037 in a separate statement.
10039 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10041 PR tree-optimization/83847
10042 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
10043 group gathers and scatters.
10045 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10047 PR rtl-optimization/86620
10048 * params.def (max-sched-ready-insns): Bump minimum value to 1.
10050 PR rtl-optimization/83213
10051 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
10052 to last if both are JUMP_INSNs.
10054 PR tree-optimization/83843
10055 * gimple-ssa-store-merging.c
10056 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
10057 store_immediate_info for bswap/nop orig_stores.
10059 2018-01-15 Andrew Waterman <andrew@sifive.com>
10061 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
10063 <UDIV>: Increase cost if !TARGET_DIV.
10065 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
10067 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
10068 (define_attr "cr_logical_3op"): New.
10069 (cceq_ior_compare): Adjust.
10070 (cceq_ior_compare_complement): Adjust.
10071 (*cceq_rev_compare): Adjust.
10072 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
10073 (is_cracked_insn): Adjust.
10074 (insn_must_be_first_in_group): Adjust.
10075 * config/rs6000/40x.md: Adjust.
10076 * config/rs6000/440.md: Adjust.
10077 * config/rs6000/476.md: Adjust.
10078 * config/rs6000/601.md: Adjust.
10079 * config/rs6000/603.md: Adjust.
10080 * config/rs6000/6xx.md: Adjust.
10081 * config/rs6000/7450.md: Adjust.
10082 * config/rs6000/7xx.md: Adjust.
10083 * config/rs6000/8540.md: Adjust.
10084 * config/rs6000/cell.md: Adjust.
10085 * config/rs6000/e300c2c3.md: Adjust.
10086 * config/rs6000/e500mc.md: Adjust.
10087 * config/rs6000/e500mc64.md: Adjust.
10088 * config/rs6000/e5500.md: Adjust.
10089 * config/rs6000/e6500.md: Adjust.
10090 * config/rs6000/mpc.md: Adjust.
10091 * config/rs6000/power4.md: Adjust.
10092 * config/rs6000/power5.md: Adjust.
10093 * config/rs6000/power6.md: Adjust.
10094 * config/rs6000/power7.md: Adjust.
10095 * config/rs6000/power8.md: Adjust.
10096 * config/rs6000/power9.md: Adjust.
10097 * config/rs6000/rs64.md: Adjust.
10098 * config/rs6000/titan.md: Adjust.
10100 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10102 * config/i386/predicates.md (indirect_branch_operand): Rewrite
10103 ix86_indirect_branch_register logic.
10105 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10107 * config/i386/constraints.md (Bs): Update
10108 ix86_indirect_branch_register check. Don't check
10109 ix86_indirect_branch_register with GOT_memory_operand.
10111 * config/i386/predicates.md (GOT_memory_operand): Don't check
10112 ix86_indirect_branch_register here.
10113 (GOT32_symbol_operand): Likewise.
10115 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10117 * config/i386/predicates.md (constant_call_address_operand):
10118 Rewrite ix86_indirect_branch_register logic.
10119 (sibcall_insn_operand): Likewise.
10121 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10123 * config/i386/constraints.md (Bs): Replace
10124 ix86_indirect_branch_thunk_register with
10125 ix86_indirect_branch_register.
10127 * config/i386/i386.md (indirect_jump): Likewise.
10128 (tablejump): Likewise.
10129 (*sibcall_memory): Likewise.
10130 (*sibcall_value_memory): Likewise.
10131 Peepholes of indirect call and jump via memory: Likewise.
10132 * config/i386/i386.opt: Likewise.
10133 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10134 (GOT_memory_operand): Likewise.
10135 (call_insn_operand): Likewise.
10136 (sibcall_insn_operand): Likewise.
10137 (GOT32_symbol_operand): Likewise.
10139 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10141 PR middle-end/83837
10142 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
10143 type rather than type addr's type points to.
10144 (expand_omp_atomic_mutex): Likewise.
10145 (expand_omp_atomic): Likewise.
10147 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10150 * config/i386/i386.c (output_indirect_thunk_function): Use
10151 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
10152 for __x86_return_thunk.
10154 2018-01-15 Richard Biener <rguenther@suse.de>
10156 PR middle-end/83850
10157 * expmed.c (extract_bit_field_1): Fix typo.
10159 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10162 * config/arm/iterators.md (VF): New mode iterator.
10163 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
10164 Remove integer-related logic from pattern.
10165 (neon_vabd<mode>_3): Likewise.
10167 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10169 PR middle-end/82694
10170 * common.opt (fstrict-overflow): No longer an alias.
10171 (fwrapv-pointer): New option.
10172 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
10173 also for pointer types based on flag_wrapv_pointer.
10174 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
10175 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
10176 opts->x_flag_wrapv got set.
10177 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
10178 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
10179 POINTER_TYPE_OVERFLOW_UNDEFINED.
10180 * match.pd: Likewise in address comparison pattern.
10181 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
10183 2018-01-15 Richard Biener <rguenther@suse.de>
10186 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
10187 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
10188 Reset type names to their identifier if their TYPE_DECL doesn't
10189 have linkage (and thus is used for ODR and devirt).
10190 (save_debug_info_for_decl): Remove.
10191 (save_debug_info_for_type): Likewise.
10192 (add_tree_to_fld_list): Adjust.
10193 * tree-pretty-print.c (dump_generic_node): Make dumping of
10194 type names more robust.
10196 2018-01-15 Richard Biener <rguenther@suse.de>
10198 * BASE-VER: Bump to 8.0.1.
10200 2018-01-14 Martin Sebor <msebor@redhat.com>
10203 * builtins.c (check_access): Avoid warning when the no-warning bit
10206 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
10208 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
10209 * ira-color (allocno_hard_regs_compare): Likewise.
10211 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
10214 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
10215 Use .pushsection/.popsection.
10217 2018-01-14 Martin Sebor <msebor@redhat.com>
10220 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
10222 2018-01-14 Jakub Jelinek <jakub@redhat.com>
10224 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
10225 entry from extra_headers.
10226 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
10227 extra_headers, make the list bitwise identical to the i?86-*-* one.
10229 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10231 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
10232 -mcmodel=large with -mindirect-branch=thunk,
10233 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
10234 -mfunction-return=thunk-extern.
10235 * doc/invoke.texi: Document -mcmodel=large is incompatible with
10236 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
10237 -mfunction-return=thunk and -mfunction-return=thunk-extern.
10239 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10241 * config/i386/i386.c (print_reg): Print the name of the full
10242 integer register without '%'.
10243 (ix86_print_operand): Handle 'V'.
10244 * doc/extend.texi: Document 'V' modifier.
10246 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10248 * config/i386/constraints.md (Bs): Disallow memory operand for
10249 -mindirect-branch-register.
10251 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10252 (GOT_memory_operand): Likewise.
10253 (call_insn_operand): Likewise.
10254 (sibcall_insn_operand): Likewise.
10255 (GOT32_symbol_operand): Likewise.
10256 * config/i386/i386.md (indirect_jump): Call convert_memory_address
10257 for -mindirect-branch-register.
10258 (tablejump): Likewise.
10259 (*sibcall_memory): Likewise.
10260 (*sibcall_value_memory): Likewise.
10261 Disallow peepholes of indirect call and jump via memory for
10262 -mindirect-branch-register.
10263 (*call_pop): Replace m with Bw.
10264 (*call_value_pop): Likewise.
10265 (*sibcall_pop_memory): Replace m with Bs.
10266 * config/i386/i386.opt (mindirect-branch-register): New option.
10267 * doc/invoke.texi: Document -mindirect-branch-register option.
10269 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10271 * config/i386/i386-protos.h (ix86_output_function_return): New.
10272 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
10273 set function_return_type.
10274 (indirect_thunk_name): Add ret_p to indicate thunk for function
10276 (output_indirect_thunk_function): Pass false to
10277 indirect_thunk_name.
10278 (ix86_output_indirect_branch_via_reg): Likewise.
10279 (ix86_output_indirect_branch_via_push): Likewise.
10280 (output_indirect_thunk_function): Create alias for function
10281 return thunk if regno < 0.
10282 (ix86_output_function_return): New function.
10283 (ix86_handle_fndecl_attribute): Handle function_return.
10284 (ix86_attribute_table): Add function_return.
10285 * config/i386/i386.h (machine_function): Add
10286 function_return_type.
10287 * config/i386/i386.md (simple_return_internal): Use
10288 ix86_output_function_return.
10289 (simple_return_internal_long): Likewise.
10290 * config/i386/i386.opt (mfunction-return=): New option.
10291 (indirect_branch): Mention -mfunction-return=.
10292 * doc/extend.texi: Document function_return function attribute.
10293 * doc/invoke.texi: Document -mfunction-return= option.
10295 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10297 * config/i386/i386-opts.h (indirect_branch): New.
10298 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
10299 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
10300 with local indirect jump when converting indirect call and jump.
10301 (ix86_set_indirect_branch_type): New.
10302 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
10303 (indirectlabelno): New.
10304 (indirect_thunk_needed): Likewise.
10305 (indirect_thunk_bnd_needed): Likewise.
10306 (indirect_thunks_used): Likewise.
10307 (indirect_thunks_bnd_used): Likewise.
10308 (INDIRECT_LABEL): Likewise.
10309 (indirect_thunk_name): Likewise.
10310 (output_indirect_thunk): Likewise.
10311 (output_indirect_thunk_function): Likewise.
10312 (ix86_output_indirect_branch_via_reg): Likewise.
10313 (ix86_output_indirect_branch_via_push): Likewise.
10314 (ix86_output_indirect_branch): Likewise.
10315 (ix86_output_indirect_jmp): Likewise.
10316 (ix86_code_end): Call output_indirect_thunk_function if needed.
10317 (ix86_output_call_insn): Call ix86_output_indirect_branch if
10319 (ix86_handle_fndecl_attribute): Handle indirect_branch.
10320 (ix86_attribute_table): Add indirect_branch.
10321 * config/i386/i386.h (machine_function): Add indirect_branch_type
10322 and has_local_indirect_jump.
10323 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
10325 (tablejump): Likewise.
10326 (*indirect_jump): Use ix86_output_indirect_jmp.
10327 (*tablejump_1): Likewise.
10328 (simple_return_indirect_internal): Likewise.
10329 * config/i386/i386.opt (mindirect-branch=): New option.
10330 (indirect_branch): New.
10333 (thunk-inline): Likewise.
10334 (thunk-extern): Likewise.
10335 * doc/extend.texi: Document indirect_branch function attribute.
10336 * doc/invoke.texi: Document -mindirect-branch= option.
10338 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
10341 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
10343 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
10345 * ipa-inline.c (want_inline_small_function_p): Return false if
10346 inlining has already failed with CIF_FINAL_ERROR.
10347 (update_caller_keys): Call want_inline_small_function_p before
10349 (update_callee_keys): Likewise.
10351 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
10353 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
10355 (rs6000_quadword_masked_address_p): Likewise.
10356 (quad_aligned_load_p): Likewise.
10357 (quad_aligned_store_p): Likewise.
10358 (const_load_sequence_p): Add comment to describe the outer-most loop.
10359 (mimic_memory_attributes_and_flags): New function.
10360 (rs6000_gen_stvx): Likewise.
10361 (replace_swapped_aligned_store): Likewise.
10362 (rs6000_gen_lvx): Likewise.
10363 (replace_swapped_aligned_load): Likewise.
10364 (replace_swapped_load_constant): Capitalize argument name in
10365 comment describing this function.
10366 (rs6000_analyze_swaps): Add a third pass to search for vector loads
10367 and stores that access quad-word aligned addresses and replace
10368 with stvx or lvx instructions when appropriate.
10369 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
10370 New function prototype.
10371 (rs6000_quadword_masked_address_p): Likewise.
10372 (rs6000_gen_lvx): Likewise.
10373 (rs6000_gen_stvx): Likewise.
10374 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
10375 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
10376 when memory address is aligned.
10377 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
10378 this split to select lvx instruction when memory address is aligned.
10379 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
10380 instruction when memory address is aligned.
10381 (*vsx_le_perm_load_v16qi): Likewise.
10382 (four unnamed splitters): Modify to select the stvx instruction
10383 when memory is aligned.
10385 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
10387 * predict.c (determine_unlikely_bbs): Handle correctly BBs
10388 which appears in the queue multiple times.
10390 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10391 Alan Hayward <alan.hayward@arm.com>
10392 David Sherwood <david.sherwood@arm.com>
10394 * tree-vectorizer.h (vec_lower_bound): New structure.
10395 (_loop_vec_info): Add check_nonzero and lower_bounds.
10396 (LOOP_VINFO_CHECK_NONZERO): New macro.
10397 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
10398 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
10399 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
10400 fields. Make seg_len the distance travelled, not including the
10402 (dr_direction_indicator): Declare.
10403 (dr_zero_step_indicator): Likewise.
10404 (dr_known_forward_stride_p): Likewise.
10405 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
10407 (runtime_alias_check_p): Allow runtime alias checks with
10409 (operator ==): Compare access_size and align.
10410 (prune_runtime_alias_test_list): Rework for new distinction between
10411 the access_size and seg_len.
10412 (create_intersect_range_checks_index): Likewise. Cope with polynomial
10414 (get_segment_min_max): New function.
10415 (create_intersect_range_checks): Use it.
10416 (dr_step_indicator): New function.
10417 (dr_direction_indicator): Likewise.
10418 (dr_zero_step_indicator): Likewise.
10419 (dr_known_forward_stride_p): Likewise.
10420 * tree-loop-distribution.c (data_ref_segment_size): Return
10421 DR_STEP * (niters - 1).
10422 (compute_alias_check_pairs): Update call to the dr_with_seg_len
10424 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
10425 (vect_preserves_scalar_order_p): New function, split out from...
10426 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
10427 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
10428 (vect_vfa_access_size): New function.
10429 (vect_vfa_align): Likewise.
10430 (vect_compile_time_alias): Take access_size_a and access_b arguments.
10431 (dump_lower_bound): New function.
10432 (vect_check_lower_bound): Likewise.
10433 (vect_small_gap_p): Likewise.
10434 (vectorizable_with_step_bound_p): Likewise.
10435 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
10436 depencies if the vectorization factor is 1. Convert the checks
10437 for nonzero steps into checks on the bounds of DR_STEP. Try using
10438 a bunds check for variable steps if the minimum required step is
10439 relatively small. Update calls to the dr_with_seg_len
10440 constructor and to vect_compile_time_alias.
10441 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
10443 (vect_loop_versioning): Call it.
10444 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
10446 (vect_estimate_min_profitable_iters): Account for any bounds checks.
10448 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10449 Alan Hayward <alan.hayward@arm.com>
10450 David Sherwood <david.sherwood@arm.com>
10452 * doc/sourcebuild.texi (vect_scatter_store): Document.
10453 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
10455 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
10457 * genopinit.c (main): Add supports_vec_scatter_store and
10458 supports_vec_scatter_store_cached to target_optabs.
10459 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
10460 IFN_MASK_SCATTER_STORE.
10461 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
10463 * internal-fn.h (internal_store_fn_p): Declare.
10464 (internal_fn_stored_value_index): Likewise.
10465 * internal-fn.c (scatter_store_direct): New macro.
10466 (expand_scatter_store_optab_fn): New function.
10467 (direct_scatter_store_optab_supported_p): New macro.
10468 (internal_store_fn_p): New function.
10469 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
10470 IFN_MASK_SCATTER_STORE.
10471 (internal_fn_mask_index): Likewise.
10472 (internal_fn_stored_value_index): New function.
10473 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
10474 for scatter stores.
10475 * optabs-query.h (supports_vec_scatter_store_p): Declare.
10476 * optabs-query.c (supports_vec_scatter_store_p): New function.
10477 * tree-vectorizer.h (vect_get_store_rhs): Declare.
10478 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
10479 true for scatter stores.
10480 (vect_gather_scatter_fn_p): Handle scatter stores too.
10481 (vect_check_gather_scatter): Consider using scatter stores if
10482 supports_vec_scatter_store_p.
10483 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
10484 scatter stores too.
10485 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10486 internal_fn_stored_value_index.
10487 (check_load_store_masking): Handle scatter stores too.
10488 (vect_get_store_rhs): Make public.
10489 (vectorizable_call): Use internal_store_fn_p.
10490 (vectorizable_store): Handle scatter store internal functions.
10491 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
10492 when deciding whether the end of the group has been reached.
10493 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
10494 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
10495 (mask_scatter_store<mode>): New insns.
10497 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10498 Alan Hayward <alan.hayward@arm.com>
10499 David Sherwood <david.sherwood@arm.com>
10501 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
10502 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
10503 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
10505 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
10506 Use vect_truncate_gather_scatter_offset if we can't treat the
10507 operation as a normal gather load or scatter store.
10508 (get_group_load_store_type): Take the gather_scatter_info
10509 as argument. Try using a gather load or scatter store for
10510 single-element groups.
10511 (get_load_store_type): Update calls to get_group_load_store_type
10512 and vect_use_strided_gather_scatters_p.
10514 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10515 Alan Hayward <alan.hayward@arm.com>
10516 David Sherwood <david.sherwood@arm.com>
10518 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
10519 optional tree argument.
10520 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
10522 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
10523 but continue to use the current value as a fallback.
10524 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
10525 to compare the updates.
10526 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
10527 (get_load_store_type): Use it when handling a strided access.
10528 (vect_get_strided_load_store_ops): New function.
10529 (vect_get_data_ptr_increment): Likewise.
10530 (vectorizable_load): Handle strided gather loads. Always pass
10531 a step to vect_create_data_ref_ptr and bump_vector_ptr.
10533 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10534 Alan Hayward <alan.hayward@arm.com>
10535 David Sherwood <david.sherwood@arm.com>
10537 * doc/md.texi (gather_load@var{m}): Document.
10538 (mask_gather_load@var{m}): Likewise.
10539 * genopinit.c (main): Add supports_vec_gather_load and
10540 supports_vec_gather_load_cached to target_optabs.
10541 * optabs-tree.c (init_tree_optimization_optabs): Use
10542 ggc_cleared_alloc to allocate target_optabs.
10543 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
10544 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
10546 * internal-fn.h (internal_load_fn_p): Declare.
10547 (internal_gather_scatter_fn_p): Likewise.
10548 (internal_fn_mask_index): Likewise.
10549 (internal_gather_scatter_fn_supported_p): Likewise.
10550 * internal-fn.c (gather_load_direct): New macro.
10551 (expand_gather_load_optab_fn): New function.
10552 (direct_gather_load_optab_supported_p): New macro.
