1 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
3 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
4 Check for subset of GENERAL_REGS and FP_REGS.
5 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
8 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
10 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
11 and wi::to_poly_offset. Add the current offset and then check
12 whether the sum fits, rather than using an unchecked addition of
13 a checked term. Check for a shwi rather than a uhwi.
14 * expr.c (get_bit_range): Use tree_to_poly_uint64.
15 (store_constructor): Use poly_int_tree_p.
16 (expand_expr_real_1): Likewise.
17 * function.c (assign_temp): Likewise.
18 * fold-const.c (const_binop): Use poly_int_tree_p and
20 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
22 * ipa-icf-gimple.c (func_checker::compare_operand): Use
23 to_poly_offset for MEM offsets.
24 * ipa-icf.c (sem_variable::equals): Likewise.
25 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
26 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
27 wi::to_poly_offset for BIT_FIELD_REF offsets.
28 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
30 * var-tracking.c (emit_note_insn_var_location): Use
33 2018-05-29 Jim Wilson <jimw@sifive.com>
35 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
37 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
40 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
41 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
43 (sse4_1_round<mode>2): Use nonimmediate_operand
44 for operand 1 predicate.
46 2018-05-29 Martin Sebor <msebor@redhat.com>
47 Richard Biener <rguenther@suse.de>
50 * calls.c (get_size_range): Call determine_value_range instead
52 * tree-vrp.h (determine_value_range): Declared new function.
53 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
55 2018-05-29 Richard Biener <rguenther@suse.de>
57 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
58 sure to use non-pattern stmts for get_earlier_stmt arguments.
59 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
60 called on pattern stmts.
61 (get_later_stmt): Likewise.
63 2018-05-29 Martin Liska <mliska@suse.cz>
66 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
69 2018-05-29 Jakub Jelinek <jakub@redhat.com>
71 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
73 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
77 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
78 VEC_PACK_FLOAT_EXPR): New tree codes.
79 * tree-pretty-print.c (op_code_prio): Handle
80 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
81 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
82 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
83 * tree-inline.c (estimate_operator_cost): Likewise.
84 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
85 * fold-const.c (const_binop): Likewise.
86 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
87 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
88 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
89 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
90 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
91 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
92 * expr.c (expand_expr_real_2): Likewise.
93 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
94 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
95 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
97 * optabs.c (expand_widen_pattern_expr): For
98 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
99 sign from result type rather than operand's type.
100 (expand_binop_directly): For vec_packu_float_optab and
101 vec_packs_float_optab allow result type to be different from operand's
103 * optabs-tree.c (optab_for_tree_code): Handle
104 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
105 VEC_PACK_FLOAT_EXPR. Formatting fixes.
106 * tree-vect-generic.c (expand_vector_operations_1): Handle
107 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
109 * tree-vect-stmts.c (supportable_widening_operation): Handle
111 (supportable_narrowing_operation): Handle FLOAT_EXPR.
112 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
113 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
114 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
115 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
117 (vec_pack<floatprefix>_float_<mode>): New expander.
118 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
120 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
121 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
122 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
123 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
124 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
126 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
127 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
128 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
129 VEC_PACK_FLOAT_EXPR): Document.
131 2018-05-29 Richard Biener <rguenther@suse.de>
133 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
135 (stmt_vec_info_vec): Make pointer.
136 (init_stmt_vec_info_vec): Remove.
137 (free_stmt_vec_info_vec): Likewise.
138 (set_stmt_vec_info_vec): New function.
139 (free_stmt_vec_infos): Likewise.
140 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
141 (set_vinfo_for_stmt): Likewise.
142 (get_earlier_stmt): Likewise.
143 (get_later_stmt): Likewise.
144 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
145 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
146 (vec_info::~vec_info): Free stmt_vec_infos.
147 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
148 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
149 (pass_slp_vectorize::execute): Likewise.
150 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
151 (free_stmt_vec_info_vec): Likewise.
152 (set_stmt_vec_info_vec): New function.
153 (free_stmt_vec_infos): Likewise.
154 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
155 the global stmt_vec_info_vec.
156 * tree-parloops.c (gather_scalar_reductions): Use
157 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
160 2018-05-29 Richard Biener <rguenther@suse.de>
162 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
164 2018-05-29 Martin Liska <mliska@suse.cz>
165 David Malcolm <dmalcolm@redhat.com>
167 * vec.c (test_reverse): New.
168 (vec_c_tests): Add new test.
169 * vec.h (vl_ptr>::reverse): New function.
171 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
173 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
175 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
178 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
180 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
182 2018-05-28 Richard Biener <rguenther@suse.de>
184 PR tree-optimization/85933
185 * tree-vect-data-refs.c (vect_record_base_alignments): Only
186 look at stmts marked as vectorizable.
188 2018-05-28 Richard Biener <rguenther@suse.de>
190 PR tree-optimization/85934
191 * tree-vect-generic.c (expand_vector_operations_1): Hoist
192 vector boolean check before scalar optimization.
194 2018-05-28 Jakub Jelinek <jakub@redhat.com>
196 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
199 2018-05-28 Mark Wielaard <mark@klomp.org>
201 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
202 if it is an expression containing a minus sign.
204 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
206 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
208 2018-05-27 Paul Koning <ni1d@arrl.net>
210 * config/pdp11/pdp11.md (truncsihi2): Remove.
212 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
213 Chung-Ju Wu <jasonwucj@gmail.com>
215 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
217 (unaligned_store_dw): Ditto.
218 * config/nds32/nds32-memory-manipulation.c
219 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
220 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
221 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
222 (emit_setmem_word_loop): Rename to ...
223 (emit_setmem_doubleword_loop): ... this.
224 (nds32_gen_dup_4_byte_to_word_value): New function.
225 (nds32_gen_dup_8_byte_to_double_word_value): New function.
226 (nds32_expand_setmem_loop): Refine implementation.
227 (nds32_expand_setmem_loop_v3m): Ditto.
228 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
231 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
233 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
235 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
237 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
238 (nds32_init_machine_status): Initialize machine->attr_naked_p and
239 machine->attr_no_prologue_p.
240 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
241 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
242 (nds32_expand_epilogue): Consider attr_naked_p.
243 (nds32_expand_epilogue_v3pop): Likewise.
244 (nds32_can_use_return_insn): Likewise.
245 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
246 attr_no_prologue_p fields.
247 * config/nds32/nds32.opt (mret-in-naked-func): New option.
249 2018-05-27 Jakub Jelinek <jakub@redhat.com>
252 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
255 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
257 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
259 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
261 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
263 (*<floatsuffix>floatv2div2sf2): Rename to ...
264 (*float<floatunssuffix>v2div2sf2): ... this.
265 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
266 (float<floatunssuffix>v2div2sf2_mask): ... this.
267 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
268 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
269 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
271 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
273 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
275 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
277 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
279 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
281 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
282 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
283 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
284 gen_ufix_truncv8dfv8si2.
285 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
286 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
287 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
288 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
289 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
290 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
291 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
292 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
294 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
298 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
300 2018-05-25 Jim Wilson <jimw@sifive.com>
302 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
303 * config/riscv/riscv.c (struct machine_function): Add
304 interrupt_handler_p and attribute_checked_p fields.
305 (riscv_attribute_table): Add interrupt.
306 (riscv_interrupt_type_p): New.
307 (riscv_save_reg_p): Save extra regs for interrupt handler.
308 (riscv_use_save_libcall): Return false for interrupt handler.
309 (riscv_first_stack_step): Add forward declaration.
310 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
311 for interrupt handler with large frame. Use it for saved reg list.
312 (riscv_expand_prologue): Move flag_stack_usage_info support to
313 eliminate duplication.
314 (riscv_expand_epilogue): Generate mret for interrupt handler.
315 (riscv_epilogue_uses): New.
316 (riscv_can_use_return_insn): Return false for interrupt handler.
317 (riscv_function_ok_for_sibcall): Likewise.
318 (riscv_set_current_function): Add interrupt handler support.
319 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
320 * config/riscv/riscv.md (UNSPECV_MRET): New.
322 (riscv_frflags, riscv_fsflags): Use tab after opcode.
324 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
326 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
328 PR tree-optimization/85712
329 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
330 this candidate has already been replaced in-situ by a copy.
332 2018-05-25 Jason Merrill <jason@redhat.com>
334 PR c++/80485 - inline function non-zero address.
335 * symtab.c (nonzero_address): Check DECL_COMDAT.
337 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
340 * config/alpha/alpha.md (ashlsi3): New insn pattern.
341 (*ashlsi_se): Rename from *ashldi_se. Define as sign
342 extension of SImode operation. Use const123_operand predicate.
344 (*saddl_se_1): Ditto.
346 (*ssubl_se_1): Ditto.
347 * config/alpha/predicates.md (const123_operand): New predicate.
348 * config/alpha/constraints.md (P): Use IN_RANGE.
350 2018-05-25 Richard Biener <rguenther@suse.de>
352 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
354 (ref_maybe_used_by_stmt_p): Likewise.
355 (stmt_may_clobber_ref_p): Likewise.
356 (stmt_may_clobber_ref_p_1): Likewise.
357 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
359 (ref_maybe_used_by_stmt_p): Likewise.
360 (stmt_may_clobber_ref_p): Likewise.
361 (stmt_may_clobber_ref_p_1): Likewise.
362 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
363 the alias oracle to disambiguate DRs with stmts DR analysis
365 (vect_analyze_data_refs): Do not give up on not analyzable
366 DRs for BB vectorization. Remove code truncating the dataref
369 2018-05-25 Jakub Jelinek <jakub@redhat.com>
372 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
373 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
374 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
376 2018-05-25 Richard Biener <rguenther@suse.de>
378 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
379 function, combining stmt data ref gathering and fatal analysis
381 (vect_analyze_data_refs): Remove now redudnant code and simplify.
382 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
383 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
384 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
385 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
387 2018-05-25 Bin Cheng <bin.cheng@arm.com>
389 PR tree-optimization/85720
390 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
391 SCC if all partitions are builtins.
392 (version_loop_by_alias_check): New parameter. Generate cancelable
393 runtime alias check if all partitions are builtins.
394 (distribute_loop): Update call to above function.
396 2018-05-25 Bin Cheng <bin.cheng@arm.com>
398 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
399 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
400 (parm_default_def_partition_arg): Ditto.
401 (set_parm_default_def_partition): Ditto.
402 (get_parm_default_def_partitions): Ditto and make it static.
403 (get_undefined_value_partitions): Ditto and make it static.
404 (remove_ssa_form): Refactor call to init_var_map here.
405 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
406 computation for loop region.
407 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
408 (register_default_def): Delete.
409 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
410 (parm_default_def_partition_arg): Ditto.
411 (set_parm_default_def_partition): Ditto.
412 (get_parm_default_def_partitions): Ditto and make it static.
413 (get_undefined_value_partitions): Ditto and make it static.
414 (coalesce_with_default, coalesce_with_default): Update comment.
415 (create_coalesce_list_for_region): New func factored out from
416 create_outofssa_var_map.
417 (populate_coalesce_list_for_outofssa): New func factored out from
418 create_outofssa_var_map and coalesce_ssa_name.
419 (create_outofssa_var_map): Delete.
420 (coalesce_ssa_name): Refactor to support live range computation.
421 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
422 (get_parm_default_def_partitions): Delete.
423 (get_undefined_value_partitions): Ditto.
424 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
425 computation for loop region.
426 (new_tree_live_info, loe_visit_block): Ditto.
427 (live_worklist, set_var_live_on_entry): Ditto.
428 (calculate_live_on_exit, verify_live_on_entry): Ditto.
429 * tree-ssa-live.h (struct _var_map): New fields.
430 (init_var_map): Change decl.
431 (region_contains_p): New.
433 2018-05-25 Bin Cheng <bin.cheng@arm.com>
435 * tree-ssa-live.h (live_merge_and_clear): Delete.
437 2018-05-25 Richard Biener <rguenther@suse.de>
440 * tree-dump.c (dequeue_and_dump): Remove access to removed
441 operand 2 of a SWITCH_EXPR.
443 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
445 * doc/sourcebuild.texi (vect_double_cond_arith): Include
446 multiplication and division.
447 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
448 (cond_udiv@var{m}, cond_umod@var{m}): Document.
449 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
450 (cond_udiv_optab, cond_umod_optab): New optabs.
451 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
452 (IFN_COND_RDIV): New internal functions.
453 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
454 TRUNC_MOD_EXPR and RDIV_EXPR.
455 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
456 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
458 (SVE_INT_BINARY): Include mult.
459 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
460 (optab, sve_int_op): Handle mult.
461 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
463 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
464 for SVE_INT_BINARY_SD.
466 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
468 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
469 (optab, sve_int_op): Handle div and udiv.
470 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
471 for SVE_INT_BINARY_SD.
472 (*<optab><mode>3): New insn for the same.
474 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
476 * tree-vect-patterns.c: Include predict.h.
477 (vect_recog_divmod_pattern): Restrict check for division support
478 to when optimizing for size.
480 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
482 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
483 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
484 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
485 (gimple_match_op::set_op): Likewise.
486 (gimple_resimplify4): Declare.
487 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
488 (expr::gen_transform): Likewise.
489 (decision_tree::gen): Generate a simplification routine for 4 operands.
490 * gimple-match-head.c (gimple_simplify): Add an overload for
491 4 operands. In the top-level function, handle up to 4 call
492 arguments and call gimple_resimplify4.
493 (gimple_resimplify4): New function.
494 (build_call_internal): Pass a fourth operand.
495 (maybe_push_to_seq): Likewise.
496 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
497 Fold VEC_COND_EXPRs of an operation and a default value into
498 an IFN_COND_* function if possible.
499 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
501 (SVE_COND_FP_BINARY): Include them.
502 (optab, sve_fp_op): Handle them.
503 (SVE_INT_BINARY_REV): New code iterator.
504 (SVE_COND_FP_BINARY_REV): New int iterator.
505 (commutative): New int attribute.
506 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
508 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
510 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
511 (*cond_<optab><mode>): New patterns for reversed operands.
513 2018-05-25 Richard Biener <rguenther@suse.de>
515 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
516 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
517 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
518 (STMT_VINFO_GROUPED_ACCESS): Adjust.
519 * tree-vect-data-refs.c (everywhere): Adjust users.
520 * tree-vect-loop.c (everywhere): Likewise.
521 * tree-vect-slp.c (everywhere): Likewise.
522 * tree-vect-stmts.c (everywhere): Likewise.
523 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
525 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
527 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
529 (gcc_cv_as_section_exclude): ... this.
530 Try Solaris as #exclude syntax.
531 * configure: Regenerate.
532 * config.in: Regenerate.
533 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
535 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
536 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
538 * varasm.c (default_elf_asm_named_section): Don't check if
539 HAVE_GAS_SECTION_EXCLUDE is defined.
541 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
543 * doc/md.texi: Update the documentation of the cond_* optabs
544 to mention the new final operand. Fix GET_MODE_NUNITS call.
545 Describe the scalar case too.
546 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
547 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
549 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
550 (get_conditional_internal_fn): Update comment.
551 * tree-vect-loop.c (vectorizable_reduction): Pass the original
552 accumulator value as a final argument to conditional functions.
553 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
554 a define_expand and add an "else" operand. Assert for now that
555 the else operand is equal to operand 2. Use SVE_INT_BINARY and
556 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
557 (*cond_<optab><mode>): New patterns.
558 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
559 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
560 (UNSPEC_COND_EOR): Delete.
561 (optab): Remove associated mappings.
562 (SVE_INT_BINARY): New code iterator.
563 (sve_int_op): Remove int attribute and add "minus" to the code
565 (SVE_COND_INT_OP): Delete.
566 (SVE_COND_FP_OP): Rename to...
567 (SVE_COND_FP_BINARY): ...this.
569 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
571 * optabs.c (can_reuse_operands_p): New function.
572 (maybe_legitimize_operands): Try to reuse the results for
575 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
577 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
578 Add {q} suffix to insn mnemonic.
580 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
582 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
583 (msp430_warn_func_return): New.
585 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
587 * fold-const.c (tree_nonzero_bits): New function.
588 * fold-const.h (tree_nonzero_bits): Likewise.
589 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
590 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
592 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
596 * varasm.c (assemble_alias): Check ifunc_resolver only on
599 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
602 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
603 when memory input operand is handled.
605 2018-05-24 Luis Machado <luis.machado@linaro.org>
607 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
609 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
611 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
613 * match.pd: Delay FMA folds until after vectorization.
615 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
618 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
619 address check not strict.
621 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
623 * gimple-match.h (gimple_match_op): New class.
624 (mprts_hook): Replace parameters with a gimple_match_op *.
625 (maybe_build_generic_op): Likewise.
626 (gimple_simplified_result_is_gimple_val): Replace parameters with
627 a const gimple_match_op *.
628 (gimple_simplify): Replace code_helper * and tree * parameters with
629 a gimple_match_op * parameter.
630 (gimple_resimplify1): Replace code_helper *, tree and tree *
631 parameters with a gimple_match_op * parameter.
632 (gimple_resimplify2): Likewise.
633 (gimple_resimplify3): Likewise.
634 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
635 parameters with a gimple_match_op * parameter.
636 * gimple-match-head.c (gimple_simplify): Change prototypes of
637 auto-generated functions to take a gimple_match_op * instead of
638 separate code_helper * and tree * parameters. Make the same
639 change in the top-level overload and update calls to the
640 gimple_resimplify routines. Update calls to the auto-generated
641 functions and to maybe_push_res_to_seq in the publicly-facing
642 operation-specific gimple_simplify overloads.
643 (gimple_match_op::MAX_NUM_OPS): Define.
644 (gimple_resimplify1): Replace rcode and ops with a single res_op
645 parameter. Update call to gimple_simplify.
646 (gimple_resimplify2): Likewise.
647 (gimple_resimplify3): Likewise.
648 (mprts_hook): Replace parameters with a gimple_match_op *.
649 (maybe_build_generic_op): Likewise.
650 (build_call_internal): Replace type, nargs and ops with
652 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
653 with a single gimple_match_op *. Update calls to mprts_hook,
654 build_call_internal and gimple_simplified_result_is_gimple_val.
655 Factor out code that is common to the tree_code and combined_fn cases.
656 * genmatch.c (expr::gen_transform): Replace tem_code and
657 tem_ops with a gimple_match_op called tem_op. Update calls
658 to the gimple_resimplify functions and maybe_push_res_to_seq.
659 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
660 res_ops. Update call to the gimple_resimplify functions.
661 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
662 (decision_tree::gen): Make the functions take a gimple_match_op *
663 called res_op instead of separate res_code and res_ops parameters.
664 Update call accordingly.
665 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
666 and ops with a single res_op parameter. Update calls to
667 maybe_build_generic_op and maybe_push_res_to_seq.
668 (fold_stmt_1): Update calls to gimple_simplify and
669 replace_stmt_with_simplification.
670 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
671 and gimple_simplified_result_is_gimple_val.
672 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
674 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
675 with a gimple_match_op *.
676 (vn_nary_build_or_lookup): Likewise. Update call to
677 vn_nary_build_or_lookup_1.
678 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
679 gimple_match_op *. Update calls to the gimple_resimplify routines
680 and to gimple_simplified_result_is_gimple_val.
681 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
682 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
683 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
684 (visit_nary_op): Likewise.
685 (visit_reference_op_load): Likewise.
687 2018-05-23 Luis Machado <luis.machado@linaro.org>
689 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
690 modifier for printing the step amount.
692 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
695 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
698 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
700 * doc/sourcebuild.texi (Endianness): New subsubsection.
702 2018-05-23 Luis Machado <luis.machado@linaro.org>
704 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
705 <prefetch_dynamic_strides>: New const bool field.
706 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
707 prefetch_dynamic_strides.
708 (exynosm1_prefetch_tune): Likewise.
709 (thunderxt88_prefetch_tune): Likewise.
710 (thunderx_prefetch_tune): Likewise.
711 (thunderx2t99_prefetch_tune): Likewise.
712 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
714 (aarch64_override_options_internal): Update to set
715 PARAM_PREFETCH_DYNAMIC_STRIDES.
716 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
717 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
718 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
719 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
720 prefetch-dynamic-strides setting.
722 2018-05-23 Luis Machado <luis.machado@linaro.org>
724 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
725 <minimum_stride>: New const int field.
726 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
727 minimum_stride field defaulting to -1.
728 (exynosm1_prefetch_tune): Likewise.
729 (thunderxt88_prefetch_tune): Likewise.
730 (thunderx_prefetch_tune): Likewise.
731 (thunderx2t99_prefetch_tune): Likewise.
732 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
733 <default_opt_level>: Set to 3.
734 (aarch64_override_options_internal): Update to set
735 PARAM_PREFETCH_MINIMUM_STRIDE.
736 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
737 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
738 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
739 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
740 stride is constant and is below the minimum stride threshold.
742 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
744 * config/arm/arm-cpus.in (mode26): Delete.
745 (armv4): Delete mode26 reference.
746 * config/arm/arm.c (arm_configure_build_target): Delete use of
749 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
751 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
753 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
754 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
756 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
757 (floatunsdidf2): Ditto.
759 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
761 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
762 (fixuns_trunc<mode>si2_avx512f): Ditto.
763 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
764 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
765 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
767 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
769 PR rtl-optimization/79985
770 * df-scan.c (df_insn_refs_collect): Remove special case for
771 global registers and asm statements.
773 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
775 * extend.texi (Global Register Variables): Rewrite the bullet list.
776 Note that the register is available for allocation. Note that access
777 via inline asm must use constraints. Add note about async-signal
778 handlers. Remove paragraph about automagic register selection.
780 2018-05-23 Richard Biener <rguenther@suse.de>
782 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
783 of fixed offset from memset VN.
785 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
787 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
789 (alloc_cand_and_find_basis): Initialize first_interp field.
790 (slsr_process_mul): Modify first_interp field.
791 (slsr_process_add): Likewise.
792 (slsr_process_cast): Modify first_interp field for each new
794 (slsr_process_copy): Likewise.
795 (dump_candidate): Dump first_interp field.
796 (replace_mult_candidate): Process all interpretations, not just
798 (replace_rhs_if_not_dup): Likewise.
799 (replace_one_candidate): Likewise.
801 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
803 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
805 (aarch64_needs_frame_chain): New function.
806 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
808 2018-05-23 Sudakshina Das <sudi.das@arm.com>
811 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
812 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
813 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
814 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
815 as true for strict-align.
816 (aarch64_can_inline_p): Perform checks even when callee has no
817 attributes to check for strict alignment.
818 * doc/extend.texi (AArch64 Function Attributes): Document
820 * doc/invoke.texi: (AArch64 Options): Likewise.
822 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
824 PR tree-optimization/85853
825 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
826 the handling of the root of the node to...
827 (vect_slp_analyze_node_operations_1): ...this new function,
828 and run the whole thing with the child nodes' def types
829 set according to their SLP node's def type.
831 2018-05-23 Richard Biener <rguenther@suse.de>
834 * tree-data-ref.c (create_runtime_alias_checks): Defer
835 and ignore overflow warnings.
837 2018-05-23 Yury Gribov <tetra2005@gmail.com>
839 PR tree-optimization/85822
840 * tree-vrp.c (is_masked_range_test): Fix handling of negative
843 2018-05-23 Richard Biener <rguenther@suse.de>
845 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
846 memset constants via native_interpret_expr.
848 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
851 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
853 (cgraph_node::create_alias): Likewise.
854 (cgraph_node::get_availability): Check ifunc_resolver instead
855 of looking up ifunc attribute.
856 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
857 * varasm.c (do_assemble_alias): Likewise.
858 (assemble_alias): Likewise.
859 (default_binds_local_p_3): Likewise.
860 * cgraph.h (cgraph_node): Add ifunc_resolver.
861 (cgraph_node::only_called_directly_or_aliased_p): Return false
863 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
865 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
866 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
867 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
868 instead of looking up ifunc attribute.
870 2018-05-22 Luis Machado <luis.machado@linaro.org>
872 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
874 2018-05-22 Martin Sebor <msebor@redhat.com>
877 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
878 only when expasion succeeds.
879 (expand_builtin_strcmp): Same.
880 (expand_builtin_strncmp): Same.
882 2018-05-22 Martin Sebor <msebor@redhat.com>
884 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
886 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
887 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
889 * config/aarch64/aarch64-ldpstp.md: Replace uses of
890 aarch64_mem_pair_operand with memory_operand and delete operand swapping
892 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
893 Add check for legitimate_address.
894 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
895 (aarch64_swap_ldrstr_operands): New.
896 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
899 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
900 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
902 * config/aarch64/aarch64.md: New patterns to generate stp
904 (store_pair_sw, store_pair_dw): New patterns to generate stp for
905 single words and double words.
906 (load_pair_sw, load_pair_dw): Likewise.
907 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
909 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
911 * config/aarch64/aarch64-ldpstp.md: Modify peephole
912 for different mode ldpstp and add peephole for merged zero stores.
914 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
916 (aarch64_gen_store_pair): Rename calls to match new patterns.
917 (aarch64_gen_load_pair): Rename calls to match new patterns.
918 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
919 (load_pair<DREG:mode><DREG2:mode>): ... This.
920 (store_pair<mode>): Rename to...
921 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
922 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
924 (V_INT_EQUIV): Handle SImode.
925 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
928 2018-05-22 Martin Sebor <msebor@redhat.com>
931 * calls.c (maybe_warn_nonstring_arg): Use string length to set
932 or ajust the presumed bound on an operation to avoid unnecessary
935 2018-05-22 Martin Sebor <msebor@redhat.com>
937 PR tree-optimization/85826
938 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
939 assuming that a DECL necesarily has a constant size.
941 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
944 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
946 2018-05-22 Richard Biener <rguenther@suse.de>
948 PR tree-optimization/85834
949 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
950 non-constant and non-zero memset arguments.
952 2018-05-22 Martin Liska <mliska@suse.cz>
955 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
957 2018-05-22 Richard Biener <rguenther@suse.de>
959 PR tree-optimization/85863
960 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
961 comparisons when vectype is specified.
962 (vectorizable_condition): Do not specify vectype for
963 vect_is_simple_cond when SLP vectorizing.
965 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
968 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
969 define __ibm128 as long double.
970 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
971 as a distinct type when IEEE 128-bit support is enabled.
972 (init_float128_ieee): Fix up conversions between IFmode and IEEE
973 128-bit types to use the correct functions.
974 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
975 convert between 128-bit floating point types that have different
976 modes but the same representation, instead of using gen_lowpart to
978 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
980 (IFKF_reg): New attributes to give the register constraints for
982 (extend<mode>tf2_internal): New insns to mark an explicit
983 conversion between 128-bit floating point types that have a
984 different mode but share the same representation.
986 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
988 PR tree-optimization/85814
989 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
990 a null return from get_strinfo when unsharing the next
991 strinfo in the chain.
993 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
996 * varasm.c (weak_finish): Clean up weak_decls.
998 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1000 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
1001 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
1002 UNSPEC_UADALP values.
1003 * config/aarch64/iterators.md (ABAL): New int iterator.
1006 (sur): Add mappings for the above.
1007 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
1009 (aarch64_<sur>abal<mode>_4): Likewise.
1010 (aarch64_<sur>adalp<mode>_3): Likewise.
1011 (<sur>sadv16qi): New define_expand.
1013 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
1015 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
1016 (*movdf_internal): Ditto.
1017 (*rcpsf2_sse): Ditto.
1018 (*rsqrtsf2_sse): Ditto.
1019 (*sqrt<mode>2_sse): Ditto.
1021 2018-05-21 Tamar Christina <tamar.christina@arm.com>
1023 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
1025 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
1026 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
1027 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
1028 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
1030 * config/aarch64/arm_neon.h: Likewise.
1031 * config/aarch64/iterators.md (VQ_I): New.
1033 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
1035 * config.gcc: Add arc/t-multilib-linux to tmake_file for
1037 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
1040 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1042 * config/nds32/constraints.md (S): New constraint.
1043 * config/nds32/nds32.md (call_internal): Use constraint S.
1044 (call_value_internal): Likewise.
1045 (sibcall_internal): Likewise.
1046 (sibcall_value_internal): Likewise.
1048 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1049 Chung-Ju Wu <jasonwucj@gmail.com>
1051 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
1054 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1055 Chung-Ju Wu <jasonwucj@gmail.com>
1057 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
1058 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
1059 (nds32_rtx_costs_impl): Simplify.
1060 (nds32_address_cost_impl): Simplify.
1061 (nds32_init_rtx_costs): New function.
1062 (nds32_rtx_costs_speed_prefer): Likewise.
1063 (nds32_rtx_costs_size_prefer): Likewise.
1064 (nds32_address_cost_speed_prefer): Likewise.
1065 (nds32_address_cost_speed_fwprop): Likewise.
1066 (nds32_address_cost_size_prefer): Likewise.
1067 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
1068 * config/nds32/nds32.c (nds32_option_override): Use
1069 nds32_init_rtx_costs function.
1071 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1073 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
1074 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
1075 (TARGET_PIPELINE_N8): Likewise.
1076 (TARGET_PIPELINE_N10): Likewise.
1077 (TARGET_PIPELINE_N13): Likewise.
1078 (TARGET_PIPELINE_GRAYWOLF): Likewise.
1080 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
1082 * config/nds32/nds32-fpu.md: Update copyright year.
1084 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1086 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
1088 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1090 * config/nds32/nds32.c
1091 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
1092 * config/nds32/nds32.opt (minline-asm-r15): New option.
1094 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1096 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
1098 * config/nds32/nds32.md (abssi2): New pattern.
1100 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
1102 * config/i386/i386.md (rex64namesuffix): New mode attribute.
1103 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
1104 Merge insn pattern from sse_cvtsi2ss<round_name> and
1105 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
1106 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
1107 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
1108 using SWI48 mode iterator.
1109 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
1110 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
1111 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
1112 pattern from sse_cvttss2si<round_saeonly_name>
1113 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
1114 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
1115 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
1116 using SWI48 mode iterator.
1117 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
1118 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
1119 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
1120 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
1121 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
1122 using SWI48 mode iterator.
1123 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
1124 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
1125 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
1126 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
1127 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
1128 SWI48 mode iterator.
1129 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
1130 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
1131 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
1132 pattern from sse_cvttsd2si<round_saeonly_name>
1133 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
1135 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1137 * config/nds32/nds32-md-auxiliary.c
1138 (nds32_valid_smw_lwm_base_p): Refine.
1139 (nds32_output_smw_single_word): Refine.
1140 (nds32_output_smw_double_word): New.
1141 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
1143 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1145 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
1146 (nds32_output_stack_pop): Refine.
1147 (nds32_expand_unaligned_load): Refine.
1148 (nds32_expand_unaligned_store): Refine.
1150 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1151 Chung-Ju Wu <jasonwucj@gmail.com>
1153 * config/nds32/constants.md: Add TP_REGNUM constant.
1154 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
1155 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
1157 * config/nds32/nds32-doubleword.md: Consider flag_pic.
1158 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
1159 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
1160 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
1161 and PIC code generation.
1162 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
1164 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
1166 * config/nds32/nds32.md: Support TLS and PIC.
1167 * config/nds32/nds32.c: Support TLS and PIC.
1168 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
1169 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
1172 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1174 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
1175 mode with E_ prefix.
1177 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1178 Chung-Ju Wu <jasonwucj@gmail.com>
1180 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
1181 * config/nds32/nds32-md-auxiliary.c
1182 (symbolic_reference_mentioned_p): New.
1183 (nds32_legitimize_ict_address): New.
1184 (nds32_expand_ict_move): New.
1185 (nds32_indirect_call_referenced_p): New.
1186 (nds32_symbol_binds_local_p): Delete.
1187 (nds32_long_call_p): Modify.
1188 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
1189 * config/nds32/nds32-protos.h
1190 (symbolic_reference_mentioned_p): Declare.
1191 (nds32_legitimize_ict_address): Declare.
1192 (nds32_expand_ict_move): Declare.
1193 (nds32_indirect_call_referenced_p): Declare.
1194 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
1195 (nds32_relax_group): Use nds32_ict_const_p as condition.
1196 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
1197 (nds32_asm_file_start): Output ict_model directive in asm code.
1198 (nds32_legitimate_address_p): Consider indirect call.
1199 (nds32_print_operand): Consider indirect call.
1200 (nds32_print_operand_address): Consider indirect call.
1201 (nds32_insert_attributes): Handle "indirect_call" attribute.
1202 (TARGET_LEGITIMATE_ADDRESS_P): Define.
1203 (TARGET_LEGITIMATE_CONSTANT_P): Define.
1204 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
1205 (TARGET_DELEGITIMIZE_ADDRESS): Define.
1206 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
1207 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
1208 (TARGET_ICT_MODEL_SMALL): Define.
1209 (TARGET_ICT_MODEL_LARGE): Define.
1210 * config/nds32/nds32.md (movsi): Consider ict model.
1211 (call, call_value): Consider ict model.
1212 (sibcall, sibcall_value): Consider ict model.
1213 * config/nds32/nds32.opt (mict-model): New option.
1214 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
1217 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
1218 Monk Chiang <sh.chiang04@gmail.com>
1219 Jim Wilson <jimw@sifive.com>
1221 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
1222 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
1223 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
1224 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
1225 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
1226 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
1227 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
1228 compute save_libcall_adjustment properly.
1229 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
1230 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
1231 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
1232 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
1233 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
1234 (ABI_SPEC): Handle mabi=ilp32e.
1235 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
1236 (RVE): Add RVE mask.
1237 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
1238 <-march>: Add rv32e as an example.
1240 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
1243 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
1244 (intra_create_variable_infos): Handle C++ constructors.
1246 2018-05-18 Martin Liska <mliska@suse.cz>
1248 * passes.def: Remove a redundant pass.
1250 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
1253 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
1255 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1257 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
1259 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
1261 (armv2, armv2a, armv3, armv3m): Delete architectures.
1262 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
1263 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
1264 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
1266 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
1267 (*mulsidi3adddi): Likewise.
1268 (mulsidi3): Likewise.
1269 (*mulsidi3_nov6): Likewise.
1270 (umulsidi3): Likewise.
1271 (umulsidi3_nov6): Likewise.
1272 (umaddsidi4): Likewise.
1273 (*umulsidi3adddi): Likewise.
1274 (smulsi3_highpart): Likewise.
1275 (*smulsi3_highpart_nov6): Likewise.
1276 (umulsi3_highpart): Likewise.
1277 (*umulsi3_highpart_nov6): Likewise.
1278 * config/arm/arm.h (arm_arch3m): Delete.
1279 * config/arm/arm.c (arm_arch3m): Delete.
1280 (arm_option_override_internal): Update armv3-related comment.
1281 (arm_configure_build_target): Delete use of isa_bit_mode32.
1282 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
1283 (arm_rtx_costs_internal): Delete check of arm_arch3m.
1284 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
1286 (mulusa3): Likewise.
1287 * config/arm/arm-protos.h (arm_arch3m): Delete.
1288 * config/arm/arm-tables.opt: Regenerate.
1289 * config/arm/arm-tune.md: Likewise.
1290 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
1291 deleted architectures.
1293 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1295 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
1296 (armv5t, armv5te): New features.
1297 (ARMv5, ARMv5e): Delete fgroups.
1298 (ARMv5t, ARMv5te): Adjust for above changes.
1300 (armv5, armv5e): Delete arches.
1301 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
1303 (*call_reg_arm): Likewise.
1304 (*call_value_reg_armv5): Likewise.
1305 (*call_value_reg_arm): Likewise.
1306 (*call_symbol): Likewise.
1307 (*call_value_symbol): Likewise.
1308 (*sibcall_insn): Likewise.
1309 (*sibcall_value_insn): Likewise.
1311 (prefetch): Likewise.