10553 (direct_internal_fn_optab): New function.
10554 (internal_load_fn_p): Likewise.
10555 (internal_gather_scatter_fn_p): Likewise.
10556 (internal_fn_mask_index): Likewise.
10557 (internal_gather_scatter_fn_supported_p): Likewise.
10558 * optabs-query.c (supports_at_least_one_mode_p): New function.
10559 (supports_vec_gather_load_p): Likewise.
10560 * optabs-query.h (supports_vec_gather_load_p): Declare.
10561 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
10562 and memory_type field.
10563 (NUM_PATTERNS): Bump to 15.
10564 * tree-vect-data-refs.c: Include internal-fn.h.
10565 (vect_gather_scatter_fn_p): New function.
10566 (vect_describe_gather_scatter_call): Likewise.
10567 (vect_check_gather_scatter): Try using internal functions for
10568 gather loads. Recognize existing calls to a gather load function.
10569 (vect_analyze_data_refs): Consider using gather loads if
10570 supports_vec_gather_load_p.
10571 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
10572 (vect_get_gather_scatter_offset_type): Likewise.
10573 (vect_convert_mask_for_vectype): Likewise.
10574 (vect_add_conversion_to_patterm): Likewise.
10575 (vect_try_gather_scatter_pattern): Likewise.
10576 (vect_recog_gather_scatter_pattern): New pattern recognizer.
10577 (vect_vect_recog_func_ptrs): Add it.
10578 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10579 internal_fn_mask_index and internal_gather_scatter_fn_p.
10580 (check_load_store_masking): Take the gather_scatter_info as an
10581 argument and handle gather loads.
10582 (vect_get_gather_scatter_ops): New function.
10583 (vectorizable_call): Check internal_load_fn_p.
10584 (vectorizable_load): Likewise. Handle gather load internal
10586 (vectorizable_store): Update call to check_load_store_masking.
10587 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
10588 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
10589 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
10590 (aarch64_gather_scale_operand_d): New predicates.
10591 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
10592 (mask_gather_load<mode>): New insns.
10594 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10595 Alan Hayward <alan.hayward@arm.com>
10596 David Sherwood <david.sherwood@arm.com>
10598 * optabs.def (fold_left_plus_optab): New optab.
10599 * doc/md.texi (fold_left_plus_@var{m}): Document.
10600 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
10601 * internal-fn.c (fold_left_direct): Define.
10602 (expand_fold_left_optab_fn): Likewise.
10603 (direct_fold_left_optab_supported_p): Likewise.
10604 * fold-const-call.c (fold_const_fold_left): New function.
10605 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
10606 * tree-parloops.c (valid_reduction_p): New function.
10607 (gather_scalar_reductions): Use it.
10608 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
10609 (vect_finish_replace_stmt): Declare.
10610 * tree-vect-loop.c (fold_left_reduction_fn): New function.
10611 (needs_fold_left_reduction_p): New function, split out from...
10612 (vect_is_simple_reduction): ...here. Accept reductions that
10613 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
10614 (vect_force_simple_reduction): Also store the reduction type in
10615 the assignment's STMT_VINFO_REDUC_TYPE.
10616 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
10617 (merge_with_identity): New function.
10618 (vect_expand_fold_left): Likewise.
10619 (vectorize_fold_left_reduction): Likewise.
10620 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
10621 scalar phi in place for it. Check for target support and reject
10622 cases that would reassociate the operation. Defer the transform
10623 phase to vectorize_fold_left_reduction.
10624 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
10625 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
10626 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
10628 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10630 * tree-if-conv.c (predicate_mem_writes): Remove redundant
10631 call to ifc_temp_var.
10633 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10634 Alan Hayward <alan.hayward@arm.com>
10635 David Sherwood <david.sherwood@arm.com>
10637 * target.def (legitimize_address_displacement): Take the original
10638 offset as a poly_int.
10639 * targhooks.h (default_legitimize_address_displacement): Update
10641 * targhooks.c (default_legitimize_address_displacement): Likewise.
10642 * doc/tm.texi: Regenerate.
10643 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
10644 as an argument, moving assert of ad->disp == ad->disp_term to...
10645 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
10646 Try calling targetm.legitimize_address_displacement before expanding
10647 the address rather than afterwards, and adjust for the new interface.
10648 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
10649 Match the new hook interface. Handle SVE addresses.
10650 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
10651 new hook interface.
10653 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10655 * Makefile.in (OBJS): Add early-remat.o.
10656 * target.def (select_early_remat_modes): New hook.
10657 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
10658 * doc/tm.texi: Regenerate.
10659 * targhooks.h (default_select_early_remat_modes): Declare.
10660 * targhooks.c (default_select_early_remat_modes): New function.
10661 * timevar.def (TV_EARLY_REMAT): New timevar.
10662 * passes.def (pass_early_remat): New pass.
10663 * tree-pass.h (make_pass_early_remat): Declare.
10664 * early-remat.c: New file.
10665 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
10667 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
10669 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10670 Alan Hayward <alan.hayward@arm.com>
10671 David Sherwood <david.sherwood@arm.com>
10673 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
10674 vfm1 with a bound_epilog parameter.
10675 (vect_do_peeling): Update calls accordingly, and move the prologue
10676 call earlier in the function. Treat the base bound_epilog as 0 for
10677 fully-masked loops and retain vf - 1 for other loops. Add 1 to
10678 this base when peeling for gaps.
10679 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
10680 with fully-masked loops.
10681 (vect_estimate_min_profitable_iters): Handle the single peeled
10682 iteration in that case.
10684 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10685 Alan Hayward <alan.hayward@arm.com>
10686 David Sherwood <david.sherwood@arm.com>
10688 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
10689 single-element interleaving even if the size is not a power of 2.
10690 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
10691 accesses for single-element interleaving if the group size is
10694 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10695 Alan Hayward <alan.hayward@arm.com>
10696 David Sherwood <david.sherwood@arm.com>
10698 * doc/md.texi (fold_extract_last_@var{m}): Document.
10699 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
10700 * optabs.def (fold_extract_last_optab): New optab.
10701 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
10702 * internal-fn.c (fold_extract_direct): New macro.
10703 (expand_fold_extract_optab_fn): Likewise.
10704 (direct_fold_extract_optab_supported_p): Likewise.
10705 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
10706 * tree-vect-loop.c (vect_model_reduction_cost): Handle
10707 EXTRACT_LAST_REDUCTION.
10708 (get_initial_def_for_reduction): Do not create an initial vector
10709 for EXTRACT_LAST_REDUCTION reductions.
10710 (vectorizable_reduction): Leave the scalar phi in place for
10711 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
10712 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
10713 epilogue code for EXTRACT_LAST_REDUCTION and defer the
10714 transform phase to vectorizable_condition.
10715 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
10717 (vect_finish_stmt_generation): ...here.
10718 (vect_finish_replace_stmt): New function.
10719 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
10720 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
10722 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
10724 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10725 Alan Hayward <alan.hayward@arm.com>
10726 David Sherwood <david.sherwood@arm.com>
10728 * doc/md.texi (extract_last_@var{m}): Document.
10729 * optabs.def (extract_last_optab): New optab.
10730 * internal-fn.def (EXTRACT_LAST): New internal function.
10731 * internal-fn.c (cond_unary_direct): New macro.
10732 (expand_cond_unary_optab_fn): Likewise.
10733 (direct_cond_unary_optab_supported_p): Likewise.
10734 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
10735 loops using EXTRACT_LAST.
10736 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
10737 (extract_last_<mode>): ...this optab.
10738 (vec_extract<mode><Vel>): Update accordingly.
10740 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10741 Alan Hayward <alan.hayward@arm.com>
10742 David Sherwood <david.sherwood@arm.com>
10744 * target.def (empty_mask_is_expensive): New hook.
10745 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
10746 * doc/tm.texi: Regenerate.
10747 * targhooks.h (default_empty_mask_is_expensive): Declare.
10748 * targhooks.c (default_empty_mask_is_expensive): New function.
10749 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
10750 if the target says that empty masks are expensive.
10751 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
10753 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
10755 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10756 Alan Hayward <alan.hayward@arm.com>
10757 David Sherwood <david.sherwood@arm.com>
10759 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
10760 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
10761 (vect_use_loop_mask_for_alignment_p): New function.
10762 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
10763 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
10764 niters_skip argument. Make sure that the first niters_skip elements
10765 of the first iteration are inactive.
10766 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
10767 Update call to vect_set_loop_masks_directly.
10768 (get_misalign_in_elems): New function, split out from...
10769 (vect_gen_prolog_loop_niters): ...here.
10770 (vect_update_init_of_dr): Take a code argument that specifies whether
10771 the adjustment should be added or subtracted.
10772 (vect_update_init_of_drs): Likewise.
10773 (vect_prepare_for_masked_peels): New function.
10774 (vect_do_peeling): Skip prologue peeling if we're using a mask
10775 instead. Update call to vect_update_inits_of_drs.
10776 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10778 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
10779 alignment. Do not include the number of peeled iterations in
10780 the minimum threshold in that case.
10781 (vectorizable_induction): Adjust the start value down by
10782 LOOP_VINFO_MASK_SKIP_NITERS iterations.
10783 (vect_transform_loop): Call vect_prepare_for_masked_peels.
10784 Take the number of skipped iterations into account when calculating
10786 * tree-vect-stmts.c (vect_gen_while_not): New function.
10788 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10789 Alan Hayward <alan.hayward@arm.com>
10790 David Sherwood <david.sherwood@arm.com>
10792 * doc/sourcebuild.texi (vect_fully_masked): Document.
10793 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
10794 default value to 0.
10795 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
10797 (vect_analyze_loop_2): ...here. Don't check the vectorization
10798 factor against the number of loop iterations if the loop is
10801 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10802 Alan Hayward <alan.hayward@arm.com>
10803 David Sherwood <david.sherwood@arm.com>
10805 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
10806 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
10807 (dump_groups): Update accordingly.
10808 (iv_use::mem_type): New member variable.
10809 (address_p): New function.
10810 (record_use): Add a mem_type argument and initialize the new
10812 (record_group_use): Add a mem_type argument. Use address_p.
10813 Remove obsolete null checks of base_object. Update call to record_use.
10814 (find_interesting_uses_op): Update call to record_group_use.
10815 (find_interesting_uses_cond): Likewise.
10816 (find_interesting_uses_address): Likewise.
10817 (get_mem_type_for_internal_fn): New function.
10818 (find_address_like_use): Likewise.
10819 (find_interesting_uses_stmt): Try find_address_like_use before
10820 calling find_interesting_uses_op.
10821 (addr_offset_valid_p): Use the iv mem_type field as the type
10822 of the addressed memory.
10823 (add_autoinc_candidates): Likewise.
10824 (get_address_cost): Likewise.
10825 (split_small_address_groups_p): Use address_p.
10826 (split_address_groups): Likewise.
10827 (add_iv_candidate_for_use): Likewise.
10828 (autoinc_possible_for_pair): Likewise.
10829 (rewrite_groups): Likewise.
10830 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
10831 (determine_group_iv_cost): Update after split of USE_ADDRESS.
10832 (get_alias_ptr_type_for_ptr_address): New function.
10833 (rewrite_use_address): Rewrite address uses in calls that were
10834 identified by find_address_like_use.
10836 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10837 Alan Hayward <alan.hayward@arm.com>
10838 David Sherwood <david.sherwood@arm.com>
10840 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
10842 * gimple-expr.h (is_gimple_addressable: Likewise.
10843 * gimple-expr.c (is_gimple_address): Likewise.
10844 * internal-fn.c (expand_call_mem_ref): New function.
10845 (expand_mask_load_optab_fn): Use it.
10846 (expand_mask_store_optab_fn): Likewise.
10848 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10849 Alan Hayward <alan.hayward@arm.com>
10850 David Sherwood <david.sherwood@arm.com>
10852 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
10853 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
10854 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
10855 (cond_umax@var{mode}): Document.
10856 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
10857 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
10858 (cond_umin_optab, cond_umax_optab): New optabs.
10859 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
10860 (COND_IOR, COND_XOR): New internal functions.
10861 * internal-fn.h (get_conditional_internal_fn): Declare.
10862 * internal-fn.c (cond_binary_direct): New macro.
10863 (expand_cond_binary_optab_fn): Likewise.
10864 (direct_cond_binary_optab_supported_p): Likewise.
10865 (get_conditional_internal_fn): New function.
10866 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
10867 Cope with reduction statements that are vectorized as calls rather
10869 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
10870 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
10871 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
10872 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
10873 (UNSPEC_COND_EOR): New unspecs.
10874 (optab): Add mappings for them.
10875 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
10876 (sve_int_op, sve_fp_op): New int attributes.
10878 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10879 Alan Hayward <alan.hayward@arm.com>
10880 David Sherwood <david.sherwood@arm.com>
10882 * optabs.def (while_ult_optab): New optab.
10883 * doc/md.texi (while_ult@var{m}@var{n}): Document.
10884 * internal-fn.def (WHILE_ULT): New internal function.
10885 * internal-fn.h (direct_internal_fn_supported_p): New override
10886 that takes two types as argument.
10887 * internal-fn.c (while_direct): New macro.
10888 (expand_while_optab_fn): New function.
10889 (convert_optab_supported_p): Likewise.
10890 (direct_while_optab_supported_p): New macro.
10891 * wide-int.h (wi::udiv_ceil): New function.
10892 * tree-vectorizer.h (rgroup_masks): New structure.
10893 (vec_loop_masks): New typedef.
10894 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
10895 and fully_masked_p.
10896 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
10897 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
10898 (vect_max_vf): New function.
10899 (slpeel_make_loop_iterate_ntimes): Delete.
10900 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
10901 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
10902 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
10903 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
10904 internal-fn.h, stor-layout.h and optabs-query.h.
10905 (vect_set_loop_mask): New function.
10906 (add_preheader_seq): Likewise.
10907 (add_header_seq): Likewise.
10908 (interleave_supported_p): Likewise.
10909 (vect_maybe_permute_loop_masks): Likewise.
10910 (vect_set_loop_masks_directly): Likewise.
10911 (vect_set_loop_condition_masked): Likewise.
10912 (vect_set_loop_condition_unmasked): New function, split out from
10913 slpeel_make_loop_iterate_ntimes.
10914 (slpeel_make_loop_iterate_ntimes): Rename to..
10915 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
10916 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
10917 (vect_do_peeling): Update call accordingly.
10918 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
10920 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10921 mask_compare_type, can_fully_mask_p and fully_masked_p.
10922 (release_vec_loop_masks): New function.
10923 (_loop_vec_info): Use it to free the loop masks.
10924 (can_produce_all_loop_masks_p): New function.
10925 (vect_get_max_nscalars_per_iter): Likewise.
10926 (vect_verify_full_masking): Likewise.
10927 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
10928 retries, and free the mask rgroups before retrying. Check loop-wide
10929 reasons for disallowing fully-masked loops. Make the final decision
10930 about whether use a fully-masked loop or not.
10931 (vect_estimate_min_profitable_iters): Do not assume that peeling
10932 for the number of iterations will be needed for fully-masked loops.
10933 (vectorizable_reduction): Disable fully-masked loops.
10934 (vectorizable_live_operation): Likewise.
10935 (vect_halve_mask_nunits): New function.
10936 (vect_double_mask_nunits): Likewise.
10937 (vect_record_loop_mask): Likewise.
10938 (vect_get_loop_mask): Likewise.
10939 (vect_transform_loop): Handle the case in which the final loop
10940 iteration might handle a partial vector. Call vect_set_loop_condition
10941 instead of slpeel_make_loop_iterate_ntimes.
10942 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
10943 (check_load_store_masking): New function.
10944 (prepare_load_store_mask): Likewise.
10945 (vectorizable_store): Handle fully-masked loops.
10946 (vectorizable_load): Likewise.
10947 (supportable_widening_operation): Use vect_halve_mask_nunits for
10949 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
10950 (vect_gen_while): New function.
10951 * config/aarch64/aarch64.md (umax<mode>3): New expander.
10952 (aarch64_uqdec<mode>): New insn.
10954 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10955 Alan Hayward <alan.hayward@arm.com>
10956 David Sherwood <david.sherwood@arm.com>
10958 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
10959 (reduc_xor_scal_optab): New optabs.
10960 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
10961 (reduc_xor_scal_@var{m}): Document.
10962 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
10963 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
10964 internal functions.
10965 * fold-const-call.c (fold_const_call): Handle them.
10966 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
10967 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
10968 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
10969 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
10970 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
10971 (UNSPEC_XORV): New unspecs.
10972 (optab): Add entries for them.
10973 (BITWISEV): New int iterator.
10974 (bit_reduc_op): New int attributes.
10976 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10977 Alan Hayward <alan.hayward@arm.com>
10978 David Sherwood <david.sherwood@arm.com>
10980 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
10981 * internal-fn.def (VEC_SHL_INSERT): New internal function.
10982 * optabs.def (vec_shl_insert_optab): New optab.
10983 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
10984 (duplicate_and_interleave): Likewise.
10985 * tree-vect-loop.c: Include internal-fn.h.
10986 (neutral_op_for_slp_reduction): New function, split out from
10987 get_initial_defs_for_reduction.
10988 (get_initial_def_for_reduction): Handle option 2 for variable-length
10989 vectors by loading the neutral value into a vector and then shifting
10990 the initial value into element 0.
10991 (get_initial_defs_for_reduction): Replace the code argument with
10992 the neutral value calculated by neutral_op_for_slp_reduction.
10993 Use gimple_build_vector for constant-length vectors.
10994 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
10995 but the first group_size elements have a neutral value.
10996 Use duplicate_and_interleave otherwise.
10997 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
10998 Update call to get_initial_defs_for_reduction. Handle SLP
10999 reductions for variable-length vectors by creating one vector
11000 result for each scalar result, with the elements associated
11001 with other scalar results stubbed out with the neutral value.
11002 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
11003 Require IFN_VEC_SHL_INSERT for double reductions on
11004 variable-length vectors, or SLP reductions that have
11005 a neutral value. Require can_duplicate_and_interleave_p
11006 support for variable-length unchained SLP reductions if there
11007 is no neutral value, such as for MIN/MAX reductions. Also require
11008 the number of vector elements to be a multiple of the number of
11009 SLP statements when doing variable-length unchained SLP reductions.
11010 Update call to vect_create_epilog_for_reduction.
11011 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
11012 and remove initial values.
11013 (duplicate_and_interleave): Make public.
11014 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
11015 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
11017 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11018 Alan Hayward <alan.hayward@arm.com>
11019 David Sherwood <david.sherwood@arm.com>
11021 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
11022 (can_duplicate_and_interleave_p): New function.
11023 (vect_get_and_check_slp_defs): Take the vector of statements
11024 rather than just the current one. Remove excess parentheses.
11025 Restriction rejectinon of vect_constant_def and vect_external_def
11026 for variable-length vectors to boolean types, or types for which
11027 can_duplicate_and_interleave_p is false.
11028 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
11029 (duplicate_and_interleave): New function.
11030 (vect_get_constant_vectors): Use gimple_build_vector for
11031 constant-length vectors and suitable variable-length constant
11032 vectors. Use duplicate_and_interleave for other variable-length
11033 vectors. Don't defer the update when inserting new statements.
11035 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11036 Alan Hayward <alan.hayward@arm.com>
11037 David Sherwood <david.sherwood@arm.com>
11039 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
11040 min_profitable_iters doesn't go negative.
11042 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11043 Alan Hayward <alan.hayward@arm.com>
11044 David Sherwood <david.sherwood@arm.com>
11046 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
11047 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
11048 * optabs.def (vec_mask_load_lanes_optab): New optab.
11049 (vec_mask_store_lanes_optab): Likewise.
11050 * internal-fn.def (MASK_LOAD_LANES): New internal function.
11051 (MASK_STORE_LANES): Likewise.
11052 * internal-fn.c (mask_load_lanes_direct): New macro.
11053 (mask_store_lanes_direct): Likewise.
11054 (expand_mask_load_optab_fn): Handle masked operations.
11055 (expand_mask_load_lanes_optab_fn): New macro.
11056 (expand_mask_store_optab_fn): Handle masked operations.
11057 (expand_mask_store_lanes_optab_fn): New macro.
11058 (direct_mask_load_lanes_optab_supported_p): Likewise.
11059 (direct_mask_store_lanes_optab_supported_p): Likewise.
11060 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
11062 (vect_load_lanes_supported): Likewise.
11063 * tree-vect-data-refs.c (strip_conversion): New function.
11064 (can_group_stmts_p): Likewise.
11065 (vect_analyze_data_ref_accesses): Use it instead of checking
11066 for a pair of assignments.
11067 (vect_store_lanes_supported): Take a masked_p parameter.
11068 (vect_load_lanes_supported): Likewise.
11069 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
11070 vect_store_lanes_supported and vect_load_lanes_supported.
11071 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11072 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
11073 parameter. Don't allow gaps for masked accesses.
11074 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
11075 and vect_load_lanes_supported.
11076 (get_load_store_type): Take a masked_p parameter and update
11077 call to get_group_load_store_type.
11078 (vectorizable_store): Update call to get_load_store_type.
11079 Handle IFN_MASK_STORE_LANES.
11080 (vectorizable_load): Update call to get_load_store_type.
11081 Handle IFN_MASK_LOAD_LANES.
11083 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11084 Alan Hayward <alan.hayward@arm.com>
11085 David Sherwood <david.sherwood@arm.com>
11087 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
11089 * config/aarch64/aarch64-protos.h
11090 (aarch64_sve_struct_memory_operand_p): Declare.
11091 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
11092 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
11093 (VPRED, vpred): Handle SVE structure modes.
11094 * config/aarch64/constraints.md (Utx): New constraint.
11095 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
11096 (aarch64_sve_struct_nonimmediate_operand): New predicates.
11097 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
11098 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
11099 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
11100 structure modes. Split into pieces after RA.
11101 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
11102 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
11104 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
11105 SVE structure modes.
11106 (aarch64_classify_address): Likewise.
11107 (sizetochar): Move earlier in file.
11108 (aarch64_print_operand): Handle SVE register lists.
11109 (aarch64_array_mode): New function.
11110 (aarch64_sve_struct_memory_operand_p): Likewise.
11111 (TARGET_ARRAY_MODE): Redefine.
11113 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11114 Alan Hayward <alan.hayward@arm.com>
11115 David Sherwood <david.sherwood@arm.com>
11117 * target.def (array_mode): New target hook.
11118 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
11119 * doc/tm.texi: Regenerate.
11120 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
11121 * hooks.c (hook_optmode_mode_uhwi_none): New function.
11122 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
11123 targetm.array_mode.
11124 * stor-layout.c (mode_for_array): Likewise. Support polynomial
11127 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11128 Alan Hayward <alan.hayward@arm.com>
11129 David Sherwood <david.sherwood@arm.com>
11131 * fold-const.c (fold_binary_loc): Check the argument types
11132 rather than the result type when testing for a vector operation.
11134 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11136 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
11137 * doc/tm.texi: Regenerate.
11139 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11140 Alan Hayward <alan.hayward@arm.com>
11141 David Sherwood <david.sherwood@arm.com>
11143 * doc/invoke.texi (-msve-vector-bits=): Document new option.
11144 (sve): Document new AArch64 extension.
11145 * doc/md.texi (w): Extend the description of the AArch64
11146 constraint to include SVE vectors.
11147 (Upl, Upa): Document new AArch64 predicate constraints.
11148 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
11150 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
11151 (msve-vector-bits=): New option.
11152 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
11153 SVE when these are disabled.
11154 (sve): New extension.
11155 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
11156 modes. Adjust their number of units based on aarch64_sve_vg.
11157 (MAX_BITSIZE_MODE_ANY_MODE): Define.
11158 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
11159 aarch64_addr_query_type.
11160 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
11161 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
11162 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
11163 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
11164 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
11165 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
11166 (aarch64_simd_imm_zero_p): Delete.
11167 (aarch64_check_zero_based_sve_index_immediate): Declare.
11168 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11169 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11170 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11171 (aarch64_sve_float_mul_immediate_p): Likewise.
11172 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11173 rather than an rtx.
11174 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
11175 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
11176 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
11177 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
11178 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
11179 (aarch64_regmode_natural_size): Likewise.
11180 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
11181 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
11183 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
11184 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
11185 for VG and the SVE predicate registers.
11186 (V_ALIASES): Add a "z"-prefixed alias.
11187 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
11188 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
11189 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
11190 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
11191 (REG_CLASS_NAMES): Add entries for them.
11192 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
11193 and the predicate registers.
11194 (aarch64_sve_vg): Declare.
11195 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
11196 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
11197 (REGMODE_NATURAL_SIZE): Define.
11198 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
11200 * config/aarch64/aarch64.c: Include cfgrtl.h.
11201 (simd_immediate_info): Add a constructor for series vectors,
11202 and an associated step field.
11203 (aarch64_sve_vg): New variable.
11204 (aarch64_dbx_register_number): Handle VG and the predicate registers.
11205 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
11206 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
11207 (VEC_ANY_DATA, VEC_STRUCT): New constants.
11208 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
11209 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
11210 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
11211 (aarch64_get_mask_mode): New functions.
11212 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
11213 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
11214 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
11215 predicate modes and predicate registers. Explicitly restrict
11216 GPRs to modes of 16 bytes or smaller. Only allow FP registers
11217 to store a vector mode if it is recognized by
11218 aarch64_classify_vector_mode.
11219 (aarch64_regmode_natural_size): New function.
11220 (aarch64_hard_regno_caller_save_mode): Return the original mode
11222 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
11223 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
11224 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
11225 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
11227 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
11228 does not overlap dest if the function is frame-related. Handle
11230 (aarch64_split_add_offset): New function.
11231 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
11232 them aarch64_add_offset.
11233 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
11234 and update call to aarch64_sub_sp.
11235 (aarch64_add_cfa_expression): New function.
11236 (aarch64_expand_prologue): Pass extra temporary registers to the
11237 functions above. Handle the case in which we need to emit new
11238 DW_CFA_expressions for registers that were originally saved
11239 relative to the stack pointer, but now have to be expressed
11240 relative to the frame pointer.
11241 (aarch64_output_mi_thunk): Pass extra temporary registers to the
11243 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
11244 IP0 and IP1 values for SVE frames.
11245 (aarch64_expand_vec_series): New function.
11246 (aarch64_expand_sve_widened_duplicate): Likewise.
11247 (aarch64_expand_sve_const_vector): Likewise.
11248 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
11249 Handle SVE constants. Use emit_move_insn to move a force_const_mem
11250 into the register, rather than emitting a SET directly.
11251 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
11252 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
11253 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
11254 (offset_9bit_signed_scaled_p): New functions.
11255 (aarch64_replicate_bitmask_imm): New function.
11256 (aarch64_bitmask_imm): Use it.
11257 (aarch64_cannot_force_const_mem): Reject expressions involving
11258 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
11259 (aarch64_classify_index): Handle SVE indices, by requiring
11260 a plain register index with a scale that matches the element size.
11261 (aarch64_classify_address): Handle SVE addresses. Assert that
11262 the mode of the address is VOIDmode or an integer mode.
11263 Update call to aarch64_classify_symbol.
11264 (aarch64_classify_symbolic_expression): Update call to
11265 aarch64_classify_symbol.
11266 (aarch64_const_vec_all_in_range_p): New function.
11267 (aarch64_print_vector_float_operand): Likewise.
11268 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
11269 "vN" for FP registers with SVE modes. Handle (const ...) vectors
11270 and the FP immediates 1.0 and 0.5.
11271 (aarch64_print_address_internal): Handle SVE addresses.
11272 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
11273 (aarch64_regno_regclass): Handle predicate registers.
11274 (aarch64_secondary_reload): Handle big-endian reloads of SVE
11276 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
11277 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
11278 (aarch64_convert_sve_vector_bits): New function.
11279 (aarch64_override_options): Use it to handle -msve-vector-bits=.
11280 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11281 rather than an rtx.
11282 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
11283 Handle SVE vector and predicate modes. Accept VL-based constants
11284 that need only one temporary register, and VL offsets that require
11285 no temporary registers.
11286 (aarch64_conditional_register_usage): Mark the predicate registers
11287 as fixed if SVE isn't available.
11288 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
11289 Return true for SVE vector and predicate modes.
11290 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
11291 rather than an unsigned int. Handle SVE modes.
11292 (aarch64_preferred_simd_mode): Update call accordingly. Handle
11294 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
11296 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11297 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11298 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11299 (aarch64_sve_float_mul_immediate_p): New functions.
11300 (aarch64_sve_valid_immediate): New function.
11301 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
11302 Explicitly reject structure modes. Check for INDEX constants.
11303 Handle PTRUE and PFALSE constants.
11304 (aarch64_check_zero_based_sve_index_immediate): New function.
11305 (aarch64_simd_imm_zero_p): Delete.
11306 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
11307 vector modes. Accept constants in the range of CNT[BHWD].
11308 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
11309 ask for an Advanced SIMD mode.
11310 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
11311 (aarch64_simd_vector_alignment): Handle SVE predicates.
11312 (aarch64_vectorize_preferred_vector_alignment): New function.
11313 (aarch64_simd_vector_alignment_reachable): Use it instead of
11315 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
11316 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
11318 (MAX_VECT_LEN): Delete.
11319 (expand_vec_perm_d): Add a vec_flags field.
11320 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
11321 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
11322 (aarch64_evpc_ext): Don't apply a big-endian lane correction
11324 (aarch64_evpc_rev): Rename to...
11325 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
11326 (aarch64_evpc_rev_global): New function.
11327 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
11328 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
11330 (aarch64_evpc_sve_tbl): New function.
11331 (aarch64_expand_vec_perm_const_1): Update after rename of
11332 aarch64_evpc_rev. Handle SVE permutes too, trying
11333 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
11334 than aarch64_evpc_tbl.
11335 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
11336 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
11337 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
11338 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
11339 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
11340 (aarch64_expand_sve_vcond): New functions.
11341 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
11342 of aarch64_vector_mode_p.
11343 (aarch64_dwarf_poly_indeterminate_value): New function.
11344 (aarch64_compute_pressure_classes): Likewise.
11345 (aarch64_can_change_mode_class): Likewise.
11346 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
11347 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
11348 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
11349 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
11350 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
11351 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
11352 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
11353 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
11355 (Dn, Dl, Dr): Accept const as well as const_vector.
11356 (Dz): Likewise. Compare against CONST0_RTX.
11357 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
11358 of "vector" where appropriate.
11359 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
11360 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
11361 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
11362 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
11363 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
11364 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
11365 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
11366 (v_int_equiv): Extend to SVE modes.
11367 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
11369 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
11370 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
11371 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
11372 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
11373 (SVE_COND_FP_CMP): New int iterators.
11374 (perm_hilo): Handle the new unpack unspecs.
11375 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
11377 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
11378 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
11379 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
11380 (aarch64_equality_operator, aarch64_constant_vector_operand)
11381 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
11382 (aarch64_sve_nonimmediate_operand): Likewise.
11383 (aarch64_sve_general_operand): Likewise.
11384 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
11385 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
11386 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
11387 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
11388 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
11389 (aarch64_sve_float_arith_immediate): Likewise.
11390 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
11391 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
11392 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
11393 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
11394 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
11395 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
11396 (aarch64_sve_float_arith_operand): Likewise.
11397 (aarch64_sve_float_arith_with_sub_operand): Likewise.
11398 (aarch64_sve_float_mul_operand): Likewise.
11399 (aarch64_sve_vec_perm_operand): Likewise.
11400 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
11401 (aarch64_mov_operand): Accept const_poly_int and const_vector.
11402 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
11403 as well as const_vector.
11404 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
11405 in file. Use CONST0_RTX and CONSTM1_RTX.
11406 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
11407 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
11408 Use aarch64_simd_imm_zero.
11409 * config/aarch64/aarch64-sve.md: New file.
11410 * config/aarch64/aarch64.md: Include it.
11411 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
11412 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
11413 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
11414 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
11415 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
11416 (sve): New attribute.
11417 (enabled): Disable instructions with the sve attribute unless
11419 (movqi, movhi): Pass CONST_POLY_INT operaneds through
11420 aarch64_expand_mov_immediate.
11421 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
11422 CNT[BHSD] immediates.
11423 (movti): Split CONST_POLY_INT moves into two halves.
11424 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
11425 Split additions that need a temporary here if the destination
11426 is the stack pointer.
11427 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
11428 (*add<mode>3_poly_1): New instruction.
11429 (set_clobber_cc): New expander.
11431 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11433 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
11434 parameter and use it instead of GET_MODE_SIZE (innermode). Use
11435 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
11436 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
11437 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
11438 Change innermode from fixed_mode_size to machine_mode.
11439 (simplify_subreg): Update call accordingly. Handle a constant-sized
11440 subreg of a variable-length CONST_VECTOR.
11442 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11443 Alan Hayward <alan.hayward@arm.com>
11444 David Sherwood <david.sherwood@arm.com>
11446 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
11447 (add_offset_to_base): New function, split out from...
11448 (create_mem_ref): ...here. When handling a scale other than 1,
11449 check first whether the address is valid without the offset.
11450 Add it into the base if so, leaving the index and scale as-is.
11452 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11455 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
11456 fold_for_warn before checking if arg2 is INTEGER_CST.
11458 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
11460 * config/rs6000/predicates.md (load_multiple_operation): Delete.
11461 (store_multiple_operation): Delete.
11462 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
11463 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
11464 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
11465 guarded by TARGET_STRING.
11466 (rs6000_output_load_multiple): Delete.
11467 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
11468 OPTION_MASK_STRING / TARGET_STRING handling.
11469 (print_operand) <'N', 'O'>: Add comment that these are unused now.
11470 (const rs6000_opt_masks) <"string">: Change mask to 0.
11471 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
11472 (MASK_STRING): Delete.
11473 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
11475 (load_multiple): Delete.
11482 (store_multiple): Delete.
11489 (movmemsi_8reg): Delete.
11490 (corresponding unnamed define_insn): Delete.
11491 (movmemsi_6reg): Delete.
11492 (corresponding unnamed define_insn): Delete.
11493 (movmemsi_4reg): Delete.
11494 (corresponding unnamed define_insn): Delete.
11495 (movmemsi_2reg): Delete.
11496 (corresponding unnamed define_insn): Delete.
11497 (movmemsi_1reg): Delete.
11498 (corresponding unnamed define_insn): Delete.
11499 * config/rs6000/rs6000.opt (mno-string): New.
11500 (mstring): Replace by deprecation warning stub.
11501 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
11503 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11505 * regrename.c (regrename_do_replace): If replacing the same
11506 reg multiple times, try to reuse last created gen_raw_REG.
11509 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
11510 main to workaround a bug in GDB.
11512 2018-01-12 Tom de Vries <tom@codesourcery.com>
11515 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
11517 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
11519 PR rtl-optimization/80481
11520 * ira-color.c (get_cap_member): New function.
11521 (allocnos_conflict_by_live_ranges_p): Use it.
11522 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
11523 (setup_slot_coalesced_allocno_live_ranges): Ditto.
11525 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
11528 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
11529 (*saddl_se_1): Ditto.
11530 (*ssubsi_1): Ditto.
11531 (*ssubl_se_1): Ditto.
11533 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11535 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
11536 rather than wi::to_widest for DR_INITs.
11537 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
11538 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
11539 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
11541 (vect_analyze_group_access_1): Note that here.
11543 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11545 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
11546 polynomial type sizes.
11548 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11550 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
11551 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
11552 (gimple_add_tmp_var): Likewise.
11554 2018-01-12 Martin Liska <mliska@suse.cz>
11556 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
11557 (gimple_alloc_sizes): Likewise.
11558 (dump_gimple_statistics): Use PRIu64 in printf format.
11559 * gimple.h: Change uint64_t to int.
11561 2018-01-12 Martin Liska <mliska@suse.cz>
11563 * tree-core.h: Use uint64_t instead of int.
11564 * tree.c (tree_node_counts): Likewise.
11565 (tree_node_sizes): Likewise.
11566 (dump_tree_statistics): Use PRIu64 in printf format.
11568 2018-01-12 Martin Liska <mliska@suse.cz>
11570 * Makefile.in: As qsort_chk is implemented in vec.c, add
11571 vec.o to linkage of gencfn-macros.
11572 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
11573 passing the info to record_node_allocation_statistics.
11574 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
11576 * ggc-common.c (struct ggc_usage): Add operator== and use
11577 it in operator< and compare function.
11578 * mem-stats.h (struct mem_usage): Likewise.
11579 * vec.c (struct vec_usage): Remove operator< and compare
11580 function. Can be simply inherited.
11582 2018-01-12 Martin Jambor <mjambor@suse.cz>
11585 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
11586 * tree-ssa-math-opts.c: Include domwalk.h.
11587 (convert_mult_to_fma_1): New function.
11588 (fma_transformation_info): New type.