1312 (define_split and define_peephole2 dependent on arm_arch5):
1314 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
1316 (TARGET_ARM_QBIT): Likewise.
1317 (TARGET_DSP_MULTIPLY): Likewise.
1318 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
1319 (arm_arch5, arm_arch5e): Delete.
1320 (arm_arch5t, arm_arch5te): Declare.
1321 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
1322 (arm_arch5t): Declare.
1323 (arm_option_reconfigure_globals): Update for the above.
1324 (arm_options_perform_arch_sanity_checks): Update comment, replace
1325 use of arm_arch5 with arm_arch5t.
1326 (use_return_insn): Likewise.
1327 (arm_emit_call_insn): Likewise.
1328 (output_return_instruction): Likewise.
1329 (arm_final_prescan_insn): Likewise.
1330 (arm_coproc_builtin_available): Likewise.
1331 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
1332 arm_arch5e with arm_arch5t and arm_arch5te.
1333 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
1334 (arm_arch5t, arm_arch5te): Declare.
1335 * config/arm/arm-tables.opt: Regenerate.
1336 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
1337 * config/arm/t-multilib: Likewise.
1338 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
1339 instead of arm_arch5.
1340 (*call_reg_thumb1): Likewise.
1341 (*call_value_reg_thumb1_v5): Likewise.
1342 (*call_value_reg_thumb1): Likewise.
1343 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
1345 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
1347 2018-05-18 Martin Liska <mliska@suse.cz>
1349 PR gcov-profile/84846
1350 * doc/gcov.texi: Document -t option of gcov tool.
1352 2018-05-18 Martin Liska <mliska@suse.cz>
1354 PR gcov-profile/84846
1355 * gcov.c (print_usage): Add new -t option.
1356 (process_args): Handle the option.
1357 (generate_results): Use stdout as output when requested by
1360 2018-05-18 Martin Liska <mliska@suse.cz>
1362 PR gcov-profile/84846
1363 * coverage.c (coverage_init): Write PWD to .gcno file.
1364 * doc/gcov.texi: Document how working directory is printed.
1365 * gcov-dump.c (dump_gcov_file): Print PWD.
1366 * gcov.c (output_intermediate_file): Likewise.
1367 (read_graph_file): Read PWD string.
1368 (output_lines): Print PWD.
1370 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1373 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
1374 for retval and return false if all args to phi are zero.
1376 2018-05-18 Richard Biener <rguenther@suse.de>
1378 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
1380 (evrp_dom_walker::before_dom_children): Call it.
1382 2018-05-18 Richard Biener <rguenther@suse.de>
1384 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
1385 results when processing array refs with variable index.
1387 2018-05-18 Toon Moene <toon@moene.org>
1389 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
1390 directly after that of -floop-interchange. Indicate that both
1391 options are enabled by default when specifying -O3.
1393 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1395 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
1396 iterator. Delete separate integer-mode vec_set<mode> expander.
1397 (aarch64_simd_vec_setv2di): Delete.
1398 (vec_setv2di): Delete.
1399 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
1400 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
1401 the "w, r" alternative.
1403 2018-05-18 Martin Liska <mliska@suse.cz>
1405 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
1406 * tree-pass.h (make_pass_lower_switch_O0): New function.
1407 * tree-switch-conversion.c (node_has_low_bound): Remove.
1408 (node_has_high_bound): Likewise.
1409 (node_is_bounded): Likewise.
1410 (class pass_lower_switch): Make it a template type and create
1412 (pass_lower_switch::execute): Add template argument.
1413 (make_pass_lower_switch): New function.
1414 (make_pass_lower_switch_O0): New function.
1415 (do_jump_if_equal): Remove.
1416 (emit_case_nodes): Simplify to just handle all 3 cases and leave
1417 all the hard work to tree optimization passes.
1419 2018-05-18 Martin Liska <mliska@suse.cz>
1421 * dbgcnt.c (limit_low): Renamed from limit.
1422 (limit_high): New variable.
1423 (dbg_cnt_is_enabled): Check for upper limit.
1424 (dbg_cnt): Adjust dumping.
1425 (dbg_cnt_set_limit_by_index): Add new argument for high
1427 (dbg_cnt_set_limit_by_name): Likewise.
1428 (dbg_cnt_process_single_pair): Parse new format.
1429 (dbg_cnt_process_opt): Use strtok.
1430 (dbg_cnt_list_all_counters): Remove 'value' and add
1432 * doc/invoke.texi: Document changes.
1434 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
1436 * doc/sourcebuild.texi (scalar_all_fma): Document.
1437 * tree.def (FMA_EXPR): Delete.
1438 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
1439 * internal-fn.c (ternary_direct): New macro.
1440 (expand_ternary_optab_fn): Likewise.
1441 (direct_ternary_optab_supported_p): Likewise.
1442 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
1443 * builtins.c (fold_builtin_fma): Delete.
1444 (fold_builtin_3): Don't call it.
1445 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
1446 * expr.c (expand_expr_real_2): Likewise.
1447 * fold-const.c (operand_equal_p): Likewise.
1448 (fold_ternary_loc): Likewise.
1449 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
1450 * gimple.c (DEFTREECODE): Likewise.
1451 * gimplify.c (gimplify_expr): Likewise.
1452 * optabs-tree.c (optab_for_tree_code): Likewise.
1453 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
1454 * tree-eh.c (operation_could_trap_p): Likewise.
1455 (stmt_could_throw_1_p): Likewise.
1456 * tree-inline.c (estimate_operator_cost): Likewise.
1457 * tree-pretty-print.c (dump_generic_node): Likewise.
1458 (op_code_prio): Likewise.
1459 * tree-ssa-loop-im.c (stmt_cost): Likewise.
1460 * tree-ssa-operands.c (get_expr_operands): Likewise.
1461 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
1462 * fold-const-call.h (fold_fma): Delete.
1463 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
1464 CFN_FNMA and CFN_FNMS.
1466 * genmatch.c (combined_fn): New enum.
1467 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
1468 (commutative_op): New function.
1469 (commutate): Use it. Handle more than 2 operands.
1470 (dt_operand::gen_gimple_expr): Use commutative_op.
1471 (parser::parse_expr): Allow :c to be used with non-binary
1472 operators if the commutative operand is known.
1473 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
1474 CFN_FMS, CFN_FNMA and CFN_FNMS.
1475 (backprop::process_assign_use): Remove FMA_EXPR handling.
1476 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
1477 (gen_hsa_fma): New function.
1478 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
1479 IFN_FNMA and IFN_FNMS.
1480 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
1481 * gimple-fold.h (follow_all_ssa_edges): Declare.
1482 * gimple-fold.c (follow_all_ssa_edges): New function.
1483 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
1484 gimple_build interface and use follow_all_ssa_edges to fold the result.
1485 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
1486 instead of checking for optabs directly.
1487 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
1488 rather than FMA_EXPRs.
1489 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
1490 call to IFN_FMA instead of an FMA_EXPR.
1492 2018-05-17 Jim Wilson <jimw@sifive.com>
1494 * expr.c (do_tablejump): When converting index to Pmode, if we have a
1495 sign extended promoted subreg, and the range does not have the sign bit
1496 set, then do a sign extend.
1498 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
1499 test, check for sign extended subreg and/or constant operands, and
1500 do a sign extend in that case.
1502 2018-05-17 Steve Ellcey <sellcey@cavium.com>
1504 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
1505 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
1507 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
1508 Change logics_shift_reg to logics_shift_imm.
1509 (thunderx2t99_fp_loadpair_basic): Delete.
1510 (thunderx2t99_fp_storepair_basic): Delete.
1511 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
1512 (thunderx2t99_asimd_polynomial): Delete.
1513 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
1514 and neon_fp_mul_d_scalar_q.
1515 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
1516 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
1517 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
1518 (thunderx2t99_asimd_lut): Add missing tbl types.
1519 (thunderx2t99_asimd_ext): Delete.
1520 (thunderx2t99_asimd_load1_1_mult): Delete.
1521 (thunderx2t99_asimd_load1_2_mult): Delete.
1522 (thunderx2t99_asimd_load1_ldp): New.
1523 (thunderx2t99_asimd_load1): New.
1524 (thunderx2t99_asimd_load2): Add missing *load2* types.
1525 (thunderx2t99_asimd_load3): New.
1526 (thunderx2t99_asimd_load4): New.
1527 (thunderx2t99_asimd_store1_1_mult): Delete.
1528 (thunderx2t99_asimd_store1_2_mult): Delete.
1529 (thunderx2t99_asimd_store2_mult): Delete.
1530 (thunderx2t99_asimd_store2_onelane): Delete.
1531 (thunderx2t99_asimd_store_stp): New.
1532 (thunderx2t99_asimd_store1): New.
1533 (thunderx2t99_asimd_store2): New.
1534 (thunderx2t99_asimd_store3): New.
1535 (thunderx2t99_asimd_store4): New.
1537 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
1539 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
1540 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
1542 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
1543 Segher Boessenkool <segher@kernel.crashing.org>
1546 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
1549 2018-05-17 Richard Biener <rguenther@suse.de>
1551 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
1552 for pruning loop and prune defs feeding only already visited PHIs.
1554 2018-05-17 Richard Biener <rguenther@suse.de>
1556 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
1558 2018-05-17 Bin Cheng <bin.cheng@arm.com>
1559 Richard Biener <rguenther@suse.de>
1561 PR tree-optimization/85793
1562 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
1563 for VMAT_ELEMENTWISE.
1565 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1567 * internal-fn.h (lookup_internal_fn): Declare
1568 * internal-fn.c (lookup_internal_fn): New function.
1569 * gimple.c (gimple_build_call_from_tree): Handle calls to
1571 * gimple-pretty-print.c (dump_gimple_call): Print "." before
1572 internal function names.
1573 * tree-pretty-print.c (dump_generic_node): Likewise.
1574 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
1576 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1578 * gimple-fold.h (gimple_build): Make the function forms take
1579 combined_fn rather than built_in_function.
1580 (gimple_simplify): Likewise.
1581 * gimple-match-head.c (gimple_simplify): Likewise.
1582 * gimple-fold.c (gimple_build): Likewise.
1583 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
1584 rather than gimple_build_call_internal.
1585 (get_initial_defs_for_reduction): Likewise.
1586 (vect_create_epilog_for_reduction): Likewise.
1587 (vectorizable_live_operation): Likewise.
1589 2018-05-17 Martin Liska <mliska@suse.cz>
1591 * gimple-ssa-sprintf.c (format_directive): Do not use
1592 space in between 'G_' and '('.
1594 2018-05-17 Jakub Jelinek <jakub@redhat.com>
1597 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
1598 even if the mask is not all ones.
1601 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
1603 (ix86_gimple_fold_builtin): Likewise.
1606 * config/i386/i386.c: Include tree-vector-builder.h.
1607 (ix86_vector_shift_count): New function.
1608 (ix86_fold_builtin): Fold shift builtins by scalar count.
1609 (ix86_gimple_fold_builtin): Likewise.
1611 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
1612 _mm512_setzero): New intrinsics.
1614 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
1615 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1617 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
1618 code generation for cases where splatting a value is not useful.
1619 * simplify-rtx.c (simplify_ternary_operation): Simplify
1620 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
1621 mode to a vec_concat.
1623 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
1625 * config.gcc: Support "goldmont-plus".
1626 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1628 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1629 PROCESSOR_GOLDMONT_PLUS.
1630 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
1631 (processor_target_table): Add "goldmont-plus".
1632 (PTA_GOLDMONT_PLUS): Define.
1633 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
1634 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
1635 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
1636 (fold_builtin_cpu): Add "goldmont-plus".
1637 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
1638 (ix86_option_override_internal): Add "goldmont-plus".
1639 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
1640 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
1641 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
1642 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
1644 2018-05-17 Richard Biener <rguenther@suse.de>
1646 PR tree-optimization/85757
1647 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
1648 remove defs that only feed that PHI from further processing.
1650 2018-05-16 Jim Wilson <jimw@sifive.com>
1652 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
1654 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
1656 2018-05-16 Mark Wielaard <mark@klomp.org>
1658 * dwarf2out.c (count_index_strings): New function.
1659 (output_indirect_strings): Call count_index_strings and generate
1660 header for dwarf_version >= 5.
1662 2018-05-16 Mark Wielaard <mark@klomp.org>
1664 * dwarf2out.c (dwarf_FORM): New function.
1665 (set_indirect_string): Use dwarf_FORM.
1666 (reset_indirect_string): Likewise.
1667 (size_of_die): Likewise.
1668 (value_format): Likewise.
1669 (output_die): Likewise.
1670 (add_skeleton_AT_string): Likewise.
1671 (output_macinfo_op): Likewise.
1672 (index_string): Likewise.
1673 (output_index_string_offset): Likewise.
1674 (output_index_string): Likewise.
1675 (count_index_strings): Likewise.
1677 2018-05-16 Carl Love <cel@us.ibm.com>
1679 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1680 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
1682 2018-05-16 Martin Jambor <mjambor@suse.cz>
1684 * ipa-prop.c (ipa_free_all_edge_args): Remove.
1685 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
1687 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
1689 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
1690 (fnma<mode>4): Likewise.
1691 (fms<mode>4): Likewise.
1692 (fnms<mode>4): Likewise.
1693 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
1694 (aarch64_fnma<mode>4): Likewise.
1695 (aarch64_fms<mode>4): Likewise.
1696 (aarch64_fnms<mode>4): Likewise.
1697 (aarch64_fnmadd<mode>4): Likewise.
1699 2018-05-16 Jason Merrill <jason@redhat.com>
1701 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
1703 2018-05-16 Richard Biener <rguenther@suse.de>
1705 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
1706 (dump_stmt_cost): Declare.
1707 (add_stmt_cost): Dump cost we add.
1708 (add_stmt_costs): New function.
1709 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
1711 (vect_analyze_stmt): Adjust prototype.
1712 (vectorizable_condition): Likewise.
1713 (vectorizable_live_operation): Likewise.
1714 (vectorizable_reduction): Likewise.
1715 (vectorizable_induction): Likewise.
1716 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
1717 cost vector to pass to vectorizable_ and record afterwards.
1718 (vect_model_reduction_cost): Take cost vector argument and adjust.
1719 (vect_model_induction_cost): Likewise.
1720 (vectorizable_reduction): Likewise.
1721 (vectorizable_induction): Likewise.
1722 (vectorizable_live_operation): Likewise.
1723 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
1724 SLP_TREE_NUMBER_OF_VEC_STMTS.
1725 (vect_analyze_slp_cost_1): Remove.
1726 (vect_analyze_slp_cost): Likewise.
1727 (vect_slp_analyze_node_operations): Take visited args and
1728 a target cost vector. Avoid processing already visited stmt sets.
1729 (vect_slp_analyze_operations): Use a local cost vector to gather
1730 costs and register those of non-discarded instances.
1731 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
1732 (vect_schedule_slp_instance): Remove copying of
1733 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
1735 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
1736 adding cost. Record cost entry location.
1737 (vect_prologue_cost_for_slp_op): Function to compute cost of
1738 a constant or invariant generated for SLP vect in the prologue,
1739 split out from vect_analyze_slp_cost_1.
1740 (vect_model_simple_cost): Make static. Adjust for SLP costing.
1741 (vect_model_promotion_demotion_cost): Likewise.
1742 (vect_model_store_cost): Likewise, make static.
1743 (vect_model_load_cost): Likewise.
1744 (vectorizable_bswap): Add cost vector arg and adjust.
1745 (vectorizable_call): Likewise.
1746 (vectorizable_simd_clone_call): Likewise.
1747 (vectorizable_conversion): Likewise.
1748 (vectorizable_assignment): Likewise.
1749 (vectorizable_shift): Likewise.
1750 (vectorizable_operation): Likewise.
1751 (vectorizable_store): Likewise.
1752 (vectorizable_load): Likewise.
1753 (vectorizable_condition): Likewise.
1754 (vectorizable_comparison): Likewise.
1755 (can_vectorize_live_stmts): Likewise.
1756 (vect_analyze_stmt): Likewise.
1757 (vect_transform_stmt): Adjust calls to vectorizable_*.
1758 * tree-vectorizer.c: Include gimple-pretty-print.h.
1759 (dump_stmt_cost): New function.
1761 2018-05-16 Richard Biener <rguenther@suse.de>
1763 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
1764 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
1765 * tree-ssa-dse.c: Include tree-ssa-loop.h.
1766 (check_name): New callback.
1767 (dse_classify_store): Track cycles via a visited bitmap of PHI
1768 defs and simplify handling of in-loop and across loop dead stores
1769 and properly fail for loop-variant refs. Handle byte-tracking with
1770 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
1773 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
1775 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
1776 (vect_get_mask_type_for_stmt): Likewise.
1777 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
1779 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
1780 to determine the statement's vector type and the vector type that
1781 should be used for calculating nunits. Deal with cases in which
1782 the type has to be deferred.
1783 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
1784 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
1785 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
1786 (vect_determine_vf_for_stmt): New functions, split out from...
1787 (vect_determine_vectorization_factor): ...here.
1788 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
1789 (vect_get_mask_type_for_stmt): New functions, split out from
1790 vect_determine_vectorization_factor.
1792 2018-05-16 Richard Biener <rguenther@suse.de>
1794 * tree-cfg.c (verify_gimple_assign_ternary): Properly
1795 verify the [VEC_]COND_EXPR embedded comparison.
1797 2018-05-15 Martin Sebor <msebor@redhat.com>
1799 PR tree-optimization/85753
1800 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
1801 RECORD_TYPE in addition to ARRAY_TYPE.
1803 2018-05-15 Martin Sebor <msebor@redhat.com>
1806 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
1808 2018-05-15 Richard Biener <rguenther@suse.de>
1810 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
1811 add by_clobber_p one. Change algorithm to collect all defs
1812 representing uses we need to walk and try reducing them to
1813 a single one before failing.
1814 (dse_dom_walker::dse_optimize_stmt): Adjust.
1816 2018-05-13 Mark Wielaard <mark@klomp.org>
1818 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
1819 (size_of_loc_descr): Likewise.
1820 (output_loc_operands): Likewise.
1821 (output_loc_operands_raw): Likewise.
1822 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
1823 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
1824 (hash_loc_operands): Likewise.
1825 (compare_loc_operands): Likewise.
1827 2018-05-14 Mark Wielaard <mark@klomp.org>
1829 * dwarf2out.c (count_index_addrs): New function.
1830 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
1832 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1834 PR tree-optimization/83648
1835 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
1836 return value as malloc candidate.
1838 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1841 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
1842 as true in call to suggest_attribute.
1844 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
1846 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
1847 -mreadonly-in-sdata.
1849 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1851 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
1853 (aarch64_crypto_aesd_fused): Likewise.
1855 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
1857 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
1858 (movsi_aarch64): Likewise.
1859 (load_pairsi): Likewise.
1860 (load_pairdi): Likewise.
1861 (store_pairsi): Likewise.
1862 (store_pairdi): Likewise.
1863 (load_pairsf): Likewise.
1864 (load_pairdf): Likewise.
1865 (store_pairsf): Likewise.
1866 (store_pairdf): Likewise.
1867 (zero_extend): Likewise.
1868 (trunc): Swap alternatives.
1869 (fcvt_target): Add '?' to prefer w over r.
1871 2018-05-14 Jakub Jelinek <jakub@redhat.com>
1874 * config/i386/i386.md: Disallow non-commutative arithmetics in
1875 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
1876 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
1877 in the peephole2 before it.
1879 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
1881 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
1882 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
1883 (ix86_handle_option): Handle -mcldemote.
1884 * config.gcc: New header.
1885 * config/i386/cldemoteintrin.h: New file.
1886 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
1887 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1889 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1890 OPTION_MASK_ISA_CLDEMOTE.
1891 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
1892 (ix86_valid_target_attribute_inner_p): Ditto.
1893 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
1894 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
1895 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
1896 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
1897 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
1899 * config/i386/i386.opt: Add -mcldemote.
1900 * config/i386/x86intrin.h: New header.
1901 * doc/invoke.texi: Add -mcldemote.
1903 2018-05-14 Richard Biener <rguenther@suse.de>
1905 * doc/match-and-simplify.texi: Adjust :s documentation.
1907 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
1909 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
1910 intended memcpy size.
1911 (REORDER_45): Likewise.
1913 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
1915 * sort.cc: New file.
1916 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
1917 * vec.c (qsort_chk): Use gcc_qsort.
1918 * Makefile.in (OBJS-libcommon): Add sort.o.
1919 (build/sort.o): New target. Use it...
1920 (BUILD_RTL): ... here, and...
1921 (build/gencfn-macros): ... here, and...
1922 (build/genmatch): ... here.
1924 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1925 Chung-Ju Wu <jasonwucj@gmail.com>
1927 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
1928 * config/nds32/nds32-graywolf.md: New file.
1929 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
1930 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
1932 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
1933 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
1934 * config/nds32/nds32.md (pipeline_model): Add graywolf.
1935 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
1936 * config/nds32/pipelines.md: Include n15 settings.
1938 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1939 Chung-Ju Wu <jasonwucj@gmail.com>
1941 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
1942 * config/nds32/nds32-n13.md: New file.
1943 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
1944 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
1946 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
1947 * config/nds32/nds32.md (pipeline_model): Add n13.
1948 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
1949 * config/nds32/pipelines.md: Include n13 settings.
1951 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1952 Chung-Ju Wu <jasonwucj@gmail.com>
1954 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
1955 * config/nds32/nds32-n10.md: New file.
1956 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
1957 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
1959 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
1960 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
1961 * config/nds32/nds32.md (pipeline_model): Add n10.
1962 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
1963 * config/nds32/pipelines.md: Include n10 settings.
1965 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
1966 Kito Cheng <kito.cheng@gmail.com>
1967 Chung-Ju Wu <jasonwucj@gmail.com>
1969 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
1970 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1971 Add enum values for DSP extension instructions.
1972 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
1974 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
1975 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
1977 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
1978 * config/nds32/nds32-dspext.md: New file for DSP implementation.
1979 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
1980 * config/nds32/nds32-intrinsic.md: Likewise.
1981 * config/nds32/nds32_intrinsic.h: Likewise.
1982 * config/nds32/nds32-md-auxiliary.c: Likewise.
1983 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
1984 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
1985 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
1986 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
1987 * config/nds32/nds32-protos.h: New declarations for DSP extension.
1988 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
1989 TYPE_DMAC in switch statement.
1990 * config/nds32/nds32.c: New checking and implementation for DSP
1991 extension instructions.
1992 * config/nds32/nds32.h: Likewise.
1993 * config/nds32/nds32.md: Likewise.
1994 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
1995 * config/nds32/predicates.md: Implement new predicates for DSP
1998 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
2000 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
2001 Reformat alternatives and attributes so it is easier to identify
2002 which constraints/attributes go with which instruction.
2003 (mov<mode>_hardfloat32, FMOVE64): Likewise.
2004 (mov<mode>_softfloat32, FMOVE64): Likewise.
2005 (mov<mode>_hardfloat64, FMOVE64): Likewise.
2006 (mov<mode>_softfloat64, FMOVE64): Likewise.
2008 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
2010 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2012 (Basic PowerPC Built-in Functions): The new name of the
2013 subsection previously known as "PowerPC Built-in Functions".
2014 (Basic PowerPC Built-in Functions Available on all Configurations):
2016 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
2017 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
2018 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
2019 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
2021 2018-05-11 Martin Jambor <mjambor@suse.cz>
2024 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
2027 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2030 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
2032 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
2034 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
2035 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
2036 (ix86_handle_option): Handle -mwaitpkg.
2037 * config.gcc: New header.
2038 * config/i386/cpuid.h (bit_WAITPKG): New bit.
2039 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
2040 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
2042 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2043 OPTION_MASK_ISA_WAITPKG.
2044 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
2045 (ix86_option_override_internal): Add PTA_WAITPKG.
2046 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
2047 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
2048 IX86_BUILTIN_TPAUSE.
2049 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
2050 __builtin_ia32_umwait and __builtin_ia32_tpause.
2051 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
2052 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
2053 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
2054 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
2055 UNSPECV_TPAUSE): New.
2056 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
2057 * config/i386/i386.opt: Add -mwaitpkg.
2058 * config/i386/waitpkgintrin.h: New file.
2059 * config/i386/x86intrin.h: New header.
2060 * doc/invoke.texi: Add -mwaitpkg.
2062 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2065 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
2067 (cortex-m0): Use armv6s-m isa.
2068 (cortex-m0plus): Likewise.
2069 (cortex-m1): Likewise.
2070 (cortex-m0.small-multiply): Likewise.
2071 (cortex-m0plus.small-multiply): Likewise.
2072 (cortex-m1.small-multiply): Likewise.
2074 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
2075 Jakub Jelinek <jakub@redhat.com>
2077 PR tree-optimization/85692
2078 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
2079 source permute as well.
2081 2018-05-11 Martin Liska <mliska@suse.cz>
2084 * doc/extend.texi: Document LLVM style format for no_sanitize
2087 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
2089 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
2090 mode_supports_vsx_dform_quad to mode_supports_dq_form.
2091 (mode_supports_vsx_dform_quad): Likewise.
2092 (mode_supports_vmx_dform): Move these functions to be next to the
2093 other mode_supports functions.
2094 (mode_supports_dq_form): Likewise.
2095 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
2096 mode_supports_dq_form.
2097 (reg_offset_addressing_ok_p): Likewise.
2098 (offsettable_ok_by_alignment): Likewise.
2099 (rs6000_legitimate_offset_address_p): Likewise.
2100 (legitimate_lo_sum_address_p): Likewise.
2101 (rs6000_legitimize_address): Likewise.
2102 (rs6000_legitimize_reload_address): Likewise.
2103 (rs6000_secondary_reload_inner): Likewise.
2104 (rs6000_preferred_reload_class): Likewise.
2105 (rs6000_output_move_128bit): Likewise.
2107 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2109 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
2110 Generate SImode target register for null target.
2111 <case IX86_BUILTIN_XGETBV>: Ditto.
2112 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
2113 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
2115 2018-05-10 Carl Love <cel@us.ibm.com>
2117 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2118 dcbtt and dcbtstt if operands[2] is 0.
2120 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2123 * config/i386/sse.md (usadv64qi): New expander.
2125 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
2127 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
2128 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
2129 -maltivec=be support.
2130 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
2131 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
2132 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
2133 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
2134 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
2135 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
2136 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
2137 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
2138 altivec_vsumsws): Adjust.
2139 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
2140 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
2141 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
2143 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
2144 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
2145 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
2146 (altivec_lve<VI_char>x): Delete expand.
2147 (*altivec_lve<VI_char>x_internal): Rename to...
2148 (altivec_lve<VI_char>x): ... this.
2149 (altivec_lvxl_<mode>): Delete expand.
2150 (*altivec_lvxl_<mode>_internal): Rename to ...
2151 (altivec_lvxl_<mode>): ... this.
2152 (altivec_stvxl_<mode>): Delete expand.
2153 (*altivec_stvxl_<mode>_internal): Rename to ...
2154 (altivec_stvxl_<mode>): ... this.
2155 (altivec_stve<VI_char>x): Delete expand.
2156 (*altivec_stve<VI_char>x_internal): Rename to ...
2157 (altivec_stve<VI_char>x): ... this.
2158 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
2159 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
2160 reduc_plus_scal_<mode>): Adjust.
2161 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
2163 (rs6000_cpu_cpp_builtins): Adjust.
2164 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
2165 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
2166 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
2167 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
2168 -maltivec=be support.
2169 (rs6000_split_vec_extract_var): Adjust.
2170 (rs6000_split_v4si_init): Adjust.
2171 (swap_selector_for_mode): Delete.
2172 (altivec_expand_lvx_be, altivec_expand_stvx_be,
2173 altivec_expand_stvex_be): Delete.
2174 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
2175 -maltivec=be support.
2176 (rs6000_gimple_fold_builtin): Ditto.
2177 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
2179 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
2180 (TARGET_DIRECT_MOVE_64BIT): Adjust.
2181 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
2182 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
2183 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
2184 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
2185 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
2186 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
2187 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
2188 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
2189 anonymous split): Adjust.
2190 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
2191 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
2193 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
2195 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
2196 when --with-gxx-include-dir is also specified.
2197 * configure: Regenerate.
2199 2018-05-09 Jim Wilson <jimw@sifive.com>
2202 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
2203 * config/riscv/t-withmultilib: New.
2204 * config/riscv/withmultilib.h: New.
2205 * doc/install.texi: Document RISC-V --with-multilib-list support.
2207 2018-05-09 Richard Biener <rguenther@suse.de>
2209 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
2211 (vect_bb_vectorization_profitable_p): Adjust. Compute
2212 actual scalar cost using the cost vector and the add_stmt_cost
2215 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2217 PR rtl-optimization/85645
2218 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
2219 in the REG_CFA_REGISTER note for LR, don't leave it empty.
2221 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2223 PR rtl-optimization/85645
2224 * shrink-wrap.c (spread_components): Return a boolean saying if
2225 anything was changed.
2226 (try_shrink_wrapping_separate): Iterate spread_components until
2227 nothing changes anymore.
2229 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2231 PR rtl-optimization/85645
2232 * regrename.c (build_def_use): Also kill the chains that include the
2233 destination of a REG_CFA_REGISTER note.
2235 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2237 PR rtl-optimization/85645
2238 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
2239 insn that has a REG_CFA_REGISTER note.
2241 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
2243 * cfgexpand.c (expand_clobber): New function.
2244 (expand_gimple_stmt_1): Use it.
2245 * tree-vect-stmts.c (vect_clobber_variable): New function,
2247 (vectorizable_simd_clone_call): ...here.
2248 (vectorizable_store): Emit a clobber either side of an
2249 IFN_STORE_LANES sequence.
2250 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
2252 2018-05-09 Tom de Vries <tom@codesourcery.com>
2255 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
2256 (define_insn "trap_if_false"): Add exit after trap.
2258 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
2260 PR rtl-optimization/85638
2261 * bb-reorder.c: Include common/common-target.h.
2262 (create_forwarder_block): New function extracted from...
2263 (fix_up_crossing_landing_pad): ...here. Rename into...
2264 (dw2_fix_up_crossing_landing_pad): ...this.
2265 (sjlj_fix_up_crossing_landing_pad): New function.
2266 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
2267 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
2268 from both partitions and exit the loop after one iteration.
2270 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2273 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2275 (Basic PowerPC Built-in Functions): The new name of the
2276 subsection previously known as "PowerPC Built-in Functions".
2277 (Basic PowerPC Built-in Functions Available on all Configurations):
2279 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2281 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2283 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2285 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2288 2018-05-08 Jim Wilson <jimw@sifive.com>
2290 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
2291 (LD_EMUL_SUFFIX): New.
2292 (LINK_SPEC): Use it.
2294 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2296 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2298 (Basic PowerPC Built-in Functions): The new name of the
2299 subsection previously known as "PowerPC Built-in Functions".
2300 (Basic PowerPC Built-in Functions Available on all Configurations):
2302 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2304 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2306 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2308 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2311 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2314 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
2315 after cmpelim optimization.
2317 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
2319 * config.gcc: Support "goldmont".
2320 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
2321 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2323 * config/i386/i386.c (m_GOLDMONT): Define.
2324 (processor_target_table): Add "goldmont".
2325 (PTA_GOLDMONT): Define.
2326 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
2327 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
2328 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
2329 (fold_builtin_cpu): Add "goldmont".
2330 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
2331 (ix86_option_override_internal): Add "goldmont".
2332 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
2333 (processor_type): Add PROCESSOR_GOLDMONT.
2334 * config/i386/i386.md: Add CPU "glm".
2335 * config/i386/glm.md: New file.
2336 * config/i386/x86-tune.def: Add m_GOLDMONT.
2337 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
2339 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2342 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
2344 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
2345 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
2346 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
2349 * config/i386/i386.c (ix86_fold_builtin): Handle
2350 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
2353 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
2354 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
2356 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
2359 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
2360 (check_arch): Likewise.
2361 (check_fpu): Return the result rather than printing it.
2362 (end arch): Fix operator precedence.
2363 (end cpu): Likewise.
2364 (END): Print the result from check_fpu.
2366 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2367 Alan Hayward <alan.hayward@arm.com>
2368 David Sherwood <david.sherwood@arm.com>
2370 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
2371 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
2372 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
2373 (*fcmuo<mode>_and): New patterns.
2375 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2377 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
2378 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
2379 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
2380 (cmp_op, sve_imm_con): New code attributes.
2381 (SVE_COND_INT_CMP, imm_con): Delete.
2382 (cmp_op): Remove above unspecs from int attribute.
2383 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
2385 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
2386 comparison-specific unspecs.
2387 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
2388 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
2389 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
2390 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
2391 (*vec_fcm<cmp_op><mode>): Rename to...
2392 (*fcm<cmp_op><mode>): ...this and adjust likewise.
2393 (*vec_fcmuo<mode>): Rename to...
2394 (*fcmuo<mode>): ...this and adjust likewise.
2395 (*pred_fcm<cmp_op><mode>): New pattern.
2396 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
2397 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
2399 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
2401 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
2402 (aarch64_emit_sve_predicated_cond): New function.
2403 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
2404 (aarch64_emit_unspec_cond_or): Replace with...
2405 (aarch64_emit_sve_or_conds): ...this new function. Use
2406 aarch64_emit_sve_ptrue_op for the individual comparisons and
2407 aarch64_emit_binop to OR them together.
2408 (aarch64_emit_inverted_unspec_cond): Replace with...
2409 (aarch64_emit_sve_inverted_cond): ...this new function. Use
2410 aarch64_emit_sve_ptrue_op for the comparison and
2411 aarch64_emit_unop to invert the result.
2412 (aarch64_expand_sve_vec_cmp_float): Update after the above
2413 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
2415 2018-05-07 Nathan Sidwell <nathan@acm.org>
2417 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
2418 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
2419 (Backwards Compatibility): Likewise.
2421 2018-05-07 Luis Machado <luis.machado@linaro.org>
2425 2018-05-07 Luis Machado <luis.machado@linaro.org>
2427 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2428 <prefetch_dynamic_strides>: New const bool field.
2429 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2430 prefetch_dynamic_strides.
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 prefetch_dynamic_strides to false.
2436 (aarch64_override_options_internal): Update to set
2437 PARAM_PREFETCH_DYNAMIC_STRIDES.
2438 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2439 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2440 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2441 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2442 prefetch-dynamic-strides setting.
2444 2018-05-07 Luis Machado <luis.machado@linaro.org>
2446 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2447 <minimum_stride>: New const int field.
2448 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2449 minimum_stride field.
2450 (exynosm1_prefetch_tune): Likewise.
2451 (thunderxt88_prefetch_tune): Likewise.
2452 (thunderx_prefetch_tune): Likewise.
2453 (thunderx2t99_prefetch_tune): Likewise.
2454 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2455 (aarch64_override_options_internal): Update to set
2456 PARAM_PREFETCH_MINIMUM_STRIDE.
2457 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2458 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2459 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2460 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2461 stride is constant and is below the minimum stride threshold.
2463 2018-05-07 Luis Machado <luis.machado@linaro.org>
2465 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
2468 2018-05-07 Luis Machado <luis.machado@linaro.org>
2470 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2471 <prefetch_dynamic_strides>: New const bool field.
2472 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2473 prefetch_dynamic_strides.
2474 (exynosm1_prefetch_tune): Likewise.
2475 (thunderxt88_prefetch_tune): Likewise.
2476 (thunderx_prefetch_tune): Likewise.
2477 (thunderx2t99_prefetch_tune): Likewise.
2478 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2479 (aarch64_override_options_internal): Update to set
2480 PARAM_PREFETCH_DYNAMIC_STRIDES.
2481 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2482 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2483 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2484 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2485 prefetch-dynamic-strides setting.
2487 2018-05-07 Luis Machado <luis.machado@linaro.org>
2489 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2490 <minimum_stride>: New const int field.
2491 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2492 minimum_stride field.
2493 (exynosm1_prefetch_tune): Likewise.
2494 (thunderxt88_prefetch_tune): Likewise.
2495 (thunderx_prefetch_tune): Likewise.
2496 (thunderx2t99_prefetch_tune): Likewise.
2497 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2498 (aarch64_override_options_internal): Update to set
2499 PARAM_PREFETCH_MINIMUM_STRIDE.
2500 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2501 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2502 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2503 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2504 stride is constant and is below the minimum stride threshold.
2506 2018-05-06 Jakub Jelinek <jakub@redhat.com>
2509 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
2510 the type is addressable. Don't force op into register if it has
2513 2018-05-05 Roland McGrath <mcgrathr@google.com>
2516 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
2517 any section for which we don't know a specific type it should have,
2518 regardless of name. Previously this was done only for the exact
2519 names ".init_array", ".fini_array", and ".preinit_array".
2520 (default_elf_asm_named_section): Add comment about
2521 relationship with default_section_type_flags and SECTION_NOTYPE.
2522 (get_section): Don't consider it a type conflict if one side has
2523 SECTION_NOTYPE and the other doesn't, as long as neither has the
2524 SECTION_BSS et al used in the default_section_type_flags logic.
2526 2018-05-05 Tom de Vries <tom@codesourcery.com>
2529 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
2530 (workaround_barsyncs): New function.
2531 (nvptx_reorg): Use workaround_barsyncs.
2532 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2533 (define_expand "nvptx_membar_cta"): New define_expand.
2534 (define_insn "*nvptx_membar_cta"): New insn.
2536 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
2538 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
2539 To improve optimization opportunities.
2540 * builtin-types.def: The new needed builtin types for the above.
2542 2018-05-04 Richard Biener <rguenther@suse.de>
2544 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
2545 * gimple-ssa-store-merging.c
2546 (imm_store_chain_info::output_merged_store): Remove redundant create,
2547 release split_store vector contents on failure.
2548 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
2549 scalar stmt vector on cache hit.
2551 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
2553 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
2555 * config.gcc (powerpc-xilinx-eabi*): Remove.
2556 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
2558 (fusion_addis_mem_combo_load): Ditto.
2559 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
2561 (rs6000_cpu_cpp_builtins): Ditto.
2562 * config/rs6000/rs6000-linux.c
2563 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
2564 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
2565 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
2567 (rs6000_setup_reg_addr_masks): Ditto.
2568 (rs6000_init_hard_regno_mode_ok): Ditto.
2569 (rs6000_option_override_internal): Ditto.
2570 (legitimate_lo_sum_address_p): Ditto.
2571 (rs6000_legitimize_address): Ditto.
2572 (rs6000_legitimize_reload_address): Ditto.
2573 (rs6000_legitimate_address_p): Ditto.
2574 (abi_v4_pass_in_fpr): Ditto.
2575 (setup_incoming_varargs): Ditto.
2576 (rs6000_gimplify_va_arg): Ditto.
2577 (rs6000_split_multireg_move): Ditto.
2578 (rs6000_savres_strategy): Ditto.
2579 (rs6000_emit_prologue_components): Ditto.
2580 (rs6000_emit_epilogue_components): Ditto.
2581 (rs6000_emit_prologue): Ditto.
2582 (rs6000_emit_epilogue): Ditto.
2583 (rs6000_elf_file_end): Ditto.
2584 (rs6000_function_value): Ditto.
2585 (rs6000_libcall_value): Ditto.
2586 * config/rs6000/rs6000.h: Ditto.
2587 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
2588 (TARGET_MINMAX): ... this. New.
2589 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
2590 * config/rs6000/rs6000.md: Remove Xilinx FP support.
2591 (*movsi_internal1_single): Delete.
2592 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
2593 mfpu=, mxilinx-fpu): Delete.
2594 * config/rs6000/singlefp.h: Delete.
2595 * config/rs6000/sysv4.h: Remove Xilinx FP support.
2596 * config/rs6000/t-rs6000: Ditto.
2597 * config/rs6000/t-xilinx: Delete.
2598 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
2599 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
2600 (VStype_simple): Delete.
2601 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
2602 * config/rs6000/xfpu.h: Delete.
2603 * config/rs6000/xfpu.md: Delete.
2604 * config/rs6000/xilinx.h: Delete.
2605 * config/rs6000/xilinx.opt: Delete.
2606 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
2607 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
2609 2018-05-04 Tom de Vries <tom@codesourcery.com>
2612 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
2615 2018-05-04 Richard Biener <rguenther@suse.de>
2618 * tree-complex.c (update_complex_assignment): We are always in SSA form.
2619 (expand_complex_div_wide): Likewise.
2620 (expand_complex_operations_1): Likewise.
2621 (expand_complex_libcall): Preserve EH info of the original stmt.
2622 (tree_lower_complex): Handle removed blocks.
2623 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
2624 on complex multiplication and division libcall builtins.
2626 2018-05-04 Richard Biener <rguenther@suse.de>
2629 * fold-const.c (negate_expr_p): Restrict negation of operand
2630 zero of a division to when we know that can happen without
2632 (fold_negate_expr_1): Likewise.
2634 2018-05-04 Jakub Jelinek <jakub@redhat.com>
2637 * real.h (real_nextafter): Declare.
2638 * real.c (real_nextafter): New function.
2639 * fold-const-call.c (fold_const_nextafter): New function.
2640 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
2641 CASE_CFN_NEXTTOWARD.
2642 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
2643 even when arg1_mode is different from arg0_mode.
2645 2018-05-03 Nathan Sidwell <nathan@acm.org>
2647 * doc/extend.texi (Deprecated Features): Remove
2649 (Backwards Compatibility): Likewise.
2650 * doc/invoke.texi (C++ Language Options): Likewise.
2651 (C++ Dialect Options): Likewise.
2653 2018-05-03 Jakub Jelinek <jakub@redhat.com>
2656 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
2657 _mm512_mask_mullox_epi64): New intrinsics.
2659 2018-05-03 Tom de Vries <tom@codesourcery.com>
2662 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2663 dump files): Add offload-tree.
2665 2018-05-03 Richard Biener <rguenther@suse.de>
2667 PR tree-optimization/85615
2668 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
2669 to loops not nested in BBs loop father to avoid creating multi-entry
2672 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2674 PR tree-optimization/70291
2675 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
2676 arguments. Change return type to tree. Emit libcall as a new
2677 statement rather than replacing existing one when inplace_p is true.
2678 (expand_complex_multiplication_components): New function.
2679 (expand_complex_multiplication): Expand floating-point complex
2680 multiplication using the above.
2681 (expand_complex_division): Rename inner_type parameter to type.
2682 Update expand_complex_libcall call-site.
2683 (expand_complex_operations_1): Update expand_complex_multiplication
2684 and expand_complex_division call-sites.
2686 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2689 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2690 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2691 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
2692 the highest significant bit of the shift count mask is clear. In
2693 check whether and[sq]i3 is needed verify that all significant bits
2694 of the shift count other than the highest are set.
2696 2018-05-02 Tom de Vries <tom@codesourcery.com>
2699 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
2700 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
2701 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
2702 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
2703 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
2704 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
2705 __builtin_goacc_parlevel_size.
2707 2018-05-02 Richard Biener <rguenther@suse.de>
2709 PR tree-optimization/85597
2710 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
2711 do not use split vect_get_vec_defs call but call vect_get_slp_defs
2714 2018-05-02 Tom de Vries <tom@codesourcery.com>
2717 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2718 dump files): Add ltrans-tree.
2720 2018-05-02 Tom de Vries <tom@codesourcery.com>
2723 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2724 dump files): Add wpa-ipa.
2726 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
2728 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
2729 powerpc*-*-linux*paired* target.
2730 * config/rs6000/750cl.h: Delete.
2731 * config/rs6000/paired.h: Delete.
2732 * config/rs6000/paired.md: Delete.
2733 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
2735 * config/rs6000/rs6000-builtin.def: Remove paired float support.
2736 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
2737 comment. Remove paired float support.
2738 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
2739 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
2741 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
2742 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
2744 * config/rs6000/rs6000.c: Remove paired float support.
2745 (paired_expand_vector_init, paired_expand_vector_move,
2746 paired_emit_vector_compare, paired_emit_vector_cond_expr,
2747 (paired_expand_lv_builtin, paired_expand_stv_builtin,
2748 paired_expand_builtin, paired_expand_predicate_builtin,
2749 paired_init_builtins): Delete.
2750 * config/rs6000/rs6000.h: Remove paired float support.
2751 * config/rs6000/rs6000.md: Remove paired float support.
2752 (move_from_CR_ov_bit): Delete.
2753 * config/rs6000/rs6000.opt (mpaired): Delete.
2754 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
2755 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
2757 2018-05-02 Richard Biener <rguenther@suse.de>
2760 * gimplify.c (gimplify_save_expr): When in SSA form allow
2761 SAVE_EXPRs to compute to SSA vars.
2763 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2766 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2767 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2768 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
2769 clobber operands[2], instead use a new pseudo. Formatting fixes.
2771 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
2773 PR tree-optimization/85586
2774 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
2775 exit early for statements in the same group if the accesses are
2778 2018-05-02 Tom de Vries <tom@codesourcery.com>
2781 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
2784 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
2786 PR tree-optimization/85143
2787 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
2789 2018-05-01 Tom de Vries <tom@codesourcery.com>
2792 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
2793 not found" error message.
2795 2018-05-01 Tom de Vries <tom@codesourcery.com>
2798 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
2799 * vec.c (test_ordered_remove_if): New function.
2800 (vec_c_tests): Call test_ordered_remove_if.
2801 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
2802 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
2803 * tree-vect-patterns.c (vect_pattern_recog_1): Use
2804 VEC_ORDERED_REMOVE_IF.
2806 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2808 PR tree-optimization/82665
2809 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
2810 pointer subtraction where arguments come from a memchr call.
2812 2018-05-01 Jakub Jelinek <jakub@redhat.com>
2814 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
2815 --push-state --as-needed and --pop-state instead of --as-needed and
2816 --no-as-needed if ld supports it.
2817 * configure: Regenerated.
2820 * doc/install.texi2html: Replace _002d with - and _002a with * in
2821 generated html files using sed.
2823 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2826 * gcc-rich-location.c (blank_line_before_p): New function.
2827 (use_new_line): New function.
2828 (gcc_rich_location::add_fixit_insert_formatted): New function.
2829 * gcc-rich-location.h
2830 (gcc_rich_location::add_fixit_insert_formatted): New function.
2832 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2834 * selftest.c (assert_streq): Rename "expected" and "actual" to
2835 "val1" and "val2". Extend NULL-handling to cover both inputs
2836 symmetrically, while still requiring both to be non-NULL for a pass.
2837 * selftest.h (assert_streq): Rename "expected" and "actual" to
2839 (ASSERT_EQ): Likewise.
2840 (ASSERT_EQ_AT): Likewise.
2841 (ASSERT_KNOWN_EQ): Likewise.
2842 (ASSERT_KNOWN_EQ_AT): Likewise.
2843 (ASSERT_NE): Likewise.
2844 (ASSERT_MAYBE_NE): Likewise.
2845 (ASSERT_MAYBE_NE_AT): Likewise.
2846 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
2847 the assertion to pass.
2848 (ASSERT_STREQ_AT): Likewise.
2850 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2852 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
2853 interaction with -pie.
2855 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2857 * selftest.h: Fix alphabetization of per-source-file selftest
2860 2018-04-30 Jason Merrill <jason@redhat.com>
2862 PR c++/61982 - dead stores to destroyed objects.
2863 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
2866 2018-04-30 Jason Merrill <jason@redhat.com>
2868 * tree.c (build_clobber): New.
2869 * tree.h: Declare it.
2870 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
2872 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2874 * diagnostic-show-locus.c (layout::layout): Update for
2875 location_get_source_line returning a char_span.
2876 (struct char_span): Move to input.h.
2877 (struct correction): Update for fields in char_span becoming
2879 (struct source_line): Update for location_get_source_line
2880 returning a char_span.
2881 (layout::print_line): Likewise.
2882 * edit-context.c (edited_file::print_content): Likewise.
2883 (edited_file::print_diff_hunk): Likewise.
2884 (edited_file::print_run_of_changed_lines): Likewise.
2885 (edited_file::get_num_lines): Likewise.
2886 (edited_line::edited_line): Likewise.
2887 * final.c (asm_show_source): Likewise.
2888 * input.c (location_get_source_line): Convert return type
2889 from const char * to char_span, losing the final "line_len"
2891 (dump_location_info): Update for the above.
2892 (get_substring_ranges_for_loc): Likewise. Use a char_span
2893 when handling the literal within the line.
2894 (test_reading_source_line): Update for location_get_source_line
2895 returning a char_span.
2896 * input.h (class char_span): Move here from
2897 diagnostic-show-locus.c, converting from a struct to a class.
2898 Make data members private.
2899 (char_span::operator bool): New.
2900 (char_span::length): New.
2901 (char_span::get_buffer): New.
2902 (char_span::operator[]): New.
2903 (char_span::subspan): Make const.
2904 (char_span::xstrdup): New.
2905 (location_get_source_line): Convert return type from const char *
2906 to char_span, losing the final "line_size" param.
2908 2018-04-30 Jan Hubicka <jh@suse.cz>
2910 * lto-wrapper.c (ltrans_priorities): New static var.
2911 (cmp_priority): New.
2912 (run_gcc): Read priorities and if doing parallel build order
2913 the Makefile by them.
2915 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2917 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
2919 2018-04-30 Richard Biener <rguenther@suse.de>
2921 * tree-cfg.c (verify_address): Remove base argument, add
2922 flag whether to check TREE_ADDRESSABLE and do that.
2923 (verify_expr): Remove.
2924 (verify_types_in_gimple_reference): Add pieces from verify_expr.
2925 (verify_gimple_assign_single): Likewise.
2926 (verify_gimple_switch): Likewise.
2927 (verify_expr_location_1): Dereference tp once. Add (disabled)
2928 piece from verify_expr.
2929 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
2931 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2933 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
2935 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2937 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
2938 (small_data_pattern): Likewise.
2939 (arc_rewrite_small_data): Likewise.
2940 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
2941 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
2942 (get_symbol_alignment): New function.
2943 (legitimate_small_data_address_p): Likewise.
2944 (legitimate_scaled_address): Update, call
2945 legitimate_small_data_address_p.
2946 (output_sdata): New static variable.
2947 (arc_print_operand): Update how we handle small data operands.
2948 (arc_print_operand_address): Likewise.
2949 (arc_legitimate_address_p): Update, use
2950 legitimate_small_data_address_p.
2951 (arc_rewrite_small_data_p): Remove.
2952 (arc_rewrite_small_data_1): Likewise.
2953 (arc_rewrite_small_data): Likewise.
2954 (small_data_pattern): Likewise.
2955 (compact_sda_memory_operand): Update to use
2956 legitimate_small_data_address_p and get_symbol_alignment.
2957 (prepare_move_operands): Don't rewite sdata pattern.
2958 (prepare_extend_operands): Remove.
2959 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
2961 (zero_extendqisi2): Likewise.
2962 (zero_extendhisi2): Likewise.
2963 (extendqihi2): Likewise.
2964 (extendqisi2): Likewise.
2965 (extendhisi2): Likewise.
2969 * config/arc/constraints.md (Usd): Change it to memory constraint.
2971 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2973 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
2974 as source of std instructions.
2975 * config/arc/arc.md (movsi_insn): Update pattern predicate to
2976 allow 6-bit constants as source for store instructions.
2977 (movdi_insn): Update instruction pattern to allow 6-bit constants
2978 as source for store instructions.
2980 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2982 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
2984 2018-04-30 Nathan Sidwell <nathan@acm.org>
2985 Sandra Loosemore <sandra@codesourcery.com>
2987 * dumpfile.c (dump_open): Allow '-' for stdout.
2988 * doc/invoke.texi (Developer Options): Document dump filename
2989 determination early. Document stdin/stdout selection.
2991 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
2993 Microblaze Target: PIC data text relative
2995 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
2996 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
2998 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
2999 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
3000 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
3001 New addressing mode for data-text relative position indepenedent code.
3002 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
3003 'ADDRESS_SYMBOLIC_TXT_REL'.
3004 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
3005 (microblaze_legitimate_pic_operand): Exclude function calls from
3006 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
3007 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
3009 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3010 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3011 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
3012 for 'address + offset'.
3013 (microblaze_expand_prologue): Add new function prologue call for
3015 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
3016 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
3017 table in case of TARGET_PIC_DATA_TEXT_REL.
3018 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
3019 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
3020 Add new macros 'UNSPEC_TEXT',
3021 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
3022 + exclude function calls from 'UNSPEC_PLT' in case of data text
3024 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3025 new target hook for generating address diff vector tables in case of
3027 * doc/tm.texi : Regenerate.
3028 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
3029 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
3030 of addr diff vector generation.
3031 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3032 target hook definition.
3033 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
3034 Add default function for generate_pic_addr_diff_vec -> flag_pic.
3035 * doc/invoke.texi (Add new pic option): Add new microblaze pic
3036 option for data text relative.
3038 2018-04-30 Richard Biener <rguenther@suse.de>
3040 * tree-chrec.h (evolution_function_is_constant_p): Remove
3042 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
3044 2018-04-30 Richard Biener <rguenther@suse.de>
3047 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
3049 2018-04-30 Richard Biener <rguenther@suse.de>
3051 PR tree-optimization/28364
3052 PR tree-optimization/85275
3053 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
3054 copying first exit test.
3056 2018-04-28 Mark Wielaard <mark@klomp.org>
3058 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
3060 (dwarf_AT): Handle DW_AT_addr_base.
3061 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
3063 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
3066 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
3067 (*ashl<dwi>3_doubleword_mask_1): Ditto.
3068 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
3069 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
3071 2018-04-28 Richard Biener <rguenther@suse.de>
3073 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
3074 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
3075 to reflect use. Only add interesting stmts.
3077 2018-04-27 Martin Jambor <mjambor@suse.cz>
3080 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
3081 the jump function allows for passing through aggregate values.
3083 2018-04-27 David Malcolm <dmalcolm@redhat.com>
3085 * input.h (in_system_header_at): Convert from macro to inline
3087 (from_macro_expansion_at): Likewise.
3088 (from_macro_definition_at): Likewise.
3090 2018-04-27 Jeff Law <law@redhat.com>
3092 * config.gcc: Mark tile* targets as deprecated/obsolete.
3094 2018-04-27 Richard Biener <rguenther@suse.de>
3096 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
3099 2018-04-27 Richard Biener <rguenther@suse.de>
3101 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
3103 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
3105 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
3106 with Yd constraint. Set "preferred_for_speed" attribute from
3107 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
3109 (*movdi_internal): Ditto.
3110 (movti_interunit splitters): Remove
3111 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
3112 (movdi_interunit splitters): Ditto.
3113 * config/i386/constraints.md (Ye): Remove.
3114 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
3116 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3119 * config/aarch64/constraints.md (Usg): Limit to 31.
3122 2018-04-27 Jakub Jelinek <jakub@redhat.com>
3124 PR tree-optimization/85529
3125 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
3126 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
3127 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
3128 zero extension or masking of the MSB bit.
3129 (optimize_range_tests): Add FIRST_BB argument, pass it through
3130 to optimize_range_tests_var_bound.
3131 (maybe_optimize_range_tests, reassociate_bb): Adjust
3132 optimize_range_tests callers.
3134 2018-04-26 Richard Biener <rguenther@suse.de>
3135 Jakub Jelinek <jakub@redhat.com>
3137 * cgraph.h (symbol_table): Just declare debug method here.
3138 * symtab.c (symbol_table::debug): Define.
3140 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
3142 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
3144 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
3146 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
3147 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
3148 (*movdi_internal): Substitute Yi and Yj constraint with x
3149 and Ym and Yn constraint with y constraint. Update "isa"
3150 attribute and set "preferred_for_speed" attribute from
3151 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
3152 (*movsi_internal): Ditto.
3153 (*movdf_internal): Ditto.
3154 (*movsf_internal): Ditto.
3155 (*zero_extendsidi2): Ditto.
3156 * config/i386/sse.md (vec_set<mode>_0): Ditto.
3157 (sse2_loadld): Ditto.
3158 (*vec_extract<ssevecmodelower>_0): Ditto.
3159 (*vec_extractv4si_0_zext_sse4): Ditto.
3160 (vec_concatv2di): Ditto.
3161 (*vec_dup<mode>): Ditto.
3162 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
3163 * config/i386/constraints.md (Yi): Remove.
3168 2018-04-26 Nathan Sidwell <nathan@acm.org>
3170 * dumpfile.c (dump_open): New.
3171 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
3172 (dump_finish): Detect stdio/stderr by value not name.
3174 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
3176 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
3178 2018-04-26 Tom de Vries <tom@codesourcery.com>
3181 * config/nvptx/nvptx.c (verify_neutering_jumps)
3182 (verify_neutering_labels): New function
3183 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
3185 2018-04-26 Tom de Vries <tom@codesourcery.com>
3188 * config/nvptx/nvptx.c (needs_neutering_p): New function.
3189 (nvptx_single): Use needs_neutering_p to skip over insns that do not
3192 2018-04-26 Richard Biener <rguenther@suse.de>
3193 Tom de Vries <tom@codesourcery.com>
3196 * lto-streamer-out.c (output_function): Fixup loops if required to match
3197 discovery done in the reader.
3199 2018-04-26 Richard Biener <rguenther@suse.de>
3201 PR tree-optimization/85116
3202 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
3203 have a loop exit from the single latch predecessor. Remove
3204 case of header with just condition.
3205 (ch_base::copy_headers): Exclude infinite loops from any
3207 (pass_ch::execute): Record exits.
3209 2018-04-26 Richard Biener <rguenther@suse.de>
3211 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
3212 prologue cost vector and pass it to vect_get_load_cost.
3213 (vect_get_peeling_costs_all_drs): Likewise.
3214 (vect_peeling_hash_get_lowest_cost): Likewise.
3215 (vect_enhance_data_refs_alignment): Likewise.
3217 2018-04-26 Richard Biener <rguenther@suse.de>
3220 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
3221 checking of integer<->pointer conversions.
3222 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
3223 sign-/zero-extending pointer types.
3224 (expand_omp_for_static_chunk): Likewise.
3226 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
3227 Jean Lee <xiaoyur347@gmail.com>
3229 * config/mips/mips.c (mips_asan_shadow_offset): New function.
3230 (TARGET_ASAN_SHADOW_OFFSET): Define.
3231 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
3232 true for -fsanitize=address.
3234 2018-04-25 Mark Wielaard <mark@klomp.org>
3236 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
3239 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3241 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
3242 than "alu", remove explicit "memory" and "imm_disp" attributes.
3243 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
3246 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
3247 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
3248 gen_lowpart_no_emit.
3250 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
3253 * config/i386/i386.c (ix86_expand_builtin): Change memory
3254 operand to XI, extend p0 to Pmode.
3255 * config/i386/i386.md: Change unspec volatile and operand
3256 1 mode to XI, change operand 0 mode to P.
3258 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3260 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
3261 GET_MODE_MASK before any checking.
3262 (nds32_can_use_bset_p): Likewise.
3263 (nds32_can_use_btgl_p): Likewise.
3265 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3267 * config/nds32/nds32-doubleword.md: New define_split pattern for
3268 illegal register number.
3270 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3272 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
3274 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3276 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
3278 2018-04-25 Richard Biener <rguenther@suse.de>
3280 * lto-streamer.h (LTO_major_version): Bump to 8.
3282 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3284 * BASE-VER: Set to 9.0.0.
3286 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
3288 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
3289 in __abskf2 and __powikf2.
3291 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3294 * config/aarch64/constraints.md (Usg, Usj): New constraints.
3295 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
3296 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
3297 Use the above on operand 2. Reindent.
3298 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
3300 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3303 * common/config/i386/i386-common.c (ix86_handle_option): Don't
3305 * config/i386/i386.opt (mcet): Removed.
3306 * doc/install.texi: Remove -mcet documentation.
3307 * doc/invoke.texi: Likewise.
3309 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3312 * doc/install.texi: Remove -mcet from bootstrap-cet.
3314 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3317 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
3318 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
3322 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
3323 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
3324 containing a CONST_VECTOR.
3326 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
3328 * doc/install.texi: Update newlib dependency for nvptx.
3330 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3333 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
3334 instead of INTVAL when shifting x left.
3336 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
3338 PR tree-optimization/85478
3339 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
3340 vect_grouped_store_supported for single element vectors.
3342 2018-04-24 Richard Biener <rguenther@suse.de>
3345 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
3346 load cost increase to the case of non-constant step.
3348 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3351 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
3352 destination if any_malformed_asm.
3354 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
3357 * expr.c (store_field): In the bitfield case, if the value comes from
3358 a function call and is returned in registers by means of a PARALLEL,
3359 do not change the mode of the temporary unless BLKmode and VOIDmode.
3361 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
3363 PR rtl-optimization/85423
3364 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
3365 dependencies to debug insns when the previous insn is non-debug.
3367 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3369 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
3370 enums into a single definition.
3371 (fls): Fix predicates and printing.
3374 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3376 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
3377 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
3378 and short u6 immediate.
3379 (check_if_valid_sleep_operand): Remove.
3380 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
3382 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3384 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
3385 flag_always_save_lp condition.
3386 * config/nds32/nds32.opt (malways-save-lp): New option.
3388 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3390 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
3391 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
3392 * config/nds32/nds32.h
3393 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
3394 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
3396 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3398 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
3399 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
3401 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3402 Chung-Ju Wu <jasonwucj@gmail.com>
3404 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
3406 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
3407 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
3409 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3411 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
3413 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3415 * config/nds32/nds32-protos.h (nds32_data_alignment,
3416 nds32_local_alignment): Declare.
3417 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
3418 nds32_local_alignment): New functions.
3419 (TARGET_CONSTANT_ALIGNMENT): Define.
3420 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
3422 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3424 * config/nds32/nds32.c
3425 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
3426 (TARGET_MODES_TIEABLE_P): Likewise.
3428 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3430 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
3433 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
3434 Chung-Ju Wu <jasonwucj@gmail.com>
3436 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3437 for unaligned access.
3438 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3440 * config/nds32/nds32-intrinsic.md: Likewise.
3441 * config/nds32/nds32_intrinsic.h: Likewise.
3442 * config/nds32/nds32.h (nds32_builtins): Likewise.
3443 * config/nds32/nds32.opt (munaligned-access): New option.
3444 * config/nds32/nds32.c (nds32_asm_file_start): Display
3445 flag_unaligned_access status.
3447 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3449 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
3450 -mno-relax is present.
3451 * config/riscv/linux.h (LINK_SPEC): Ditto.
3453 2018-04-20 Martin Sebor <msebor@redhat.com>
3456 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
3458 (gimple_fold_builtin_stxcpy_chk): Same.
3459 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
3461 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
3464 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
3465 __powikf2 when long double is IEEE 128-bit.
3467 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3469 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
3470 step to make sure stack always aligned.
3472 2018-04-20 Carl Love <cel@us.ibm.com>
3475 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
3476 size check for arg0.
3478 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
3479 Tom de Vries <tom@codesourcery.com>
3482 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
3483 Emit insns for calls too.
3484 (nvptx_find_par): Always look for worker-level predecessor insn.
3485 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
3487 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
3488 (nvptx_process_pars): Propagate frames for calls.
3490 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
3493 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
3495 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
3496 (ix86_handle_option): Don't handle OPT_mibt.
3497 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
3499 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
3501 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
3502 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
3503 (ix86_target_macros): Define __CET__ with flag_cf_protection
3504 for -fcf-protection.
3505 * config/i386/i386.c (isa2_opts): Remove -mibt.
3506 * config/i386/i386.h (TARGET_IBT): Removed.
3507 (TARGET_IBT_P): Likewise.
3508 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
3509 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
3510 * config/i386/i386.opt (mcet): Update help message.
3513 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
3514 -mcet as an alias for -mshstk.
3516 2018-04-20 Richard Biener <rguenther@suse.de>
3519 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
3520 complexity by forcing a single use of the multiply operand.
3522 2018-04-20 Martin Jambor <mjambor@suse.cz>
3525 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
3526 recursion dependency to only apply to non-clones.
3528 2018-04-20 Martin Jambor <mjambor@suse.cz>
3531 * ipa-cp.c (create_specialized_node): Check that clones of
3532 self-recursive edges exist during IPA-CP.
3534 2018-04-19 Toon Moene <toon@moene.org>
3536 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
3539 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3541 PR tree-optimization/85467
3542 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
3543 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
3544 VECTOR_CST element to type.
3546 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3549 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
3550 * config/i386/i386.md (builtin_setjmp_setup): Removed.
3551 (builtin_longjmp): Likewise.
3552 (save_stack_nonlocal): New pattern.
3553 (restore_stack_nonlocal): Likewise.
3555 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3558 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3559 Replace ASM_OUTPUT_LABEL with fprintf.
3561 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3564 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3565 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
3566 * config/i386/i386-c.c (ix86_target_macros_internal): Also
3567 define __IBT__ and __SHSTK__ for -fcf-protection.
3568 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
3570 (ix86_trampoline_init): Likewise.
3571 (x86_output_mi_thunk): Likewise.
3572 (ix86_notrack_prefixed_insn_p): Likewise.
3573 (ix86_option_override_internal): Don't disallow -fcf-protection.
3574 * config/i386/i386.md (rdssp<mode>): Also enable for
3576 (incssp<mode>): Likewise.
3577 (nop_endbr): Likewise.
3578 * config/i386/i386.opt (mcet): Change help message to built-in
3582 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
3583 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
3584 enable CET built-in functions.
3586 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
3588 * common/config/i386/i386-common.c
3589 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
3590 OPTION_MASK_ISA_MOVDIRI_UNSET,
3591 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
3592 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
3593 * config.gcc (movdirintrin.h): New header.
3594 * config/i386/cpuid.h (bit_MOVDIRI,
3595 bit_MOVDIR64B): New bits.
3596 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
3598 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
3599 (VOID, PVOID, PCVOID)): New function types.
3600 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
3601 __builtin_ia32_directstoreu_u64,
3602 __builtin_ia32_movdir64b): New builtins.
3603 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
3604 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
3606 (ix86_valid_target_attribute_inner_p): Ditto.
3607 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
3608 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
3609 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
3610 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
3611 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
3612 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
3613 (movdiri<mode>, movdir64b_<mode>): New.
3614 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
3615 * config/i386/immintrin.h: Include movdirintrin.h.
3616 * config/i386/movdirintrin.h: New file.
3617 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
3619 2018-04-19 Richard Biener <rguenther@suse.de>
3622 * cfg.c (clear_bb_flags): When loop state says we have
3623 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
3625 2018-04-19 Richard Biener <rguenther@suse.de>
3627 PR tree-optimization/84737
3628 * tree-vect-data-refs.c (vect_copy_ref_info): New function
3629 copying restrict info.
3630 (vect_setup_realignment): Use it.
3631 * tree-vectorizer.h (vect_copy_ref_info): Declare.
3632 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
3633 the first DR to all generated stores.
3634 (vectorizable_load): Likewise for loads.
3636 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3638 PR tree-optimization/85446
3639 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
3640 the integral and pointer types to have the same precision.
3642 * doc/install.texi: Document --disable-cet being the default and
3645 2018-04-18 Martin Liska <mliska@suse.cz>
3647 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
3650 2018-04-18 Martin Liska <mliska@suse.cz>
3653 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3656 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3657 arguments if they are comparable.
3659 2018-04-18 Martin Liska <mliska@suse.cz>
3662 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3665 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3668 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
3671 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
3672 ENDBR after calling __morestack.
3674 2018-04-18 David Malcolm <dmalcolm@redhat.com>
3677 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
3678 by using gcc_base_ver to generate a gcc_driver_version, and use
3679 it when generating GCC_DRIVER_NAME.
3680 * configure: Regenerate.
3682 2018-04-18 Jakub Jelinek <jakub@redhat.com>
3685 * config.gcc: Obsolete powerpc*-*-*spe*.
3687 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3690 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
3691 (stabstr_D): Change type of unum from unsigned int to
3692 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
3695 2018-04-17 Jim Wilson <jimw@sifive.com>
3698 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
3699 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
3700 Set arg_pointer_offset after using pretend_args_size.
3702 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3704 PR rtl-optimization/85431
3705 * dse.c (record_store): Ignore zero width stores.
3708 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
3709 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
3710 __builtin_stack_restore rather than after it.
3711 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
3712 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
3713 argument instead of virtual_dynamic_stack_rtx.
3715 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
3717 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
3719 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
3720 Add note to error message to explain internal mapping of overloaded
3721 built-in function name to non-overloaded built-in function name.
3722 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
3725 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3728 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
3729 where the inputs overlap with the output.
3731 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3733 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
3734 (=v, v) alternative and explicit "memory" attribute.
3735 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
3736 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3738 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
3739 "sselog1" type instead of "sselog".
3740 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
3741 "sselog". Remove explicit "memory" attribute.
3742 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
3743 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3745 (vec_extract_hi_v32hi): Merge all alternatives into one, use
3746 "sselog1" type instead of "sselog". Remove explicit "memory"
3748 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
3749 use "sselog1" type instead of "sselog". Remove explicit "memory"
3751 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
3752 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3754 (vec_extract_hi_v64qi): Merge all alternatives into one, use
3755 "sselog1" type instead of "sselog". Remove explicit "memory"
3757 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
3758 use "sselog1" type instead of "sselog". Remove explicit "memory"
3762 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
3765 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
3768 2018-04-17 Martin Jambor <mjambor@suse.cz>
3771 * ipa-cp.c (create_specialized_node): Call
3772 expand_all_artificial_thunks if necessary.
3774 2018-04-17 Martin Liska <mliska@suse.cz>
3777 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
3778 in message, remote space in between '_G' and '('.
3780 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3783 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
3784 avx512f_vmcmp<mode>3<round_saeonly_name>,
3785 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
3786 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
3787 avx512f_rndscale<mode><round_saeonly_name>,
3788 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
3789 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
3790 Use %<iptr>2 instead of %2 for -masm=intel.
3791 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
3792 avx512f_vcvttss2usi<round_saeonly_name>,
3793 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
3795 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
3796 avx512f_vcvttsd2usi<round_saeonly_name>,
3797 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
3798 Use %q1 instead of %1 for -masm=intel.
3799 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
3800 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
3801 of %3 for -masm=intel.
3802 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
3804 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
3806 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
3808 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
3809 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
3811 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
3813 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
3814 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
3815 %g1 and one with %0 and %1.
3816 (avx512er_vmrcp28<mode><round_saeonly_name>,
3817 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
3819 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
3820 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
3821 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
3822 of %0 and %{%4%} for -masm=intel.
3823 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
3824 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
3825 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
3826 order of %0 and %{%5%}%{z%} for -masm=intel.
3828 2018-04-17 Jan Hubicka <jh@suse.cz>
3831 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
3833 2018-04-17 Martin Liska <mliska@suse.cz>
3836 * multiple_target.c (create_dispatcher_calls): Set apostrophes
3837 for target_clone error message. Make default implementation
3838 clone to be a local declaration.
3839 (separate_attrs): Add new argument and check for an empty
3841 (expand_target_clones): Handle it.
3842 (ipa_target_clone): Make redirection just for target_clones
3845 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
3846 Tom de Vries <tom@codesourcery.com>
3849 * omp-expand.c (expand_oacc_for): Add dummy false branch for
3850 tiled basic blocks without omp continue statements.