11589 (fma_deferring_state): Likewise.
11590 (cancel_fma_deferring): New function.
11591 (result_of_phi): Likewise.
11592 (last_fma_candidate_feeds_initial_phi): Likewise.
11593 (convert_mult_to_fma): Added deferring logic, split actual
11594 transformation to convert_mult_to_fma_1.
11595 (math_opts_dom_walker): New type.
11596 (math_opts_dom_walker::after_dom_children): New method, body moved
11597 here from pass_optimize_widening_mul::execute, added deferring logic
11599 (pass_optimize_widening_mul::execute): Moved most of code to
11600 math_opts_dom_walker::after_dom_children.
11601 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
11602 * config/i386/i386.c (ix86_option_override_internal): Added
11603 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
11605 2018-01-12 Richard Biener <rguenther@suse.de>
11608 * dwarf2out.c (gen_variable_die): Do not reset old_die for
11609 inline instance vars.
11611 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
11614 * config/rx/rx.c (rx_is_restricted_memory_address):
11615 Handle SUBREG case.
11617 2018-01-12 Richard Biener <rguenther@suse.de>
11619 PR tree-optimization/80846
11620 * target.def (split_reduction): New target hook.
11621 * targhooks.c (default_split_reduction): New function.
11622 * targhooks.h (default_split_reduction): Declare.
11623 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
11624 target requests first reduce vectors by combining low and high
11626 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
11627 (get_vectype_for_scalar_type_and_size): Export.
11628 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
11629 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
11630 * doc/tm.texi: Regenerate.
11631 * config/i386/i386.c (ix86_split_reduction): Implement
11632 TARGET_VECTORIZE_SPLIT_REDUCTION.
11634 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11637 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
11638 in PIC mode except for TARGET_VXWORKS_RTP.
11639 * config/sparc/sparc.c: Include cfgrtl.h.
11640 (TARGET_INIT_PIC_REG): Define.
11641 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
11642 (sparc_pic_register_p): New predicate.
11643 (sparc_legitimate_address_p): Use it.
11644 (sparc_legitimize_pic_address): Likewise.
11645 (sparc_delegitimize_address): Likewise.
11646 (sparc_mode_dependent_address_p): Likewise.
11647 (gen_load_pcrel_sym): Remove 4th parameter.
11648 (load_got_register): Adjust call to above. Remove obsolete stuff.
11649 (sparc_expand_prologue): Do not call load_got_register here.
11650 (sparc_flat_expand_prologue): Likewise.
11651 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
11652 (sparc_use_pseudo_pic_reg): New function.
11653 (sparc_init_pic_reg): Likewise.
11654 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11655 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
11657 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
11659 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
11660 Add item for branch_cost.
11662 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11664 PR rtl-optimization/83565
11665 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
11666 not extend the result to a larger mode for rotate operations.
11667 (num_sign_bit_copies1): Likewise.
11669 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11672 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
11674 Use values-Xc.o for -pedantic.
11675 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
11677 2018-01-12 Martin Liska <mliska@suse.cz>
11680 * ipa-devirt.c (final_warning_record::grow_type_warnings):
11682 (possible_polymorphic_call_targets): Use it.
11683 (ipa_devirt): Likewise.
11685 2018-01-12 Martin Liska <mliska@suse.cz>
11687 * profile-count.h (enum profile_quality): Use 0 as invalid
11688 enum value of profile_quality.
11690 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
11692 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
11693 -mext-string options.
11695 2018-01-12 Richard Biener <rguenther@suse.de>
11697 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
11698 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
11699 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
11701 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
11703 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
11705 * configure.ac (--with-long-double-format): Add support for the
11706 configuration option to change the default long double format on
11708 * config.gcc (powerpc*-linux*-*): Likewise.
11709 * configure: Regenerate.
11710 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
11711 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
11712 used without modification.
11714 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11716 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
11717 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
11718 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
11719 MISC_BUILTIN_SPEC_BARRIER.
11720 (rs6000_init_builtins): Likewise.
11721 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
11723 (speculation_barrier): New define_insn.
11724 * doc/extend.texi: Document __builtin_speculation_barrier.
11726 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11729 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
11730 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
11731 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
11733 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
11734 integral modes instead of "ss" and "sd".
11735 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
11736 vectors with 32-bit and 64-bit elements.
11737 (vecdupssescalarmodesuffix): New mode attribute.
11738 (vec_dup<mode>): Use it.
11740 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
11743 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
11744 frame if argument is passed on stack.
11746 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11749 * ree.c (combine_reaching_defs): Optimize also
11750 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
11751 reg2=any_extend(exp); reg1=reg2;, formatting fix.
11753 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11755 PR middle-end/83189
11756 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
11758 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11760 PR middle-end/83718
11761 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
11762 after they are computed.
11764 2018-01-11 Bin Cheng <bin.cheng@arm.com>
11766 PR tree-optimization/83695
11767 * gimple-loop-linterchange.cc
11768 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
11769 reset cached scev information after interchange.
11770 (pass_linterchange::execute): Remove call to scev_reset_htab.
11772 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11774 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
11775 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
11776 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
11777 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
11778 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
11779 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
11780 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
11781 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
11782 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
11783 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
11784 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
11785 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
11786 (V_lane_reg): Likewise.
11787 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
11789 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
11790 (vfmal_lane_low<mode>_intrinsic,
11791 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
11792 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
11793 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
11794 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
11795 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
11796 vfmsl_lane_high<mode>_intrinsic): New define_insns.
11798 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11800 * config/arm/arm-cpus.in (fp16fml): New feature.
11801 (ALL_SIMD): Add fp16fml.
11802 (armv8.2-a): Add fp16fml as an option.
11803 (armv8.3-a): Likewise.
11804 (armv8.4-a): Add fp16fml as part of fp16.
11805 * config/arm/arm.h (TARGET_FP16FML): Define.
11806 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
11808 * config/arm/arm-modes.def (V2HF): Define.
11809 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
11810 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
11811 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
11812 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
11813 vfmsl_low, vfmsl_high): New set of builtins.
11814 * config/arm/iterators.md (PLUSMINUS): New code iterator.
11815 (vfml_op): New code attribute.
11816 (VFMLHALVES): New int iterator.
11817 (VFML, VFMLSEL): New mode attributes.
11818 (V_reg): Define mapping for V2HF.
11819 (V_hi, V_lo): New mode attributes.
11820 (VF_constraint): Likewise.
11821 (vfml_half, vfml_half_selector): New int attributes.
11822 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
11824 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
11825 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
11827 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
11828 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
11829 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
11830 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
11832 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
11833 Document new effective target and option set.
11835 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11837 * config/arm/arm-cpus.in (armv8_4): New feature.
11838 (ARMv8_4a): New fgroup.
11839 (armv8.4-a): New arch.
11840 * config/arm/arm-tables.opt: Regenerate.
11841 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
11842 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
11843 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
11844 Add matching rules for -march=armv8.4-a and extensions.
11845 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
11847 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
11850 * config/rx/rx.md (BW): New mode attribute.
11851 (sync_lock_test_and_setsi): Add mode suffix to insn output.
11853 2018-01-11 Richard Biener <rguenther@suse.de>
11855 PR tree-optimization/83435
11856 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
11857 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
11858 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
11860 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11861 Alan Hayward <alan.hayward@arm.com>
11862 David Sherwood <david.sherwood@arm.com>
11864 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
11866 (aarch64_classify_address): Initialize it. Track polynomial offsets.
11867 (aarch64_print_address_internal): Use it to check for a zero offset.
11869 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11870 Alan Hayward <alan.hayward@arm.com>
11871 David Sherwood <david.sherwood@arm.com>
11873 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
11874 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
11875 Return a poly_int64 rather than a HOST_WIDE_INT.
11876 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
11877 rather than a HOST_WIDE_INT.
11878 * config/aarch64/aarch64.h (aarch64_frame): Protect with
11879 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
11880 hard_fp_offset, frame_size, initial_adjust, callee_offset and
11881 final_offset from HOST_WIDE_INT to poly_int64.
11882 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
11883 to_constant when getting the number of units in an Advanced SIMD
11885 (aarch64_builtin_vectorized_function): Check for a constant number
11887 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
11889 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
11890 attribute instead of GET_MODE_NUNITS.
11891 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
11892 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
11893 GET_MODE_SIZE for fixed-size registers.
11894 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
11895 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
11896 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
11897 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
11898 (aarch64_print_operand, aarch64_print_address_internal)
11899 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
11900 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
11901 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
11902 Handle polynomial GET_MODE_SIZE.
11903 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
11904 wider than SImode without modification.
11905 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
11906 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
11907 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
11908 passing and returning SVE modes.
11909 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
11910 rather than GEN_INT.
11911 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
11912 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
11913 (aarch64_allocate_and_probe_stack_space): Likewise.
11914 (aarch64_layout_frame): Cope with polynomial offsets.
11915 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
11916 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
11917 polynomial offsets.
11918 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
11919 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
11920 poly_int64 rather than a HOST_WIDE_INT.
11921 (aarch64_get_separate_components, aarch64_process_components)
11922 (aarch64_expand_prologue, aarch64_expand_epilogue)
11923 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
11924 (aarch64_anchor_offset): New function, split out from...
11925 (aarch64_legitimize_address): ...here.
11926 (aarch64_builtin_vectorization_cost): Handle polynomial
11927 TYPE_VECTOR_SUBPARTS.
11928 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
11930 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
11931 number of elements from the PARALLEL rather than the mode.
11932 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
11933 rather than GET_MODE_BITSIZE.
11934 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
11935 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
11936 (aarch64_expand_vec_perm_const_1): Handle polynomial
11937 d->perm.length () and d->perm elements.
11938 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
11939 Apply to_constant to d->perm elements.
11940 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
11941 polynomial CONST_VECTOR_NUNITS.
11942 (aarch64_move_pointer): Take amount as a poly_int64 rather
11944 (aarch64_progress_pointer): Avoid temporary variable.
11945 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
11946 the mode attribute instead of GET_MODE.
11948 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11949 Alan Hayward <alan.hayward@arm.com>
11950 David Sherwood <david.sherwood@arm.com>
11952 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
11953 x exists before using it.
11954 (aarch64_add_constant_internal): Rename to...
11955 (aarch64_add_offset_1): ...this. Replace regnum with separate
11956 src and dest rtxes. Handle the case in which they're different,
11957 including when the offset is zero. Replace scratchreg with an rtx.
11958 Use 2 additions if there is no spare register into which we can
11959 move a 16-bit constant.
11960 (aarch64_add_constant): Delete.
11961 (aarch64_add_offset): Replace reg with separate src and dest
11962 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
11963 Use aarch64_add_offset_1.
11964 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
11965 an rtx rather than an int. Take the delta as a poly_int64
11966 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
11967 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
11968 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
11969 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
11970 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
11971 and aarch64_add_sp.
11972 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
11973 aarch64_add_constant.
11975 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11977 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
11978 Use scalar_float_mode.
11980 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11982 * config/aarch64/aarch64-simd.md
11983 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
11984 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
11985 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
11986 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
11987 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
11988 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
11989 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
11990 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
11991 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
11992 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
11994 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11997 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
11998 targ_options->x_arm_arch_string is non NULL.
12000 2018-01-11 Tamar Christina <tamar.christina@arm.com>
12002 * config/aarch64/aarch64.h
12003 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
12005 2018-01-11 Sudakshina Das <sudi.das@arm.com>
12008 * expmed.c (emit_store_flag_force): Swap if const op0
12009 and change VOIDmode to mode of op0.
12011 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12013 PR rtl-optimization/83761
12014 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
12015 than bytes to mode_for_size.
12017 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12019 PR middle-end/83189
12020 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
12021 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
12024 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12026 PR middle-end/83575
12027 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
12028 when in layout mode.
12029 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
12030 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
12033 2018-01-10 Michael Collison <michael.collison@arm.com>
12035 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
12036 * config/aarch64/aarch64-option-extension.def: Add
12037 AARCH64_OPT_EXTENSION of 'fp16fml'.
12038 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12039 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
12040 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
12041 * config/aarch64/constraints.md (Ui7): New constraint.
12042 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
12043 (VFMLA_SEL_W): Ditto.
12046 (VFMLA16_LOW): New int iterator.
12047 (VFMLA16_HIGH): Ditto.
12048 (UNSPEC_FMLAL): New unspec.
12049 (UNSPEC_FMLSL): Ditto.
12050 (UNSPEC_FMLAL2): Ditto.
12051 (UNSPEC_FMLSL2): Ditto.
12052 (f16mac): New code attribute.
12053 * config/aarch64/aarch64-simd-builtins.def
12054 (aarch64_fmlal_lowv2sf): Ditto.
12055 (aarch64_fmlsl_lowv2sf): Ditto.
12056 (aarch64_fmlalq_lowv4sf): Ditto.
12057 (aarch64_fmlslq_lowv4sf): Ditto.
12058 (aarch64_fmlal_highv2sf): Ditto.
12059 (aarch64_fmlsl_highv2sf): Ditto.
12060 (aarch64_fmlalq_highv4sf): Ditto.
12061 (aarch64_fmlslq_highv4sf): Ditto.
12062 (aarch64_fmlal_lane_lowv2sf): Ditto.
12063 (aarch64_fmlsl_lane_lowv2sf): Ditto.
12064 (aarch64_fmlal_laneq_lowv2sf): Ditto.
12065 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
12066 (aarch64_fmlalq_lane_lowv4sf): Ditto.
12067 (aarch64_fmlsl_lane_lowv4sf): Ditto.
12068 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
12069 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
12070 (aarch64_fmlal_lane_highv2sf): Ditto.
12071 (aarch64_fmlsl_lane_highv2sf): Ditto.
12072 (aarch64_fmlal_laneq_highv2sf): Ditto.
12073 (aarch64_fmlsl_laneq_highv2sf): Ditto.
12074 (aarch64_fmlalq_lane_highv4sf): Ditto.
12075 (aarch64_fmlsl_lane_highv4sf): Ditto.
12076 (aarch64_fmlalq_laneq_highv4sf): Ditto.
12077 (aarch64_fmlsl_laneq_highv4sf): Ditto.
12078 * config/aarch64/aarch64-simd.md:
12079 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
12080 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12081 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
12082 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12083 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
12084 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
12085 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
12086 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
12087 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
12088 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
12089 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
12090 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
12091 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
12092 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
12093 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
12094 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
12095 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
12096 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
12097 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
12098 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
12099 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
12100 (vfmlsl_low_u32): Ditto.
12101 (vfmlalq_low_u32): Ditto.
12102 (vfmlslq_low_u32): Ditto.
12103 (vfmlal_high_u32): Ditto.
12104 (vfmlsl_high_u32): Ditto.
12105 (vfmlalq_high_u32): Ditto.
12106 (vfmlslq_high_u32): Ditto.
12107 (vfmlal_lane_low_u32): Ditto.
12108 (vfmlsl_lane_low_u32): Ditto.
12109 (vfmlal_laneq_low_u32): Ditto.
12110 (vfmlsl_laneq_low_u32): Ditto.
12111 (vfmlalq_lane_low_u32): Ditto.
12112 (vfmlslq_lane_low_u32): Ditto.
12113 (vfmlalq_laneq_low_u32): Ditto.
12114 (vfmlslq_laneq_low_u32): Ditto.
12115 (vfmlal_lane_high_u32): Ditto.
12116 (vfmlsl_lane_high_u32): Ditto.
12117 (vfmlal_laneq_high_u32): Ditto.
12118 (vfmlsl_laneq_high_u32): Ditto.
12119 (vfmlalq_lane_high_u32): Ditto.
12120 (vfmlslq_lane_high_u32): Ditto.
12121 (vfmlalq_laneq_high_u32): Ditto.
12122 (vfmlslq_laneq_high_u32): Ditto.
12123 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
12124 (AARCH64_FL_FOR_ARCH8_4): New.
12125 (AARCH64_ISA_F16FML): New ISA flag.
12126 (TARGET_F16FML): New feature flag for fp16fml.
12127 (doc/invoke.texi): Document new fp16fml option.
12129 2018-01-10 Michael Collison <michael.collison@arm.com>
12131 * config/aarch64/aarch64-builtins.c:
12132 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
12133 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12134 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
12135 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
12136 (AARCH64_ISA_SHA3): New ISA flag.
12137 (TARGET_SHA3): New feature flag for sha3.
12138 * config/aarch64/iterators.md (sha512_op): New int attribute.
12139 (CRYPTO_SHA512): New int iterator.
12140 (UNSPEC_SHA512H): New unspec.
12141 (UNSPEC_SHA512H2): Ditto.
12142 (UNSPEC_SHA512SU0): Ditto.
12143 (UNSPEC_SHA512SU1): Ditto.
12144 * config/aarch64/aarch64-simd-builtins.def
12145 (aarch64_crypto_sha512hqv2di): New builtin.
12146 (aarch64_crypto_sha512h2qv2di): Ditto.
12147 (aarch64_crypto_sha512su0qv2di): Ditto.
12148 (aarch64_crypto_sha512su1qv2di): Ditto.
12149 (aarch64_eor3qv8hi): Ditto.
12150 (aarch64_rax1qv2di): Ditto.
12151 (aarch64_xarqv2di): Ditto.
12152 (aarch64_bcaxqv8hi): Ditto.
12153 * config/aarch64/aarch64-simd.md:
12154 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
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/arm_neon.h (vsha512hq_u64): New intrinsic.
12162 (vsha512h2q_u64): Ditto.
12163 (vsha512su0q_u64): Ditto.
12164 (vsha512su1q_u64): Ditto.
12165 (veor3q_u16): Ditto.
12166 (vrax1q_u64): Ditto.
12167 (vxarq_u64): Ditto.
12168 (vbcaxq_u16): Ditto.
12169 * config/arm/types.md (crypto_sha512): New type attribute.
12170 (crypto_sha3): Ditto.
12171 (doc/invoke.texi): Document new sha3 option.
12173 2018-01-10 Michael Collison <michael.collison@arm.com>
12175 * config/aarch64/aarch64-builtins.c:
12176 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
12177 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12178 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
12179 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
12180 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
12181 (AARCH64_ISA_SM4): New ISA flag.
12182 (TARGET_SM4): New feature flag for sm4.
12183 * config/aarch64/aarch64-simd-builtins.def
12184 (aarch64_sm3ss1qv4si): Ditto.
12185 (aarch64_sm3tt1aq4si): Ditto.
12186 (aarch64_sm3tt1bq4si): Ditto.
12187 (aarch64_sm3tt2aq4si): Ditto.