3852 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
3855 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
3856 vec_extract expression as having side effects to make sure it gets
3859 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
3862 * config/i386/i386.c (get_builtin_code_for_version): Check
3865 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
3868 * gcc/config.gcc: Support "skylake".
3869 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
3871 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
3872 (processor_target_table): Add "skylake".
3873 (ix86_option_override_internal): Add "skylake".
3874 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
3875 PROCESSOR_CANNONLAKE.
3876 (get_builtin_code_for_version): Fix priority for
3877 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
3878 PROCESSOR_SKYLAKE-AVX512.
3879 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
3880 (processor_type): Add PROCESSOR_SKYLAKE.
3882 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
3883 Jason Merrill <jason@redhat.com>
3886 * convert.c (convert_to_integer_1): Use direct recursion for
3887 enumeral types and types with a precision less than the number
3888 of bits in their mode.
3890 2018-04-16 Julia Koval <julia.koval@intel.com>
3893 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
3894 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
3896 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
3899 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
3900 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
3901 and -mno-direct-move.
3903 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
3906 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
3907 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
3908 Ensure negative shifts result in {0}.
3910 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
3912 PR rtl-optimization/79916
3913 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
3914 regs (if any) to define how to gnerate SD moves when LRA is in
3917 2018-04-13 Jakub Jelinek <jakub@redhat.com>
3919 PR rtl-optimization/85393
3920 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
3921 * except.c (expand_dw2_landing_pad_for_region): Make static.
3922 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
3923 a label and unconditional jump to old_bb, rather than
3924 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
3927 PR rtl-optimization/85376
3928 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
3929 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
3930 instead of a specific value.
3932 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3933 Bin Cheng <bin.cheng@arm.com>
3935 PR tree-optimization/82965
3936 PR tree-optimization/83991
3937 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
3938 by_profile_only parameter.
3939 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
3940 information if the loop was predicted to iterate too many times.
3941 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
3943 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3946 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
3949 2018-04-13 Martin Liska <mliska@suse.cz>
3950 Jakub Jelinek <jakub@redhat.com>
3953 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
3954 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
3955 * builtins.c (expand_builtin_memory_copy_args): Use
3956 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
3957 handle dest_addr == pc_rtx.
3959 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
3962 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
3963 asked to not generate direct moves.
3964 (fix_trunc<mode>si2_stfiwx): Similar.
3965 (fix_trunc<mode>si2_internal): Similar.
3967 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3970 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
3971 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
3972 lookup if dest in some wider mode is known to be const0_rtx and
3973 if so, record permanent equivalence for it to be ZERO_EXTEND of
3974 the narrower mode destination.
3976 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3978 * lto-streamer-out.c (output_function): Revert 259346.
3979 * omp-expand.c (expand_oacc_for): Likewise.
3981 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
3983 PR rtl-optimization/85354
3984 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
3985 * sel-sched.c (sel_global_init): ... here.
3987 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
3990 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
3991 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
3992 mode for PE-COFF targets.
3993 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
3994 (i386_pe_asm_lto_end): Likewise.
3995 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
3996 (TARGET_ASM_LTO_END): Likewise.
3997 * config/i386/winnt.c (saved_debug_info_level): New static variable.
3998 (i386_pe_asm_lto_start): New function.
3999 (i386_pe_asm_lto_end): Likewise.
4001 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4002 Richard Biener <rguenther@suse.de>
4005 * lto-streamer-out.c (output_function): Fix CFG loop state before
4007 * omp-expand.c (expand_oacc_for): Handle calls to internal
4008 functions like regular functions.
4010 2018-04-12 Richard Biener <rguenther@suse.de>
4013 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
4014 for the early LTO debug to properly generate references to it
4015 during DIE emission. Do not re-use that for the skeleton for
4017 (dwarf2out_early_finish): Likewise.
4019 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4022 * config/i386/sse.md
4023 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
4024 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
4025 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
4026 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
4027 and output is a reg, avoid creating invalid lowpart subreg, but
4028 instead split into a 512-bit move. Don't split if not AVX512VL,
4029 input is xmm16+ reg and output is a mem.
4030 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
4031 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
4032 xmm16+ reg and output is a mem.
4034 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4036 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
4037 also for flag_dwarf2_cfi_asm.
4039 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4041 PR rtl-optimization/85342
4042 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
4043 a bool scalar var inside of the loop instead. Don't try to update
4044 recog_data.operand after failed apply_change_group.
4046 2018-04-12 Tom de Vries <tom@codesourcery.com>
4049 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
4050 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
4051 array with flexible array member as array without given dimension.
4052 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
4053 argument for undefined param to true.
4055 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
4058 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4059 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
4060 from PowerPC section.
4061 * config/rs6000/sysv4.opt (mcall-): Improve help text.
4062 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
4063 help text that is too long.
4064 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
4065 help text that is too long.
4066 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
4067 help text that is too long.
4069 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
4071 * config/alpha/alpha.md (stack_probe_internal): Rename
4072 from "probe_stack". Update all callers.
4074 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4076 PR rtl-optimization/84566
4077 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
4078 sched_macro_fuse_insns.
4080 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4083 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
4084 (compute_block_dependences): ... from here.
4086 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4088 PR tree-optimization/85331
4089 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
4090 from int to HOST_WIDE_INT.
4092 2018-04-11 Martin Jambor <mjambor@suse.cz>
4095 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
4096 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
4097 not the same as the source val.
4098 (cgraph_edge_brings_value_p): New parameter.
4099 (gather_edges_for_value): Pass destination value to
4100 cgraph_edge_brings_value_p.
4101 (perhaps_add_new_callers): Likewise.
4102 (get_info_about_necessary_edges): Likewise and exclude values brought
4103 only by self-recursive edges.
4104 (create_specialized_node): Redirect only clones of self-calling edges.
4105 (+self_recursive_pass_through_p): New function.
4106 (find_more_scalar_values_for_callers_subset): Use it.
4107 (find_aggregate_values_for_callers_subset): Likewise.
4108 (known_aggs_to_agg_replacement_list): Removed.
4109 (decide_whether_version_node): Re-calculate known constants for all
4110 remaining context clones.
4112 2018-04-11 Richard Biener <rguenther@suse.de>
4115 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
4116 from early DWARF output.
4117 (dwarf2out_early_finish): Output line info unconditionally into
4118 early DWARF and add reference to it.
4120 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4123 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
4124 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
4125 other than V2DFmode using iptr mode attribute.
4126 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
4128 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4130 PR rtl-optimization/84659
4131 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
4133 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4136 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
4138 (output_loc_list): Pass address of a dummy size variable even in the
4139 locview handling loop.
4140 (index_location_lists): Add comment on why skip_loc_list_entry can't
4143 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
4146 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
4149 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
4152 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4153 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
4154 and -mstring-compare-inline-limit.
4156 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4159 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
4160 for stack clash protection in a register whenever we need it to be in
4163 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4165 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
4166 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
4168 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4171 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
4173 (mlong-double-): Ditto.
4174 * config/rs6000/sysv4.opt (msdata=): Ditto.
4175 (mtls-size=): Ditto.
4177 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4179 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4180 erroneous entries for
4181 "vector int vec_ldl (int, long int *)", and
4182 "vector unsigned int vec_ldl (int, unsigned long int *)".
4183 Add comments and entries for
4184 "vector bool char vec_ldl (int, bool char *)",
4185 "vector bool short vec_ldl (int, bool short *)",
4186 "vector bool int vec_ldl (int, bool int *)",
4187 "vector bool long long vec_ldl (int, bool long long *)",
4188 "vector pixel vec_ldl (int, pixel *)",
4189 "vector long long vec_ldl (int, long long *)",
4190 "vector unsigned long long vec_ldl (int, unsigned long long *)".
4191 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
4192 type tree bool_long_long_type_node and correct definition of
4193 bool_V2DI_type_node to make reference to this new type tree.
4194 (rs6000_mangle_type): Replace erroneous reference to
4195 bool_long_type_node with bool_long_long_type_node.
4196 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
4197 comments to emphasize sign distinctions for char and int types and
4198 replace RS6000_BTI_bool_long constant with
4199 RS6000_BTI_bool_long_long constant. Also add comment to restrict
4200 use of RS6000_BTI_pixel.
4201 (bool_long_type_node): Remove this macro definition.
4202 (bool_long_long_type_node): New macro definition
4204 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4206 PR rtl-optimization/85300
4207 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
4208 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
4209 simplify_unary_operation fails.
4211 2018-04-10 Martin Liska <mliska@suse.cz>
4213 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
4214 cgraph_edge and ipa_ref.
4216 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4220 * config/i386/sse.md
4221 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
4222 computation of the VEC_MERGE selector from mask.
4223 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
4224 Fix decoding of the VEC_MERGE selector into mask.
4226 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4228 PR tree-optimization/85286
4229 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
4231 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4233 * final.c (final_1): Set insn_last_address as well as
4234 insn_current_address.
4236 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4239 * explow.c (emit_stack_probe): Call validize_mem on memory location
4240 before passing it to gen_probe_stack. Create address operand and
4241 legitimize it for the probe_stack_address case.
4243 2018-04-09 Jan Hubicka <jh@suse.cz>
4246 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
4247 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
4248 * tree.c (free_lang_data_in_type): Fix handling of binfos;
4250 (free_lang_data): Rebuild type inheritance graph.
4252 2018-04-09 Martin Sebor <msebor@redhat.com>
4254 * invoke.texi (-finline-small-functions): Mention other optimization
4256 (-findirect-inlining, -fpartial-inlining): Same.
4257 (-finline-functions-called-once): Same.
4258 (-freorder-blocks-and-partition): Same.
4260 2018-04-09 Jan Hubicka <jh@suse.cz>
4263 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
4264 jumps; choose last target that matches the criteria (i.e.
4265 no partition changes for non-crossing jumps).
4266 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
4267 support for redirecting crossing jumps to non-crossing.
4269 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
4271 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
4272 also for naked functions.
4274 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
4276 * config/arc/arc.md (add_shift): New pattern.
4277 (add_shift2): Likewise.
4278 (sub_shift): Likewise.
4279 (sub_shift_cmp0_noout): Likewise.
4280 (compare_si_ashiftsi): Likewise.
4281 (xbfu_cmp0_noout): New combine pattern.
4282 (xbfu_cmp0"): Likewise.
4283 (movsi_set_cc_insn): Place the predicable variant first.
4284 (commutative_binary_cmp0_noout): Remove clobber.
4285 (commutative_binary_cmp0): New pattern.
4286 (noncommutative_binary_cmp0): Likewise.
4287 (noncommutative_binary_cmp0_noout): Likewise.
4288 (noncommutative_binary_comparison_result_used): Removed.
4289 (rsub_cmp0): New pattern.
4290 (rsub_cmp0_noout): Likewise.
4291 (extzvsi): Changed, keep only meaningful variants.
4292 (SQH, SEZ): New iterators.
4293 (SQH_postfix): New mode attribute.
4294 (SEZ_prefix): New code attribute.
4295 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
4296 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
4297 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
4299 (noncommutative_operator): Check the availability of barrel
4302 2018-04-09 Richard Biener <rguenther@suse.de>
4304 PR tree-optimization/85284
4305 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
4306 Only use the niter constraining form of simple_iv when the exit
4309 2018-04-09 Tom de Vries <tom@codesourcery.com>
4312 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4313 (define_expand "*memory_barrier"): New define_expand.
4314 (define_insn "memory_barrier"): New insn.
4316 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4318 PR rtl-optimization/80463
4319 PR rtl-optimization/83972
4320 PR rtl-optimization/83480
4322 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
4323 correct producer for the insn.
4324 (tidy_control_flow): Fixup seqnos in case of debug insns.
4326 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4328 PR rtl-optimization/83913
4330 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
4331 different sched-times when merging exprs.
4333 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4335 PR rtl-optimization/83962
4337 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
4338 tidy_fallthru_edge and tidy_control_flow.
4340 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4342 PR rtl-optimization/83530
4344 * sel-sched.c (force_next_insn): New global variable.
4345 (remove_insn_for_debug): When force_next_insn is true, also leave only
4346 next insn in the ready list.
4347 (sel_sched_region): When the region wasn't scheduled, make another pass
4348 over it with force_next_insn set to 1.
4350 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4352 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
4354 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4355 for interrupt control.
4356 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4357 functions for interrupt control.
4358 * config/nds32/nds32-intrinsic.md: Likewise.
4359 * config/nds32/nds32_intrinsic.h: Likewise.
4360 * config/nds32/nds32.h (nds32_builtins): Likewise.
4362 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4364 * config/nds32/nds32.c (nds32_init_machine_status,
4365 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
4366 strict_aligned_p field.
4367 (nds32_expand_to_rtl_hook): New function.
4368 (TARGET_EXPAND_TO_RTL_HOOK): Define.
4369 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
4371 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4372 Chung-Ju Wu <jasonwucj@gmail.com>
4374 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
4375 * config/nds32/nds32-n7.md: New file.
4376 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
4377 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
4379 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
4380 * config/nds32/nds32.md (pipeline_model): Add n7.
4381 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
4382 * config/nds32/pipelines.md: Include n7 settings.
4384 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4385 Chung-Ju Wu <jasonwucj@gmail.com>
4387 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
4388 * config/nds32/nds32-e8.md: New file.
4389 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
4390 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
4392 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
4393 * config/nds32/nds32.md (pipeline_model): Add e8.
4394 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
4395 * config/nds32/pipelines.md: Include e8 settings.
4397 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4398 Chung-Ju Wu <jasonwucj@gmail.com>
4400 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
4401 * config/nds32/nds32-n8.md: New file.
4402 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
4403 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
4405 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
4406 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
4407 * config/nds32/nds32.md (pipeline_model): Add n8.
4408 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
4409 * config/nds32/pipelines.md: Include n8 settings.
4411 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4412 Chung-Ju Wu <jasonwucj@gmail.com>
4414 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
4415 * config/nds32/nds32-n9-2r1w.md: New file.
4416 * config/nds32/nds32-n9-3r2w.md: New file.
4417 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
4418 nds32_register_ports): New or modify for cpu n9.
4419 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
4421 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
4422 * config/nds32/nds32-utils.c: New file.
4423 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
4424 TARGET_MUL_SLOW): Define.
4425 * config/nds32/nds32.md (pipeline_model): New attribute.
4426 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
4427 New options that support cpu n9.
4428 * config/nds32/pipelines.md: Include n9 settings.
4429 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
4431 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4433 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
4434 information if necessary.
4435 (output_cond_branch_compare_zero): Likewise.
4436 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
4437 (nds32_target_alignment): Refine for alignment.
4438 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
4439 (FUNCTION_BOUNDARY): Modify.
4440 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
4442 * config/nds32/nds32.opt (malways-align, malign-functions): New.
4444 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4446 * config/nds32/constants.md (unspec_volatile_element): Add values for
4447 TLB operation and data prefetch.
4448 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4449 functions for TLB operation and data prefetch.
4450 * config/nds32/nds32-intrinsic.md: Likewise.
4451 * config/nds32/nds32_intrinsic.h: Likewise.
4452 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
4453 (nds32_print_operand): Likewise.
4454 * config/nds32/nds32.h (nds32_builtins): Likewise.
4456 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
4457 Andrew Pinski <pinsika@gcc.gnu.org>
4460 * match.pd: Use constant_boolean_node of correct type instead of
4461 boolean_true_node or boolean_false_node for simplifying
4462 pointer comparisons to zero.
4464 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4466 PR tree-optimization/80021
4467 * tree.c (verify_type_variant): Make error call in verify_variant_match
4468 translatable and remove final full stop.
4470 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4472 * config/nds32/constants.md (unspec_volatile_element): Add
4473 UNSPEC_VOLATILE_EH_RETURN.
4474 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
4475 nds32_output_stack_pop): Support dwarf exception handling process.
4476 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
4477 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
4478 exception handling process.
4479 (nds32_compute_stack_frame): Likewise.
4480 (nds32_return_addr_rtx): Likewise.
4481 (nds32_initial_elimination_offset): Likewise.
4482 (nds32_expand_prologue): Likewise.
4483 (nds32_expand_epilogue): Likewise.
4484 (nds32_dynamic_chain_address): New function.
4485 * config/nds32/nds32.h (machine_function): Add fields for dwarf
4487 (DYNAMIC_CHAIN_ADDRESS): Define.
4488 (EH_RETURN_DATA_REGNO): Define.
4489 (EH_RETURN_STACKADJ_RTX): Define.
4490 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
4491 patterns for dwarf exception handling.
4493 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4495 * config/nds32/nds32.h: Clean up obsolete macros.
4497 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4499 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4500 Add enum values for particular instructions.
4501 * config/nds32/nds32-intrinsic.c: Implementation of expanding
4502 particular intrinsic functions.
4503 * config/nds32/nds32-intrinsic.md: Likewise.
4504 * config/nds32/nds32_intrinsic.h: Likewise.
4505 * config/nds32/nds32.h (nds32_builtins): Likewise.
4506 * config/nds32/nds32.md (type): Add pbsad and pbsada.
4507 (btst, ave): New patterns for particular instructions.
4509 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4511 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4512 Add enum values for atomic load/store and memory sync.
4513 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
4515 * config/nds32/nds32-intrinsic.md: Likewise.
4516 * config/nds32/nds32_intrinsic.h: Likewise.
4517 * config/nds32/nds32.h (nds32_builtins): Likewise.
4519 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4521 PR tree-optimization/85257
4522 * fold-const.c (native_encode_vector): If not all elts could fit
4523 and off is -1, return 0 rather than offset.
4524 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
4525 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
4526 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
4527 adjust buffer in native_interpret_expr call.
4529 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4531 * config/nds32/constants.md (unspec_volatile_element): Add cache
4532 control enum values.
4533 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
4534 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
4535 * config/nds32/nds32.c (nds32_cctl_names): New.
4536 (nds32_print_operand): Handle cache control register names.
4537 * config/nds32/nds32.h (nds32_builtins): New enum values.
4538 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
4540 * config/nds32/nds32.md (type): Add mmu.
4541 * config/nds32/pipelines.md (simple_insn): Add mmu.
4543 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4545 * config/nds32/nds32.md (type): Remove call.
4546 * config/nds32/pipelines.md (simple_insn): Likewise.
4548 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4550 * config/nds32/constants.md (unspec_volatile_element): Add
4551 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
4552 UNSPEC_VOLATILE_FMFCFG.
4553 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
4554 description for fmfcfg and fmfcsr.
4555 (bdesc_1arg): Add fmtcsr.
4556 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
4557 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
4558 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
4559 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
4560 unspec_fmfcfg): New patterns.
4561 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
4562 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
4563 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
4564 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
4565 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
4566 __nds32__fmfcfg): Define.
4568 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4570 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
4571 intrinsic register names.
4572 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
4573 intrinsic register enum values and macros.
4575 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4577 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
4578 for load/store addressing form.
4579 (nds32_print_operand_address): Likewise.
4581 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
4584 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
4585 based on LABEL_REF. Remove useless assertion.
4586 (pic_address_needs_scratch): Fix formatting.
4587 (sparc_legitimize_pic_address): Minor tweaks.
4588 (sparc_delegitimize_address): Adjust assertion accordingly.
4589 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
4590 into symbolic_operand.
4591 (movsi_high_pic_label_ref): Likewise.
4592 (movsi_lo_sum_pic_label_ref): Likewise.
4593 (movdi_pic_label_ref): Likewise.
4594 (movdi_high_pic_label_ref): Likewise.
4595 (movdi_lo_sum_pic_label_ref): Likewise.
4597 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
4599 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
4600 custom LIB_SPEC setup.
4602 2018-04-06 Ruslan Bukin <br@bsdpad.com>
4603 Kito Cheng <kito.cheng@gmail.com>
4605 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
4606 * config/riscv/freebsd.h: New.
4608 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4610 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
4611 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
4614 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4615 Kito Cheng <kito.cheng@gmail.com>
4617 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
4618 nds32_output_call, nds32_symbol_binds_local_p): New functions.
4619 * config/nds32/nds32-protos.h (nds32_output_call,
4620 nds32_output_return): Declare.
4621 * config/nds32/nds32.md: Refine all the call and return patterns.
4623 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4626 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
4627 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
4629 PR rtl-optimization/84872
4630 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
4631 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
4634 2018-04-06 Tamar Christina <tamar.christina@arm.com>
4636 * expr.c (copy_blkmode_to_reg): Revert 254862.
4637 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
4639 2018-04-06 Richard Biener <rguenther@suse.de>
4642 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
4643 after seeing a component reference with an adjacent field. Treat
4644 refs to arrays at struct end of external decls similar to
4645 refs to unconstrained commons.
4647 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4650 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
4651 look through SAVE_EXPRs with non-side-effects argument. Adjust
4653 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
4656 2018-04-06 Richard Biener <rguenther@suse.de>
4659 * alias.c (find_base_term): New wrapper around find_base_term
4660 unwinding CSELIB_VAL_PTR changes.
4661 (find_base_term): Do not restore CSELIB_VAL_PTR during the
4664 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4666 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
4668 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
4669 constant definitions.
4670 ("nop"): lr 0,0 -> nopr r0
4671 ("nop_lr0", "nop_lr1"): New insn definitions.
4673 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4675 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
4676 NDS32_V3PUSH_AVAILABLE_P macro.
4678 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
4679 Chung-Ju Wu <jasonwucj@gmail.com>
4681 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
4682 (nds32*-*-*): Add float and fpu_config into supported_defaults.
4683 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
4684 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
4685 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
4686 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
4687 * config/nds32/constraints.md: New constraints and checking for hard
4688 float configuration.
4689 * config/nds32/iterators.md: New mode iterator and attribute for hard
4690 float configuration.
4691 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
4693 * config/nds32/nds32-fpu.md: New file.
4694 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
4695 deal with hard float code generation.
4696 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
4698 (abi_type, float_reg_number): New enum type.
4699 * config/nds32/nds32-predicates.c: New predicates for hard float.
4700 * config/nds32/nds32-protos.h: Declare functions for hard float.
4701 * config/nds32/nds32.c: Implementation for hard float configuration.
4702 * config/nds32/nds32.h: Definitions for hard float configuration.
4703 * config/nds32/nds32.md: Include hard float machine description and
4704 modify patterns for hard float configuration.
4705 * config/nds32/nds32.opt: New options for hard float configuration.
4706 * config/nds32/predicates.md: New predicates for hard float
4709 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4711 * common/config/nds32/nds32-common.c
4712 (nds32_option_optimization_table): Enable -mreleax-hint by default.
4714 2018-04-05 Jakub Jelinek <jakub@redhat.com>
4717 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
4718 CONSTRUCTOR_ELT (ctor, ...)->value.
4720 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
4723 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
4725 2018-04-05 Tom de Vries <tom@codesourcery.com>
4728 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
4731 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4732 Kito Cheng <kito.cheng@gmail.com>
4734 * config/nds32/constraints.md (U33): Fine-tune checking condition.
4735 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
4736 * config/nds32/nds32.h (nds32_16bit_address_type): Add
4737 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
4739 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4740 Kito Cheng <kito.cheng@gmail.com>
4742 * config/nds32/constraints.md (Ufe): New memory constraint.
4743 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
4744 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
4745 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
4747 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
4748 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
4750 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4752 * config/nds32/nds32.md: Use optimize_size in the condition for
4753 alu-shift instructions.
4755 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4757 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
4759 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4761 * config/nds32/nds32.md (negsi2): Refine pattern.
4763 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4764 Chung-Ju Wu <jasonwucj@gmail.com>
4766 * config/nds32/iterators.md (shift_rotate): New code iterator.
4767 (shift): New code attribute.
4768 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
4769 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
4770 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
4771 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
4772 bit-wise operations.
4773 (andsi3, *andsi3): Ditto.
4774 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
4775 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
4776 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
4777 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
4778 nds32_ior_operand, nds32_xor_operand): New predicates.
4780 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4782 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
4783 (addsi3, subsi3): ... this.
4785 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4787 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
4789 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4791 * config/nds32/nds32.md: Adjust indention.
4793 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4795 * config/nds32/nds32.md (feature): New attribute.
4797 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4799 * config/nds32/nds32.md (subtype): New attribute.
4801 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
4804 * config/arm/arm-builtins.c (arm_expand_builtin): Change
4805 expansion to perform a bitwise AND of the argument followed by a
4806 boolean negation of the result.
4808 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
4810 PR rtl-optimization/84878
4811 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
4812 the basic block. Assert the use reference is not artificial and that
4813 it has an associated insn.
4815 2018-04-04 Michael Matz <matz@suse.de>
4817 * builtins.c (compute_objsize): Pass correct operand
4818 to array_at_struct_end_p.
4820 2018-04-04 Richard Biener <rguenther@suse.de>
4823 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
4824 from contexts for DINFO_LEVEL_TERSE and below.
4826 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4828 * config/nds32/nds32-doubleword.md (move_<mode>): Require
4829 resiter_operand condition.
4830 * config/nds32/nds32.md (*move<mode>): Ditto.
4832 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4833 Monk Chiang <sh.chiang04@gmail.com>
4835 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
4837 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4839 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
4841 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4842 Kito Cheng <kito.cheng@gmail.com>
4844 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
4845 nds32_cond_code_str, output_cond_branch,
4846 output_cond_branch_compare_zero, nds32_expand_cbranch,
4847 nds32_expand_cstore, nds32_expand_movcc,
4848 nds32_output_cbranchsi4_equality_zero,
4849 nds32_output_cbranchsi4_equality_reg,
4850 nds32_output_cbranchsi4_equality_reg_or_const_int,
4851 nds32_output_cbranchsi4_greater_less_zero: New functions.
4852 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
4853 nds32_expand_cstore, nds32_expand_movcc,
4854 nds32_output_cbranchsi4_equality_zero,
4855 nds32_output_cbranchsi4_equality_reg,
4856 nds32_output_cbranchsi4_equality_reg_or_const_int,
4857 nds32_output_cbranchsi4_greater_less_zero): Declare.
4858 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
4859 nds32_rimm11s_operand): New predicates.
4860 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
4861 * config/nds32/nds32.md: Rewrite all the branch and conditional move
4864 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4866 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
4867 * config/nds32/nds32.md: Ditto.
4868 * config/nds32/pipelines.md: Ditto.
4870 2018-04-04 Richard Biener <rguenther@suse.de>
4872 PR tree-optimization/85168
4873 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
4874 propagating abnormals.
4876 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4878 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
4880 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4881 Kito Cheng <kito.cheng@gmail.com>
4883 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
4884 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
4885 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
4886 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
4887 * config/nds32/nds32.md (sibcall_internal): New.
4888 (sibcall_register): Remove.
4889 (sibcall_immediate): Remove.
4890 (sibcall_value_internal): New.
4891 (sibcall_value_register): Remove.
4892 (sibcall_value_immediate): Remove.
4893 * config/nds32/predicates.md (nds32_general_register_operand): New.
4894 (nds32_call_address_operand): New.
4896 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4898 PR rtl-optimization/85167
4899 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
4900 bb_defs if *split_p, instead preinitialize it to NULL.
4902 PR tree-optimization/85156
4903 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
4904 evaluating the argument multiple times.
4906 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
4908 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
4910 (_mm_cvtpd_ps): Likewise.
4911 (_mm_cvttpd_epi32): Likewise.
4912 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
4913 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
4914 vector, pixel, and bool following altivec.h include.
4916 2018-04-03 Martin Sebor <msebor@redhat.com>
4918 * doc/extend.texi (Common Function Attributes): Clarify.
4919 (const attribute): Likewise.
4920 (pure attribute): Likewise.
4922 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4925 * config/i386/i386.c (ix86_expand_vector_set): Use
4926 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
4928 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
4930 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
4931 instructions when changing rounding bits to preserve precision bits
4932 in the x87 control word.
4934 2018-04-03 Martin Liska <mliska@suse.cz>
4936 PR tree-optimization/82491
4937 * rtl.h (strip_offset_and_add): Replace += suboffset with
4938 poly_uint64 () + suboffset.
4940 2018-03-29 Martin Liska <mliska@suse.cz>
4941 Martin Jambor <mjambor@suse.cz>
4944 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
4945 param_type is not an integral or pointer type.
4947 2018-04-03 Richard Biener <rguenther@suse.de>
4949 * sese.h (recompute_all_dominators): Remove.
4951 2018-04-02 Martin Sebor <msebor@redhat.com>
4953 * doc/invoke.texi (-Wrestrict): Fix typos.
4955 2018-04-02 Jim Wilson <jimw@sifive.com>
4957 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
4958 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
4959 (<optab>di3, <optab>si3_extend): Likewise.
4960 (<optab>si3_mask, <optab>si3_mask_1): New.
4961 (<optab>di3_mask, <optab>di3_mask_1): New.
4962 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
4963 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
4964 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
4966 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
4968 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
4971 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
4973 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
4974 (nds32_canonicalize_comparison): New function.
4976 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4977 Kito Cheng <kito.cheng@gmail.com>
4978 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4980 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
4981 * config/nds32/constants.md (unspec_volatile_element): Add
4982 UNSPEC_VOLATILE_RELAX_GROUP.
4983 * config/nds32/nds32-relax-opt.c: New file.
4984 * config/nds32/nds32-predicates.c
4985 (nds32_symbol_load_store_p): New function.
4986 * config/nds32/nds32-protos.h
4987 (nds32_symbol_load_store_p): Declare function.
4988 (make_pass_nds32_relax_opt): Declare new rtl pass function.
4989 * config/nds32/nds32.c
4990 (nds32_register_pass): New function to register pass.
4991 (nds32_register_passes): New function to register passes.
4992 * config/nds32/nds32.md (relax_group): New pattern.
4993 * config/nds32/nds32.opt (mrelax-hint): New option.
4994 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
4996 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
4998 * config/nds32/t-nds32: Modify files dependency.
5000 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5002 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
5003 (PROFILE_HOOK): Define its implementation.
5005 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5007 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
5008 type and 32-bit size.
5010 2018-04-01 Jakub Jelinek <jakub@redhat.com>
5013 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
5014 (V_128_256): New mode iterator.
5015 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
5016 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
5017 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
5019 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
5020 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
5022 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
5025 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
5026 NaN inputs correctly.
5028 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
5031 * config/rs6000/vsx.md (??r): New mode attribute.
5032 (*vsx_mov<mode>_64bit): Use it.
5033 (*vsx_mov<mode>_32bit): Likewise.
5035 2018-03-30 Martin Sebor <msebor@redhat.com>
5037 PR tree-optimization/84818
5038 * builtins.c (check_access): Use warning_n.
5040 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5043 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
5045 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
5048 2018-03-30 Julia Koval <julia.koval@intel.com>
5051 * x86-tune.def (movx, partial_reg_dependency): Enable for
5054 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
5057 * lra-constraints.c (process_alt_operands): Move setting
5058 this_alternative_matches below.
5060 2018-03-29 Martin Liska <mliska@suse.cz>
5063 * doc/invoke.texi: Document how LTO works with debug info.
5064 Describe auto-load support of binutils. Mention 'x86-64'
5065 as valid option value of -march option.
5067 2018-03-29 Jakub Jelinek <jakub@redhat.com>
5069 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
5072 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
5073 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
5074 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
5077 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5080 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
5081 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
5082 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
5083 for RS6000_BTM_POWERPC64.
5084 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
5085 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
5086 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
5091 2018-03-28 Carl Love <cel@us.ibm.com>
5094 2017-09-27 Carl Love <cel@us.ibm.com>
5096 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
5097 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
5098 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
5101 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5103 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
5104 instead of __vector bool.
5105 (_mm_max_pu8): Likewise.
5106 (_mm_min_pi16): Likewise.
5108 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5111 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
5112 (DIVWEUO): Likewise.
5114 (DIVDEUO): Likewise.
5115 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
5116 DIVWEUO and DIVDEUO.
5117 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
5118 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
5119 (div_extend): Likewise.
5120 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
5122 (__builtin_divweuo): Likewise.
5123 (__builtin_divdeo): Likewise.
5124 (__builtin_divdeuo): Likewise.
5126 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5129 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
5130 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
5132 PR tree-optimization/82004
5133 * gimple-match-head.c (optimize_pow_to_exp): New function.
5134 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
5135 Don't fold to exp if optimize_pow_to_exp is false.
5137 2018-03-28 Martin Liska <mliska@suse.cz>
5140 * calls.c (initialize_argument_information): Fix trailing space.
5141 * common.opt: Fix typo and provide better explanation for
5142 -fsanitize-coverage option.
5143 * config/i386/i386.opt: Fix typo.
5145 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5146 Martin Liska <mliska@suse.cz>
5149 * gimplify.c (asan_poison_variable): Don't do the check for
5150 gimplify_omp_ctxp here.
5151 (gimplify_decl_expr): Do it here.
5152 (gimplify_target_expr): Likewise.
5154 2018-03-28 Martin Liska <mliska@suse.cz>
5157 * config/i386/i386.c (ix86_function_arg_advance): Do not call
5158 chkp_type_bounds_count if MPX is not enabled.
5160 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
5162 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
5164 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
5167 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
5168 function to create the function decl for complex long double
5169 multiply and divide for -mabi=ieeelongdouble.
5170 (init_float128_ieee): Call it.
5172 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
5175 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
5176 -fcf-protection=branch -mibt.
5177 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
5179 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5182 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
5184 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
5187 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
5188 extern array declarations.
5190 2018-03-27 Richard Biener <rguenther@suse.de>
5193 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
5194 explicit single_use checks.
5196 2018-03-27 Richard Biener <rguenther@suse.de>
5198 PR tree-optimization/85082
5199 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
5202 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5204 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
5205 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
5206 Turn on fasynchronous-unwind-tables and funwind-tables.
5208 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
5211 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
5212 (*bmi_blsr_<mode>_ccz): Ditto.
5214 2018-03-26 Tom de Vries <tom@codesourcery.com>
5216 PR tree-optimization/85063
5217 * omp-general.c (offloading_function_p): New function. Factor out
5219 * omp-offload.c (pass_omp_target_link::gate): ... here.
5220 * omp-general.h (offloading_function_p): Declare.
5221 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
5222 with attribute omp declare target for offloading functions.
5224 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
5226 PR tree-optimization/84005
5227 * tree-data-ref.h (get_base_for_alignment): Declare.
5228 * tree-data-ref.c (get_base_for_alignment_1): New function.
5229 (get_base_for_alignment): Likewise.
5230 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
5231 get_base_for_alignment to find a suitable base object, instead
5232 of always using drb->base_address.
5234 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5237 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
5238 known size by default.
5240 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
5243 * lra-constraints.c (process_alt_operands): Don't match BLKmode
5244 and non BLKmode operands.
5246 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5249 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
5250 Clean up attributes.
5252 2018-03-23 Richard Biener <rguenther@suse.de>
5255 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
5256 we are going to emit early debug for LTO.
5258 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5261 * function.c (match_asm_constraints_1): Don't optimize if input
5262 doesn't satisfy general_operand predicate for output's mode.
5265 * alias.c (write_dependence_p): Don't require for x_canonicalized
5266 non-VOIDmode if x has VOIDmode.
5269 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
5270 just don't try to optimize it rather than assert it never happens.
5272 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5274 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
5275 macro expansions for definition of ST_INTERNAL_<mode> and
5276 LD_INTERNAL_<mode> builtins.
5277 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
5279 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
5281 (altivec_expand_st_builtin): Likewise.
5282 (altivec_expand_builtin): Remove calls to deleted functions.
5283 (rs6000_address_for_altivec): Delete this function.
5284 * config/rs6000/vector.md: Remove expands for
5285 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
5287 2018-03-22 Sudakshina Das <sudi.das@arm.com>
5290 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
5291 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
5292 re-computing once computed.
5293 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
5294 (arm_init_machine_status): Initialize
5295 machine->static_chain_stack_bytes.
5297 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5300 * doc/extend.texi: Add four new prototypes for vec_ld.
5301 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
5302 definitions for more logical presentation.
5303 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
5304 entries for V1TI variants of __builtin_altivec_ld builtin.
5305 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
5306 handling of V1TI variant of LVX icode pattern.
5307 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
5308 (rs6000_gimple_fold_builtin): Likewise.
5309 (altivec_init_builtins): Add code to define
5310 __builtin_altivec_lvx_v1ti function.
5312 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5315 * function.c (match_asm_constraints_1): Don't do the optimization
5316 if input isn't a REG, SUBREG, MEM or constant.
5318 2018-03-22 Tom de Vries <tom@codesourcery.com>
5320 PR tree-optimization/84956
5321 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
5322 bb_has_abnormal_pred.
5324 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5327 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
5328 DECL_INITIAL (decl) to decl at the end.
5329 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
5332 2018-03-21 Joseph Myers <joseph@codesourcery.com>
5334 * doc/extend.texi (__builtin_tgmath): Document when complex
5335 integer types are treated as _Complex _Float64.
5337 2018-03-21 Tom de Vries <tom@codesourcery.com>
5339 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
5341 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5343 PR tree-optimization/84960
5344 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
5345 if it is ENTRY block, move them into single succ of ENTRY in that case.
5347 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
5349 PR tree-optimization/84811
5350 * poly-int.h (poly_span_traits): Remove the T3 parameter and
5351 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
5352 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
5353 (known_subrange_p): Update accordingly. Cast each value involved
5354 in the size comparison, rather than casting the result of the
5357 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5359 PR tree-optimization/84982
5360 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
5361 by flipping the least significant bit rather than all bits from
5362 bitpos to bitpos + bitsize - 1.
5364 2018-03-21 Nathan Sidwell <nathan@acm.org>
5366 * doc/extend.texi (Deprecated Features): Remove mention of
5367 long-deleted deprecations.
5369 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5372 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
5373 * configure: Regenerate.
5375 2018-03-21 Tom de Vries <tom@codesourcery.com>
5377 PR tree-optimization/83126
5378 * tree-parloops.c (num_phis): New function.
5379 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
5381 2018-03-21 Nathan Sidwell <nathan@acm.org>
5383 * doc/extend.texi (Deprecated Features): Update deprecated flags,
5384 mention anon-struct/union members and trailing attributes.
5386 2018-03-21 Bin Cheng <bin.cheng@arm.com>
5388 PR tree-optimization/84969
5389 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
5390 builtin memset partitions if they set different rhs values.
5392 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5394 PR rtl-optimization/84989
5395 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
5396 VEC_DUPLICATE with scalar result mode.
5398 2018-03-21 Martin Liska <mliska@suse.cz>
5401 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
5402 not intended return statement.
5404 2018-03-21 Martin Liska <mliska@suse.cz>
5407 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
5408 (chkp_find_bound_slots_1): Limit number of iterations.
5410 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
5413 * Minor grammar fixes for x86 options.
5415 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5418 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
5419 holding REG_CFA_RESTORE notes, instead turn them into a USE.
5421 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
5424 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
5425 (altivec_lvx_<mode>_1op): Likewise.
5426 (altivec_stvx_<mode>_2op): Likewise.
5427 (altivec_stvx_<mode>_1op): Likewise.
5428 (altivec_lvx_<VM2:mode>): New define_expand.
5429 (altivec_stvx_<VM2:mode>): Likewise.
5430 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
5431 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5432 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
5433 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5434 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
5435 (rs6000_gen_lvx): Likewise.
5436 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
5437 (altivec_expand_stv_builtin): Likewise.
5438 (altivec_expand_builtin): Likewise.
5439 * config/rs6000/vector.md: Likewise.
5441 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5444 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
5447 2018-03-20 Richard Biener <rguenther@suse.de>
5450 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
5451 sign-conversions as zero, fall back to standard scalar_stmt
5454 2018-03-20 Martin Liska <mliska@suse.cz>
5457 * predict.c (rebuild_frequencies): Handle case when we have
5458 PROFILE_ABSENT, but flag_guess_branch_prob is false.
5460 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5463 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
5464 flag_section_anchors.
5465 * varasm.c (use_blocks_for_decl_p): Remove hack for
5466 dw2_force_const_mem.
5469 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
5471 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
5472 be created, use lowpart_subreg of operands[0] rather than operands[0]
5474 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
5475 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
5476 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
5477 and n constraint instead of aarch64_shift_imm_di and Usd.
5478 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
5479 (*aarch64_<optab>_reg_minus<mode>3): ... this.
5481 2018-03-20 Sudakshina Das <sudi.das@arm.com>
5484 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
5485 to favor GPR over NEON registers.
5486 (<shift>di3_neon): Likewise.
5488 2018-03-20 Tom de Vries <tom@codesourcery.com>
5491 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
5492 (nvptx_process_pars): Emit bar.sync asap and alap.
5494 2018-03-20 Tom de Vries <tom@codesourcery.com>
5497 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
5498 seen_label if seen_label is already set.
5500 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5503 * config/i386/i386.c (fold_builtin_cpu): For features above 31
5504 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
5505 Use 1U instead of 1. Formatting fixes.
5508 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
5509 instead of TREE_TYPE (s1) for the return value.
5511 2018-03-19 Jakub Jelinek <jakub@redhat.com>
5513 PR tree-optimization/84946
5514 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
5515 bitsize + bitsize in poly_uint64 rather than poly_int64.
5518 * dwarf2asm.c: Include fold-const.c.
5519 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
5520 of decl rather than decl itself.
5522 PR rtl-optimization/84643
5523 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
5525 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
5528 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
5529 calling assemble_variable.
5531 2018-03-19 Sudakshina Das <sudi.das@arm.com>
5534 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
5535 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
5537 2018-03-19 Jim Wilson <jimw@sifive.com>
5540 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
5541 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
5542 (riscv_first_stack_step): Likewise.
5543 (riscv_option_override): Use STACK_BOUNDARY instead of
5545 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
5547 (BIGGEST_ALIGNMENT): Set to 128.
5548 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
5549 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
5552 2018-03-19 Richard Biener <rguenther@suse.de>
5554 PR tree-optimization/84933
5555 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
5556 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
5558 2018-03-19 Richard Biener <rguenther@suse.de>
5560 PR tree-optimization/84859
5561 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
5562 (cond_if_else_store_replacement): Perform sinking operation on
5563 single-store BBs regardless of MAX_STORES_TO_SINK setting.
5564 Generalize what a BB with a single eligible store is.
5566 2018-03-19 Richard Biener <rguenther@suse.de>
5568 PR tree-optimization/84929
5569 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
5570 chrec_is_positive against non-chrec arg.
5572 2018-03-19 Tamar Christina <tamar.christina@arm.com>
5575 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
5577 2018-03-18 Martin Liska <mliska@suse.cz>
5579 PR rtl-optimization/84635
5580 * regrename.c (build_def_use): Use matches_mode only when
5583 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
5585 PR tree-optimization/84913
5586 * tree-vect-loop.c (vectorizable_reduction): Don't try to
5587 vectorize chains of COND_EXPRs.
5589 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5591 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
5593 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5595 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
5597 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5599 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
5601 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
5602 Kito Cheng <kito.cheng@gmail.com>
5604 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
5605 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
5606 (nds32_adjust_reg_alloc_order): New function.
5607 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
5609 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
5611 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
5612 nds32_print_operand, nds32_print_operand_address): Use
5613 HOST_WIDE_INT_PRINT_DEC instead.
5615 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
5617 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
5619 2018-03-17 Jakub Jelinek <jakub@redhat.com>
5622 * config/i386/i386.c (initial_ix86_tune_features,
5623 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
5625 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
5626 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
5627 rather than 1u << ix86_tune. Formatting fix.
5628 (ix86_option_override_internal): Change ix86_arch_mask from
5629 unsigned int to unsigned HOST_WIDE_INT, initialize to
5630 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
5631 (ix86_function_specific_restore): Likewise.
5633 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5636 * postreload.c (reload_combine_recognize_pattern): Perform
5637 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
5638 truncate_int_for_mode the result for the destination's mode.
5641 * hsa-gen.c (mem_type_for_type): Fix comment typo.
5642 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
5644 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5647 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
5650 * lra-assigns.c (lra_split_hard_reg_for): Don't use
5651 regno_allocno_class_array and sorted_pseudos.
5652 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
5653 insns where regno is used.
5655 2018-03-16 Martin Liska <mliska@suse.cz>
5658 * multiple_target.c (create_dispatcher_calls): Redirect
5659 reference in the symbol table.
5661 2018-03-16 Martin Liska <mliska@suse.cz>
5664 * multiple_target.c (create_dispatcher_calls): Redirect also
5667 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5671 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
5672 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
5673 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
5675 2018-03-16 Julia Koval <julia.koval@intel.com>
5677 * doc/invoke.texi (Skylake Server): Add CLWB.
5678 Cannonlake): Remove CLWB.
5680 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5682 PR tree-optimization/84841
5683 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
5685 (FLOAT_ONE_CONST_TYPE): Define.
5686 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
5687 (sort_by_operand_rank): Put entries with higher constant_type last
5688 rather than first to match comments.
5690 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
5692 * config/nios2/nios2.md (movsi_internal): Fix thinko in
5695 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5698 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
5699 check and use address of target always.
5701 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
5704 * config/i386/i386.c (indirect_thunk_needed): Update comments.
5705 (indirect_thunk_bnd_needed): Likewise.
5706 (indirect_thunks_used): Likewise.
5707 (indirect_thunks_bnd_used): Likewise.
5708 (indirect_return_needed): New.
5709 (indirect_return_bnd_needed): Likewise.
5710 (output_indirect_thunk_function): Add a bool argument for
5712 (output_indirect_thunk_function): Don't generate alias for
5713 function return thunk.
5714 (ix86_code_end): Call output_indirect_thunk_function to generate
5715 function return thunks.
5716 (ix86_output_function_return): Set indirect_return_bnd_needed
5717 and indirect_return_needed instead of indirect_thunk_bnd_needed
5718 and indirect_thunk_needed.
5720 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
5722 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
5723 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
5724 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
5726 2018-03-15 David Malcolm <dmalcolm@redhat.com>
5727 Paul Hua <paul.hua.gm@gmail.com>
5730 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
5732 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
5734 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
5735 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
5738 2018-03-15 Tamar Christina <tamar.christina@arm.com>
5741 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
5742 instead of GET_MODE_SIZE when comparing Units.
5744 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
5747 * varasm.c (hash_section): Return an unchangeble hash value
5748 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
5749 Return !aarch64_can_use_per_function_literal_pools_p ().
5751 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5754 * optabs.c (emit_conditional_move): Pass address of cmode's copy
5755 rather than address of cmode as last argument to prepare_cmp_insn.
5757 2018-03-15 Julia Koval <julia.koval@intel.com>
5759 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
5760 F_AVX512VNNI, F_AVX512BITALG): New.
5762 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
5765 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
5766 insn for floating-point loads and stores.
5768 2018-03-14 Carl Love <cel@us.ibm.com>
5770 * config/rs6000/rs6000-c.c: Add macro definitions for
5771 ALTIVEC_BUILTIN_VEC_PERMXOR.
5772 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
5773 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
5774 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
5775 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
5777 * config/doc/extend.texi: Add prototypes for vec_permxor.
5779 2018-03-14 David Malcolm <dmalcolm@redhat.com>
5782 * diagnostic-show-locus.c (class layout_point): Convert m_line
5783 from int to linenum_type.
5784 (line_span::comparator): Use linenum "compare" function when
5785 comparing line numbers.
5786 (test_line_span): New function.
5787 (layout_range::contains_point): Convert param "row" from int to
5789 (layout_range::intersects_line_p): Likewise.
5790 (layout::will_show_line_p): Likewise.
5791 (layout::print_source_line): Likewise.
5792 (layout::should_print_annotation_line_p): Likewise.
5793 (layout::print_annotation_line): Likewise.
5794 (layout::print_leading_fixits): Likewise.
5795 (layout::annotation_line_showed_range_p): Likewise.
5796 (struct line_corrections): Likewise for field m_row.
5797 (line_corrections::line_corrections): Likewise for param "row".
5798 (layout::print_trailing_fixits): Likewise.
5799 (layout::get_state_at_point): Likewise.
5800 (layout::get_x_bound_for_row): Likewise.
5801 (layout::print_line): Likewise.
5802 (diagnostic_show_locus): Likewise for locals "last_line" and
5804 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
5805 * input.c (selftest::test_linenum_comparisons): New function.
5806 (selftest::input_c_tests): Call it.
5807 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
5808 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
5809 * selftest.h (ASSERT_GT): New macro.
5810 (ASSERT_GT_AT): New macro.
5811 (ASSERT_LT): New macro.
5812 (ASSERT_LT_AT): New macro.
5814 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
5816 PR rtl-optimization/84780
5817 * combine.c (distribute_links): Don't make a link based on pc_rtx.
5819 2018-03-14 Martin Liska <mliska@suse.cz>
5821 * tree.c (record_node_allocation_statistics): Use
5822 get_stats_node_kind.
5823 (get_stats_node_kind): New function extracted from
5824 record_node_allocation_statistics.
5825 (free_node): Use get_stats_node_kind.
5827 2018-03-14 Richard Biener <rguenther@suse.de>
5829 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
5830 that the value-set of ANTIC_IN doesn't grow.
5833 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5835 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5836 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5838 2018-03-14 Julia Koval <julia.koval@intel.com>
5840 * config.gcc (icelake-client, icelake-server): New.
5842 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
5843 (initial_ix86_arch_features): Ditto.
5844 (PTA_SKYLAKE): Add SGX.
5845 (PTA_ICELAKE): Remove.
5846 (PTA_ICELAKE_CLIENT): New.
5847 (PTA_ICELAKE_SERVER): New.
5848 (ix86_option_override_internal): Split up icelake on icelake client and
5850 (get_builtin_code_for_version): Ditto.
5851 (fold_builtin_cpu): Ditto.
5852 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
5853 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
5854 * config/i386/i386.h (processor_type): Ditto.
5855 * doc/invoke.texi: Ditto.
5857 2018-03-14 Jakub Jelinek <jakub@redhat.com>
5860 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
5861 INTEGER_CST offset, add it together with bitpos / 8 and
5862 sign extend based on POINTER_SIZE.
5866 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
5869 * config/i386/constraints.md (Yc): New register constraint.
5870 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
5871 Use Yc constraint for alternative 2 of operand 0. Remove
5872 preferred_for_speed attribute.
5874 2018-03-14 Richard Biener <rguenther@suse.de>
5876 PR tree-optimization/84830
5877 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
5878 with the old one to avoid oscillations.
5880 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
5883 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
5885 (assign_by_spills): Return a flag of reload assignment failure.
5886 Do not process the reload assignment failures. Do not spill other
5887 reload pseudos if they has the same reg class. Update n if
5889 (lra_assign): Add a return arg. Set up from the result of
5890 assign_by_spills call.
5891 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5892 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5893 usage_insns if it is not NULL.
5894 (spill_hard_reg_in_range): New function.
5895 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5896 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5897 function prototypes.
5898 (lra_assign): Change prototype.
5899 * lra.c (lra): Add code to deal with fails by splitting hard reg
5902 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
5904 * config/riscv/riscv.opt (mrelax): New option.
5905 * config/riscv/riscv.c (riscv_file_start): Emit ".option
5906 "norelax" when riscv_mrelax is disabled.
5907 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
5909 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5912 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
5913 reassociation for int modes.
5915 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5917 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
5918 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
5920 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
5921 * config/aarch64/aarch64-sve.md
5922 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
5923 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
5924 (*extend<mode><Vwide>2): Rename to...
5925 (aarch64_sve_extend<mode><Vwide>2): ...this.
5926 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
5927 renaming the old pattern to...
5928 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
5930 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
5931 define_expand, renaming the old pattern to...
5932 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
5933 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
5934 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
5935 account when deciding which SVE instruction the optab should use.
5936 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
5938 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5940 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
5941 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
5942 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
5943 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
5944 (tlsdesc_small_<mode>): Turn a define_expand and use
5945 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
5946 (tlsdesc_small_advsimd_<mode>): ...this.
5947 (tlsdesc_small_sve_<mode>): New pattern.
5949 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5951 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
5952 (UNSPEC_UMUL_HIGHPART): New constants.
5953 (MUL_HIGHPART): New int iteraor.
5954 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
5955 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
5957 (*<su>mul<mode>3_highpart): New define_insn.
5959 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
5962 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
5965 2018-03-13 Martin Liska <mliska@suse.cz>
5968 * (sem_item_optimizer::sem_item_optimizer): Initialize new
5970 (sem_item_optimizer::~sem_item_optimizer): Release it.
5971 (sem_item_optimizer::merge_classes): Register variable aliases.
5972 (sem_item_optimizer::fixup_pt_set): New function.
5973 (sem_item_optimizer::fixup_points_to_sets): Likewise.
5974 * ipa-icf.h: Declare new variables and functions.
5976 2018-03-13 Jakub Jelinek <jakub@redhat.com>
5979 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
5980 integer_pow2p@2 and test integer_pow2p in condition.
5981 (A < 0 ? C : 0): Similarly for @1.
5984 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
5985 characters starting at p contain '\0' character, don't look beyond
5989 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
5990 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
5993 * reg-stack.c (change_stack): Change update_end var from int to
5994 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
5995 also call set_block_for_insn on the newly added insns and rescan.
5998 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
5999 on the last operand.
6002 * tree.c (stabilize_reference_1): Return save_expr (e) for
6003 STATEMENT_LIST even if it doesn't have side-effects.
6005 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
6007 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
6009 2018-03-12 Renlin Li <renlin.li@arm.com>
6011 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
6012 aarch64_output_scalar_simd_mov_immediate.
6014 2018-03-12 Martin Sebor <msebor@redhat.com>
6016 PR tree-optimization/83456
6017 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
6018 for perfectly overlapping calls to memcpy.
6019 (gimple_fold_builtin_memory_chk): Same.
6020 (gimple_fold_builtin_strcpy): Handle no-warning.
6021 (gimple_fold_builtin_stxcpy_chk): Same.
6022 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
6024 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6026 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
6027 parameter. Use it for SFmode.
6028 (rs6000_function_arg_advance_1): Adjust.
6029 (rs6000_function_arg): Adjust.
6030 (rs6000_gimplify_va_arg): Pass false for that new parameter.
6032 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6034 PR rtl-optimization/84169
6035 PR rtl-optimization/84780
6036 * combine.c (can_combine_p): Check for a 2-insn combination whether
6037 the destination register is used between the two insns, too.
6039 2018-03-12 Richard Biener <rguenther@suse.de>
6041 PR tree-optimization/84803
6042 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
6043 for refs DR analysis didn't process.
6045 2018-03-12 Richard Biener <rguenther@suse.de>
6047 PR tree-optimization/84777
6048 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
6049 force-vectorize loops ignore whether we are optimizing for size.
6051 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
6053 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
6054 (TARGET_MD_ASM_ADJUST): Define.
6056 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
6057 Kito Cheng <kito.cheng@gmail.com>
6058 Chung-Ju Wu <jasonwucj@gmail.com>
6060 * config/nds32/nds32.c (nds32_compute_stack_frame,
6061 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
6062 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
6063 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
6064 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
6065 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
6066 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
6067 * config/nds32/nds32.md (prologue, epilogue): Use macro
6068 NDS32_V3PUSH_AVAILABLE_P to do checking.
6070 2018-03-11 Jakub Jelinek <jakub@redhat.com>
6073 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
6074 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
6075 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
6076 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
6077 addition of most attributes on !orig_type_die or the attribute not
6078 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
6080 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6081 Chung-Ju Wu <jasonwucj@gmail.com>
6083 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
6085 * config/nds32/nds32.opt (mvh): New option.
6087 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6088 Chung-Ju Wu <jasonwucj@gmail.com>
6090 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
6092 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
6093 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
6096 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6097 Chung-Ju Wu <jasonwucj@gmail.com>
6099 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
6101 * config/nds32/nds32-multiple.md (strlensi): New pattern.
6102 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
6104 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
6105 Kito Cheng <kito.cheng@gmail.com>
6106 Chung-Ju Wu <jasonwucj@gmail.com>
6108 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
6109 UNSPEC_FFMISM and UNSPEC_FLMISM.
6110 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
6111 for ffb, ffmism and flmism.
6112 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
6113 (unspec_ffmism): Ditto.
6114 (unspec_flmism): Ditto.
6115 (nds32_expand_builtin_impl): Check if string extension is available.
6116 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
6117 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
6119 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
6122 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6125 * lra-assigns.c (assign_by_spills): Return a flag of reload
6126 assignment failure. Do not process the reload assignment
6127 failures. Do not spill other reload pseudos if they has the same
6129 (lra_assign): Add a return arg. Set up from the result of
6130 assign_by_spills call.
6131 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6132 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6133 usage_insns if it is not NULL.
6134 (spill_hard_reg_in_range): New function.
6135 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6136 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6137 function prototypes.
6138 (lra_assign): Change prototype.
6139 * lra.c (lra): Add code to deal with fails by splitting hard reg
6142 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
6145 * config/i386/i386.opt: Replace Enforcment with Enforcement.
6147 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
6150 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
6151 (dw_val_node): Add val_symbolic_view.
6152 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
6153 (symview_upper_bound): New.
6154 (new_line_info_table): Initialize symviews_since_reset.
6155 (dwarf2out_source_line): Count symviews_since_reset and set
6156 symview_upper_bound.
6157 (dw_val_equal_p): Handle symview.
6158 (add_AT_symview): New.
6159 (print_dw_val): Handle symview.
6160 (attr_checksum, attr_checksum_ordered): Likewise.
6161 (same_dw_val_p, size_of_die): Likewise.
6162 (value_format, output_die): Likewise.
6163 (add_high_low_attributes): Use add_AT_symview for entry_view.
6164 (dwarf2out_finish): Reset symview_upper_bound, clear
6167 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
6170 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
6171 Add strict argument and use it.
6172 (rs6000_split_multireg_move): Update for new strict argument.
6173 (mem_operand_gpr): Disallow all non-offsettable addresses.
6174 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
6176 2018-03-09 Jakub Jelinek <jakub@redhat.com>
6179 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
6180 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
6181 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
6184 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
6185 decl, use remap_type if we want to use the type.
6187 2018-03-09 Martin Sebor <msebor@redhat.com>
6189 PR tree-optimization/84526
6190 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6192 (builtin_access::generic_overlap): Be prepared to handle non-array
6195 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
6197 PR rtl-optimization/84682
6198 * lra-constraints.c (process_address_1): Check is_address flag
6199 for address constraints.
6200 (process_alt_operands): Likewise.
6201 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
6202 preprocess_constraints.
6203 * recog.h (preprocess_constraints): Add oploc parameter.
6205 * recog.c (preprocess_constraints): Test address_operand for
6206 CT_ADDRESS constraints.
6208 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6211 * lra-assigns.c (assign_by_spills): Return a flag of reload
6212 assignment failure. Do not process the reload assignment
6213 failures. Do not spill other reload pseudos if they has the same
6215 (lra_assign): Add a return arg. Set up from the result of
6216 assign_by_spills call.
6217 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6218 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6219 usage_insns if it is not NULL.
6220 (spill_hard_reg_in_range): New function.
6221 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6222 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6223 function prototypes.
6224 (lra_assign): Change prototype.
6225 * lra.c (lra): Add code to deal with fails by splitting hard reg
6228 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6231 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
6232 Accept complain bool parameter. Only emit errors if it is true.
6233 (arm_parse_cpu_option_name): Likewise.
6234 (arm_target_thumb_only): Adjust callers of the above.
6235 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
6236 prototype to take a default true bool parameter.
6237 (arm_parse_arch_option_name): Likewise.
6239 2018-03-09 David Malcolm <dmalcolm@redhat.com>
6240 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
6244 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
6245 * configure: Regenerate.
6246 * configure.ac ("linker --version-script option"): New.
6247 ("linker soname option"): New.
6249 2018-03-09 Richard Biener <rguenther@suse.de>
6251 PR tree-optimization/84775
6252 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
6253 immediate uses of predicate stmts and mark them modified.
6256 PR tree-optimization/84178
6257 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6259 (version_loop_for_if_conversion): Delay update_ssa call.
6260 (tree_if_conversion): Delay update_ssa until after predicate
6263 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
6266 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
6267 when the function accesses prior frames.
6269 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6272 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
6273 gen_llsym, otherwise call maybe_gen_llsym.
6276 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
6277 has ',' character inside of it.
6279 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6282 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
6283 as clobbering CC_REGNUM.
6285 2018-03-08 Richard Biener <rguenther@suse.de>
6288 * tree-scalar-evolution.c: Include tree-into-ssa.h.
6289 (follow_copies_to_constant): Do not follow SSA names registered
6292 2018-03-08 Richard Biener <rguenther@suse.de>
6294 PR tree-optimization/84178
6295 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6297 (version_loop_for_if_conversion): Delay update_ssa call.
6298 (tree_if_conversion): Delay update_ssa until after predicate
6301 2018-03-08 David Malcolm <dmalcolm@redhat.com>
6303 PR tree-optimization/84178
6304 * tree-if-conv.c (release_bb_predicate): Remove the
6305 the assertion that the stmts have NULL use_ops.
6306 Discard the statements, asserting that they haven't
6307 yet been added to a BB.
6309 2018-03-08 Richard Biener <rguenther@suse.de>
6311 PR tree-optimization/84746
6312 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
6313 (phi_translate): Pass in destination ANTIC_OUT set.
6314 (phi_translate_1): Likewise. For a simplified result lookup
6315 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
6316 (phi_translate_set): Adjust.
6317 (do_pre_regular_insertion): Likewise.
6318 (do_pre_partial_partial_insertion): Likewise.
6320 2018-03-08 Martin Liska <mliska@suse.cz>
6322 PR gcov-profile/84735
6323 * doc/gcov.texi: Document usage of profile files.
6324 * gcov-io.h: Document changes in the format.
6326 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
6330 * dwarf2out.c (struct dw_line_info_table): Update comments for
6332 (FORCE_RESET_NEXT_VIEW): New.
6333 (FORCE_RESETTING_VIEW_P): New.
6334 (RESETTING_VIEW_P): Check for -1 too.
6335 (ZERO_VIEW_P): Likewise.
6336 (new_line_info_table): Force-reset next view.
6337 (dwarf2out_begin_function): Likewise.
6338 (dwarf2out_source_line): Simplify zero_view_p initialization.
6339 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
6340 view directly. Omit view when omitting .loc at line 0.
6342 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6344 PR tree-optimization/84740
6345 * tree-switch-conversion.c (process_switch): Call build_constructors
6346 only if info.phi_count is non-zero.
6348 PR tree-optimization/84739
6349 * tree-tailcall.c (find_tail_calls): Check call arguments against
6350 DECL_ARGUMENTS (current_function_decl) rather than
6351 DECL_ARGUMENTS (func) when checking for tail recursion.
6353 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6355 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
6356 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
6357 Volker Reichelt's entry and add entries for people that perform
6358 GCC fuzzy testing and report numerous bugs.
6360 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
6363 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
6364 readonly data in sdata, if that is disabled.
6365 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
6366 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
6367 -mreadonly-in-sdata option.
6369 2018-03-07 Martin Sebor <msebor@redhat.com>
6371 PR tree-optimization/84468
6372 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
6373 basic block when looking for nul assignment.
6375 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
6378 * except.h (output_function_exception_table): Adjust prototype.
6379 * except.c (output_function_exception_table): Remove FNNAME parameter
6380 and add SECTION parameter. Ouput one part of the table at a time.
6381 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
6382 the first part of the exception table and emit unwind directives.
6383 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
6384 (i386_pe_seh_cold_init): Likewise.
6385 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
6386 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
6387 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
6388 (ix86_output_call_insn): Emit a nop in one more case for SEH.
6389 * config/i386/winnt.c: Include except.h.
6390 (struct seh_frame_state): Add reg_offset, after_prologue and
6391 in_cold_section fields.
6392 (i386_pe_seh_end_prologue): Set seh->after_prologue.
6393 (i386_pe_seh_cold_init): New function.
6394 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
6395 to seh->in_cold_section.
6396 (seh_emit_push): Record the offset of the push.
6397 (seh_emit_save): Record the offet of the save.
6398 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
6399 Test seh->after_prologue to disregard the epilogue.
6400 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
6401 (i386_pe_end_cold_function): New function.
6403 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6406 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
6407 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
6410 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
6412 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
6413 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
6416 * multiple_target.c: Include tree-inline.h and intl.h.
6417 (expand_target_clones): Diagnose and fail if node->definition and
6418 !tree_versionable_function_p (node->decl).
6420 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
6422 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
6424 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
6425 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
6426 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
6428 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6431 * combine.c (try_combine): Use reg_or_subregno instead of handling
6432 just paradoxical SUBREGs and REGs.
6434 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
6436 * config/arc/arc.c (arc_finalize_pic): Remove function.
6437 (arc_must_save_register): We use single base PIC register, remove
6438 checks to save/restore the PIC register.
6439 (arc_expand_prologue): Likewise.
6440 * config/arc/arc-protos.h (arc_set_default_type_attributes):
6442 (arc_verify_short): Likewise.
6443 (arc_attr_type): Likewise.
6444 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
6445 (walk_stores): Likewise.
6446 (arc_address_cost): Make it static.
6447 (arc_verify_short): Likewise.
6448 (branch_dest): Likewise.
6449 (arc_attr_type): Likewise.
6450 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
6451 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
6452 (arc_final_prescan_insn): Remove inserting the nops due to
6453 hardware hazards. It is done in reorg step.
6454 (insn_length_variant_t): Remove.
6455 (insn_length_parameters_t): Likewise.
6456 (arc_insn_length_parameters): Likewise.
6457 (arc_get_insn_variants): Likewise.
6458 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
6460 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6463 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
6466 PR tree-optimization/84687
6467 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
6469 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
6471 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6473 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
6474 Rename to ppc_speculation_barrier.
6475 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
6476 __builtin_ppc_speculation_barrier.
6478 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6481 * combine.c (combine_simplify_rtx): Don't try to simplify if
6482 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
6485 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
6487 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
6488 to 32 bytes when compiling for POWER9.
6490 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6493 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
6494 regparm >= 3 with no arg reg available also for calls with
6495 flag_force_indirect_call. Pass decl to ix86_function_regparm.
6498 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
6500 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
6502 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
6505 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
6507 2018-03-05 Richard Biener <rguenther@suse.de>
6509 PR tree-optimization/84486
6510 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
6511 When inserting a __builtin_assume_aligned call set the LHS
6512 SSA name alignment info accordingly.
6514 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
6516 PR tree-optimization/84114
6517 * config/aarch64/aarch64.c (aarch64_reassociation_width)
6518 Avoid reassociation of FLOAT_MODE addition.
6520 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
6522 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
6523 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
6524 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
6525 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
6526 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
6527 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
6528 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
6530 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
6531 __builtin_ia32_wbinvd): New builtins.
6532 (SPECIAL_ARGS2): New.
6533 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
6534 (SPECIAL_ARGS2): New.
6535 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
6536 (ix86_valid_target_attribute_inner_p): Ditto.
6537 (ix86_init_mmx_sse_builtins): Add special_args2.
6538 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
6539 TARGET_WBNOINVD_P): New.
6540 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
6541 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
6542 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
6543 * config/i386/immintrin.h (_wbinvd): New intrinsic.
6544 * config/i386/pconfigintrin.h: New file.
6545 * config/i386/wbnoinvdintrin.h: Ditto.
6546 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
6547 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
6549 2018-03-05 Richard Biener <rguenther@suse.de>
6551 PR tree-optimization/84670
6552 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6554 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6555 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6556 (compute_antic_aux): Only assert the number of values in ANTIC_IN
6557 doesn't grow if all successors (recursively) were visited at least
6560 2018-03-05 Richard Biener <rguenther@suse.de>
6562 PR tree-optimization/84650
6563 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
6564 if executed in the loop pipeline.
6566 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
6568 * doc/configfiles.texi (Configuration Files): Move info about
6569 conditionalizing $target-protos.h to...
6570 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
6571 differs from $target-protos.h.
6573 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
6574 Chung-Ju Wu <jasonwucj@gmail.com>
6576 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
6577 * config/nds32/nds32-multiple.md (setmemsi): Define.
6578 * config/nds32/nds32-memory-manipulation.c
6579 (nds32_gen_dup_4_byte_to_word_value): New.
6580 (emit_setmem_word_loop): New.
6581 (emit_setmem_byte_loop): New.
6582 (nds32_expand_setmem_loop): New.
6583 (nds32_expand_setmem_loop_v3m): New.
6584 (nds32_expand_setmem_unroll): New.
6585 (nds32_expand_setmem): New.
6587 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6588 Chung-Ju Wu <jasonwucj@gmail.com>
6590 * config/nds32/nds32-memory-manipulation.c
6591 (nds32_emit_load_store): New.
6592 (nds32_emit_post_inc_load_store): New.
6593 (nds32_emit_mem_move): New.
6594 (nds32_emit_mem_move_block): New.
6595 (nds32_expand_movmemsi_loop_unknown_size): New.
6596 (nds32_expand_movmemsi_loop_known_size): New.
6597 (nds32_expand_movmemsi_loop): New.
6598 (nds32_expand_movmemsi_unroll): New.
6599 (nds32_expand_movmemqi): Rename ...
6600 (nds32_expand_movmemsi): ... to this.
6601 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
6602 (movmemsi): ... to this.
6603 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
6604 (nds32_expand_movmemsi): ... to this.
6606 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6607 Monk Chiang <sh.chiang04@gmail.com>
6608 Chung-Ju Wu <jasonwucj@gmail.com>
6610 * config/nds32/nds32-protos.h
6611 (nds32_expand_load_multiple): New arguments.
6612 (nds32_expand_store_multiple): Ditto.
6613 (nds32_valid_multiple_load_store): Rename ...
6614 (nds32_valid_multiple_load_store_p): ... to this.
6615 * config/nds32/nds32-memory-manipulation.c
6616 (nds32_expand_load_multiple): Refine implementation.
6617 (nds32_expand_store_multiple): Ditto.
6618 * config/nds32/nds32-multiple.md
6619 (load_multiple): Update nds32_expand_load_multiple interface.
6620 (store_multiple): Update nds32_expand_store_multiple interface.
6621 * config/nds32/nds32-predicates.c
6622 (nds32_valid_multiple_load_store): Rename ...
6623 (nds32_valid_multiple_load_store_p): ... to this and refine
6625 * config/nds32/predicates.md
6626 (nds32_load_multiple_and_update_address_operation): New predicate.
6627 (nds32_store_multiple_and_update_address_operation): New predicate.
6629 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6630 Chung-Ju Wu <jasonwucj@gmail.com>
6632 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
6633 (combo): New attribute.
6634 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
6636 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
6638 * config/nds32/nds32.opt: Change -mcmodel= default value.
6640 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
6641 Monk Chiang <sh.chiang04@gmail.com>
6642 Chung-Ju Wu <jasonwucj@gmail.com>
6644 * config/nds32/constants.md (unspec_element): New enum.
6645 * config/nds32/constraints.md (Umw): New constraint.
6646 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
6647 * config/nds32/nds32-intrinsic.md: Likewise.
6648 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
6649 (nds32_valid_smw_lwm_base_p): New.
6650 (nds32_output_smw_single_word): New.
6651 (nds32_output_lmw_single_word): New.
6652 (nds32_expand_unaligned_load): New.
6653 (nds32_expand_unaligned_store): New.
6654 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
6655 (nds32_output_smw_single_word): Declare.
6656 (nds32_output_lmw_single_word): Declare.
6657 (nds32_expand_unaligned_load): Declare.
6658 (nds32_expand_unaligned_store): Declare.