12188 (aarch64_sm3tt2bq4si): Ditto.
12189 (aarch64_sm3partw1qv4si): Ditto.
12190 (aarch64_sm3partw2qv4si): Ditto.
12191 (aarch64_sm4eqv4si): Ditto.
12192 (aarch64_sm4ekeyqv4si): Ditto.
12193 * config/aarch64/aarch64-simd.md:
12194 (aarch64_sm3ss1qv4si): Ditto.
12195 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
12196 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
12197 (aarch64_sm4eqv4si): Ditto.
12198 (aarch64_sm4ekeyqv4si): Ditto.
12199 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
12200 (sm3part_op): Ditto.
12201 (CRYPTO_SM3TT): Ditto.
12202 (CRYPTO_SM3PART): Ditto.
12203 (UNSPEC_SM3SS1): New unspec.
12204 (UNSPEC_SM3TT1A): Ditto.
12205 (UNSPEC_SM3TT1B): Ditto.
12206 (UNSPEC_SM3TT2A): Ditto.
12207 (UNSPEC_SM3TT2B): Ditto.
12208 (UNSPEC_SM3PARTW1): Ditto.
12209 (UNSPEC_SM3PARTW2): Ditto.
12210 (UNSPEC_SM4E): Ditto.
12211 (UNSPEC_SM4EKEY): Ditto.
12212 * config/aarch64/constraints.md (Ui2): New constraint.
12213 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
12214 * config/arm/types.md (crypto_sm3): New type attribute.
12215 (crypto_sm4): Ditto.
12216 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
12217 (vsm3tt1aq_u32): Ditto.
12218 (vsm3tt1bq_u32): Ditto.
12219 (vsm3tt2aq_u32): Ditto.
12220 (vsm3tt2bq_u32): Ditto.
12221 (vsm3partw1q_u32): Ditto.
12222 (vsm3partw2q_u32): Ditto.
12223 (vsm4eq_u32): Ditto.
12224 (vsm4ekeyq_u32): Ditto.
12225 (doc/invoke.texi): Document new sm4 option.
12227 2018-01-10 Michael Collison <michael.collison@arm.com>
12229 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
12230 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
12231 (AARCH64_FL_FOR_ARCH8_4): New.
12232 (AARCH64_FL_V8_4): New flag.
12233 (doc/invoke.texi): Document new armv8.4-a option.
12235 2018-01-10 Michael Collison <michael.collison@arm.com>
12237 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12238 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
12239 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
12240 * config/aarch64/aarch64-option-extension.def: Add
12241 AARCH64_OPT_EXTENSION of 'sha2'.
12242 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
12243 (crypto): Disable sha2 and aes if crypto disabled.
12244 (crypto): Enable aes and sha2 if enabled.
12245 (simd): Disable sha2 and aes if simd disabled.
12246 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
12248 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
12249 (TARGET_SHA2): New feature flag for sha2.
12250 (TARGET_AES): New feature flag for aes.
12251 * config/aarch64/aarch64-simd.md:
12252 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
12253 conditional on TARGET_AES.
12254 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
12255 (aarch64_crypto_sha1hsi): Make pattern conditional
12257 (aarch64_crypto_sha1hv4si): Ditto.
12258 (aarch64_be_crypto_sha1hv4si): Ditto.
12259 (aarch64_crypto_sha1su1v4si): Ditto.
12260 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
12261 (aarch64_crypto_sha1su0v4si): Ditto.
12262 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
12263 (aarch64_crypto_sha256su0v4si): Ditto.
12264 (aarch64_crypto_sha256su1v4si): Ditto.
12265 (doc/invoke.texi): Document new aes and sha2 options.
12267 2018-01-10 Martin Sebor <msebor@redhat.com>
12269 PR tree-optimization/83781
12270 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
12273 2018-01-11 Martin Sebor <msebor@gmail.com>
12274 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12276 PR tree-optimization/83501
12277 PR tree-optimization/81703
12279 * tree-ssa-strlen.c (get_string_cst): Rename...
12280 (get_string_len): ...to this. Handle global constants.
12281 (handle_char_store): Adjust.
12283 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
12284 Jim Wilson <jimw@sifive.com>
12286 * config/riscv/riscv-protos.h (riscv_output_return): New.
12287 * config/riscv/riscv.c (struct machine_function): New naked_p field.
12288 (riscv_attribute_table, riscv_output_return),
12289 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
12290 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
12291 (riscv_compute_frame_info): Only compute frame->mask if not a naked
12293 (riscv_expand_prologue): Add early return for naked function.
12294 (riscv_expand_epilogue): Likewise.
12295 (riscv_function_ok_for_sibcall): Return false for naked function.
12296 (riscv_set_current_function): New.
12297 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
12298 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
12299 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
12300 * doc/extend.texi (RISC-V Function Attributes): New.
12302 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
12304 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
12305 check for 128-bit long double before checking TCmode.
12306 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
12307 128-bit long doubles before checking TFmode or TCmode.
12308 (FLOAT128_IBM_P): Likewise.
12310 2018-01-10 Martin Sebor <msebor@redhat.com>
12312 PR tree-optimization/83671
12313 * builtins.c (c_strlen): Unconditionally return zero for the empty
12315 Use -Warray-bounds for warnings.
12316 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
12317 for non-constant array indices with COMPONENT_REF, arrays of
12318 arrays, and pointers to arrays.
12319 (gimple_fold_builtin_strlen): Determine and set length range for
12320 non-constant character arrays.
12322 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
12324 PR middle-end/81897
12325 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
12328 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
12330 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
12332 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12335 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
12336 VECTOR_MEM_ALTIVEC_OR_VSX_P.
12337 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
12338 indexed_or_indirect_operand predicate.
12339 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
12340 (*vsx_le_perm_load_v8hi): Likewise.
12341 (*vsx_le_perm_load_v16qi): Likewise.
12342 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
12343 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
12344 (*vsx_le_perm_store_v8hi): Likewise.
12345 (*vsx_le_perm_store_v16qi): Likewise.
12346 (eight unnamed splitters): Likewise.
12348 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12350 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
12351 * config/rs6000/emmintrin.h: Likewise.
12352 * config/rs6000/mmintrin.h: Likewise.
12353 * config/rs6000/xmmintrin.h: Likewise.
12355 2018-01-10 David Malcolm <dmalcolm@redhat.com>
12358 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
12360 * tree.c (tree_nop_conversion): Return true for location wrapper
12362 (maybe_wrap_with_location): New function.
12363 (selftest::check_strip_nops): New function.
12364 (selftest::test_location_wrappers): New function.
12365 (selftest::tree_c_tests): Call it.
12366 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
12367 (maybe_wrap_with_location): New decl.
12368 (EXPR_LOCATION_WRAPPER_P): New macro.
12369 (location_wrapper_p): New inline function.
12370 (tree_strip_any_location_wrapper): New inline function.
12372 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
12375 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
12376 stack_realign_offset for the largest alignment of stack slot
12378 (ix86_find_max_used_stack_alignment): New function.
12379 (ix86_finalize_stack_frame_flags): Use it. Set
12380 max_used_stack_alignment if we don't realign stack.
12381 * config/i386/i386.h (machine_function): Add
12382 max_used_stack_alignment.
12384 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
12386 * config/arm/arm.opt (-mbranch-cost): New option.
12387 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
12390 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
12393 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
12394 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
12396 2018-01-10 Richard Biener <rguenther@suse.de>
12399 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
12400 early out so it also covers the case where we have a non-NULL
12403 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12405 PR tree-optimization/83753
12406 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
12407 for non-strided grouped accesses if the number of elements is 1.
12409 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12412 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
12413 * i386.h (TARGET_USE_GATHER): Define.
12414 * x86-tune.def (X86_TUNE_USE_GATHER): New.
12416 2018-01-10 Martin Liska <mliska@suse.cz>
12419 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
12420 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
12422 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
12423 CLEANUP_NO_PARTITIONING is not set.
12425 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12427 * doc/rtl.texi: Remove documentation of (const ...) wrappers
12428 for vectors, as a partial revert of r254296.
12429 * rtl.h (const_vec_p): Delete.
12430 (const_vec_duplicate_p): Don't test for vector CONSTs.
12431 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
12432 * expmed.c (make_tree): Likewise.
12435 * common.md (E, F): Use CONSTANT_P instead of checking for
12437 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
12438 checking for CONST_VECTOR.
12440 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12442 PR middle-end/83575
12443 * predict.c (force_edge_cold): Handle in more sane way edges
12444 with no prediction.
12446 2018-01-09 Carl Love <cel@us.ibm.com>
12448 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
12450 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
12451 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
12452 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
12453 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
12454 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
12455 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
12456 * config/rs6000/rs6000-protos.h: Add extern defition for
12457 rs6000_generate_float2_double_code.
12458 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
12460 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
12461 (float2_v2df): Add define_expand.
12463 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
12466 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
12467 op_mode in the force_to_mode call.
12469 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12471 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
12472 instead of checking each element individually.
12473 (aarch64_evpc_uzp): Likewise.
12474 (aarch64_evpc_zip): Likewise.
12475 (aarch64_evpc_ext): Likewise.
12476 (aarch64_evpc_rev): Likewise.
12477 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
12478 instead of checking each element individually. Return true without
12480 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
12481 whether all selected elements come from the same input, instead of
12482 checking each element individually. Remove calls to gen_rtx_REG,
12483 start_sequence and end_sequence and instead assert that no rtl is
12486 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12488 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
12489 order of HIGH and CONST checks.
12491 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12493 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
12494 if the destination isn't an SSA_NAME.
12496 2018-01-09 Richard Biener <rguenther@suse.de>
12498 PR tree-optimization/83668
12499 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
12501 (canonicalize_loop_form): ... here, renamed from ...
12502 (canonicalize_loop_closed_ssa_form): ... this and amended to
12503 swap successor edges for loop exit blocks to make us use
12504 the RPO order we need for initial schedule generation.
12506 2018-01-09 Joseph Myers <joseph@codesourcery.com>
12508 PR tree-optimization/64811
12509 * match.pd: When optimizing comparisons with Inf, avoid
12510 introducing or losing exceptions from comparisons with NaN.
12512 2018-01-09 Martin Liska <mliska@suse.cz>
12515 * asan.c (shadow_mem_size): Add gcc_assert.
12517 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
12519 Don't save registers in main().
12522 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
12523 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
12524 * config/avr/avr.c (avr_set_current_function): Don't error if
12525 naked, OS_task or OS_main are specified at the same time.
12526 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
12528 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
12530 * common/config/avr/avr-common.c (avr_option_optimization_table):
12531 Switch on -mmain-is-OS_task for optimizing compilations.
12533 2018-01-09 Richard Biener <rguenther@suse.de>
12535 PR tree-optimization/83572
12536 * graphite.c: Include cfganal.h.
12537 (graphite_transform_loops): Connect infinite loops to exit
12538 and remove fake edges at the end.
12540 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12542 * ipa-inline.c (edge_badness): Revert accidental checkin.
12544 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12547 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
12548 symbols; not inline clones.
12550 2018-01-09 Jakub Jelinek <jakub@redhat.com>
12553 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
12554 hard registers. Formatting fixes.
12556 PR preprocessor/83722
12557 * gcc.c (try_generate_repro): Pass
12558 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
12559 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
12562 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
12563 Kito Cheng <kito.cheng@gmail.com>
12565 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
12566 (riscv_leaf_function_p): Delete.
12567 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
12569 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12571 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
12573 (do_ifelse): New function.
12574 (do_isel): New function.
12575 (do_sub3): New function.
12576 (do_add3): New function.
12577 (do_load_mask_compare): New function.
12578 (do_overlap_load_compare): New function.
12579 (expand_compare_loop): New function.
12580 (expand_block_compare): Call expand_compare_loop() when appropriate.
12581 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
12582 option description.
12583 (-mblock-compare-inline-loop-limit): New option.
12585 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12588 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
12589 Reverse order of second and third operands in first alternative.
12590 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
12591 of first and second elements in UNSPEC_VPERMR vector.
12592 (altivec_expand_vec_perm_le): Likewise.
12594 2018-01-08 Jeff Law <law@redhat.com>
12596 PR rtl-optimizatin/81308
12597 * tree-switch-conversion.c (cfg_altered): New file scoped static.
12598 (process_switch): If group_case_labels makes a change, then set
12600 (pass_convert_switch::execute): If a switch is converted, then
12601 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
12603 PR rtl-optimization/81308
12604 * recog.c (split_all_insns): Conditionally cleanup the CFG after
12607 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
12609 PR target/83663 - Revert r255946
12610 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
12611 generation for cases where splatting a value is not useful.
12612 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
12613 across a vec_duplicate and a paradoxical subreg forming a vector
12614 mode to a vec_concat.
12616 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12618 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
12619 -march=armv8.3-a variants.
12620 * config/arm/t-multilib: Likewise.
12621 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
12623 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12625 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
12627 (cceq_ior_compare_complement): Give it a name so I can use it, and
12628 change boolean_or_operator predicate to boolean_operator so it can
12629 be used to generate a crand.
12630 (eqne): New code iterator.
12631 (bd/bd_neg): New code_attrs.
12632 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
12633 a single define_insn.
12634 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
12635 decrement (bdnzt/bdnzf/bdzt/bdzf).
12636 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
12637 with the new names of the branch decrement patterns, and added the
12638 names of the branch decrement conditional patterns.
12640 2018-01-08 Richard Biener <rguenther@suse.de>
12642 PR tree-optimization/83563
12643 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
12646 2018-01-08 Richard Biener <rguenther@suse.de>
12648 PR middle-end/83713
12649 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
12651 2018-01-08 Richard Biener <rguenther@suse.de>
12653 PR tree-optimization/83685
12654 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
12655 references to abnormals.
12657 2018-01-08 Richard Biener <rguenther@suse.de>
12660 * dwarf2out.c (output_indirect_strings): Handle empty
12661 skeleton_debug_str_hash.
12662 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
12664 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12666 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
12667 (emit_store_direct): Likewise.
12668 (arc_trampoline_adjust_address): Likewise.
12669 (arc_asm_trampoline_template): New function.
12670 (arc_initialize_trampoline): Use asm_trampoline_template.
12671 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
12672 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
12673 * config/arc/arc.md (flush_icache): Delete pattern.
12675 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12677 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
12678 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
12681 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12684 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
12685 by not USED_FOR_TARGET.
12686 (make_pass_resolve_sw_modes): Likewise.
12688 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12690 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
12693 2018-01-08 Richard Biener <rguenther@suse.de>
12695 PR middle-end/83580
12696 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
12698 2018-01-08 Richard Biener <rguenther@suse.de>
12700 PR middle-end/83517
12701 * match.pd ((t * 2) / 2) -> t): Add missing :c.
12703 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
12705 PR middle-end/81897
12706 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
12707 basic blocks with a small number of successors.
12708 (convert_control_dep_chain_into_preds): Improve handling of
12710 (dump_predicates): Split apart into...
12711 (dump_pred_chain): ...here...
12712 (dump_pred_info): ...and here.
12713 (can_one_predicate_be_invalidated_p): Add debugging printfs.
12714 (can_chain_union_be_invalidated_p): Improve check for invalidation
12716 (uninit_uses_cannot_happen): Avoid unnecessary if
12717 convert_control_dep_chain_into_preds yielded nothing.
12719 2018-01-06 Martin Sebor <msebor@redhat.com>
12721 PR tree-optimization/83640
12722 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
12723 subtracting negative offset from size.
12724 (builtin_access::overlap): Adjust offset bounds of the access to fall
12725 within the size of the object if possible.
12727 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
12729 PR rtl-optimization/83699
12730 * expmed.c (extract_bit_field_1): Restrict the vector usage of
12731 extract_bit_field_as_subreg to cases in which the extracted
12732 value is also a vector.
12734 * lra-constraints.c (process_alt_operands): Test for the equivalence
12735 substitutions when detecting a possible reload cycle.
12737 2018-01-06 Jakub Jelinek <jakub@redhat.com>
12740 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
12741 by default if flag_selective_schedling{,2}. Formatting fixes.
12743 PR rtl-optimization/83682
12744 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
12745 if it has non-VECTOR_MODE element mode.
12746 (vec_duplicate_p): Likewise.
12748 PR middle-end/83694
12749 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
12750 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
12752 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12755 * config/i386/i386-builtin.def
12756 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
12757 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
12758 Require also OPTION_MASK_ISA_AVX512F in addition to
12759 OPTION_MASK_ISA_GFNI.
12760 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
12761 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
12762 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
12763 to OPTION_MASK_ISA_GFNI.
12764 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
12765 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
12766 OPTION_MASK_ISA_AVX512BW.
12767 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
12768 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
12769 addition to OPTION_MASK_ISA_GFNI.
12770 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
12771 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
12772 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
12773 to OPTION_MASK_ISA_GFNI.
12774 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
12775 a requirement for all ISAs rather than any of them with a few
12777 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
12779 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
12780 bitmasks to be enabled with 3 exceptions, instead of requiring any
12781 enabled ISA with lots of exceptions.
12782 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
12783 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
12784 Change avx512bw in isa attribute to avx512f.
12785 * config/i386/sgxintrin.h: Add license boilerplate.
12786 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
12787 to __AVX512F__ and __AVX512VL to __AVX512VL__.
12788 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
12789 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
12791 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
12792 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
12793 temporarily sse2 rather than sse if not enabled already.
12796 * config/i386/sse.md (VI248_VLBW): Rename to ...
12797 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
12798 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
12799 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
12800 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
12801 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
12802 mode iterator instead of VI248_VLBW.
12804 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
12806 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
12807 (record_modified): Skip clobbers; add debug output.
12808 (param_change_prob): Use sreal frequencies.
12810 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12812 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
12813 punt for user-aligned variables.
12815 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12817 * tree-chrec.c (chrec_contains_symbols): Return true for
12820 2018-01-05 Sudakshina Das <sudi.das@arm.com>
12823 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
12824 of (x|y) == x for BICS pattern.
12826 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12828 PR tree-optimization/83605
12829 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
12830 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
12833 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
12835 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
12836 * config/epiphany/rtems.h: New file.
12838 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12839 Uros Bizjak <ubizjak@gmail.com>
12842 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
12843 QIreg_operand instead of register_operand predicate.
12844 * config/i386/i386.c (ix86_rop_should_change_byte_p,
12845 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
12846 comments instead of -fmitigate[-_]rop.
12848 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12851 * cgraphunit.c (symbol_table::compile): Switch to text_section
12852 before calling assembly_start debug hook.
12853 * run-rtl-passes.c (run_rtl_passes): Likewise.