6659 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
6660 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
6661 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
6662 NDS32_BUILTIN_UASTORE_DW.
6663 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
6666 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
6667 Kito Cheng <kito.cheng@gmail.com>
6668 Chung-Ju Wu <jasonwucj@gmail.com>
6670 * config/nds32/nds32-intrinsic.c
6671 (nds32_expand_builtin_null_ftype_reg): Delete.
6672 (nds32_expand_builtin_reg_ftype_imm): Ditto.
6673 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
6674 (nds32_read_argument): New.
6675 (nds32_legitimize_target): Ditto.
6676 (nds32_legitimize_argument): Ditto.
6677 (nds32_check_constant_argument): Ditto.
6678 (nds32_expand_unop_builtin): Ditto.
6679 (nds32_expand_unopimm_builtin): Ditto.
6680 (nds32_expand_binop_builtin): Ditto.
6681 (nds32_builtin_decl_impl): Ditto.
6682 (builtin_description): Ditto.
6683 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
6684 (nds32_init_builtins_impl): Ditto.
6685 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
6686 (nds32_builtin_decl): New.
6687 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
6688 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
6690 2018-03-02 Jeff Law <law@redhat.com>
6692 * reorg.c (stop_search_p): Handle DEBUG_INSN.
6693 (redundant_insn, fill_simple_delay_slots): Likewise.
6694 (fill_slots_from_thread): Likewise.
6695 * resource.c (mark_referenced_resources): Likewise.
6696 (mark_set_resources, find_dead_or_set_registers): Likewise.
6698 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6700 * substring-locations.h (format_warning_va): Formatting fix for
6702 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
6704 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
6705 * substring-locations.c: Include intl.h.
6706 (format_warning_va): Turned into small wrapper around
6707 format_warning_n_va, renamed to ...
6708 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
6709 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
6711 (format_warning_at_substring_n): New function.
6712 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
6713 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
6714 format_warning_at_substring with just a shorter name instead of
6715 const function pointer.
6716 (fmtwarn_n): New function.
6717 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
6718 appropriate, get rid of all the fmtstr temporaries, move conditionals
6719 with G_() wrapped string literals directly into fmtwarn arguments,
6720 cast dir.len to (int), formatting fixes.
6722 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
6724 * doc/invoke.texi: Remove "Cilk Plus" references.
6726 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6727 Richard Biener <rguenther@suse.de>
6730 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
6731 for error or warning attributes if CALL_FROM_THUNK_P is set.
6734 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6737 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
6738 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
6741 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
6742 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
6744 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
6745 instead of -1U in last predictors element's probability member.
6747 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
6750 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
6751 arguments if they are comparable.
6753 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6755 PR tree-optimization/84634
6756 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
6757 masks and masked_loop_p with a single loop_masks, making sure it's
6758 null for bb vectorization.
6760 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6762 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
6763 (vect_analyze_data_ref_access): Use loop->safe_len rather than
6764 loop->force_vectorize to check whether there is no alias.
6766 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6769 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
6771 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
6773 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
6774 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
6775 instead of a loop around prev_real_insn.
6776 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
6780 * config/i386/i386.c (ix86_print_operand): Use conditional
6781 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
6784 2018-03-02 Richard Biener <rguenther@suse.de>
6786 PR tree-optimization/84427
6787 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
6788 (bitmap_set_subtract_values): Rewrite to handle multiple
6791 (prune_clobbered_mems): Likewise.
6792 (phi_translate): Take edge instead of pred/phiblock.
6793 (phi_translate_1): Likewise.
6794 (phi_translate_set): Likewise. Insert all translated
6795 exprs for a value into the set, keeping possibly multiple
6796 expressions per value.
6797 (compute_antic_aux): Adjust for phi_translate changes.
6798 When intersecting union the expressions and prune those
6799 not in the final value set, keeping possibly multiple
6800 expressions per value. Do not use value-insertion
6801 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
6802 all expressions. Add verification that the value-sets
6803 only shrink during iteration.
6804 (compute_partial_antic_aux): Adjust for the phi_translate changes.
6805 (do_pre_regular_insertion): Likewise.
6806 (do_pre_partial_partial_insertion): Likewise.
6808 2018-03-02 Richard Biener <rguenther@suse.de>
6811 * config/darwin.c (saved_debug_info_level): New static global.
6812 (darwin_asm_lto_start): Disable debug info generation for LTO out.
6813 (darwin_asm_lto_end): Restore debug info generation settings.
6815 2018-03-01 Martin Liska <mliska@suse.cz>
6818 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
6821 2018-03-01 Richard Biener <rguenther@suse.de>
6824 * dwarf2out.c (gen_variable_die): Properly handle late VLA
6825 type annotation with LTO when debug was disabled at compile-time.
6827 2018-03-01 Matthew Fortune <mfortune@gmail.com>
6829 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
6831 (mips_final_postscan_insn): Likewise.
6833 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
6835 PR rtl-optimization/84528
6836 * alias.c (init_alias_target): Add commentary.
6837 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
6838 a unique base value if the frame pointer is not eliminated
6839 to the stack pointer.
6841 2018-03-01 Tom de Vries <tom@codesourcery.com>
6843 PR rtl-optimization/83327
6844 * lra-int.h (hard_regs_spilled_into): Declare.
6845 * lra.c (hard_regs_spilled_into): Define.
6846 (init_reg_info): Init hard_regs_spilled_into.
6847 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
6848 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
6849 (process_bb_lives): Handle hard_regs_spilled_into.
6850 (lra_create_live_ranges_1): Before doing liveness propagation, clear
6851 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
6853 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
6855 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
6856 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
6857 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
6858 * config/rs6000/aix72.h: New file.
6860 2018-02-28 Jakub Jelinek <jakub@redhat.com>
6862 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
6863 instead of warning_at with conditional singular and plural messages
6867 * stor-layout.c (update_alignment_for_field): For
6868 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
6869 && !DECL_PACKED (field), do the alignment update, just use
6870 only desired_align instead of MAX (type_align, desired_align)
6872 (place_field): Don't do known_align < desired_align handling
6873 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
6874 is non-NULL, instead do it after rli->prev_field handling and
6875 only if not within a bitfield word. For DECL_PACKED (field)
6876 use type_align of BITS_PER_UNIT.
6878 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
6880 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
6881 superfluous parentheses and trailing spaces.
6883 2018-02-28 Richard Biener <rguenther@suse.de>
6885 PR tree-optimization/84584
6886 * graphite-scop-detection.c (scop_detection::add_scop): Discard
6887 SCoPs with fake exit edge.
6889 2018-02-28 Martin Liska <mliska@suse.cz>
6892 * timevar.c (timer::print): Fix format to properly print 100%
6895 2018-02-28 Richard Biener <rguenther@suse.de>
6898 * genmatch.c (capture_info::walk_match): Do not mark
6899 captured expressions without operands as expr_p given
6900 they act more like predicates and should be subject to
6901 "lost tail" side-effect preserving.
6903 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
6905 PR rtl-optimization/81611
6906 * auto-inc-dec.c (attempt_change): Move dead note from
6907 mem_insn if it's the next use of regno
6908 (find_address): Take address use of reg holding
6909 non-incremented value. Add parm to limit search to the named
6911 (merge_in_block): Attempt to use a mem insn that is the next
6912 use of the original regno.
6914 2018-02-27 Martin Sebor <msebor@redhat.com>
6917 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
6918 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
6920 2018-02-27 Martin Sebor <msebor@redhat.com>
6922 PR translation/84207
6923 * diagnostic-core.h (warning_n, error_n, inform_n): Change
6924 n argument to unsigned HOST_WIDE_INT.
6925 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
6926 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
6927 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
6928 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
6930 2018-02-27 Richard Biener <rguenther@suse.de>
6932 PR tree-optimization/84512
6933 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
6934 Do not use the estimate returned from record_stmt_cost for
6935 the scalar iteration cost but sum properly using add_stmt_cost.
6937 2018-02-27 Richard Biener <rguenther@suse.de>
6939 PR tree-optimization/84466
6940 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6941 Adjust last change to less strictly validate use operands.
6943 2018-02-27 Martin Liska <mliska@suse.cz>
6945 PR gcov-profile/84548
6946 * gcov.c (process_file): Allow partial overlap and consider it
6947 also as group functions.
6948 (output_lines): Properly calculate range of lines for a group.
6950 2018-02-27 Martin Liska <mliska@suse.cz>
6952 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
6953 'ggc' suffixes. Change first column width.
6954 (timer::print): Fix formatting of the column.
6956 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
6958 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
6959 preserve inline entry blocks for the sake of debug inline
6960 entry point markers alone.
6961 (remove_unused_locals): Suggest in comments a better place to
6962 force the preservation of inline entry blocks that are
6963 otherwise unused, but do not preserve them.
6965 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6967 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
6969 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6972 * config/i386/constraints.md (Bs): Replace
6973 ix86_indirect_branch_register with
6974 TARGET_INDIRECT_BRANCH_REGISTER.
6976 * config/i386/i386.md (indirect_jump): Likewise.
6977 (tablejump): Likewise.
6978 (*sibcall_memory): Likewise.
6979 (*sibcall_value_memory): Likewise.
6980 Peepholes of indirect call and jump via memory: Likewise.
6981 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
6982 (*sibcall_value_GOT_32): Likewise.
6983 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6984 (GOT_memory_operand): Likewise.
6985 (call_insn_operand): Likewise.
6986 (sibcall_insn_operand): Likewise.
6987 (GOT32_symbol_operand): Likewise.
6988 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
6990 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
6992 PR rtl-optimization/83496
6993 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
6994 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
6995 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
6996 redundant insn, if any.
6997 (relax_delay_slots): Likewise.
6998 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
7000 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
7002 PR tree-optimization/83965
7003 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
7004 that grouped statements are part of a reduction chain. Return
7005 true if the statement is not marked as a reduction itself but
7007 (vect_recog_dot_prod_pattern): Don't check whether the statement
7008 is part of a group here.
7009 (vect_recog_sad_pattern): Likewise.
7010 (vect_recog_widen_sum_pattern): Likewise.
7012 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7015 * final.c (rest_of_clean_state): Also look for calls inside sequences.
7017 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7020 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
7022 (ix86_output_indirect_function_return): New prototype.
7023 (ix86_split_simple_return_pop_internal): Likewise.
7024 * config/i386/i386.c (indirect_return_via_cx): New.
7025 (indirect_return_via_cx_bnd): Likewise.
7026 (indirect_thunk_name): Handle return va CX_REG.
7027 (output_indirect_thunk_function): Create alias for
7028 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
7029 (ix86_output_indirect_jmp): Remove the bool argument.
7030 (ix86_output_indirect_function_return): New function.
7031 (ix86_split_simple_return_pop_internal): Likewise.
7032 * config/i386/i386.md (*indirect_jump): Don't pass false
7033 to ix86_output_indirect_jmp.
7034 (*tablejump_1): Likewise.
7035 (simple_return_pop_internal): Change it to define_insn_and_split.
7036 Call ix86_split_simple_return_pop_internal to split it for
7038 (simple_return_indirect_internal): Call
7039 ix86_output_indirect_function_return instead of
7040 ix86_output_indirect_jmp.
7042 2018-02-26 Jakub Jelinek <jakub@redhat.com>
7045 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
7046 memset and value initialization afterwards.
7048 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
7050 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
7052 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7055 * common/config/aarch64/aarch64-common.c
7056 (aarch_option_optimization_table[]): Switch
7057 off fomit-frame-pointer
7059 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7060 Chung-Ju Wu <jasonwucj@gmail.com>
7062 * config/nds32/nds32-multiple.md (load_multiple): Disallow
7064 (store_multiple): Ditto.
7066 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7068 * config.gcc: Add --with-cpu support for nds32 target.
7069 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
7070 * config/nds32/nds32.opt: Add -mcpu= option.
7072 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
7074 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
7075 isel=yes): Warn for these deprecated options.
7077 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
7079 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
7080 ISA_2_5_MASKS_EMBEDDED.
7082 2018-02-23 Jakub Jelinek <jakub@redhat.com>
7084 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
7085 p->max as pointers rather than using iterative_hash_expr.
7087 2018-02-23 Carl Love <cel@us.ibm.com>
7089 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
7090 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
7092 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
7093 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
7094 P8V_BUILTIN_VEC_VUNSIGNED2.
7096 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
7099 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
7100 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
7102 * lra-constraints.c (curr_insn_transform): Set up
7103 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
7105 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
7106 * lra-eliminations.c (spill_pseudos): Ditto.
7107 (process_insn_for_elimination): Ditto.
7108 * lra-lives.c (reg_early_clobber_p): Use the new macros.
7109 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
7110 LRA_NON_CLOBBERED_ALT.
7112 2018-02-22 Martin Sebor <msebor@redhat.com>
7114 PR tree-optimization/84480
7115 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
7116 to maybe_diag_stxncpy_trunc. Call it.
7117 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
7118 from gimple_fold_builtin_strcpy. Print inlining stack.
7119 (handle_builtin_stxncpy): Print inlining stack.
7120 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
7122 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
7125 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
7126 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
7127 and -fcheck-pointer-bounds are used together.
7128 (indirect_thunk_prefix): New enum.
7129 (indirect_thunk_need_prefix): New function.
7130 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
7131 "_nt" instead of "_bnd" for NOTRACK prefix.
7132 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
7133 (output_indirect_thunk_function): Likewise.
7135 (ix86_code_end): Update output_indirect_thunk_function calls.
7136 (ix86_output_indirect_branch_via_reg): Replace
7137 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
7138 (ix86_output_indirect_branch_via_push): Likewise.
7139 (ix86_output_function_return): Likewise.
7140 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
7141 incompatible with -fcf-protection=branch and
7142 -fcheck-pointer-bounds.
7144 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7147 * config/aarch64/aarch64.c (aarch64_print_address_internal):
7148 Change gcc_assert call to output_operand_lossage.
7150 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7152 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
7154 2018-02-22 DJ Delorie <dj@redhat.com>
7155 Sebastian Perta <sebastian.perta@renesas.com>
7156 Oleg Endo <olegendo@gcc.gnu.org>
7158 * config/rx/rx.c (rx_rtx_costs): New function.
7159 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
7161 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
7163 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
7165 2018-02-22 Martin Liska <mliska@suse.cz>
7168 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
7169 Add "native" as a possible value.
7171 2018-02-22 Martin Liska <mliska@suse.cz>
7174 * config/i386/i386.c (ix86_option_override_internal):
7175 Add "native" as a possible value for -march and -mtune.
7177 2018-02-22 Jakub Jelinek <jakub@redhat.com>
7180 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
7181 to all type variants.
7183 PR tree-optimization/84503
7184 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
7185 width as info->bitpos + info->bitsize - start.
7186 (merged_store_group::merge_overlapping): Simplify width computation.
7187 (check_no_overlap): New function.
7188 (imm_store_chain_info::try_coalesce_bswap): Compute expected
7189 start + width and last_order of the group, fail if check_no_overlap
7191 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
7192 to group if check_no_overlap fails.
7194 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7196 * config/rs6000/altivec.md: Delete contraint arguments to
7197 define_expand, define_split, and define_peephole2, and in
7198 define_insn_and_split if always unused.
7199 * config/rs6000/darwin.md: Ditto.
7200 * config/rs6000/dfp.md: Ditto.
7201 * config/rs6000/rs6000.md: Ditto.
7202 * config/rs6000/sync.md: Ditto.
7203 * config/rs6000/vector.md: Ditto.
7204 * config/rs6000/vsx.md: Ditto.
7206 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7208 * config/rs6000/altivec.md: Write output control strings as braced
7209 blocks instead of double-quoted strings.
7210 * config/rs6000/darwin.md: Ditto.
7211 * config/rs6000/rs6000.md: Ditto.
7212 * config/rs6000/vector.md: Ditto.
7213 * config/rs6000/vsx.md: Ditto.
7215 2018-02-21 Jason Merrill <jason@redhat.com>
7217 PR c++/84314 - ICE with templates and fastcall attribute.
7218 * attribs.c (build_type_attribute_qual_variant): Remove assert.
7220 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7222 * ipa-cp.c (determine_versionability): Fix comment typos.
7224 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7227 * ipa-cp.c (determine_versionability): Do not version functions caling
7230 2018-02-21 Martin Liska <mliska@suse.cz>
7233 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
7234 Add "native" as a possible value.
7235 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
7236 the macro when native cpu detection is available.
7238 2018-02-21 Martin Liska <mliska@suse.cz>
7241 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
7242 Add "native" as a possible value.
7243 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
7244 when native cpu detection is available.
7246 2018-02-21 Jakub Jelinek <jakub@redhat.com>
7247 Martin Sebor <msebor@redhat.com>
7249 PR tree-optimization/84478
7250 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
7252 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
7253 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
7254 support which is conservatively correct, for 2 only stay conservative
7255 for maxlen. Formatting and comment capitalization fixes. Add STRICT
7256 argument to the 2 argument get_range_strlen, adjust 6 arg
7257 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
7259 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
7260 (gimple_fold_builtin_strlen): Pass true as last argument to
7263 2018-02-20 Martin Sebor <msebor@redhat.com>
7266 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
7267 (builtin_memref::set_base_and_offset): Same. Handle inner references.
7268 (builtin_memref::builtin_memref): Factor out parts into
7269 set_base_and_offset and call it.
7271 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7274 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
7275 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
7276 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
7277 search at the associated MODE_INT.
7279 2018-02-20 Jeff Law <law@redhat.com>
7282 PR tree-optimization/81592
7284 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
7285 for range data rather than using global data.
7286 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
7287 range data rather than using global data.
7288 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
7289 pass it to children as needed.
7290 (struct directive::fmtresult): Similarly.
7291 (struct directive::set_width): Similarly.
7292 (struct directive::set_precision): Similarly.
7293 (format_integer, format_directive, parse_directive): Similarly.
7294 (format_none): Accept unnamed vr_values parameter.
7295 (format_percent, format_floating, format_character): Similarly.
7296 (format_string, format_plain): Similarly.
7297 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
7298 the EVRP range analyzer for range data rather than using global data.
7299 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
7300 gimple-ssa-evrp-analyze.h
7301 (class sprintf_dom_walker): Add after_dom_children member function.
7302 Add evrp_range_analyzer member.
7303 (sprintf_dom_walker::before_dom_children): Call into the EVRP
7304 range analyzer as needed.
7305 (sprintf_dom_walker::after_dom_children): New member function.
7306 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
7308 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
7309 (evrp_range_analyzer::pop_to_marker): Likewise.
7311 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7313 PR tree-optimization/84419
7314 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
7315 with the required type if its current type is compatible but
7318 2018-02-20 Jakub Jelinek <jakub@redhat.com>
7321 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
7322 after vectorization.
7324 2018-02-20 Martin Liska <mliska@suse.cz>
7327 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
7328 possible values if we don't have a hint.
7330 2018-02-20 Martin Liska <mliska@suse.cz>
7334 * final.c (shorten_branches): Build align_tab array with one
7336 * opts.c (finish_options): Add alignment option limit check.
7337 (MAX_CODE_ALIGN): Likewise.
7338 (MAX_CODE_ALIGN_VALUE): Likewise.
7339 * doc/invoke.texi: Document maximum allowed option value for
7340 all -falign-* options.
7342 2018-02-19 Jakub Jelinek <jakub@redhat.com>
7345 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
7346 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
7347 * var-tracking.c (emit_note_insn_var_location): Remove all references
7348 to NOTE_INSN_CALL_ARG_LOCATION.
7349 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
7350 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
7351 Use copy_rtx_if_shared.
7352 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
7353 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
7354 (dwarf2out_var_location): Remove handling of
7355 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
7357 * final.c (final_scan_insn): Remove all references to
7358 NOTE_INSN_CALL_ARG_LOCATION.
7359 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
7360 before dumping final insns.
7361 * except.c (emit_note_eh_region_end): Remove all references to
7362 NOTE_INSN_CALL_ARG_LOCATION.
7363 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
7364 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
7365 * config/arc/arc.c (hwloop_optimize): Likewise.
7366 * config/arm/arm.c (create_fix_barrier): Likewise.
7367 * config/s390/s390.c (s390_chunkify_start): Likewise.
7368 * config/sh/sh.c (find_barrier): Likewise.
7369 * config/i386/i386.c (rest_of_insert_endbranch,
7370 ix86_seh_fixup_eh_fallthru): Likewise.
7371 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
7372 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
7373 * config/frv/frv.c (frv_function_prologue): Likewise.
7374 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
7376 (note_outside_basic_block_p): Remove all references to
7377 NOTE_INSN_CALL_ARG_LOCATION.
7378 * gengtype.c (adjust_field_rtx_def): Likewise.
7379 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
7381 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
7382 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
7385 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
7388 PR tree-optimization/84452
7389 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
7390 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
7393 2018-02-19 Martin Liska <mliska@suse.cz>
7396 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
7398 2018-02-19 Martin Liska <mliska@suse.cz>
7399 Richard Sandiford <richard.sandiford@linaro.org>
7401 PR tree-optimization/82491
7402 * gimple-fold.c (get_base_constructor): Make earlier bail out
7405 2018-02-19 Carl Love <cel@us.ibm.com>
7407 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
7408 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
7410 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
7411 P8V_BUILTIN_VEC_NEG.
7413 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
7415 * config/rl78/rl78.md (movdf): New define expand.
7417 2018-02-19 Martin Liska <mliska@suse.cz>
7420 * doc/invoke.texi: Fix typo.
7421 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
7423 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
7425 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
7426 handle rs6000_single_float and rs6000_double_float specially for
7429 2018-02-16 Jeff Law <law@redhat.com>
7431 * config/rx/rx.c (add_pop_cfi_notes): New function.;
7434 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7437 * ipa-inline.c (inline_small_functions): Fix a typo.
7439 2018-02-16 Nathan Sidwell <nathan@acm.org>
7441 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
7443 2018-02-16 Carl Love <cel@us.ibm.com>
7445 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
7446 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
7447 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
7448 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
7449 expansion to P8V_BUILTIN_VEC_FLOAT2.
7451 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
7453 PR rtl-optimization/70023
7454 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
7455 src_regno into account.
7457 2018-02-16 Carl Love <cel@us.ibm.com>
7459 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
7460 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
7461 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
7462 * config/rs6000/rs6000.c: Remove case statements for
7463 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
7464 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
7465 and P9V_BUILTIN_VEC_VINSERT4B.
7466 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
7467 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
7468 * config/rs6000/vsx.md:
7469 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
7472 2018-02-16 Carl Love <cel@us.ibm.com>
7474 * config/rs6000/altivec.h: Add builtin names vec_extract4b
7476 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
7478 * config/rs6000/rs6000-c.c: Add the definitions for
7479 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
7480 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
7481 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
7482 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
7483 definition for insert4b and define insn *insert3b_internal.
7484 * doc/extend.texi: Add documentation for vec_extract4b.
7486 2018-02-16 Nathan Sidwell <nathan@acm.org>
7488 * doc/extend.texi (Backwards Compatibility): Mention friend
7489 injection. Note for-scope is deprecated.
7490 * doc/invoke.texi (-ffriend-injection): Deprecate.
7492 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
7494 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
7495 that moved to I2, also allow destinations that are a paradoxical
7496 subreg (instead of a normal reg).
7498 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
7501 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
7504 2018-02-16 Richard Biener <rguenther@suse.de>
7506 PR tree-optimization/84037
7507 PR tree-optimization/84016
7509 * config/i386/i386.c (ix86_builtin_vectorization_cost):
7510 Adjust vec_construct for the fact we need additional higher latency
7511 128bit inserts for AVX256 and AVX512 vector builds.
7512 (ix86_add_stmt_cost): Scale vector construction cost for
7515 2018-02-16 Richard Biener <rguenther@suse.de>
7517 PR tree-optimization/84417
7518 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
7519 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
7520 (non_rewritable_lvalue_p): Likewise, use poly-ints.
7522 2018-02-16 Martin Liska <mliska@suse.cz>
7525 * internal-fn.def (ASAN_CHECK): Set proper flags.
7526 (ASAN_MARK): Likewise.
7528 2018-02-16 Julia Koval <julia.koval@intel.com>
7530 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
7531 from PTA_CANNONLAKE.
7533 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7536 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
7537 Use ++iter rather than iter++ for std::list iterators.
7538 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
7539 defer deleting them until all nodes in the forest are processed. Do
7540 free even leaf nodes. Change to_process into auto_vec.
7543 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
7544 * vec.h (vec_default_construct): Use memset instead of placement new
7545 if BROKEN_VALUE_INITIALIZATION is defined.
7546 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
7547 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
7550 PR rtl-optimization/83723
7551 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
7552 * lra.c (lra_substitute_pseudo): Likewise. If true, use
7553 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
7555 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
7557 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
7559 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
7561 PR rtl-optimization/81443
7562 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
7563 from inner REGs to paradoxical SUBREGs.
7565 2018-02-16 Richard Biener <rguenther@suse.de>
7567 PR tree-optimization/84399
7568 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7569 For operands we can analyze at their definition make sure we can
7570 analyze them at each use as well.
7572 2018-02-16 Richard Biener <rguenther@suse.de>
7574 PR tree-optimization/84190
7575 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
7576 volatile accesses if the decl isn't volatile.
7578 2018-02-15 Jason Merrill <jason@redhat.com>
7580 PR c++/84314 - ICE with templates and fastcall attribute.
7581 * attribs.c (build_type_attribute_qual_variant): Don't clobber
7582 TYPE_CANONICAL on an existing type.
7584 2018-02-15 Jakub Jelinek <jakub@redhat.com>
7586 PR tree-optimization/84383
7587 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
7588 dstoff nor call operand_equal_p if dstbase is NULL.
7590 PR tree-optimization/84334
7591 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
7592 also a CONSTANT_CLASS_P, punt.
7594 2018-02-14 Jim Wilson <jimw@sifive.com>
7596 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
7597 first SMALL_OPERAND check. New local min_second_step. Move assert
7598 to where locals are set. Add TARGET_RVC support.
7599 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
7601 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
7603 * doc/invoke.texi: Correct -Wformat-overflow code sample.
7605 2018-02-14 Martin Sebor <msebor@redhat.com>
7607 PR tree-optimization/83698
7608 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
7609 arrays constrain the offset range to their bounds.
7610 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
7611 (builtin_access::overlap): Avoid setting the size of overlap if it's
7613 (maybe_diag_overlap): Also consider arrays when deciding what values
7614 of offsets to include in diagnostics.
7616 2018-02-14 Martin Sebor <msebor@redhat.com>
7619 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
7620 that correspond to the kind of a declaration.
7622 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
7625 * config/pa/pa.md: Load address of PIC label using the linkage table
7626 if the label is nonlocal.
7628 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
7630 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
7631 warning message if user requests -maltivec=be.
7632 * doc/invoke.texi: Document deprecation of -maltivec=be.
7634 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
7637 * config/rs6000/rs6000-c.c: Update definitions for
7638 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
7639 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
7641 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7644 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
7645 add _get_ssp intrinsics. Remove argument from
7646 __builtin_ia32_rdssp[d|q].
7647 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
7648 * config/i386/i386-builtin.def: Remove argument from
7649 __builtin_ia32_rdssp[d|q].
7650 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
7651 ix86_expand_special_args_builtin for _rdssp[d|q].
7652 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
7653 Clear register before usage.
7654 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
7655 Add documentation for new _get_ssp and _inc_ssp intrinsics.
7657 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
7659 PR tree-optimization/84357
7660 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
7661 operand 1 of an ARRAY_REF too.
7663 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
7666 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
7667 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7669 (set_of_reg): New struct.
7670 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
7671 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
7672 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7674 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
7675 Split into bitclr, bitset, bitinvert patterns if appropriate.
7676 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
7677 use rx_fuse_in_memory_bitop.
7678 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
7679 to named insn, correct maximum insn length.
7681 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
7684 * machmode.def: Define a complex mode for PARTIAL_INT.
7685 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
7687 * doc/rtl.texi: Document CSPImode.
7688 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
7690 (msp430_hard_regno_nregs_with_padding): Likewise.
7692 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
7695 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
7697 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
7699 PR rtl-optimization/84169
7700 * combine.c (try_combine): New variable split_i2i3. Set it to true if
7701 we generated a parallel as new i3 and we split that to new i2 and i3
7702 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
7703 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
7704 those to i2, not i1. Partially rewrite this scan code.
7706 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7709 * stor-layout.c (place_field): For variable length fields, adjust
7710 offset_align afterwards not just based on the field's alignment,
7711 but also on the size.
7714 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
7715 of exps and logs in the use_exp2 case.
7717 2018-02-13 Jeff Law <law@redhat.com>
7719 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
7722 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
7725 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
7729 * common.opt (gas-loc-support, gas-locview-support): New.
7730 (ginline-points, ginternal-reset-location-views): New.
7731 * doc/invoke.texi: Document them. Use @itemx where intended.
7732 (gvariable-location-views): Adjust.
7733 * target.def (reset_location_view): New.
7734 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
7735 (TARGET_RESET_LOCATION_VIEW): New.
7736 * doc/tm.texi: Rebuilt.
7737 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
7738 (dwarf2out_default_as_locview_support): New.
7739 (output_asm_line_debug_info): Use option variables.
7740 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
7741 (output_loc_list): Likewise.
7742 (add_high_low_attributes): Check option variables.
7743 Don't output entry view attribute in strict mode.
7744 (gen_inlined_subroutine_die): Check option variables.
7745 (dwarf2out_inline_entry): Likewise.
7746 (init_sections_and_labels): Likewise.
7747 (dwarf2out_early_finish): Likewise.
7748 (maybe_reset_location_view): New, from...
7749 (dwarf2out_var_location): ... here. Call it.
7750 * debug.h (dwarf2out_default_as_loc_support): Declare.
7751 (dwarf2out_default_as_locview_support): Declare.
7752 * hooks.c (hook_int_rtx_insn_0): New.
7753 * hooks.h (hook_int_rtx_insn_0): Declare.
7754 * toplev.c (process_options): Take -gas-loc-support and
7755 -gas-locview-support from dwarf2out. Enable
7756 -gvariable-location-views by default only with locview
7757 assembler support. Enable -ginternal-reset-location-views by
7758 default only if the target defines the corresponding hook.
7759 Enable -ginline-points by default if location views are
7760 enabled; force it disabled if statement frontiers are
7762 * tree-inline.c (expand_call_inline): Check option variables.
7763 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
7765 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
7767 PR tree-optimization/84321
7768 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
7769 handling. Also check whether the anti-range contains any values
7770 that satisfy the mask; switch to a VR_RANGE if not.
7772 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
7775 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
7777 2018-02-13 Martin Jambor <mjambor@suse.cz>
7780 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
7781 of call statements, also set location of a load to a temporary.
7783 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
7785 * config/rl78/rl78.c (add_vector_labels): New function.
7786 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
7787 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
7788 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
7789 which checks that no arguments are passed.
7790 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
7791 * doc/extend.texi: Documentation for the new attribute.
7793 2018-02-13 Andreas Schwab <schwab@suse.de>
7795 * config/riscv/linux.h (CPP_SPEC): Define.
7797 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7800 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
7801 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
7802 OPTION_MASK_ISA_AES as first argument to def_builtin_const
7803 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
7804 instead of OPTION_MASK_ISA_PCLMUL as first argument to
7805 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
7806 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
7807 temporarily for AES and PCLMUL builtins.
7809 PR tree-optimization/84339
7810 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
7811 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
7815 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
7816 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
7817 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
7819 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
7821 * omp-simd-clone.h: New file.
7822 * omp-simd-clone.c: Include omp-simd-clone.h.
7823 (expand_simd_clones): No longer static.
7824 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
7825 cgraph.h and omp-simd-clone.h.
7826 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
7827 (vect_recog_widen_shift_pattern): Formatting fix.
7828 (vect_pattern_recog_1): Don't check optab for calls.
7831 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
7832 operands[2] into a REG before using gen_lowpart on it.
7834 2018-02-12 Jeff Law <law@redhat.com>
7837 * config/sh/sh.c (find_barrier): Consider a sibling call
7840 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
7841 successfully back substituting a reg.
7843 2018-02-12 Richard Biener <rguenther@suse.de>
7845 PR tree-optimization/84037
7846 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
7847 parameter, move visited init to caller.
7848 (vect_slp_analyze_operations): Separate cost from validity
7849 check, initialize visited once for all instances.
7850 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
7852 * tree-vect-stmts.c (vect_model_simple_cost): Make early
7854 (vect_model_promotion_demotion_cost): Likewise.
7855 (vectorizable_bswap): Guard cost modeling with !slp_node
7856 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
7858 (vectorizable_call): Likewise.
7859 (vectorizable_conversion): Likewise.
7860 (vectorizable_assignment): Likewise.
7861 (vectorizable_shift): Likewise.
7862 (vectorizable_operation): Likewise.
7863 (vectorizable_store): Likewise.
7864 (vectorizable_load): Likewise.
7865 (vectorizable_condition): Likewise.
7866 (vectorizable_comparison): Likewise.
7868 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
7871 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
7872 (ASAN_MARK): Fix fnspec to account for return value, change pointer
7873 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
7875 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7878 * params.def (inline-min-speedup): Increase from 8 to 15.
7879 (max-inline-insns-auto): Decrease from 40 to 30.
7880 * ipa-split.c (consider_split): Add some buffer for function to
7881 be considered inlining candidate.
7882 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
7885 2018-02-12 Richard Biener <rguenther@suse.de>
7887 PR tree-optimization/84037
7888 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
7889 matched stmts if we cannot swap the non-matched ones.
7891 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7893 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
7894 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
7895 _mm_maskz_scalef_round_ss): New intrinsics.
7896 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
7897 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
7898 __builtin_ia32_scalefss_round): Remove.
7899 (__builtin_ia32_scalefsd_mask_round,
7900 __builtin_ia32_scalefss_mask_round): New intrinsics.
7901 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
7902 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
7903 ((match_operand:VF_128 2 "<round_nimm_predicate>"
7904 "<round_constraint>")): Changed to ...
7905 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
7906 "<round_scalar_constraint>")): ... this.
7907 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
7908 %0, %1, %2<round_op3>}"): Changed to ...
7909 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
7910 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
7911 %2<round_scalar_mask_op3>}"): ... this.
7912 * config/i386/subst.md (round_scalar_nimm_predicate): New.
7914 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7916 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
7917 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
7918 (_mm_maskz_sqrt_round_ss): New intrinsics.
7919 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
7920 (__builtin_ia32_sqrtsd_mask_round)
7921 (__builtin_ia32_sqrtss_mask_round): New builtins.
7922 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
7923 (__builtin_ia32_sqrtss_round): Remove.
7924 (__builtin_ia32_sqrtsd_mask_round)
7925 (__builtin_ia32_sqrtss_mask_round): New builtins.
7926 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
7927 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
7928 ((match_operand:VF_128 1 "vector_operand"
7929 "xBm,<round_constraint>")): Changed to ...
7930 ((match_operand:VF_128 1 "vector_operand"
7931 "xBm,<round_scalar_constraint>")): ... this.
7932 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
7933 %0, %2, %<iptr>1<round_op3>}): Changed to ...
7934 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
7935 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
7936 %<iptr>1<round_scalar_mask_op3>}): ... this.
7937 ((set_attr "prefix" "<round_prefix>")): Changed to ...
7938 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
7940 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
7943 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
7944 Cast vec_cmpeq result to correct type.
7945 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
7946 Cast vec_cmpgt result to correct type.
7948 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
7950 * final.c (final_scan_insn_1): Renamed from...
7951 (final_scan_insn): ... this. New wrapper, to recover
7952 seen from the outermost call in recursive ones.
7953 * config/sparc/sparc.c (output_return): Drop seen from call.
7954 (output_sibcall): Likewise.
7955 * config/visium/visium.c (output_branch): Likewise.
7957 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
7959 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
7962 2018-02-10 Alan Modra <amodra@gmail.com>
7965 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
7966 Specify LR as an input.
7968 2018-02-10 Jakub Jelinek <jakub@redhat.com>
7971 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
7972 remove_member_access_dummy_vars): New functions.
7973 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
7974 lower_omp_1, execute_lower_omp): Use them.
7976 PR rtl-optimization/84308
7977 * shrink-wrap.c (spread_components): Release todo vector.
7979 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
7981 PR rtl-optimization/57193
7982 * ira-color.c (struct allocno_color_data): Add member
7983 conflict_allocno_hard_prefs.
7984 (update_conflict_allocno_hard_prefs): New.
7985 (bucket_allocno_compare_func): Add a preference based on
7986 conflict_allocno_hard_prefs.
7987 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
7988 (color_allocnos): Remove a dead code. Initiate
7989 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
7991 2018-02-09 Jakub Jelinek <jakub@redhat.com>
7994 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
7995 constraint from =wa to wa. Avoid a subreg on the output operand,
7996 instead use a pseudo and subreg it in a move.
7997 (p9_xxbrd_<mode>): Changed to ...
7998 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
7999 (p9_xxbrd_v2df): New expander.
8000 (p9_xxbrw_<mode>): Changed to ...
8001 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
8002 (p9_xxbrw_v4sf): New expander.
8004 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8006 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
8008 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
8011 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
8012 multiply in 32-bit mode.
8013 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
8014 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
8017 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8019 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
8020 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
8021 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
8022 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
8024 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
8027 * dwarf2out.c (is_trivial_indirect_ref): New function.
8028 (dwarf2out_late_global_decl): Do not generate a location
8029 attribute for variables that have a non-trivial DECL_VALUE_EXPR
8030 and that are not defined in the current unit.
8032 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8034 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
8035 instead of a libcall for UNORDERED.
8037 2018-02-09 Tamar Christina <tamar.christina@arm.com>
8040 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
8041 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
8043 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8046 * config/s390/s390.c (s390_set_current_function): Invoke
8047 s390_indirect_branch_settings also if fndecl didn't change.
8049 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8051 * config/rs6000/rs6000.md (blockage): Set length to zero.
8053 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8055 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
8057 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8060 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
8061 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
8065 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
8066 PARALLEL incoming that failed vt_get_decl_and_offset check.
8069 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
8070 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
8072 (get_variable_section): For decls in named .bss* sections pass true as
8073 second argument to bss_initializer_p.
8075 2018-02-09 Marek Polacek <polacek@redhat.com>
8076 Jakub Jelinek <jakub@redhat.com>
8079 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
8080 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
8081 Sync some changes from cxx_fold_indirect_ref.
8083 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8085 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
8087 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
8088 (BLOCK_INLINE_ENTRY_LABEL): New.
8089 (dwarf2out_var_location): Disregard inline entry markers.
8090 (inline_entry_data): New struct.
8091 (inline_entry_data_hasher): New hashtable type.
8092 (inline_entry_data_hasher::hash): New.
8093 (inline_entry_data_hasher::equal): New.
8094 (inline_entry_data_table): New variable.
8095 (add_high_low_attributes): Add DW_AT_entry_pc and
8096 DW_AT_GNU_entry_view attributes if a pending entry is found
8097 in inline_entry_data_table. Add old entry_pc attribute only
8098 if debug nonbinding markers are disabled.
8099 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
8100 markers are enabled.
8101 (block_within_block_p, dwarf2out_inline_entry): New.
8102 (dwarf2out_finish): Check that no entries remained in
8103 inline_entry_data_table.
8104 * final.c (reemit_insn_block_notes): Handle inline entry notes.
8105 (final_scan_insn, notice_source_line): Likewise.
8106 (rest_of_clean_state): Skip inline entry markers.
8107 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
8109 * gimple.c (gimple_build_debug_inline_entry): New.
8110 * gimple.h (enum gimple_debug_subcode): Add
8111 GIMPLE_DEBUG_INLINE_ENTRY.
8112 (gimple_build_debug_inline_entry): Declare.
8113 (gimple_debug_inline_entry_p): New.
8114 (gimple_debug_nonbind_marker_p): Adjust.
8115 * insn-notes.def (INLINE_ENTRY): New.
8116 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
8117 inline entry marker notes.
8118 (print_insn): Likewise.
8119 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
8120 (INSN_DEBUG_MARKER_KIND): Likewise.
8121 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
8122 * tree-inline.c (expand_call_inline): Build and insert
8123 debug_inline_entry stmt.
8124 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
8125 inline entry blocks early, if nonbind markers are enabled.
8126 (dump_scope_block): Dump fragment info.
8127 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
8128 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
8129 (gimple_build_debug_inline_entry): New.
8130 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
8131 Enable/disable inline entry points too.
8132 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
8133 (DEBUG_INSN): Describe inline entry markers.
8135 * common.opt (gvariable-location-views): New.
8136 (gvariable-location-views=incompat5): New.
8137 * config.in: Rebuilt.
8138 * configure: Rebuilt.
8139 * configure.ac: Test assembler for view support.
8140 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
8141 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
8142 * dwarf2out.c (var_loc_view): New typedef.
8143 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
8144 (dwarf2out_locviews_in_attribute): New.
8145 (dwarf2out_locviews_in_loclist): New.
8146 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
8147 (enum dw_line_info_opcode): Add LI_adv_address.
8148 (struct dw_line_info_table): Add view.
8149 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
8150 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
8151 (zero_view_p): New variable.
8152 (ZERO_VIEW_P): New macro.
8153 (output_asm_line_debug_info): New.
8154 (struct var_loc_node): Add view.
8155 (add_AT_view_list, AT_loc_list): New.
8156 (add_var_loc_to_decl): Add view param. Test it against last.
8157 (new_loc_list): Add view params. Record them.
8158 (AT_loc_list_ptr): Handle loc and view lists.
8159 (view_list_to_loc_list_val_node): New.
8160 (print_dw_val): Handle dw_val_class_view_list.
8161 (size_of_die): Likewise.
8162 (value_format): Likewise.
8163 (loc_list_has_views): New.
8164 (gen_llsym): Set vl_symbol too.
8165 (maybe_gen_llsym, skip_loc_list_entry): New.
8166 (dwarf2out_maybe_output_loclist_view_pair): New.
8167 (output_loc_list): Output view list or entries too.
8168 (output_view_list_offset): New.
8169 (output_die): Handle dw_val_class_view_list.
8170 (output_dwarf_version): New.
8171 (output_compilation_unit_header): Use it.
8172 (output_skeleton_debug_sections): Likewise.
8173 (output_rnglists, output_line_info): Likewise.
8174 (output_pubnames, output_aranges): Update version comments.
8175 (output_one_line_info_table): Output view numbers in asm comments.
8176 (dw_loc_list): Determine current endview, pass it to new_loc_list.
8177 Call maybe_gen_llsym.
8178 (loc_list_from_tree_1): Adjust.
8179 (add_AT_location_description): Create view list attribute if
8180 needed, check it's absent otherwise.
8181 (convert_cfa_to_fb_loc_list): Adjust.
8182 (maybe_emit_file): Call output_asm_line_debug_info for test.
8183 (dwarf2out_var_location): Reset views as needed. Precompute
8184 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
8185 attribute. Set view.
8186 (new_line_info_table): Reset next view.
8187 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
8188 (dwarf2out_source_line): Likewise. Output view resets and labels to
8189 the assembler, or select appropriate line info opcodes.
8190 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
8191 (optimize_string_length): Catch it. Adjust.
8192 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
8193 dw_val_class_view_list, and remove it if no longer needed.
8194 (hash_loc_list): Hash view numbers.
8195 (loc_list_hasher::equal): Compare them.
8196 (optimize_location_lists): Check whether a view list symbol is
8197 needed, and whether the locview attribute is present, and
8198 whether they match. Remove the locview attribute if no longer
8200 (index_location_lists): Call skip_loc_list_entry for test.
8201 (dwarf2out_finish): Call output_asm_line_debug_info for test.
8202 Use output_dwarf_version.
8203 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
8204 (struct dw_val_node): Add val_view_list.
8205 * final.c (SEEN_NEXT_VIEW): New.
8206 (set_next_view_needed): New.
8207 (clear_next_view_needed): New.
8208 (maybe_output_next_view): New.
8209 (final_start_function): Rename to...
8210 (final_start_function_1): ... this. Take pointer to FIRST,
8211 add SEEN parameter. Emit param bindings in the initial view.
8212 (final_start_function): Reintroduce SEEN-less interface.
8213 (final): Rename to...
8214 (final_1): ... this. Take SEEN parameter. Output final pending
8215 next view at the end.
8216 (final): Reintroduce seen-less interface.
8217 (final_scan_insn): Output pending next view before switching
8218 sections or ending a block. Mark the next view as needed when
8219 outputting variable locations. Notify debug backend of section
8220 changes, and of location view changes.
8221 (rest_of_handle_final): Adjust.
8222 * toplev.c (process_options): Autodetect value for debug variable
8223 location views option. Warn on incompat5 without -gdwarf-5.
8224 * doc/invoke.texi (gvariable-location-views): New.
8225 (gvariable-location-views=incompat5): New.
8226 (gno-variable-location-views): New.
8228 2018-02-08 David Malcolm <dmalcolm@redhat.com>
8230 PR tree-optimization/84136
8231 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
8232 that the result of find_edge is non-NULL.
8234 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8237 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
8238 storing integer register in SImode. Fix cost of 256 and 512
8239 byte aligned SSE register store.
8241 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8243 * config/i386/i386.c (ix86_multiplication_cost): Fix
8244 multiplication cost for TARGET_AVX512DQ.
8246 2018-02-08 Marek Polacek <polacek@redhat.com>
8248 PR tree-optimization/84238
8249 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
8252 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8254 PR tree-optimization/84265
8255 * tree-vect-stmts.c (vectorizable_store): Don't treat
8256 VMAT_CONTIGUOUS accesses as grouped.
8257 (vectorizable_load): Likewise.
8259 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8261 PR tree-optimization/81635
8262 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
8263 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
8264 (test_round_for_mask): New functions.
8265 (wide_int_cc_tests): Call test_round_for_mask.
8266 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
8267 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
8268 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
8269 range returned by get_range_info.
8271 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8274 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
8275 * symtab.c: Include builtins.h
8276 (symtab_node::output_to_lto_symbol_table_p): Move here
8277 from lto-streamer-out.c:output_symbol_p.
8278 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
8279 (output_symbol_p): Move all logic to symtab.c
8280 (produce_symtab): Update.
8282 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8284 * config/s390/s390-opts.h (enum indirect_branch): Define.
8285 * config/s390/s390-protos.h (s390_return_addr_from_memory)
8286 (s390_indirect_branch_via_thunk)
8287 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
8288 (enum s390_indirect_branch_type): Define.
8289 * config/s390/s390.c (struct s390_frame_layout, struct
8290 machine_function): Remove.
8291 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
8292 (indirect_branch_table_label_no, indirect_branch_table_name):
8294 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
8295 (enum s390_indirect_branch_option): Define.
8296 (s390_return_addr_from_memory): New function.
8297 (s390_handle_string_attribute): New function.
8298 (s390_attribute_table): Add new attribute handler.
8299 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
8300 (s390_indirect_branch_via_thunk): New function.
8301 (s390_indirect_branch_via_inline_thunk): New function.
8302 (s390_function_ok_for_sibcall): When jumping via thunk disallow
8303 sibling call optimization for non z10 compiles.
8304 (s390_emit_call): Force indirect branch target to be a single
8305 register. Add r1 clobber for non-z10 compiles.
8306 (s390_emit_epilogue): Emit return jump via return_use expander.
8307 (s390_reorg): Handle JUMP_INSNs as execute targets.
8308 (s390_option_override_internal): Perform validity checks for the
8309 new command line options.
8310 (s390_indirect_branch_attrvalue): New function.
8311 (s390_indirect_branch_settings): New function.
8312 (s390_set_current_function): Invoke s390_indirect_branch_settings.
8313 (s390_output_indirect_thunk_function): New function.
8314 (s390_code_end): Implement target hook.
8315 (s390_case_values_threshold): Implement target hook.
8316 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
8318 * config/s390/s390.h (struct s390_frame_layout)
8319 (struct machine_function): Move here from s390.c.
8320 (TARGET_INDIRECT_BRANCH_NOBP_RET)
8321 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
8322 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
8323 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
8324 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
8325 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
8326 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
8327 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
8328 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
8329 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
8330 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
8331 (mnemonic attribute): Add values which aren't recognized
8333 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
8334 pattern for branch conversion. Fix mnemonic attribute.
8335 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
8336 indirect branch via thunk if requested.
8337 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
8338 ("*indirect_jump"): Disable for branch conversion using out of
8340 ("indirect_jump_via_thunk<mode>_z10")
8341 ("indirect_jump_via_thunk<mode>")
8342 ("indirect_jump_via_inlinethunk<mode>_z10")
8343 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
8344 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
8345 ("casesi_jump_via_inlinethunk<mode>_z10")
8346 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
8347 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
8348 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
8349 ("*indirect2_jump"): Disable for branch conversion.
8350 ("casesi_jump"): Turn into expander and expand patterns for branch
8352 ("return_use"): New expander.
8353 ("*return"): Emit return via thunk and rename it to ...
8354 ("*return<mode>"): ... this one.
8355 * config/s390/s390.opt: Add new options and and enum for the
8358 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8360 * lra-constraints.c (match_reload): Unconditionally use
8361 gen_lowpart_SUBREG, rather than selecting between that
8362 and equivalent gen_rtx_SUBREG code.
8364 2018-02-08 Richard Biener <rguenther@suse.de>
8366 PR tree-optimization/84233
8367 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
8368 changed flag instead of boguously re-using phi_inserted.
8370 2018-02-08 Martin Jambor <mjambor@suse.cz>
8372 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
8373 static local variables.
8375 2018-02-08 Richard Biener <rguenther@suse.de>
8377 PR tree-optimization/84278
8378 * tree-vect-stmts.c (vectorizable_store): When looking for
8379 smaller vector types to perform grouped strided loads/stores
8380 make sure the mode is supported by the target.
8381 (vectorizable_load): Likewise.
8383 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8385 * config/aarch64/aarch64.c (aarch64_components_for_bb):
8386 Increase LDP/STP opportunities by adding adjacent callee-saves.
8388 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8390 PR rtl-optimization/84068
8391 PR rtl-optimization/83459
8392 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
8394 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
8396 PR tree-optimization/84224
8397 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
8398 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
8401 2018-02-07 Iain Sandoe <iain@codesourcery.com>
8404 * config/rs6000/altivec.md (*restore_world): Remove LR use.
8405 * config/rs6000/predicates.md (restore_world_operation): Adjust op
8406 count, remove one USE.
8408 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
8410 * doc/install.texi (Configuration): Document the
8411 --with-long-double-format={ibm,ieee} PowerPC configuration
8415 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
8416 Convert from define_expand to be define_insn_and_split. Rework
8417 float/double/_Float128 conversions to QI/HI/SImode to work with
8418 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
8419 conversions to QI/HImode types did a store and then a load to
8420 truncate the value. For conversions to VSX registers, don't split
8421 the insn, instead emit the code directly. Use the code iterator
8422 any_fix to combine signed and unsigned conversions.
8423 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
8424 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
8425 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8426 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
8427 (fix_<mode>di2_hw): Likewise.
8428 (fixuns_<mode>di2_hw): Likewise.
8429 (fix_<mode>si2_hw): Likewise.
8430 (fixuns_<mode>si2_hw): Likewise.
8431 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
8432 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8433 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
8434 fix<uns>_trunc<SFDF:mode>si2_p8.
8435 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
8437 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
8438 (fix<uns>_<mode>_mem): Likewise.
8439 (fctiw<u>z_<mode>_mem): Likewise.
8440 (fix<uns>_<mode>_mem): Likewise.
8441 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
8442 the register allocator from doing a direct move to the GPRs to do
8443 a store, and instead use the ISA 3.0 store byte/half-word from
8444 vector register instruction. For IEEE 128-bit floating point,
8445 also optimize stores of 32-bit ints.
8446 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
8448 2018-02-07 Alan Hayward <alan.hayward@arm.com>
8450 * genextract.c (push_pathstr_operand): New function to support
8452 (walk_rtx): Call push_pathstr_operand.
8453 (print_path): Support [a-zA-Z].
8455 2018-02-07 Richard Biener <rguenther@suse.de>
8457 PR tree-optimization/84037
8458 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
8459 (cse_and_gimplify_to_preheader): Declare.
8460 (vect_get_place_in_interleaving_chain): Likewise.
8461 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8463 (_loop_vec_info::~_loop_vec_info): Delete it.
8464 (cse_and_gimplify_to_preheader): New function.
8465 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
8466 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
8467 (vectorizable_load): Likewise. For grouped stores always base
8468 the IV on the first element.
8469 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
8470 condition before gimplifying.
8472 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8474 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
8475 *DIV_EXPR and *MOD_EXPR.
8477 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
8480 * config/i386/i386.c (ix86_option_override_internal): Mask out
8481 the CF_SET bit when checking -fcf-protection.
8483 2018-02-07 Tom de Vries <tom@codesourcery.com>
8486 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
8489 2018-02-07 Richard Biener <rguenther@suse.de>
8491 PR tree-optimization/84204
8492 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
8495 PR tree-optimization/84205
8496 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
8497 special-case isl_ast_op_zdiv_r.
8499 PR tree-optimization/84223
8500 * graphite-scop-detection.c (gather_bbs::before_dom_children):
8501 Only add conditions from within the region.
8502 (gather_bbs::after_dom_children): Adjust.
8504 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
8507 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
8508 * config/avr/avr.md: Only post-reload split REG-REG moves if
8509 either register is GENERAL_REG_P.
8511 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8513 PR tree-optimization/84235
8514 * tree-ssa-scopedtables.c
8515 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
8516 if the subtraction is performed in floating point type where NaNs are
8517 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
8518 build 1. Formatting fix.
8520 2018-02-06 Jakub Jelinek <jakub@redhat.com>
8523 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
8524 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
8525 and skip it regardless of bb boundaries. Use CALL_P macro,
8526 don't test INSN_P (insn) together with CALL_P or JUMP_P check
8527 unnecessarily, formatting fix.
8529 2018-02-06 Michael Collison <michael.collison@arm.com>
8531 * config/arm/thumb2.md:
8532 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
8533 (*thumb_mov_notscc): Ditto.
8535 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
8538 * config/rs6000/rs6000.md (su code attribute): Use "u" for
8539 unsigned_fix, not "s".
8541 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8543 * configure.ac (gcc_fn_eh_frame_ro): New function.
8544 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
8545 correct .eh_frame permissions.
8546 * configure: Regenerate.
8548 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
8550 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
8553 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8555 * config/rs6000/rs6000.c (rs6000_option_override_internal):
8556 Display warning message for -mno-speculate-indirect-jumps.
8558 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
8560 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
8562 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
8564 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
8566 PR tree-optimization/84225
8567 * tree-eh.c (find_trapping_overflow): Only call
8568 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
8570 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8573 * config/i386/i386.c: Reimplement the check of possible options
8574 -mibt/-mshstk conbination. Change error messages.
8575 * doc/invoke.texi: Fix a typo: remove extra '='.
8577 2018-02-06 Marek Polacek <polacek@redhat.com>
8579 PR tree-optimization/84228
8580 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
8582 2018-02-06 Tamar Christina <tamar.christina@arm.com>
8585 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
8586 emitted arch directives.
8587 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
8588 __ARM_FEATURE_COPROC before changing architectures.
8590 2018-02-06 Richard Biener <rguenther@suse.de>
8592 * config/i386/i386.c (print_reg): Fix typo.
8593 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
8595 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
8597 * configure: Regenerate.
8599 2018-02-05 Martin Sebor <msebor@redhat.com>
8601 PR tree-optimization/83369
8602 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
8605 2018-02-05 Martin Liska <mliska@suse.cz>
8607 * doc/invoke.texi: Cherry-pick upstream r323995.
8609 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
8611 * ira.c (ira_init_register_move_cost): Adjust comment.
8613 2018-02-05 Martin Liska <mliska@suse.cz>
8615 PR gcov-profile/84137
8616 * doc/gcov.texi: Fix typo in documentation.
8618 2018-02-05 Martin Liska <mliska@suse.cz>
8620 PR gcov-profile/83879
8621 * doc/gcov.texi: Document necessity of --dynamic-list-data when
8622 using dlopen functionality.
8624 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
8626 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
8627 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
8628 _mm_maskz_range_ss, _mm_mask_range_round_ss,
8629 _mm_maskz_range_round_ss): New intrinsics.
8630 (__builtin_ia32_rangesd128_round)
8631 (__builtin_ia32_rangess128_round): Remove.
8632 (__builtin_ia32_rangesd128_mask_round,
8633 __builtin_ia32_rangess128_mask_round): New builtins.
8634 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
8635 __builtin_ia32_rangess128_round): Remove.
8636 (__builtin_ia32_rangesd128_mask_round,
8637 __builtin_ia32_rangess128_mask_round): New builtins.
8638 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
8639 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
8640 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
8641 "<round_saeonly_constraint>")): Changed to ...
8642 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
8643 "<round_saeonly_scalar_constraint>")): ... this.
8644 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
8645 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
8646 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
8647 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
8648 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
8650 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
8652 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
8654 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
8655 Remove all values except native, 8540 and 8548.
8657 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
8659 * config/i386/i386.c (ix86_output_function_return): Pass
8660 INVALID_REGNUM, instead of -1, as invalid register number to
8661 indirect_thunk_name and output_indirect_thunk.
8663 2018-02-02 Julia Koval <julia.koval@intel.com>
8665 * config.gcc: Add -march=icelake.
8666 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
8667 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
8668 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
8669 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
8670 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
8671 (processor_target_table): Add icelake.
8672 (ix86_option_override_internal): Handle new PTAs.
8673 (get_builtin_code_for_version): Handle icelake.
8674 (M_INTEL_COREI7_ICELAKE): New.
8675 (fold_builtin_cpu): Handle icelake.
8676 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
8677 * doc/invoke.texi: Add -march=icelake.
8679 2018-02-02 Julia Koval <julia.koval@intel.com>
8681 * config/i386/i386.c (ix86_option_override_internal): Change flags type
8682 to wide_int_bitmask.
8683 * wide-int-bitmask.h: New.
8685 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8688 * config/i386/i386.md: Replace Pmode with word_mode in
8689 builtin_setjmp_setup and builtin_longjmp to support x32.
8691 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
8695 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
8697 (rs6000_supported_cpu_names): New static variable.
8698 (linux_cpu_translation_table): Likewise.
8699 (elf_platform) <cpu>: Define new static variable and use it.
8700 Translate kernel AT_PLATFORM name to canonical name if needed.
8701 Error if platform name is unknown.
8703 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
8706 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
8708 2018-02-01 Jeff Law <law@redhat.com>
8711 * config/i386/i386.c (release_scratch_register_on_entry): Add new
8712 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
8713 the scratch if RELEASE_VIA_POP is false.
8714 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
8715 If we have to save a temporary register, decrement SIZE appropriately.
8716 Pass new arguments to release_scratch_register_on_entry.
8717 (ix86_adjust_stack_and_probe): Likewise.
8718 (ix86_emit_probe_stack_range): Pass new arguments to
8719 release_scratch_register_on_entry.
8721 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
8723 PR rtl-optimization/84157
8724 * combine.c (change_zero_ext): Use REG_P predicate in
8725 front of HARD_REGISTER_P predicate.
8727 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
8729 * config/avr/avr.c (avr_option_override): Move disabling of
8730 -fdelete-null-pointer-checks to...
8731 * common/config/avr/avr-common.c (avr_option_optimization_table):
8734 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8736 PR tree-optimization/81635
8737 * tree-data-ref.c (split_constant_offset_1): For types that
8738 wrap on overflow, try to use range info to prove that wrapping
8741 2018-02-01 Renlin Li <renlin.li@arm.com>
8744 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
8746 (aarch64_register_move_cost): Likewise.
8747 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
8749 (REG_CLASS_NAMES): Likewise.
8750 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
8751 TAILCALL_ADDR_REGS. Remove IP registers.
8752 * config/aarch64/aarch64.md (Ucs): Update register constraint.
8754 2018-02-01 Richard Biener <rguenther@suse.de>
8756 * domwalk.h (dom_walker::dom_walker): Add additional constructor
8757 for specifying RPO order and allow NULL for that.
8758 * domwalk.c (dom_walker::dom_walker): Likewise.
8759 (dom_walker::walk): Handle NULL RPO order.
8760 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
8762 (rewrite_update_dom_walker): Likewise.
8763 (mark_def_dom_walker): Likewise.
8765 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8767 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
8768 (aarch64_maybe_expand_sve_subreg_move): Declare.
8769 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
8770 * config/aarch64/predicates.md (aarch64_any_register_operand): New
8772 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
8773 that are semantically a reverse operation.
8774 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
8775 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
8776 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
8778 (aarch64_can_change_mode_class): For big-endian, forbid changes
8779 between two SVE modes if they have different element sizes.
8781 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8783 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
8784 the TImode handling for big-endian targets.
8786 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8788 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
8789 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
8791 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
8792 Remove BSWAP handing for big-endian targets and use the form of
8793 LD1RQ appropariate for the mode.
8795 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8797 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
8798 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
8801 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8804 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
8805 check for operands that need to go through aarch64_sve_reload_be.
8807 2018-02-01 Jakub Jelinek <jakub@redhat.com>
8809 PR tree-optimization/81661
8810 PR tree-optimization/84117
8811 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
8812 * tree-eh.c: Include gimplify.h.
8813 (find_trapping_overflow, replace_trapping_overflow,
8814 rewrite_to_non_trapping_overflow): New functions.
8815 * tree-vect-loop.c: Include tree-eh.h.
8816 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
8817 * tree-data-ref.c: Include tree-eh.h.
8818 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
8820 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
8822 PR rtl-optimization/84123
8823 * combine.c (change_zero_ext): Check if hard register satisfies
8824 can_change_dest_mode before calling gen_lowpart_SUBREG.
8826 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
8829 * ira.c (ira_init_register_move_cost): Remove assert.
8831 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8833 PR rtl-optimization/84071
8834 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
8835 * doc/tm.texi: Regenerate.
8837 2018-01-31 Richard Biener <rguenther@suse.de>
8839 PR tree-optimization/84132
8840 * tree-data-ref.c (analyze_miv_subscript): Properly
8841 check whether evolution_function_is_affine_multivariate_p
8842 before calling gcd_of_steps_may_divide_p.
8844 2018-01-31 Julia Koval <julia.koval@intel.com>
8847 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
8848 * config/i386/i386.md (rdpid_rex64) New.
8849 (rdpid): Make 32bit only.
8851 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
8854 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
8855 an IDENTIFIER_NODE for FUNCTION_TYPE's.
8857 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8860 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8862 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8864 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8866 PR rtl-optimization/84071
8867 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
8868 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
8870 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8872 * config/arc/arc.c (arc_handle_aux_attribute): New function.
8873 (arc_attribute_table): Add 'aux' attribute.
8874 (arc_in_small_data_p): Consider aux like variables.
8875 (arc_is_aux_reg_p): New function.
8876 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
8877 (arc_get_aux_arg): New function.
8878 (prepare_move_operands): Handle aux-register access.
8879 (arc_handle_aux_attribute): New function.
8880 * doc/extend.texi (ARC Variable attributes): Add subsection.
8882 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8884 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
8885 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
8886 (arc_attribute_table): Add 'uncached' attribute.
8887 (arc_print_operand): Print '.di' flag for uncached memory
8889 (arc_in_small_data_p): Do not consider for small data the uncached
8891 (arc_is_uncached_mem_p): New function.
8892 * config/arc/predicates.md (compact_store_memory_operand): Check
8893 for uncached memory accesses.
8894 (nonvol_nonimm_operand): Likewise.
8895 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
8897 2018-01-31 Jakub Jelinek <jakub@redhat.com>
8900 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
8901 falign-loops=): Add Optimization flag.
8903 2018-01-30 Jeff Law <law@redhat.com>
8906 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
8907 INT_REGISTERS_SAVED. Check it prior to calling
8908 get_scratch_register_on_entry.
8909 (ix86_adjust_stack_and_probe): Similarly.
8910 (ix86_emit_probe_stack_range): Similarly.
8911 (ix86_expand_prologue): Corresponding changes.
8913 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8916 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
8917 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
8919 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
8922 * lra-constraints.c (curr_insn_transform): Process AND in the
8925 2018-01-30 Jakub Jelinek <jakub@redhat.com>
8927 PR rtl-optimization/83986
8928 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
8929 dependence against last_pending_memory_flush in addition to
8932 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
8934 PR tree-optimization/81611
8935 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
8938 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8941 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
8944 2018-01-30 Richard Biener <rguenther@suse.de>
8945 Jakub Jelinek <jakub@redhat.com>
8947 PR tree-optimization/84111
8948 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
8949 inner loops added during recursion, as they don't have up-to-date
8952 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8955 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
8956 (can_inline_edge_by_limits_p): ... here.
8957 (can_early_inline_edge_p, check_callers,
8958 update_caller_keys, update_callee_keys, recursive_inlining,
8959 add_new_edges_to_heap, speculation_useful_p,
8960 inline_small_functions,
8961 inline_small_functions, flatten_function,
8962 inline_to_all_callers_1): Update.
8964 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8966 * profile-count.c (profile_count::combine_with_ipa_count): Handle
8969 2018-01-30 Richard Biener <rguenther@suse.de>
8971 PR tree-optimization/83008
8972 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
8973 invariant and constant vector uses in stmts when they need
8976 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8979 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
8980 * configure: Regenerate.
8982 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8984 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
8986 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
8987 Use gen_rtx_REG rather than gen_lowpart.
8989 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8991 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
8992 rather than 0 when creating partial subregs.
8994 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8996 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
8999 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
9002 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
9003 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
9004 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
9005 flags. This restores the settings used before the 2017-07-24.
9006 Turning off pre increment/decrement/modify allows IVOPTS to
9007 optimize DF/SF loops where the index is an int.
9009 2018-01-29 Richard Biener <rguenther@suse.de>
9010 Kelvin Nilsen <kelvin@gcc.gnu.org>
9013 * tree-vect-stmts.c (vectorizable_call): Don't call
9014 targetm.vectorize_builtin_md_vectorized_function if callee is
9017 2018-01-22 Carl Love <cel@us.ibm.com>
9019 * doc/extend.tex: Fix typo in second arg in
9020 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
9022 2018-01-29 Richard Biener <rguenther@suse.de>
9024 PR tree-optimization/84086
9025 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
9026 (flush_ssaname_freelist): When SSA names were released reset
9027 the SCEV hash table.
9029 2018-01-29 Richard Biener <rguenther@suse.de>
9031 PR tree-optimization/84057
9032 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
9033 removed paths when removing edges.
9035 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
9037 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
9038 -mfunction-return=@var{choice}.
9040 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
9043 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
9045 (layout::print_source_line): Likewise.
9046 (test_get_line_width_without_trailing_whitespace): Add test cases.
9048 2018-01-27 Jakub Jelinek <jakub@redhat.com>
9051 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
9054 2018-01-26 Jim Wilson <jimw@sifive.com>
9056 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
9058 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
9061 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9063 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
9064 and CMP + SUB-immediate -> SUBS.
9066 2018-01-26 Martin Sebor <msebor@redhat.com>
9068 PR tree-optimization/83896
9069 * tree-ssa-strlen.c (get_string_len): Rename...
9070 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
9071 Avoid assuming length is constant.
9072 (handle_char_store): Use HOST_WIDE_INT for string length.
9074 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
9077 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
9078 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
9080 2018-01-26 Richard Biener <rguenther@suse.de>
9082 PR rtl-optimization/84003
9083 * dse.c (record_store): Only record redundant stores when
9084 the earlier store aliases at least all accesses the later one does.
9086 2018-01-26 Jakub Jelinek <jakub@redhat.com>
9088 PR rtl-optimization/83985
9089 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
9090 REG_CFA_RESTORE insns.
9091 (delete_unmarked_insns): Don't ignore separate shrink wrapping
9092 REG_CFA_RESTORE insns here.
9095 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
9096 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
9098 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9100 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
9101 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
9102 (arc_init): Likewise.
9103 (arc_override_options): Likewise.
9104 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
9106 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
9108 * config/arc/arc.h (TARGET_DBNZ): Define.
9109 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
9110 properly set the tune attribute.
9111 (dbnz): Use TARGET_DBNZ guard.
9112 * config/arc/arc.opt (mtune): Add core3 option.
9114 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9116 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
9117 recognize new pic like addresses.
9118 (arc_delegitimize_address): Clean up.
9120 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9122 * config/arc/arc-arches.def: Option mrf16 valid for all
9124 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
9125 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
9126 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
9127 * config/arc/arc-tables.opt: Regenerate.
9128 * config/arc/arc.c (arc_conditional_register_usage): Handle
9129 reduced register file case.
9130 (arc_file_start): Set must have build attributes.
9131 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
9133 * config/arc/arc.opt (mrf16): Add new option.
9134 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
9135 * config/arc/genmultilib.awk: Handle new mrf16 option.
9136 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
9137 * config/arc/t-multilib: Regenerate.
9138 * doc/invoke.texi (ARC Options): Document mrf16 option.
9140 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9142 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
9143 * config/arc/arc.c (arc_handle_secure_attribute): New function.
9144 (arc_attribute_table): Add 'secure_call' attribute.
9145 (arc_print_operand): Print secure call operand.
9146 (arc_function_ok_for_sibcall): Don't optimize tail calls when
9148 (arc_is_secure_call_p): New function. * config/arc/arc.md
9149 (call_i): Add support for sjli instruction.
9150 (call_value_i): Likewise.
9151 * config/arc/constraints.md (Csc): New constraint.
9153 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9154 John Eric Martin <John.Martin@emmicro-us.com>
9156 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
9157 * config/arc/arc.c (_arc_jli_section): New struct.
9158 (arc_jli_section): New type.
9159 (rc_jli_sections): New static variable.
9160 (arc_handle_jli_attribute): New function.
9161 (arc_attribute_table): Add jli_always and jli_fixed attribute.
9162 (arc_file_end): New function.
9163 (TARGET_ASM_FILE_END): Define.
9164 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
9165 (arc_add_jli_section): New function.
9166 (jli_call_scan): Likewise.
9167 (arc_reorg): Call jli_call_scan.
9168 (arc_output_addsi): Remove 'S' from printing asm operand.
9169 (arc_is_jli_call_p): New function.
9170 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
9172 (movhi_insn): Likewise.
9173 (movsi_insn): Likewise.
9174 (movsi_set_cc_insn): Likewise.
9175 (loadqi_update): Likewise.
9176 (load_zeroextendqisi_update): Likewise.
9177 (load_signextendqisi_update): Likewise.
9178 (loadhi_update): Likewise.
9179 (load_zeroextendhisi_update): Likewise.
9180 (load_signextendhisi_update): Likewise.
9181 (loadsi_update): Likewise.
9182 (loadsf_update): Likewise.
9183 (movsicc_insn): Likewise.
9184 (bset_insn): Likewise.
9185 (bxor_insn): Likewise.
9186 (bclr_insn): Likewise.
9187 (bmsk_insn): Likewise.
9188 (bicsi3_insn): Likewise.
9189 (cmpsi_cc_c_insn): Likewise.
9190 (movsi_ne): Likewise.
9191 (movsi_cond_exec): Likewise.
9192 (clrsbsi2): Likewise.
9203 (call_i): Remove 'S' asm letter, add jli instruction.
9204 (call_value_i): Likewise.
9205 * config/arc/arc.op (mjli-always): New option.