12856 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12858 * tree-vrp.c (extract_range_from_binary_expr_1): Check
12859 range_int_cst_p rather than !symbolic_range_p before calling
12860 extract_range_from_multiplicative_op_1.
12862 2018-01-04 Jeff Law <law@redhat.com>
12864 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
12865 redundant test in assertion.
12867 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12869 * doc/rtl.texi: Document machine_mode wrapper classes.
12871 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12873 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
12874 using tree_to_uhwi.
12876 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12878 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
12879 the VEC_PERM_EXPR fold to fail.
12881 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12884 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
12885 to switched_sections.
12887 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12890 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
12891 test for d.testing.
12893 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
12896 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
12897 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
12899 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12902 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
12903 is BLKmode and bitpos not zero or mode change is needed.
12905 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12908 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
12911 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
12914 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
12915 instead of MULT rtx. Update all corresponding splitters.
12916 (*saddl_se): Ditto.
12917 (*ssub<modesuffix>): Ditto.
12918 (*ssubl_se): Ditto.
12919 (*cmp_sadd_di): Update split patterns.
12920 (*cmp_sadd_si): Ditto.
12921 (*cmp_sadd_sidi): Ditto.
12922 (*cmp_ssub_di): Ditto.
12923 (*cmp_ssub_si): Ditto.
12924 (*cmp_ssub_sidi): Ditto.
12925 * config/alpha/predicates.md (const23_operand): New predicate.
12926 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
12927 Look for ASHIFT, not MULT inner operand.
12928 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
12930 2018-01-04 Martin Liska <mliska@suse.cz>
12932 PR gcov-profile/83669
12933 * gcov.c (output_intermediate_file): Add version to intermediate
12935 * doc/gcov.texi: Document new field 'version' in intermediate
12936 file format. Fix location of '-k' option of gcov command.
12938 2018-01-04 Martin Liska <mliska@suse.cz>
12941 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
12943 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12945 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
12947 2018-01-03 Martin Sebor <msebor@redhat.com>
12949 PR tree-optimization/83655
12950 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
12951 checking calls with invalid arguments.
12953 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12955 * tree-vect-stmts.c (vect_get_store_rhs): New function.
12956 (vectorizable_mask_load_store): Delete.
12957 (vectorizable_call): Return false for masked loads and stores.
12958 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
12959 instead of gimple_assign_rhs1.
12960 (vectorizable_load): Handle IFN_MASK_LOAD.
12961 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
12963 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12965 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
12967 (vectorizable_mask_load_store): ...here.
12968 (vectorizable_load): ...and here.
12970 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12972 * tree-vect-stmts.c (vect_build_all_ones_mask)
12973 (vect_build_zero_merge_argument): New functions, split out from...
12974 (vectorizable_load): ...here.
12976 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12978 * tree-vect-stmts.c (vect_check_store_rhs): New function,
12980 (vectorizable_mask_load_store): ...here.
12981 (vectorizable_store): ...and here.
12983 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12985 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
12987 (vectorizable_mask_load_store): ...here.
12989 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12991 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
12992 (vect_model_store_cost): Take a vec_load_store_type instead of a
12994 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
12995 (vect_model_store_cost): Take a vec_load_store_type instead of a
12997 (vectorizable_mask_load_store): Update accordingly.
12998 (vectorizable_store): Likewise.
12999 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
13001 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13003 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
13004 IFN_MASK_LOAD calls here rather than...
13005 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
13007 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13008 Alan Hayward <alan.hayward@arm.com>
13009 David Sherwood <david.sherwood@arm.com>
13011 * expmed.c (extract_bit_field_1): For vector extracts,
13012 fall back to extract_bit_field_as_subreg if vec_extract
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 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
13020 they are variable or constant sized.
13021 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
13022 slots for constant-sized data.
13024 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13025 Alan Hayward <alan.hayward@arm.com>
13026 David Sherwood <david.sherwood@arm.com>
13028 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
13029 handling COND_EXPRs with boolean comparisons, try to find a better
13030 basis for the mask type than the boolean itself.
13032 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13034 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
13035 is calculated and how it can be overridden.
13036 * genmodes.c (max_bitsize_mode_any_mode): New variable.
13037 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
13039 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
13042 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13043 Alan Hayward <alan.hayward@arm.com>
13044 David Sherwood <david.sherwood@arm.com>
13046 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
13047 Remove the mode argument.
13048 (aarch64_simd_valid_immediate): Remove the mode and inverse
13050 * config/aarch64/iterators.md (bitsize): New iterator.
13051 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
13052 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
13053 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
13054 aarch64_simd_valid_immediate.
13055 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
13056 (aarch64_reg_or_bic_imm): Likewise.
13057 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
13058 with an insn_type enum and msl with a modifier_type enum.
13059 Replace element_width with a scalar_mode. Change the shift
13060 to unsigned int. Add constructors for scalar_float_mode and
13061 scalar_int_mode elements.
13062 (aarch64_vect_float_const_representable_p): Delete.
13063 (aarch64_can_const_movi_rtx_p)
13064 (aarch64_simd_scalar_immediate_valid_for_move)
13065 (aarch64_simd_make_constant): Update call to
13066 aarch64_simd_valid_immediate.
13067 (aarch64_advsimd_valid_immediate_hs): New function.
13068 (aarch64_advsimd_valid_immediate): Likewise.
13069 (aarch64_simd_valid_immediate): Remove mode and inverse
13070 arguments. Rewrite to use the above. Use const_vec_duplicate_p
13071 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
13072 and aarch64_float_const_representable_p on the result.
13073 (aarch64_output_simd_mov_immediate): Remove mode argument.
13074 Update call to aarch64_simd_valid_immediate and use of
13075 simd_immediate_info.
13076 (aarch64_output_scalar_simd_mov_immediate): Update call
13079 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13080 Alan Hayward <alan.hayward@arm.com>
13081 David Sherwood <david.sherwood@arm.com>
13083 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
13084 (mode_nunits): Likewise CONST_MODE_NUNITS.
13085 * machmode.def (ADJUST_NUNITS): Document.
13086 * genmodes.c (mode_data::need_nunits_adj): New field.
13087 (blank_mode): Update accordingly.
13088 (adj_nunits): New variable.
13089 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
13091 (emit_mode_size_inline): Set need_bytesize_adj for all modes
13092 listed in adj_nunits.
13093 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
13094 listed in adj_nunits. Don't emit case statements for such modes.
13095 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
13096 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
13097 nothing if adj_nunits is nonnull.
13098 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
13099 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
13100 (emit_mode_fbit): Update use of print_maybe_const_decl.
13101 (emit_move_size): Likewise. Treat the array as non-const
13103 (emit_mode_adjustments): Handle adj_nunits.
13105 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13107 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
13108 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
13109 (VECTOR_MODES): Use it.
13110 (make_vector_modes): Take the prefix as an argument.
13112 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13113 Alan Hayward <alan.hayward@arm.com>
13114 David Sherwood <david.sherwood@arm.com>
13116 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
13117 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
13118 for MODE_VECTOR_BOOL.
13119 * machmode.def (VECTOR_BOOL_MODE): Document.
13120 * genmodes.c (VECTOR_BOOL_MODE): New macro.
13121 (make_vector_bool_mode): New function.
13122 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
13124 * lto-streamer-in.c (lto_input_mode_table): Likewise.
13125 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
13127 * stor-layout.c (int_mode_for_mode): Likewise.
13128 * tree.c (build_vector_type_for_mode): Likewise.
13129 * varasm.c (output_constant_pool_2): Likewise.
13130 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
13131 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
13132 for MODE_VECTOR_BOOL.
13133 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
13134 of mode class checks.
13135 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
13136 instead of a list of mode class checks.
13137 (expand_vector_scalar_condition): Likewise.
13138 (type_for_widest_vector_mode): Handle BImode as an inner mode.
13140 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13141 Alan Hayward <alan.hayward@arm.com>
13142 David Sherwood <david.sherwood@arm.com>
13144 * machmode.h (mode_size): Change from unsigned short to
13146 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
13147 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13148 or if measurement_type is not polynomial.
13149 (fixed_size_mode::includes_p): Check for constant-sized modes.
13150 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
13151 return a poly_uint16 rather than an unsigned short.
13152 (emit_mode_size): Change the type of mode_size from unsigned short
13153 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
13154 (emit_mode_adjustments): Cope with polynomial vector sizes.
13155 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13157 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13159 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
13160 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
13161 * caller-save.c (setup_save_areas): Likewise.
13162 (replace_reg_with_saved_mem): Likewise.
13163 * calls.c (emit_library_call_value_1): Likewise.
13164 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
13165 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
13166 (gen_lowpart_for_combine): Likewise.
13167 * convert.c (convert_to_integer_1): Likewise.
13168 * cse.c (equiv_constant, cse_insn): Likewise.
13169 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
13170 (cselib_subst_to_values): Likewise.
13171 * dce.c (word_dce_process_block): Likewise.
13172 * df-problems.c (df_word_lr_mark_ref): Likewise.
13173 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
13174 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
13175 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
13176 (rtl_for_decl_location): Likewise.
13177 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
13178 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
13179 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
13180 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
13181 (expand_expr_real_1): Likewise.
13182 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
13183 (pad_below): Likewise.
13184 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13185 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
13186 * ira.c (get_subreg_tracking_sizes): Likewise.
13187 * ira-build.c (ira_create_allocno_objects): Likewise.
13188 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
13189 (ira_sort_regnos_for_alter_reg): Likewise.
13190 * ira-costs.c (record_operand_costs): Likewise.
13191 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
13192 (resolve_simple_move): Likewise.
13193 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
13194 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
13195 (lra_constraints): Likewise.
13196 (CONST_POOL_OK_P): Reject variable-sized modes.
13197 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
13198 (add_pseudo_to_slot, lra_spill): Likewise.
13199 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13200 * optabs-query.c (get_best_extraction_insn): Likewise.
13201 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13202 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
13203 (expand_mult_highpart, valid_multiword_target_p): Likewise.
13204 * recog.c (offsettable_address_addr_space_p): Likewise.
13205 * regcprop.c (maybe_mode_change): Likewise.
13206 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
13207 * regrename.c (build_def_use): Likewise.
13208 * regstat.c (dump_reg_info): Likewise.
13209 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
13210 (find_reloads, find_reloads_subreg_address): Likewise.
13211 * reload1.c (eliminate_regs_1): Likewise.
13212 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
13213 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
13214 (simplify_binary_operation_1, simplify_subreg): Likewise.
13215 * targhooks.c (default_function_arg_padding): Likewise.
13216 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
13217 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
13218 (verify_gimple_assign_ternary): Likewise.
13219 * tree-inline.c (estimate_move_cost): Likewise.
13220 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13221 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
13222 (get_address_cost_ainc): Likewise.
13223 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
13224 (vect_supportable_dr_alignment): Likewise.
13225 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13226 (vectorizable_reduction): Likewise.
13227 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
13228 (vectorizable_operation, vectorizable_load): Likewise.
13229 * tree.c (build_same_sized_truth_vector_type): Likewise.
13230 * valtrack.c (cleanup_auto_inc_dec): Likewise.
13231 * var-tracking.c (emit_note_insn_var_location): Likewise.
13232 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
13233 (ADDR_VEC_ALIGN): Likewise.
13235 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13236 Alan Hayward <alan.hayward@arm.com>
13237 David Sherwood <david.sherwood@arm.com>
13239 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
13241 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13242 or if measurement_type is polynomial.
13243 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
13244 * combine.c (make_extraction): Likewise.
13245 * dse.c (find_shift_sequence): Likewise.
13246 * dwarf2out.c (mem_loc_descriptor): Likewise.
13247 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
13248 (extract_bit_field, extract_low_bits): Likewise.
13249 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
13250 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
13251 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
13252 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
13253 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13254 * reload.c (find_reloads): Likewise.
13255 * reload1.c (alter_reg): Likewise.
13256 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
13257 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
13258 * tree-if-conv.c (predicate_mem_writes): Likewise.
13259 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
13260 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
13261 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
13262 * valtrack.c (dead_debug_insert_temp): Likewise.
13263 * varasm.c (mergeable_constant_section): Likewise.
13264 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
13266 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13267 Alan Hayward <alan.hayward@arm.com>
13268 David Sherwood <david.sherwood@arm.com>
13270 * expr.c (expand_assignment): Cope with polynomial mode sizes
13271 when assigning to a CONCAT.
13273 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13274 Alan Hayward <alan.hayward@arm.com>
13275 David Sherwood <david.sherwood@arm.com>
13277 * machmode.h (mode_precision): Change from unsigned short to
13279 (mode_to_precision): Return a poly_uint16 rather than an unsigned
13281 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
13282 or if measurement_type is not polynomial.
13283 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
13284 in which the mode is already known to be a scalar_int_mode.
13285 * genmodes.c (emit_mode_precision): Change the type of mode_precision
13286 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
13288 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13289 for GET_MODE_PRECISION.
13290 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13291 for GET_MODE_PRECISION.
13292 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
13294 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
13295 (expand_field_assignment, make_extraction): Likewise.
13296 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
13297 (get_last_value): Likewise.
13298 * convert.c (convert_to_integer_1): Likewise.
13299 * cse.c (cse_insn): Likewise.
13300 * expr.c (expand_expr_real_1): Likewise.
13301 * lra-constraints.c (simplify_operand_subreg): Likewise.
13302 * optabs-query.c (can_atomic_load_p): Likewise.
13303 * optabs.c (expand_atomic_load): Likewise.
13304 (expand_atomic_store): Likewise.
13305 * ree.c (combine_reaching_defs): Likewise.
13306 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
13307 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
13308 * tree.h (type_has_mode_precision_p): Likewise.
13309 * ubsan.c (instrument_si_overflow): Likewise.
13311 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13312 Alan Hayward <alan.hayward@arm.com>
13313 David Sherwood <david.sherwood@arm.com>
13315 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
13316 polynomial numbers of units.
13317 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
13318 (valid_vector_subparts_p): New function.
13319 (build_vector_type): Remove temporary shim and take the number
13320 of units as a poly_uint64 rather than an int.
13321 (build_opaque_vector_type): Take the number of units as a
13322 poly_uint64 rather than an int.
13323 * tree.c (build_vector_from_ctor): Handle polynomial
13324 TYPE_VECTOR_SUBPARTS.
13325 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
13326 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
13327 (build_vector_from_val): If the number of units is variable,
13328 use build_vec_duplicate_cst for constant operands and
13329 VEC_DUPLICATE_EXPR otherwise.
13330 (make_vector_type): Remove temporary is_constant ().
13331 (build_vector_type, build_opaque_vector_type): Take the number of
13332 units as a poly_uint64 rather than an int.
13333 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
13335 * cfgexpand.c (expand_debug_expr): Likewise.
13336 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
13337 (store_constructor, expand_expr_real_1): Likewise.
13338 (const_scalar_mask_from_tree): Likewise.
13339 * fold-const-call.c (fold_const_reduction): Likewise.
13340 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
13341 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
13342 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
13343 (fold_relational_const): Likewise.
13344 (native_interpret_vector): Likewise. Change the size from an
13345 int to an unsigned int.
13346 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
13347 TYPE_VECTOR_SUBPARTS.
13348 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
13349 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
13350 duplicating a non-constant operand into a variable-length vector.
13351 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
13352 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
13353 * ipa-icf.c (sem_variable::equals): Likewise.
13354 * match.pd: Likewise.
13355 * omp-simd-clone.c (simd_clone_subparts): Likewise.
13356 * print-tree.c (print_node): Likewise.
13357 * stor-layout.c (layout_type): Likewise.
13358 * targhooks.c (default_builtin_vectorization_cost): Likewise.
13359 * tree-cfg.c (verify_gimple_comparison): Likewise.
13360 (verify_gimple_assign_binary): Likewise.
13361 (verify_gimple_assign_ternary): Likewise.
13362 (verify_gimple_assign_single): Likewise.
13363 * tree-pretty-print.c (dump_generic_node): Likewise.
13364 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13365 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
13366 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
13367 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
13368 (vect_shift_permute_load_chain): Likewise.
13369 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
13370 (expand_vector_condition, optimize_vector_constructor): Likewise.
13371 (lower_vec_perm, get_compute_type): Likewise.
13372 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13373 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
13374 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
13375 (vect_recog_mask_conversion_pattern): Likewise.
13376 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
13377 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
13378 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13379 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
13380 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
13381 (vectorizable_shift, vectorizable_operation, vectorizable_store)
13382 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
13383 (supportable_widening_operation): Likewise.
13384 (supportable_narrowing_operation): Likewise.
13385 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
13387 * varasm.c (output_constant): Likewise.
13389 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13390 Alan Hayward <alan.hayward@arm.com>
13391 David Sherwood <david.sherwood@arm.com>
13393 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
13394 so that both the length == 3 and length != 3 cases set up their
13395 own permute vectors. Add comments explaining why we know the
13396 number of elements is constant.
13397 (vect_permute_load_chain): Likewise.
13399 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13400 Alan Hayward <alan.hayward@arm.com>
13401 David Sherwood <david.sherwood@arm.com>
13403 * machmode.h (mode_nunits): Change from unsigned char to
13405 (ONLY_FIXED_SIZE_MODES): New macro.
13406 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
13407 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
13408 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
13410 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
13411 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
13412 or if measurement_type is not polynomial.
13413 * genmodes.c (ZERO_COEFFS): New macro.
13414 (emit_mode_nunits_inline): Make mode_nunits_inline return a
13416 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
13417 Use ZERO_COEFFS when emitting initializers.
13418 * data-streamer.h (bp_pack_poly_value): New function.
13419 (bp_unpack_poly_value): Likewise.
13420 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13421 for GET_MODE_NUNITS.
13422 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13423 for GET_MODE_NUNITS.
13424 * tree.c (make_vector_type): Remove temporary shim and make
13425 the real function take the number of units as a poly_uint64
13426 rather than an int.
13427 (build_vector_type_for_mode): Handle polynomial nunits.
13428 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
13429 * emit-rtl.c (const_vec_series_p_1): Likewise.
13430 (gen_rtx_CONST_VECTOR): Likewise.
13431 * fold-const.c (test_vec_duplicate_folding): Likewise.
13432 * genrecog.c (validate_pattern): Likewise.
13433 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
13434 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13435 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
13436 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
13437 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
13438 * rtlanal.c (subreg_get_info): Likewise.
13439 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13440 (vect_grouped_load_supported): Likewise.
13441 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
13442 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
13443 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
13444 (simplify_const_unary_operation, simplify_binary_operation_1)
13445 (simplify_const_binary_operation, simplify_ternary_operation)
13446 (test_vector_ops_duplicate, test_vector_ops): Likewise.
13447 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
13448 instead of CONST_VECTOR_NUNITS.
13449 * varasm.c (output_constant_pool_2): Likewise.
13450 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
13451 explicit-encoded elements in the XVEC for variable-length vectors.
13453 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13455 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
13457 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13458 Alan Hayward <alan.hayward@arm.com>
13459 David Sherwood <david.sherwood@arm.com>
13461 * coretypes.h (fixed_size_mode): Declare.
13462 (fixed_size_mode_pod): New typedef.
13463 * builtins.h (target_builtins::x_apply_args_mode)
13464 (target_builtins::x_apply_result_mode): Change type to
13465 fixed_size_mode_pod.
13466 * builtins.c (apply_args_size, apply_result_size, result_vector)
13467 (expand_builtin_apply_args_1, expand_builtin_apply)
13468 (expand_builtin_return): Update accordingly.
13470 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13472 * cse.c (hash_rtx_cb): Hash only the encoded elements.
13473 * cselib.c (cselib_hash_rtx): Likewise.
13474 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
13475 CONST_VECTOR encoding.
13477 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13478 Jeff Law <law@redhat.com>
13481 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
13482 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
13483 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
13484 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
13487 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
13488 explicitly probe *sp in a noreturn function if there were any callee
13489 register saves or frame pointer is needed.
13491 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13494 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
13495 BLKmode for ternary, binary or unary expressions.
13498 * var-tracking.c (delete_vta_debug_insn): New inline function.
13499 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
13500 insns from get_insns () to NULL instead of each bb separately.
13501 Use delete_vta_debug_insn. No longer static.
13502 (vt_debug_insns_local, variable_tracking_main_1): Adjust
13503 delete_vta_debug_insns callers.
13504 * rtl.h (delete_vta_debug_insns): Declare.
13505 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
13506 instead of variable_tracking_main.
13508 2018-01-03 Martin Sebor <msebor@redhat.com>
13510 PR tree-optimization/83603
13511 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
13512 arguments past the endof the argument list in functions declared
13513 without a prototype.
13514 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
13515 Avoid checking when arguments are null.
13517 2018-01-03 Martin Sebor <msebor@redhat.com>
13520 * doc/extend.texi (attribute const): Fix a typo.
13521 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
13522 issuing -Wsuggest-attribute for void functions.
13524 2018-01-03 Martin Sebor <msebor@redhat.com>
13526 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
13527 offset_int::from instead of wide_int::to_shwi.
13528 (maybe_diag_overlap): Remove assertion.
13529 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
13530 * gimple-ssa-sprintf.c (format_directive): Same.
13531 (parse_directive): Same.
13532 (sprintf_dom_walker::compute_format_length): Same.
13533 (try_substitute_return_value): Same.
13535 2018-01-03 Jeff Law <law@redhat.com>
13537 PR middle-end/83654
13538 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
13539 non-constant residual for zero at runtime and avoid probing in
13540 that case. Reorganize code for trailing problem to mirror handling
13543 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13545 PR tree-optimization/83501
13546 * tree-ssa-strlen.c (get_string_cst): New.
13547 (handle_char_store): Call get_string_cst.
13549 2018-01-03 Martin Liska <mliska@suse.cz>
13551 PR tree-optimization/83593
13552 * tree-ssa-strlen.c: Include tree-cfg.h.
13553 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
13554 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
13555 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
13557 (strlen_dom_walker::before_dom_children): Call
13558 gimple_purge_dead_eh_edges. Dump tranformation with details
13560 (strlen_dom_walker::before_dom_children): Update call by adding
13561 new argument cleanup_eh.
13562 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
13564 2018-01-03 Martin Liska <mliska@suse.cz>
13567 * cif-code.def (VARIADIC_THUNK): New enum value.
13568 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
13571 2018-01-03 Jan Beulich <jbeulich@suse.com>
13573 * sse.md (mov<mode>_internal): Tighten condition for when to use
13574 vmovdqu<ssescalarsize> for TI and OI modes.
13576 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13578 Update copyright years.
13580 2018-01-03 Martin Liska <mliska@suse.cz>
13583 * ipa-visibility.c (function_and_variable_visibility): Skip
13584 functions with noipa attribure.
13586 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13588 * gcc.c (process_command): Update copyright notice dates.
13589 * gcov-dump.c (print_version): Ditto.
13590 * gcov.c (print_version): Ditto.
13591 * gcov-tool.c (print_version): Ditto.
13592 * gengtype.c (create_file): Ditto.
13593 * doc/cpp.texi: Bump @copying's copyright year.
13594 * doc/cppinternals.texi: Ditto.
13595 * doc/gcc.texi: Ditto.
13596 * doc/gccint.texi: Ditto.
13597 * doc/gcov.texi: Ditto.
13598 * doc/install.texi: Ditto.
13599 * doc/invoke.texi: Ditto.
13601 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13603 * vector-builder.h (vector_builder::m_full_nelts): Change from
13604 unsigned int to poly_uint64.
13605 (vector_builder::full_nelts): Update prototype accordingly.
13606 (vector_builder::new_vector): Likewise.
13607 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
13608 (vector_builder::operator ==): Likewise.
13609 (vector_builder::finalize): Likewise.
13610 * int-vector-builder.h (int_vector_builder::int_vector_builder):
13611 Take the number of elements as a poly_uint64 rather than an
13613 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
13614 from unsigned int to poly_uint64.
13615 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
13616 (vec_perm_indices::new_vector): Likewise.
13617 (vec_perm_indices::length): Likewise.
13618 (vec_perm_indices::nelts_per_input): Likewise.
13619 (vec_perm_indices::input_nelts): Likewise.
13620 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
13621 number of elements per input as a poly_uint64 rather than an
13622 unsigned int. Use the original encoding for variable-length
13623 vectors, rather than clamping each individual element.
13624 For the second and subsequent elements in each pattern,
13625 clamp the step and base before clamping their sum.
13626 (vec_perm_indices::series_p): Handle polynomial element counts.
13627 (vec_perm_indices::all_in_range_p): Likewise.
13628 (vec_perm_indices_to_tree): Likewise.
13629 (vec_perm_indices_to_rtx): Likewise.
13630 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
13631 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
13632 (tree_vector_builder::new_binary_operation): Handle polynomial
13633 element counts. Return false if we need to know the number
13634 of elements at compile time.
13635 * fold-const.c (fold_vec_perm): Punt if the number of elements
13636 isn't known at compile time.
13638 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13640 * vec-perm-indices.h (vec_perm_builder): Change element type
13641 from HOST_WIDE_INT to poly_int64.
13642 (vec_perm_indices::element_type): Update accordingly.
13643 (vec_perm_indices::clamp): Handle polynomial element_types.
13644 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13645 (vec_perm_indices::all_in_range_p): Likewise.
13646 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
13648 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
13649 polynomial vec_perm_indices element types.
13650 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
13651 * fold-const.c (fold_vec_perm): Likewise.
13652 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
13653 * tree-vect-generic.c (lower_vec_perm): Likewise.
13654 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13655 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
13656 element type to HOST_WIDE_INT.
13658 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13659 Alan Hayward <alan.hayward@arm.com>
13660 David Sherwood <david.sherwood@arm.com>
13662 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
13663 rather than an int. Use plus_constant.
13664 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
13665 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
13667 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13668 Alan Hayward <alan.hayward@arm.com>
13669 David Sherwood <david.sherwood@arm.com>
13671 * calls.c (emit_call_1, expand_call): Change struct_value_size from
13672 a HOST_WIDE_INT to a poly_int64.
13674 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13675 Alan Hayward <alan.hayward@arm.com>
13676 David Sherwood <david.sherwood@arm.com>
13678 * calls.c (load_register_parameters): Cope with polynomial
13679 mode sizes. Require a constant size for BLKmode parameters
13680 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
13681 forces a parameter to be padded at the lsb end in order to
13682 fill a complete number of words, require the parameter size
13683 to be ordered wrt UNITS_PER_WORD.
13685 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13686 Alan Hayward <alan.hayward@arm.com>
13687 David Sherwood <david.sherwood@arm.com>
13689 * reload1.c (spill_stack_slot_width): Change element type
13690 from unsigned int to poly_uint64_pod.
13691 (alter_reg): Treat mode sizes as polynomial.
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 * reload.c (complex_word_subreg_p): New function.
13698 (reload_inner_reg_of_subreg, push_reload): Use it.
13700 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13701 Alan Hayward <alan.hayward@arm.com>
13702 David Sherwood <david.sherwood@arm.com>
13704 * lra-constraints.c (process_alt_operands): Reject matched
13705 operands whose sizes aren't ordered.
13706 (match_reload): Refer to this check here.
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 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
13713 that the mode size is in the set {1, 2, 4, 8, 16}.
13715 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13716 Alan Hayward <alan.hayward@arm.com>
13717 David Sherwood <david.sherwood@arm.com>
13719 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
13720 Use plus_constant instead of gen_rtx_PLUS.
13722 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13723 Alan Hayward <alan.hayward@arm.com>
13724 David Sherwood <david.sherwood@arm.com>
13726 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
13727 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
13728 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
13729 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
13730 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
13731 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
13732 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
13733 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
13734 * config/i386/i386.c (ix86_push_rounding): ...this new function.
13735 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
13737 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
13738 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
13739 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
13740 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
13741 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
13742 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
13743 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
13744 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
13745 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
13746 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
13748 * expr.c (emit_move_resolve_push): Treat the input and result
13749 of PUSH_ROUNDING as a poly_int64.
13750 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
13751 (emit_push_insn): Likewise.
13752 * lra-eliminations.c (mark_not_eliminable): Likewise.
13753 * recog.c (push_operand): Likewise.
13754 * reload1.c (elimination_effects): Likewise.
13755 * rtlanal.c (nonzero_bits1): Likewise.
13756 * calls.c (store_one_arg): Likewise. Require the padding to be
13757 known at compile time.
13759 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13760 Alan Hayward <alan.hayward@arm.com>
13761 David Sherwood <david.sherwood@arm.com>
13763 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
13764 Use plus_constant instead of gen_rtx_PLUS.
13766 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13767 Alan Hayward <alan.hayward@arm.com>
13768 David Sherwood <david.sherwood@arm.com>
13770 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
13771 rather than an int.
13773 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13774 Alan Hayward <alan.hayward@arm.com>
13775 David Sherwood <david.sherwood@arm.com>
13777 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
13778 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
13779 via stack temporaries. Treat the mode size as polynomial too.
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_2): When handling conversions involving
13786 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
13787 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
13788 as a poly_uint64 too.
13790 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13791 Alan Hayward <alan.hayward@arm.com>
13792 David Sherwood <david.sherwood@arm.com>
13794 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
13796 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13797 Alan Hayward <alan.hayward@arm.com>
13798 David Sherwood <david.sherwood@arm.com>
13800 * combine.c (can_change_dest_mode): Handle polynomial
13801 REGMODE_NATURAL_SIZE.
13802 * expmed.c (store_bit_field_1): Likewise.
13803 * expr.c (store_constructor): Likewise.
13804 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
13805 and polynomial REGMODE_NATURAL_SIZE.
13806 (gen_lowpart_common): Likewise.
13807 * reginfo.c (record_subregs_of_mode): Likewise.
13808 * rtlanal.c (read_modify_subreg_p): Likewise.
13810 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13811 Alan Hayward <alan.hayward@arm.com>
13812 David Sherwood <david.sherwood@arm.com>
13814 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
13815 numbers of elements.
13817 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13818 Alan Hayward <alan.hayward@arm.com>
13819 David Sherwood <david.sherwood@arm.com>
13821 * match.pd: Cope with polynomial numbers of vector elements.
13823 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13824 Alan Hayward <alan.hayward@arm.com>
13825 David Sherwood <david.sherwood@arm.com>
13827 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
13828 in a POINTER_PLUS_EXPR.
13830 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13831 Alan Hayward <alan.hayward@arm.com>
13832 David Sherwood <david.sherwood@arm.com>
13834 * omp-simd-clone.c (simd_clone_subparts): New function.
13835 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
13836 (ipa_simd_modify_function_body): Likewise.
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 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
13843 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
13844 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
13845 (expand_vector_condition, vector_element): Likewise.
13846 (subparts_gt): New function.
13847 (get_compute_type): Use subparts_gt.
13848 (count_type_subparts): Delete.
13849 (expand_vector_operations_1): Use subparts_gt instead of
13850 count_type_subparts.
13852 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13853 Alan Hayward <alan.hayward@arm.com>
13854 David Sherwood <david.sherwood@arm.com>
13856 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
13857 (vect_compile_time_alias): ...this new function. Do the calculation
13858 on poly_ints rather than trees.
13859 (vect_prune_runtime_alias_test_list): Update call accordingly.
13861 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13862 Alan Hayward <alan.hayward@arm.com>
13863 David Sherwood <david.sherwood@arm.com>
13865 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
13867 (vect_schedule_slp_instance): Likewise.
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_get_and_check_slp_defs): Reject
13874 constant and extern definitions for variable-length vectors.
13875 (vect_get_constant_vectors): Note that the number of units
13876 is known to be constant.
13878 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13879 Alan Hayward <alan.hayward@arm.com>
13880 David Sherwood <david.sherwood@arm.com>
13882 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
13883 of units as polynomial. Choose between WIDE and NARROW based
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 (simd_clone_subparts): New function.
13891 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
13893 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13894 Alan Hayward <alan.hayward@arm.com>
13895 David Sherwood <david.sherwood@arm.com>
13897 * tree-vect-stmts.c (vectorizable_call): Treat the number of
13898 vectors as polynomial. Use build_index_vector for
13899 IFN_GOMP_SIMD_LANE.
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 (get_load_store_type): Treat the number of
13906 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
13907 for variable-length vectors.
13908 (vectorizable_mask_load_store): Treat the number of units as
13909 polynomial, asserting that it is constant if the condition has
13910 already been enforced.
13911 (vectorizable_store, vectorizable_load): Likewise.
13913 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13914 Alan Hayward <alan.hayward@arm.com>
13915 David Sherwood <david.sherwood@arm.com>
13917 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
13918 of units as polynomial. Punt if we can't tell at compile time
13919 which vector contains the final result.
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_induction): Treat the number
13926 of units as polynomial. Punt on SLP inductions. Use an integer
13927 VEC_SERIES_EXPR for variable-length integer reductions. Use a
13928 cast of such a series for variable-length floating-point
13931 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13932 Alan Hayward <alan.hayward@arm.com>
13933 David Sherwood <david.sherwood@arm.com>
13935 * tree.h (build_index_vector): Declare.
13936 * tree.c (build_index_vector): New function.
13937 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
13938 of units as polynomial, forcibly converting it to a constant if
13939 vectorizable_reduction has already enforced the condition.
13940 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
13941 to create a {1,2,3,...} vector.
13942 (vectorizable_reduction): Treat the number of units as polynomial.
13943 Choose vectype_in based on the largest scalar element size rather
13944 than the smallest number of units. Enforce the restrictions
13947 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13948 Alan Hayward <alan.hayward@arm.com>
13949 David Sherwood <david.sherwood@arm.com>
13951 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
13952 number of units as polynomial.
13954 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13955 Alan Hayward <alan.hayward@arm.com>
13956 David Sherwood <david.sherwood@arm.com>
13958 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
13959 * target.def (autovectorize_vector_sizes): Return the vector sizes
13960 by pointer, using vector_sizes rather than a bitmask.
13961 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
13962 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
13963 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
13965 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
13966 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
13967 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
13968 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
13969 * omp-general.c (omp_max_vf): Likewise.
13970 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13971 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
13972 * tree-vect-loop.c (vect_analyze_loop): Likewise.
13973 * tree-vect-slp.c (vect_slp_bb): Likewise.
13974 * doc/tm.texi: Regenerate.
13975 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
13977 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
13978 the vector size as a poly_uint64 rather than an unsigned int.
13979 (current_vector_size): Change from an unsigned int to a poly_uint64.
13980 (get_vectype_for_scalar_type): Update accordingly.
13981 * tree.h (build_truth_vector_type): Take the size and number of
13982 units as a poly_uint64 rather than an unsigned int.
13983 (build_vector_type): Add a temporary overload that takes
13984 the number of units as a poly_uint64 rather than an unsigned int.
13985 * tree.c (make_vector_type): Likewise.
13986 (build_truth_vector_type): Take the number of units as a poly_uint64
13987 rather than an unsigned int.
13989 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13990 Alan Hayward <alan.hayward@arm.com>
13991 David Sherwood <david.sherwood@arm.com>
13993 * target.def (get_mask_mode): Take the number of units and length
13994 as poly_uint64s rather than unsigned ints.
13995 * targhooks.h (default_get_mask_mode): Update accordingly.
13996 * targhooks.c (default_get_mask_mode): Likewise.
13997 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
13998 * doc/tm.texi: Regenerate.
14000 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14001 Alan Hayward <alan.hayward@arm.com>
14002 David Sherwood <david.sherwood@arm.com>
14004 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
14005 * omp-general.c (omp_max_vf): Likewise.
14006 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
14007 (expand_omp_simd): Handle polynomial safelen.
14008 * omp-low.c (omplow_simd_context): Add a default constructor.
14009 (omplow_simd_context::max_vf): Change from int to poly_uint64.
14010 (lower_rec_simd_input_clauses): Update accordingly.
14011 (lower_rec_input_clauses): Likewise.
14013 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14014 Alan Hayward <alan.hayward@arm.com>
14015 David Sherwood <david.sherwood@arm.com>
14017 * tree-vectorizer.h (vect_nunits_for_cost): New function.
14018 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
14019 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
14020 (vect_analyze_slp_cost): Likewise.
14021 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
14022 (vect_model_load_cost): Likewise.
14024 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14025 Alan Hayward <alan.hayward@arm.com>
14026 David Sherwood <david.sherwood@arm.com>
14028 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
14029 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
14030 from an unsigned int * to a poly_uint64_pod *.
14031 (calculate_unrolling_factor): New function.
14032 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
14034 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14035 Alan Hayward <alan.hayward@arm.com>
14036 David Sherwood <david.sherwood@arm.com>
14038 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
14039 from an unsigned int to a poly_uint64.
14040 (_loop_vec_info::slp_unrolling_factor): Likewise.
14041 (_loop_vec_info::vectorization_factor): Change from an int
14043 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
14044 (vect_get_num_vectors): New function.
14045 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
14046 (vect_get_num_copies): Use vect_get_num_vectors.
14047 (vect_analyze_data_ref_dependences): Change max_vf from an int *
14048 to an unsigned int *.
14049 (vect_analyze_data_refs): Change min_vf from an int * to a
14051 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14052 than an unsigned HOST_WIDE_INT.
14053 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
14054 (vect_analyze_data_ref_dependence): Change max_vf from an int *
14055 to an unsigned int *.
14056 (vect_analyze_data_ref_dependences): Likewise.
14057 (vect_compute_data_ref_alignment): Handle polynomial vf.
14058 (vect_enhance_data_refs_alignment): Likewise.
14059 (vect_prune_runtime_alias_test_list): Likewise.
14060 (vect_shift_permute_load_chain): Likewise.
14061 (vect_supportable_dr_alignment): Likewise.
14062 (dependence_distance_ge_vf): Take the vectorization factor as a
14063 poly_uint64 rather than an unsigned HOST_WIDE_INT.
14064 (vect_analyze_data_refs): Change min_vf from an int * to a
14066 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
14067 vfm1 as a poly_uint64 rather than an int. Make the same change
14068 for the returned bound_scalar.
14069 (vect_gen_vector_loop_niters): Handle polynomial vf.
14070 (vect_do_peeling): Likewise. Update call to
14071 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
14072 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
14074 * tree-vect-loop.c (vect_determine_vectorization_factor)
14075 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
14076 (vect_get_known_peeling_cost): Likewise.
14077 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
14078 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
14079 (vect_transform_loop): Likewise. Use the lowest possible VF when
14080 updating the upper bounds of the loop.
14081 (vect_min_worthwhile_factor): Make static. Return an unsigned int
14082 rather than an int.
14083 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
14084 polynomial unroll factors.
14085 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
14086 (vect_make_slp_decision): Likewise.
14087 (vect_supported_load_permutation_p): Likewise, and polynomial
14089 (vect_analyze_slp_cost): Handle polynomial vf.
14090 (vect_slp_analyze_node_operations): Likewise.
14091 (vect_slp_analyze_bb_1): Likewise.
14092 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14093 than an unsigned HOST_WIDE_INT.
14094 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
14095 (vectorizable_load): Handle polynomial vf.
14096 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
14098 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
14100 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14101 Alan Hayward <alan.hayward@arm.com>
14102 David Sherwood <david.sherwood@arm.com>
14104 * match.pd: Handle bit operations involving three constants
14105 and try to fold one pair.
14107 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14109 * tree-vect-loop-manip.c: Include gimple-fold.h.
14110 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
14111 niters_maybe_zero parameters. Handle other cases besides a step of 1.
14112 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
14113 Add a path that uses a step of VF instead of 1, but disable it
14115 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
14116 and niters_no_overflow parameters. Update calls to
14117 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
14118 Create a new SSA name if the latter choses to use a ste other
14119 than zero, and return it via niters_vector_mult_vf_var.
14120 * tree-vect-loop.c (vect_transform_loop): Update calls to
14121 vect_do_peeling, vect_gen_vector_loop_niters and
14122 slpeel_make_loop_iterate_ntimes.
14123 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
14124 (vect_gen_vector_loop_niters): Update declarations after above changes.
14126 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
14128 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
14129 128-bit round to integer instructions.
14130 (ceil<mode>2): Likewise.
14131 (btrunc<mode>2): Likewise.
14132 (round<mode>2): Likewise.
14134 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14136 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
14137 unaligned VSX load/store on P8/P9.
14138 (expand_block_clear): Allow the use of unaligned VSX
14139 load/store on P8/P9.
14141 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14143 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
14145 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
14146 swap associated with both a load and a store.
14148 2018-01-02 Andrew Waterman <andrew@sifive.com>
14150 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
14151 * config/riscv/riscv.md (clear_cache): Use it.
14153 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
14155 * web.c: Remove out-of-date comment.
14157 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14159 * expr.c (fixup_args_size_notes): Check that any existing
14160 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
14161 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
14162 (emit_single_push_insn): ...here.
14164 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14166 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
14167 (const_vector_encoded_nelts): New function.
14168 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
14169 (const_vector_int_elt, const_vector_elt): Declare.
14170 * emit-rtl.c (const_vector_int_elt_1): New function.
14171 (const_vector_elt): Likewise.
14172 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
14173 of CONST_VECTOR_ELT.
14175 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14177 * expr.c: Include rtx-vector-builder.h.
14178 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
14179 directly on the tree encoding.
14180 (const_vector_from_tree): Likewise.
14181 * optabs.c: Include rtx-vector-builder.h.
14182 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
14183 sequence of "u" values.
14184 * vec-perm-indices.c: Include rtx-vector-builder.h.
14185 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
14186 directly on the vec_perm_indices encoding.
14188 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14190 * doc/rtl.texi (const_vector): Describe new encoding scheme.
14191 * Makefile.in (OBJS): Add rtx-vector-builder.o.
14192 * rtx-vector-builder.h: New file.
14193 * rtx-vector-builder.c: Likewise.
14194 * rtl.h (rtx_def::u2): Add a const_vector field.
14195 (CONST_VECTOR_NPATTERNS): New macro.
14196 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
14197 (CONST_VECTOR_DUPLICATE_P): Likewise.
14198 (CONST_VECTOR_STEPPED_P): Likewise.
14199 (CONST_VECTOR_ENCODED_ELT): Likewise.
14200 (const_vec_duplicate_p): Check for a duplicated vector encoding.
14201 (unwrap_const_vec_duplicate): Likewise.
14202 (const_vec_series_p): Check for a non-duplicated vector encoding.
14203 Say that the function only returns true for integer vectors.
14204 * emit-rtl.c: Include rtx-vector-builder.h.
14205 (gen_const_vec_duplicate_1): Delete.
14206 (gen_const_vector): Call gen_const_vec_duplicate instead of
14207 gen_const_vec_duplicate_1.
14208 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
14209 (gen_const_vec_duplicate): Use rtx_vector_builder.
14210 (gen_const_vec_series): Likewise.
14211 (gen_rtx_CONST_VECTOR): Likewise.
14212 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
14213 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14214 Build a new vector rather than modifying a CONST_VECTOR in-place.
14215 (handle_special_swappables): Update call accordingly.
14216 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
14217 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14218 Build a new vector rather than modifying a CONST_VECTOR in-place.
14219 (handle_special_swappables): Update call accordingly.
14221 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14223 * simplify-rtx.c (simplify_const_binary_operation): Use
14224 CONST_VECTOR_ELT instead of XVECEXP.
14226 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14228 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
14229 the selector elements to be different from the data elements
14230 if the selector is a VECTOR_CST.
14231 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
14232 ssizetype for the selector.
14234 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14236 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
14237 before testing each element individually.
14238 * tree-vect-generic.c (lower_vec_perm): Likewise.
14240 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14242 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
14243 * selftest-run-tests.c (selftest::run_tests): Call it.
14244 * vector-builder.h (vector_builder::operator ==): New function.
14245 (vector_builder::operator !=): Likewise.
14246 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
14247 (vec_perm_indices::all_from_input_p): New function.
14248 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14249 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
14250 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
14251 instead of reading the VECTOR_CST directly. Detect whether both
14252 vector inputs are the same before constructing the vec_perm_indices,
14253 and update the number of inputs argument accordingly. Use the
14254 utility functions added above. Only construct sel2 if we need to.
14256 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14258 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
14259 the broadcast of the low byte.
14260 (expand_mult_highpart): Use an explicit encoding for the permutes.
14261 * optabs-query.c (can_mult_highpart_p): Likewise.
14262 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
14263 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14264 (vectorizable_bswap): Likewise.
14265 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
14266 explicit encoding for the power-of-2 permutes.
14267 (vect_permute_store_chain): Likewise.
14268 (vect_grouped_load_supported): Likewise.
14269 (vect_permute_load_chain): Likewise.
14271 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14273 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
14274 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
14275 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
14276 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14277 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
14278 (vect_gen_perm_mask_any): Likewise.
14280 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14282 * int-vector-builder.h: New file.
14283 * vec-perm-indices.h: Include int-vector-builder.h.
14284 (vec_perm_indices): Redefine as an int_vector_builder.
14285 (auto_vec_perm_indices): Delete.
14286 (vec_perm_builder): Redefine as a stand-alone class.
14287 (vec_perm_indices::vec_perm_indices): New function.
14288 (vec_perm_indices::clamp): Likewise.
14289 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
14290 (vec_perm_indices::new_vector): New function.
14291 (vec_perm_indices::new_expanded_vector): Update for new
14292 vec_perm_indices class.
14293 (vec_perm_indices::rotate_inputs): New function.
14294 (vec_perm_indices::all_in_range_p): Operate directly on the
14295 encoded form, without computing elided elements.
14296 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
14297 encoding. Update for new vec_perm_indices class.
14298 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
14299 the given vec_perm_builder.
14300 (expand_vec_perm_var): Update vec_perm_builder constructor.
14301 (expand_mult_highpart): Use vec_perm_builder instead of
14302 auto_vec_perm_indices.
14303 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
14304 vec_perm_indices instead of auto_vec_perm_indices. Use a single
14305 or double series encoding as appropriate.
14306 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
14307 vec_perm_indices instead of auto_vec_perm_indices.
14308 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14309 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14310 (vect_permute_store_chain): Likewise.
14311 (vect_grouped_load_supported): Likewise.
14312 (vect_permute_load_chain): Likewise.
14313 (vect_shift_permute_load_chain): Likewise.
14314 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14315 (vect_transform_slp_perm_load): Likewise.
14316 (vect_schedule_slp_instance): Likewise.
14317 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14318 (vectorizable_mask_load_store): Likewise.
14319 (vectorizable_bswap): Likewise.
14320 (vectorizable_store): Likewise.
14321 (vectorizable_load): Likewise.
14322 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
14323 vec_perm_indices instead of auto_vec_perm_indices. Use
14324 tree_to_vec_perm_builder to read the vector from a tree.
14325 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
14326 vec_perm_builder instead of a vec_perm_indices.
14327 (have_whole_vector_shift): Use vec_perm_builder and
14328 vec_perm_indices instead of auto_vec_perm_indices. Leave the
14329 truncation to calc_vec_perm_mask_for_shift.
14330 (vect_create_epilog_for_reduction): Likewise.
14331 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
14332 from auto_vec_perm_indices to vec_perm_indices.
14333 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14334 instead of changing individual elements.
14335 (aarch64_vectorize_vec_perm_const): Use new_vector to install
14336 the vector in d.perm.
14337 * config/arm/arm.c (expand_vec_perm_d::perm): Change
14338 from auto_vec_perm_indices to vec_perm_indices.
14339 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14340 instead of changing individual elements.
14341 (arm_vectorize_vec_perm_const): Use new_vector to install
14342 the vector in d.perm.
14343 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
14344 Update vec_perm_builder constructor.
14345 (rs6000_expand_interleave): Likewise.
14346 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
14347 (rs6000_expand_interleave): Likewise.
14349 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14351 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
14352 to qimode could truncate the indices.
14353 * optabs.c (expand_vec_perm_var): Likewise.
14355 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14357 * Makefile.in (OBJS): Add vec-perm-indices.o.
14358 * vec-perm-indices.h: New file.
14359 * vec-perm-indices.c: Likewise.
14360 * target.h (vec_perm_indices): Replace with a forward class
14362 (auto_vec_perm_indices): Move to vec-perm-indices.h.
14363 * optabs.h: Include vec-perm-indices.h.
14364 (expand_vec_perm): Delete.
14365 (selector_fits_mode_p, expand_vec_perm_var): Declare.
14366 (expand_vec_perm_const): Declare.
14367 * target.def (vec_perm_const_ok): Replace with...
14368 (vec_perm_const): ...this new hook.
14369 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
14370 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
14371 * doc/tm.texi: Regenerate.
14372 * optabs.def (vec_perm_const): Delete.
14373 * doc/md.texi (vec_perm_const): Likewise.
14374 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
14375 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
14376 expand_vec_perm for constant permutation vectors. Assert that
14377 the mode of variable permutation vectors is the integer equivalent
14378 of the mode that is being permuted.
14379 * optabs-query.h (selector_fits_mode_p): Declare.
14380 * optabs-query.c: Include vec-perm-indices.h.
14381 (selector_fits_mode_p): New function.
14382 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
14383 is defined, instead of checking whether the vec_perm_const_optab
14384 exists. Use targetm.vectorize.vec_perm_const instead of
14385 targetm.vectorize.vec_perm_const_ok. Check whether the indices
14386 fit in the vector mode before using a variable permute.
14387 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
14388 vec_perm_indices instead of an rtx.
14389 (expand_vec_perm): Replace with...
14390 (expand_vec_perm_const): ...this new function. Take the selector
14391 as a vec_perm_indices rather than an rtx. Also take the mode of
14392 the selector. Update call to shift_amt_for_vec_perm_mask.
14393 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
14394 Use vec_perm_indices::new_expanded_vector to expand the original
14395 selector into bytes. Check whether the indices fit in the vector
14396 mode before using a variable permute.
14397 (expand_vec_perm_var): Make global.
14398 (expand_mult_highpart): Use expand_vec_perm_const.
14399 * fold-const.c: Includes vec-perm-indices.h.
14400 * tree-ssa-forwprop.c: Likewise.
14401 * tree-vect-data-refs.c: Likewise.
14402 * tree-vect-generic.c: Likewise.
14403 * tree-vect-loop.c: Likewise.
14404 * tree-vect-slp.c: Likewise.
14405 * tree-vect-stmts.c: Likewise.
14406 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
14408 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
14409 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
14410 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
14411 (aarch64_vectorize_vec_perm_const): ...this new function.
14412 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14413 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14414 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
14415 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
14416 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14417 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14418 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
14420 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
14421 check for NEON modes.
14422 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
14423 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
14424 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
14425 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
14427 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
14428 the old VEC_PERM_CONST conditions.
14429 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
14430 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
14431 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
14432 (ia64_vectorize_vec_perm_const_ok): Merge into...
14433 (ia64_vectorize_vec_perm_const): ...this new function.
14434 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
14435 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
14436 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
14437 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
14438 * config/mips/mips.c (mips_expand_vec_perm_const)
14439 (mips_vectorize_vec_perm_const_ok): Merge into...
14440 (mips_vectorize_vec_perm_const): ...this new function.
14441 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
14442 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
14443 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
14444 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
14445 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
14446 (rs6000_expand_vec_perm_const): Delete.
14447 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
14449 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14450 (altivec_expand_vec_perm_const_le): Take each operand individually.
14451 Operate on constant selectors rather than rtxes.
14452 (altivec_expand_vec_perm_const): Likewise. Update call to
14453 altivec_expand_vec_perm_const_le.
14454 (rs6000_expand_vec_perm_const): Delete.
14455 (rs6000_vectorize_vec_perm_const_ok): Delete.
14456 (rs6000_vectorize_vec_perm_const): New function.
14457 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14458 an element count and rtx array.
14459 (rs6000_expand_extract_even): Update call accordingly.
14460 (rs6000_expand_interleave): Likewise.
14461 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
14462 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
14463 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
14464 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
14465 (rs6000_expand_vec_perm_const): Delete.
14466 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14467 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14468 (altivec_expand_vec_perm_const_le): Take each operand individually.
14469 Operate on constant selectors rather than rtxes.
14470 (altivec_expand_vec_perm_const): Likewise. Update call to
14471 altivec_expand_vec_perm_const_le.
14472 (rs6000_expand_vec_perm_const): Delete.
14473 (rs6000_vectorize_vec_perm_const_ok): Delete.
14474 (rs6000_vectorize_vec_perm_const): New function. Remove stray
14475 reference to the SPE evmerge intructions.
14476 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14477 an element count and rtx array.
14478 (rs6000_expand_extract_even): Update call accordingly.
14479 (rs6000_expand_interleave): Likewise.
14480 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
14481 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
14483 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14485 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14487 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
14488 vector mode and that that mode matches the mode of the data
14490 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
14491 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
14492 directly using expand_vec_perm_1 when forcing selectors into
14494 (expand_vec_perm_var): New function, split out from expand_vec_perm.
14496 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14498 * optabs-query.h (can_vec_perm_p): Delete.
14499 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
14500 * optabs-query.c (can_vec_perm_p): Split into...
14501 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
14502 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
14503 particular selector is valid.
14504 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14505 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14506 (vect_grouped_load_supported): Likewise.
14507 (vect_shift_permute_load_chain): Likewise.
14508 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14509 (vect_transform_slp_perm_load): Likewise.
14510 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14511 (vectorizable_bswap): Likewise.
14512 (vect_gen_perm_mask_checked): Likewise.
14513 * fold-const.c (fold_ternary_loc): Likewise. Don't take
14514 implementations of variable permutation vectors into account
14515 when deciding which selector to use.
14516 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
14517 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
14518 with a false third argument.
14519 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
14520 to test whether the constant selector is valid and can_vec_perm_var_p
14521 to test whether a variable selector is valid.
14523 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14525 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
14526 * optabs-query.c (can_vec_perm_p): Likewise.
14527 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
14528 instead of vec_perm_indices.
14529 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
14530 (vect_gen_perm_mask_checked): Likewise,
14531 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
14532 (vect_gen_perm_mask_checked): Likewise,
14534 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14536 * optabs-query.h (qimode_for_vec_perm): Declare.
14537 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
14538 (qimode_for_vec_perm): ...this new function.
14539 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
14541 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14543 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
14544 does not have a conditional at the top.
14546 2018-01-02 Richard Biener <rguenther@suse.de>
14548 * ipa-inline.c (big_speedup_p): Fix expression.
14550 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14553 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
14556 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14560 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
14561 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
14562 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
14563 cond_taken_branch_cost 3->4.
14565 2018-01-01 Jakub Jelinek <jakub@redhat.com>
14567 PR tree-optimization/83581
14568 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
14569 TODO_cleanup_cfg if any changes have been made.
14571 PR middle-end/83608
14572 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
14573 convert_modes if target mode has the right side, but different mode
14576 PR middle-end/83609
14577 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
14578 last argument when extracting from CONCAT. If either from_real or
14579 from_imag is NULL, use expansion through memory. If result is not
14580 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
14581 the parts directly to inner mode, if even that fails, use expansion
14584 PR middle-end/83623
14585 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
14586 check for bswap in mode rather than HImode and use that in expand_unop
14589 Copyright (C) 2018 Free Software Foundation, Inc.
14591 Copying and distribution of this file, with or without modification,
14592 are permitted in any medium without royalty provided the copyright
14593 notice and this notice are preserved.