9206 * config/arc/constraints.md (Cji): New constraint.
9207 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
9209 (subsf3_fpx): Likewise.
9210 (mulsf3_fpx): Likewise.
9211 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
9213 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
9215 * doc/invoke.texi (ARC): Document mjli-always option.
9217 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
9219 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
9220 avoid addition with 0 and use incw and decw where possible.
9222 2018-01-26 Richard Biener <rguenther@suse.de>
9224 PR tree-optimization/81082
9225 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
9226 association if it requires casting to unsigned.
9227 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
9228 from fold_plusminus_mult_expr to catch important cases late when
9229 range info is available.
9231 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9233 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
9234 * configure.ac (hidden_linkonce): New test.
9235 * configure: Regenerate.
9236 * config.in: Regenerate.
9238 2018-01-26 Julia Koval <julia.koval@intel.com>
9240 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
9241 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
9242 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
9243 _mm_mask_bitshuffle_epi64_mask): Fix type.
9244 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
9245 USI_FTYPE_V4DI_V4DI_USI): Remove.
9246 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
9247 __builtin_ia32_vpshufbitqmb256_mask,
9248 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
9249 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
9250 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
9252 2018-01-26 Alan Modra <amodra@gmail.com>
9255 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
9256 UNSPEC_VBPERMQ. Sort other unspecs.
9258 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
9260 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
9262 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
9265 * predict.c (drop_profile): Do not push/pop cfun; update also
9267 (handle_missing_profiles): Fix logic looking for zero profiles.
9269 2018-01-25 Jakub Jelinek <jakub@redhat.com>
9272 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
9273 on functions with #pragma omp declare simd or functions with simd
9275 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
9276 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
9277 Remove trailing \n from warning_at calls.
9279 2018-01-25 Tom de Vries <tom@codesourcery.com>
9282 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9283 for neutered workers.
9285 2018-01-24 Joseph Myers <joseph@codesourcery.com>
9288 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
9289 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
9291 2018-01-24 Jeff Law <law@redhat.com>
9294 * i386.c (get_probe_interval): Move to earlier point.
9295 (ix86_compute_frame_layout): If -fstack-clash-protection and
9296 the frame is larger than the probe interval, then use pushes
9297 to save registers rather than reg->mem moves.
9298 (ix86_expand_prologue): Remove conditional for int_registers_saved
9301 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
9304 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
9305 min/max for never referenced object.
9307 2018-01-24 Jakub Jelinek <jakub@redhat.com>
9310 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
9312 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
9313 attributes from DECL_ATTRIBUTES (decl) without affecting
9314 DECL_ATTRIBUTES (current_function_decl).
9315 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
9316 functions with non-NULL DECL_ABSTRACT_ORIGIN.
9318 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
9320 PR tree-optimization/83979
9321 * fold-const.c (fold_comparison): Use constant_boolean_node
9322 instead of boolean_{true,false}_node.
9324 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
9326 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
9329 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9331 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9332 Simplify the clause that sets the length attribute.
9333 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9334 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
9335 clause that sets the length attribute.
9336 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9338 2018-01-24 Tom de Vries <tom@codesourcery.com>
9341 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
9342 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
9343 Add strict parameter.
9344 (prevent_branch_around_nothing): Insert dummy insn between branch to
9345 label and label with no ptx insn inbetween.
9346 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
9348 2018-01-24 Tom de Vries <tom@codesourcery.com>
9351 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9352 for neutered threads in warp.
9353 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
9355 2018-01-24 Richard Biener <rguenther@suse.de>
9357 PR tree-optimization/83176
9358 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
9361 2018-01-24 Richard Biener <rguenther@suse.de>
9363 PR tree-optimization/82819
9364 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
9365 code generating pluses that are no-ops in the target precision.
9367 2018-01-24 Richard Biener <rguenther@suse.de>
9370 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
9372 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9374 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
9375 to merge probabilities.
9376 * predict.c (probably_never_executed): Also mark as cold functions
9377 with global 0 profile and guessed local profile.
9378 * profile-count.c (profile_probability::combine_with_count): New
9380 * profile-count.h (profile_probability::operator*,
9381 profile_probability::operator*=, profile_probability::operator/,
9382 profile_probability::operator/=): Reduce precision to adjusted
9383 and set value to guessed on contradictory divisions.
9384 (profile_probability::combine_with_freq): Remove.
9385 (profile_probability::combine_wiht_count): Declare.
9386 (profile_count::force_nonzero):: Set to adjusted.
9387 (profile_count::probability_in):: Set quality to adjusted.
9388 * tree-ssa-tail-merge.c (replace_block_by): Use
9391 2018-01-23 Andrew Waterman <andrew@sifive.com>
9392 Jim Wilson <jimw@sifive.com>
9394 * config/riscv/riscv.c (riscv_stack_boundary): New.
9395 (riscv_option_override): Set riscv_stack_boundary. Handle
9396 riscv_preferred_stack_boundary_arg.
9397 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
9398 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
9399 (STACK_BOUNDARY): Set to riscv_stack_boundary.
9400 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
9401 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
9402 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
9404 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
9407 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
9408 of struct ix86_frame.
9409 (ix86_expand_epilogue): Likewise. Add a local variable for
9410 the reg_save_offset field in struct ix86_frame.
9412 2018-01-23 Bin Cheng <bin.cheng@arm.com>
9414 PR tree-optimization/82604
9415 * tree-loop-distribution.c (enum partition_kind): New enum item
9416 PKIND_PARTIAL_MEMSET.
9417 (partition_builtin_p): Support above new enum item.
9418 (generate_code_for_partition): Ditto.
9419 (compute_access_range): Differentiate cases that equality can be
9420 proven at all loops, the innermost loops or no loops.
9421 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
9422 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
9423 (finalize_partitions, distribute_loop): Don't fuse partition of
9424 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
9425 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
9428 2018-01-23 Martin Liska <mliska@suse.cz>
9430 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
9431 order to ignore the predictor.
9432 (PRED_POLYMORPHIC_CALL): Likewise.
9433 (PRED_RECURSIVE_CALL): Likewise.
9435 2018-01-23 Martin Liska <mliska@suse.cz>
9437 * tree-profile.c (tree_profiling): Print function header to
9438 aware reader which function we are working on.
9439 * value-prof.c (gimple_find_values_to_profile): Do not print
9440 not interesting value histograms.
9442 2018-01-23 Martin Liska <mliska@suse.cz>
9444 * profile-count.h (enum profile_quality): Add
9445 profile_uninitialized as the first value. Do not number values
9446 as they are zero based.
9447 (profile_count::verify): Update sanity check.
9448 (profile_probability::verify): Likewise.
9450 2018-01-23 Nathan Sidwell <nathan@acm.org>
9452 * doc/invoke.texi (ffor-scope): Deprecate.
9454 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9456 PR tree-optimization/83510
9457 * domwalk.c (set_all_edges_as_executable): New function.
9458 (dom_walker::dom_walker): Convert bool param
9459 "skip_unreachable_blocks" to enum reachability. Move setup of
9460 edge flags to set_all_edges_as_executable and only do it when
9461 reachability is REACHABLE_BLOCKS.
9462 * domwalk.h (enum dom_walker::reachability): New enum.
9463 (dom_walker::dom_walker): Convert bool param
9464 "skip_unreachable_blocks" to enum reachability.
9465 (set_all_edges_as_executable): New decl.
9466 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
9467 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
9469 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
9470 but converting true to REACHABLE_BLOCKS.
9471 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
9473 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
9474 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
9475 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
9477 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
9478 if check_all_array_refs will be called.
9480 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9482 * tree.c (selftest::test_location_wrappers): Add more test
9485 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9487 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
9488 (selftest::test_bit_in_range): Likewise.
9490 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9493 * doc/sourcebuild.texi (vect_float): Say that the selector
9494 only describes the situation when -funsafe-math-optimizations is on.
9495 (vect_float_strict): Document.
9497 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9499 PR tree-optimization/83965
9500 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
9501 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
9502 instead of checking only for a reduction.
9503 (vect_recog_widen_sum_pattern): Likewise.
9505 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9507 * predict.c (probably_never_executed): Only use precise profile info.
9508 (compute_function_frequency): Skip after inlining hack since we now
9509 have quality checking.
9511 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9513 * profile-count.h (profile_probability::very_unlikely,
9514 profile_probability::unlikely, profile_probability::even): Set
9515 precision to guessed.
9517 2018-01-23 Richard Biener <rguenther@suse.de>
9519 PR tree-optimization/83963
9520 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
9521 Properly terminate dominator walk when crossing the exit edge not
9522 when visiting its source block.
9524 2018-01-23 Jakub Jelinek <jakub@redhat.com>
9527 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
9528 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
9530 2018-01-22 Jakub Jelinek <jakub@redhat.com>
9532 PR tree-optimization/83957
9533 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
9534 semicolon after for body surrounded by braces.
9536 PR tree-optimization/83081
9537 * profile-count.h (profile_probability::split): New method.
9538 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
9539 Use profile_probability::split.
9540 (do_compare_rtx_and_jump): Fix adjustment of probabilities
9541 when splitting a single conditional jump into 2.
9543 2018-01-22 David Malcolm <dmalcolm@redhat.com>
9545 PR tree-optimization/69452
9546 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
9549 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9551 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
9552 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
9553 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
9555 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9557 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
9558 * config/rl78/rl78.md (movdi): New define_expand.
9559 * config/rl78/rl78.c (rl78_split_movdi): New function.
9561 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
9564 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
9566 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
9567 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
9568 128-bit to produce an UNSPEC move to get the double word with the
9569 signbit and then a shift directly to do signbit.
9570 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
9571 implementation with a new version that just does either a direct
9572 move or a regular move. Move memory interface to separate insns.
9573 Move insns so they are next to the expander.
9574 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
9575 with signbit move. Split big and little endian case.
9576 (signbit<mode>2_dm_mem_le): Likewise.
9577 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
9578 (signbit<mode>2_dm2): Likewise.
9580 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9582 * config/rl78/rl78.md (anddi3): New define_expand.
9584 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9586 * config/rl78/rl78.md (umindi3): New define_expand.
9588 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9590 * config/rl78/rl78.md (smindi3): New define_expand.
9592 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9594 * config/rl78/rl78.md (smaxdi3): New define_expand.
9596 2018-01-22 Carl Love <cel@us.ibm.com>
9598 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
9599 LVX_V1TI): Add macro expansion.
9600 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
9601 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
9602 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
9603 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
9604 Change check to determine if the instruction is a byte reversing
9605 entry. Fix typo in comment.
9606 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
9607 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
9608 Add def_builtin calls for new builtins.
9609 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
9610 Add define_insn expansion.
9612 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9614 * config/rl78/rl78.md (umaxdi3): New define_expand.
9616 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9618 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
9619 for non-QImode registers.
9621 2018-01-22 Richard Biener <rguenther@suse.de>
9623 PR tree-optimization/83963
9624 * graphite-scop-detection.c (scop_detection::get_sese): Delay
9625 including the loop exit block.
9626 (scop_detection::merge_sese): Likewise.
9627 (scop_detection::add_scop): Do it here instead.
9629 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9631 * doc/sourcebuild.texi (arm_softfloat): Document.
9633 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
9636 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
9637 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
9638 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
9640 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9641 David Edelsohn <dje.gcc@gmail.com>
9644 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9645 Change "crset eq" to "crset 2".
9646 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9647 (*call_indirect_aix<mode>_nospec): Likewise.
9648 (*call_value_indirect_aix<mode>_nospec): Likewise.
9649 (*call_indirect_elfv2<mode>_nospec): Likewise.
9650 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
9651 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
9652 change assembly output from . to $.
9653 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9654 (indirect_jump<mode>_nospec): Change assembly output from . to $.
9655 (*tablejump<mode>_internal1_nospec): Likewise.
9657 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
9660 * config/sh/sh_optimize_sett_clrt.cc:
9661 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
9663 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
9665 PR tree-optimization/83940
9666 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
9667 offset_dt to vect_constant_def rather than vect_unknown_def_type.
9668 (vect_check_load_store_mask): Add a mask_dt_out parameter and
9669 use it to pass back the definition type.
9670 (vect_check_store_rhs): Likewise rhs_dt_out.
9671 (vect_build_gather_load_calls): Add a mask_dt argument and use
9672 it instead of a call to vect_is_simple_use.
9673 (vectorizable_store): Update calls to vect_check_load_store_mask
9674 and vect_check_store_rhs. Use the dt returned by the latter instead
9675 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
9676 instead of calls to vect_is_simple_use. Pass the scalar rather
9677 than the vector operand to vect_is_simple_use when handling
9678 second and subsequent copies of an rhs value.
9679 (vectorizable_load): Update calls to vect_check_load_store_mask
9680 and vect_build_gather_load_calls. Use the cached mask_dt and
9681 gs_info.offset_dt instead of calls to vect_is_simple_use.
9683 2018-01-20 Jakub Jelinek <jakub@redhat.com>
9686 * tree-emutls.c: Include gimplify.h.
9687 (lower_emutls_2): New function.
9688 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
9689 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
9690 it before further processing.
9693 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
9694 UINTVAL (trueop1) instead of INTVAL (op1).
9696 2018-01-19 Jakub Jelinek <jakub@redhat.com>
9700 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
9701 INCOMING_FRAME_SP_OFFSET if not defined.
9702 (scan_trace): Add ENTRY argument. If true and
9703 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
9704 emit a note to adjust the CFA offset.
9705 (create_cfi_notes): Adjust scan_trace callers.
9706 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
9707 INCOMING_FRAME_SP_OFFSET in the CIE.
9708 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
9709 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
9711 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
9712 * doc/tm.texi: Regenerated.
9714 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9716 PR rtl-optimization/83147
9717 * lra-constraints.c (remove_inheritance_pseudos): Use
9718 lra_substitute_pseudo_within_insn.
9720 2018-01-19 Tom de Vries <tom@codesourcery.com>
9721 Cesar Philippidis <cesar@codesourcery.com>
9724 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
9726 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
9729 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
9730 spaces for function labels.
9732 2018-01-19 Martin Liska <mliska@suse.cz>
9734 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
9735 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
9736 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
9737 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
9738 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
9739 (PRED_CONST_RETURN): Change from 69 to 65.
9740 (PRED_NULL_RETURN): Change from 91 to 71.
9741 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
9742 (PRED_LOOP_GUARD): Change from 66 to 73.
9744 2018-01-19 Martin Liska <mliska@suse.cz>
9746 * predict.c (predict_insn_def): Add new assert.
9747 (struct branch_predictor): Change type to signed integer.
9748 (test_prediction_value_range): Amend test to cover
9750 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
9751 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
9752 (PRED_LOOP_ITERATIONS_MAX): Likewise.
9753 (PRED_LOOP_IV_COMPARE): Likewise.
9754 * predict.h (PROB_UNINITIALIZED): Define new constant.
9756 2018-01-19 Martin Liska <mliska@suse.cz>
9758 * predict.c (dump_prediction): Add new format for
9759 analyze_brprob.py script which is enabled with -details
9761 * profile-count.h (precise_p): New function.
9763 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9765 PR tree-optimization/83922
9766 * tree-vect-loop.c (vect_verify_full_masking): Return false if
9767 there are no statements that need masking.
9768 (vect_active_double_reduction_p): New function.
9769 (vect_analyze_loop_operations): Use it when handling phis that
9770 are not in the loop header.
9772 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9774 PR tree-optimization/83914
9775 * tree-vect-loop.c (vectorizable_induction): Don't convert
9776 init_expr or apply the peeling adjustment for inductions
9777 that are nested within the vectorized loop.
9779 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9781 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
9784 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9788 * function.h (gimplify_parameters): Add gimple_seq * argument.
9789 * function.c: Include gimple.h and options.h.
9790 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
9791 for the added local temporaries if needed.
9792 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
9793 if there are any parameter cleanups, wrap whole body into a
9794 try/finally with the cleanups.
9796 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
9799 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
9800 Use GET_MODE_CLASS for scalar floating point.
9802 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9806 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
9807 Fix call of call_cgraph_insertion_hooks.
9809 2018-01-18 Martin Sebor <msebor@redhat.com>
9811 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
9813 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9816 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
9819 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
9822 * common.opt: (-ffile-prefix-map): New option.
9823 * opts.c (common_handle_option): Defer it.
9824 * opts-global.c (handle_common_deferred_options): Handle it.
9825 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
9826 * file-prefix-map.h: New file.
9827 (remap_debug_filename, add_debug_prefix_map): ...here.
9828 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
9829 * final.c (debug_prefix_map, add_debug_prefix_map
9830 remap_debug_filename): Move to...
9831 * file-prefix-map.c: New file.
9832 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
9833 generalize, get rid of alloca(), use strrchr() instead of strchr().
9834 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
9835 Implement in terms of add_prefix_map().
9836 (remap_macro_filename, remap_debug_filename): Implement in term of
9838 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
9839 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
9840 * dbxout.c: Include file-prefix-map.h.
9841 * varasm.c: Likewise.
9842 * vmsdbgout.c: Likewise.
9843 * xcoffout.c: Likewise.
9844 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
9845 * doc/cppopts.texi (-fmacro-prefix-map): Document.
9846 * doc/invoke.texi (-ffile-prefix-map): Document.
9847 (-fdebug-prefix-map): Update description.
9849 2018-01-18 Martin Liska <mliska@suse.cz>
9851 * config/i386/i386.c (indirect_thunk_name): Document that also
9853 (output_indirect_thunk): Document why both instructions
9854 (pause and lfence) are generated.
9856 2018-01-18 Richard Biener <rguenther@suse.de>
9858 PR tree-optimization/83887
9859 * graphite-scop-detection.c
9860 (scop_detection::get_nearest_dom_with_single_entry): Remove.
9861 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
9862 (scop_detection::merge_sese): Re-implement with a flood-fill
9863 algorithm that properly finds a SESE region if it exists.
9865 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9868 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
9869 pointer_diff optimizations use view_convert instead of convert.
9871 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9873 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9874 Generate different code for -mno-speculate-indirect-jumps.
9875 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9876 (*call_indirect_aix<mode>): Disable for
9877 -mno-speculate-indirect-jumps.
9878 (*call_indirect_aix<mode>_nospec): New define_insn.
9879 (*call_value_indirect_aix<mode>): Disable for
9880 -mno-speculate-indirect-jumps.
9881 (*call_value_indirect_aix<mode>_nospec): New define_insn.
9882 (*sibcall_nonlocal_sysv<mode>): Generate different code for
9883 -mno-speculate-indirect-jumps.
9884 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9886 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
9888 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
9889 long double type, set the flags for noting the default long double
9890 type, even if we don't pass or return a long double type.
9892 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
9895 * ipa-inline.c (flatten_function): Do not overwrite final inlining
9898 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
9900 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
9901 support for merge[hl].
9902 (fold_mergehl_helper): New helper function.
9903 (tree-vector-builder.h): New #include for tree_vector_builder usage.
9904 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
9905 (altivec_vmrglw_direct): Add xxmrglw insn.
9907 2018-01-17 Andrew Waterman <andrew@sifive.com>
9909 * config/riscv/riscv.c (riscv_conditional_register_usage): If
9910 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
9912 2018-01-17 David Malcolm <dmalcolm@redhat.com>
9915 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
9916 call the lto_location_cache before reading the
9917 DECL_SOURCE_LOCATION of the types.
9919 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
9920 Richard Sandiford <richard.sandiford@linaro.org>
9922 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
9923 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
9924 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
9925 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
9926 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
9928 * config/aarch64/constraints.md (aarch64_movti_operand):
9930 * config/aarch64/predicates.md (Uti): Add new constraint.
9932 2018-01-17 Carl Love <cel@us.ibm.com>
9934 * config/rs6000/vsx.md (define_expand xl_len_r,
9935 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
9936 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
9938 (define_expand, define_insn): Move the shift left from the
9939 define_insn to the define_expand for lxvl and stxvl instructions.
9940 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
9941 and XL_LEN_R definitions to PURE.
9943 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
9945 * config/i386/i386.c (indirect_thunk_name): Declare regno
9946 as unsigned int. Compare regno with INVALID_REGNUM.
9947 (output_indirect_thunk): Ditto.
9948 (output_indirect_thunk_function): Ditto.
9949 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
9950 in the call to output_indirect_thunk_function.
9952 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
9955 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
9956 rather than the size of inner_type to determine the stack slot size
9957 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
9959 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
9962 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
9965 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
9967 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
9968 endian Linux systems to optionally enable multilibs for selecting
9969 the long double type if the user configured an explicit type.
9970 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
9971 have no long double multilibs if not defined.
9972 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
9973 warn if the user used -mabi={ieee,ibm}longdouble and we built
9974 multilibs for long double.
9975 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
9976 appropriate multilib option.
9977 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
9979 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
9980 for building long double multilibs.
9981 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
9983 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
9985 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
9988 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
9990 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
9993 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
9996 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
9999 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10001 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
10002 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
10004 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
10006 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
10007 different rtl trees depending on TARGET_64BIT.
10008 (rs6000_gen_lvx): Likewise.
10010 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10012 * config/visium/visium.md (nop): Tweak comment.
10013 (hazard_nop): Likewise.
10015 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10017 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
10018 -mspeculate-indirect-jumps.
10019 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
10020 for -mno-speculate-indirect-jumps.
10021 (*call_indirect_elfv2<mode>_nospec): New define_insn.
10022 (*call_value_indirect_elfv2<mode>): Disable for
10023 -mno-speculate-indirect-jumps.
10024 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
10025 (indirect_jump): Emit different RTL for
10026 -mno-speculate-indirect-jumps.
10027 (*indirect_jump<mode>): Disable for
10028 -mno-speculate-indirect-jumps.
10029 (*indirect_jump<mode>_nospec): New define_insn.
10030 (tablejump): Emit different RTL for
10031 -mno-speculate-indirect-jumps.
10032 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
10033 (tablejumpsi_nospec): New define_expand.
10034 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
10035 (tablejumpdi_nospec): New define_expand.
10036 (*tablejump<mode>_internal1): Disable for
10037 -mno-speculate-indirect-jumps.
10038 (*tablejump<mode>_internal1_nospec): New define_insn.
10039 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
10042 2018-01-16 Artyom Skrobov tyomitch@gmail.com
10044 * caller-save.c (insert_save): Drop unnecessary parameter. All
10047 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10048 Richard Biener <rguenth@suse.de>
10051 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
10052 return early, inline manually is_gimple_sizepos. Make sure if we
10053 call gimplify_expr we don't end up with a gimple constant.
10054 * tree.c (variably_modified_type_p): Don't return true for
10055 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
10056 * gimplify.h (is_gimple_sizepos): Remove.
10058 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10060 PR tree-optimization/83857
10061 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
10062 vectorizable_live_operation for pure SLP statements.
10063 (vectorizable_live_operation): Handle PHIs.
10065 2018-01-16 Richard Biener <rguenther@suse.de>
10067 PR tree-optimization/83867
10068 * tree-vect-stmts.c (vect_transform_stmt): Precompute
10069 nested_in_vect_loop_p since the scalar stmt may get invalidated.
10071 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10074 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
10075 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
10076 If off is not INTEGER_CST, issue a may not be aligned warning
10077 rather than isn't aligned. Use isn%'t rather than isn't.
10078 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
10080 <case MULT_EXPR>: Improve the case when bottom and one of the
10081 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
10082 operand, in that case check if the other operand is multiple of
10083 bottom divided by the INTEGER_CST operand.
10085 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10088 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
10089 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
10090 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
10091 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
10092 * config/pa/pa.c (pa_function_arg_advance): Likewise.
10093 (pa_function_arg, pa_arg_partial_bytes): Likewise.
10094 (pa_function_arg_size): New function.
10096 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10098 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
10099 in a separate statement.
10101 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10103 PR tree-optimization/83847
10104 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
10105 group gathers and scatters.
10107 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10109 PR rtl-optimization/86620
10110 * params.def (max-sched-ready-insns): Bump minimum value to 1.
10112 PR rtl-optimization/83213
10113 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
10114 to last if both are JUMP_INSNs.
10116 PR tree-optimization/83843
10117 * gimple-ssa-store-merging.c
10118 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
10119 store_immediate_info for bswap/nop orig_stores.
10121 2018-01-15 Andrew Waterman <andrew@sifive.com>
10123 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
10125 <UDIV>: Increase cost if !TARGET_DIV.
10127 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
10129 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
10130 (define_attr "cr_logical_3op"): New.
10131 (cceq_ior_compare): Adjust.
10132 (cceq_ior_compare_complement): Adjust.
10133 (*cceq_rev_compare): Adjust.
10134 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
10135 (is_cracked_insn): Adjust.
10136 (insn_must_be_first_in_group): Adjust.
10137 * config/rs6000/40x.md: Adjust.
10138 * config/rs6000/440.md: Adjust.
10139 * config/rs6000/476.md: Adjust.
10140 * config/rs6000/601.md: Adjust.
10141 * config/rs6000/603.md: Adjust.
10142 * config/rs6000/6xx.md: Adjust.
10143 * config/rs6000/7450.md: Adjust.
10144 * config/rs6000/7xx.md: Adjust.
10145 * config/rs6000/8540.md: Adjust.
10146 * config/rs6000/cell.md: Adjust.
10147 * config/rs6000/e300c2c3.md: Adjust.
10148 * config/rs6000/e500mc.md: Adjust.
10149 * config/rs6000/e500mc64.md: Adjust.
10150 * config/rs6000/e5500.md: Adjust.
10151 * config/rs6000/e6500.md: Adjust.
10152 * config/rs6000/mpc.md: Adjust.
10153 * config/rs6000/power4.md: Adjust.
10154 * config/rs6000/power5.md: Adjust.
10155 * config/rs6000/power6.md: Adjust.
10156 * config/rs6000/power7.md: Adjust.
10157 * config/rs6000/power8.md: Adjust.
10158 * config/rs6000/power9.md: Adjust.
10159 * config/rs6000/rs64.md: Adjust.
10160 * config/rs6000/titan.md: Adjust.
10162 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10164 * config/i386/predicates.md (indirect_branch_operand): Rewrite
10165 ix86_indirect_branch_register logic.
10167 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10169 * config/i386/constraints.md (Bs): Update
10170 ix86_indirect_branch_register check. Don't check
10171 ix86_indirect_branch_register with GOT_memory_operand.
10173 * config/i386/predicates.md (GOT_memory_operand): Don't check
10174 ix86_indirect_branch_register here.
10175 (GOT32_symbol_operand): Likewise.
10177 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10179 * config/i386/predicates.md (constant_call_address_operand):
10180 Rewrite ix86_indirect_branch_register logic.
10181 (sibcall_insn_operand): Likewise.
10183 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10185 * config/i386/constraints.md (Bs): Replace
10186 ix86_indirect_branch_thunk_register with
10187 ix86_indirect_branch_register.
10189 * config/i386/i386.md (indirect_jump): Likewise.
10190 (tablejump): Likewise.
10191 (*sibcall_memory): Likewise.
10192 (*sibcall_value_memory): Likewise.
10193 Peepholes of indirect call and jump via memory: Likewise.
10194 * config/i386/i386.opt: Likewise.
10195 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10196 (GOT_memory_operand): Likewise.
10197 (call_insn_operand): Likewise.
10198 (sibcall_insn_operand): Likewise.
10199 (GOT32_symbol_operand): Likewise.
10201 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10203 PR middle-end/83837
10204 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
10205 type rather than type addr's type points to.
10206 (expand_omp_atomic_mutex): Likewise.
10207 (expand_omp_atomic): Likewise.
10209 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10212 * config/i386/i386.c (output_indirect_thunk_function): Use
10213 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
10214 for __x86_return_thunk.
10216 2018-01-15 Richard Biener <rguenther@suse.de>
10218 PR middle-end/83850
10219 * expmed.c (extract_bit_field_1): Fix typo.
10221 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10224 * config/arm/iterators.md (VF): New mode iterator.
10225 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
10226 Remove integer-related logic from pattern.
10227 (neon_vabd<mode>_3): Likewise.
10229 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10231 PR middle-end/82694
10232 * common.opt (fstrict-overflow): No longer an alias.
10233 (fwrapv-pointer): New option.
10234 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
10235 also for pointer types based on flag_wrapv_pointer.
10236 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
10237 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
10238 opts->x_flag_wrapv got set.
10239 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
10240 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
10241 POINTER_TYPE_OVERFLOW_UNDEFINED.
10242 * match.pd: Likewise in address comparison pattern.
10243 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
10245 2018-01-15 Richard Biener <rguenther@suse.de>
10248 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
10249 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
10250 Reset type names to their identifier if their TYPE_DECL doesn't
10251 have linkage (and thus is used for ODR and devirt).
10252 (save_debug_info_for_decl): Remove.
10253 (save_debug_info_for_type): Likewise.
10254 (add_tree_to_fld_list): Adjust.
10255 * tree-pretty-print.c (dump_generic_node): Make dumping of
10256 type names more robust.
10258 2018-01-15 Richard Biener <rguenther@suse.de>
10260 * BASE-VER: Bump to 8.0.1.
10262 2018-01-14 Martin Sebor <msebor@redhat.com>
10265 * builtins.c (check_access): Avoid warning when the no-warning bit
10268 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
10270 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
10271 * ira-color (allocno_hard_regs_compare): Likewise.
10273 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
10276 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
10277 Use .pushsection/.popsection.
10279 2018-01-14 Martin Sebor <msebor@redhat.com>
10282 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
10284 2018-01-14 Jakub Jelinek <jakub@redhat.com>
10286 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
10287 entry from extra_headers.
10288 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
10289 extra_headers, make the list bitwise identical to the i?86-*-* one.
10291 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10293 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
10294 -mcmodel=large with -mindirect-branch=thunk,
10295 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
10296 -mfunction-return=thunk-extern.
10297 * doc/invoke.texi: Document -mcmodel=large is incompatible with
10298 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
10299 -mfunction-return=thunk and -mfunction-return=thunk-extern.
10301 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10303 * config/i386/i386.c (print_reg): Print the name of the full
10304 integer register without '%'.
10305 (ix86_print_operand): Handle 'V'.
10306 * doc/extend.texi: Document 'V' modifier.
10308 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10310 * config/i386/constraints.md (Bs): Disallow memory operand for
10311 -mindirect-branch-register.
10313 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10314 (GOT_memory_operand): Likewise.
10315 (call_insn_operand): Likewise.
10316 (sibcall_insn_operand): Likewise.
10317 (GOT32_symbol_operand): Likewise.
10318 * config/i386/i386.md (indirect_jump): Call convert_memory_address
10319 for -mindirect-branch-register.
10320 (tablejump): Likewise.
10321 (*sibcall_memory): Likewise.
10322 (*sibcall_value_memory): Likewise.
10323 Disallow peepholes of indirect call and jump via memory for
10324 -mindirect-branch-register.
10325 (*call_pop): Replace m with Bw.
10326 (*call_value_pop): Likewise.
10327 (*sibcall_pop_memory): Replace m with Bs.
10328 * config/i386/i386.opt (mindirect-branch-register): New option.
10329 * doc/invoke.texi: Document -mindirect-branch-register option.
10331 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10333 * config/i386/i386-protos.h (ix86_output_function_return): New.
10334 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
10335 set function_return_type.
10336 (indirect_thunk_name): Add ret_p to indicate thunk for function
10338 (output_indirect_thunk_function): Pass false to
10339 indirect_thunk_name.
10340 (ix86_output_indirect_branch_via_reg): Likewise.
10341 (ix86_output_indirect_branch_via_push): Likewise.
10342 (output_indirect_thunk_function): Create alias for function
10343 return thunk if regno < 0.
10344 (ix86_output_function_return): New function.
10345 (ix86_handle_fndecl_attribute): Handle function_return.
10346 (ix86_attribute_table): Add function_return.
10347 * config/i386/i386.h (machine_function): Add
10348 function_return_type.
10349 * config/i386/i386.md (simple_return_internal): Use
10350 ix86_output_function_return.
10351 (simple_return_internal_long): Likewise.
10352 * config/i386/i386.opt (mfunction-return=): New option.
10353 (indirect_branch): Mention -mfunction-return=.
10354 * doc/extend.texi: Document function_return function attribute.
10355 * doc/invoke.texi: Document -mfunction-return= option.
10357 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10359 * config/i386/i386-opts.h (indirect_branch): New.
10360 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
10361 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
10362 with local indirect jump when converting indirect call and jump.
10363 (ix86_set_indirect_branch_type): New.
10364 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
10365 (indirectlabelno): New.
10366 (indirect_thunk_needed): Likewise.
10367 (indirect_thunk_bnd_needed): Likewise.
10368 (indirect_thunks_used): Likewise.
10369 (indirect_thunks_bnd_used): Likewise.
10370 (INDIRECT_LABEL): Likewise.
10371 (indirect_thunk_name): Likewise.
10372 (output_indirect_thunk): Likewise.
10373 (output_indirect_thunk_function): Likewise.
10374 (ix86_output_indirect_branch_via_reg): Likewise.
10375 (ix86_output_indirect_branch_via_push): Likewise.
10376 (ix86_output_indirect_branch): Likewise.
10377 (ix86_output_indirect_jmp): Likewise.
10378 (ix86_code_end): Call output_indirect_thunk_function if needed.
10379 (ix86_output_call_insn): Call ix86_output_indirect_branch if
10381 (ix86_handle_fndecl_attribute): Handle indirect_branch.
10382 (ix86_attribute_table): Add indirect_branch.
10383 * config/i386/i386.h (machine_function): Add indirect_branch_type
10384 and has_local_indirect_jump.
10385 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
10387 (tablejump): Likewise.
10388 (*indirect_jump): Use ix86_output_indirect_jmp.
10389 (*tablejump_1): Likewise.
10390 (simple_return_indirect_internal): Likewise.
10391 * config/i386/i386.opt (mindirect-branch=): New option.
10392 (indirect_branch): New.
10395 (thunk-inline): Likewise.
10396 (thunk-extern): Likewise.
10397 * doc/extend.texi: Document indirect_branch function attribute.
10398 * doc/invoke.texi: Document -mindirect-branch= option.
10400 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
10403 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
10405 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
10407 * ipa-inline.c (want_inline_small_function_p): Return false if
10408 inlining has already failed with CIF_FINAL_ERROR.
10409 (update_caller_keys): Call want_inline_small_function_p before
10411 (update_callee_keys): Likewise.
10413 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
10415 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
10417 (rs6000_quadword_masked_address_p): Likewise.
10418 (quad_aligned_load_p): Likewise.
10419 (quad_aligned_store_p): Likewise.
10420 (const_load_sequence_p): Add comment to describe the outer-most loop.
10421 (mimic_memory_attributes_and_flags): New function.
10422 (rs6000_gen_stvx): Likewise.
10423 (replace_swapped_aligned_store): Likewise.
10424 (rs6000_gen_lvx): Likewise.
10425 (replace_swapped_aligned_load): Likewise.
10426 (replace_swapped_load_constant): Capitalize argument name in
10427 comment describing this function.
10428 (rs6000_analyze_swaps): Add a third pass to search for vector loads
10429 and stores that access quad-word aligned addresses and replace
10430 with stvx or lvx instructions when appropriate.
10431 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
10432 New function prototype.
10433 (rs6000_quadword_masked_address_p): Likewise.
10434 (rs6000_gen_lvx): Likewise.
10435 (rs6000_gen_stvx): Likewise.
10436 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
10437 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
10438 when memory address is aligned.
10439 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
10440 this split to select lvx instruction when memory address is aligned.
10441 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
10442 instruction when memory address is aligned.
10443 (*vsx_le_perm_load_v16qi): Likewise.
10444 (four unnamed splitters): Modify to select the stvx instruction
10445 when memory is aligned.
10447 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
10449 * predict.c (determine_unlikely_bbs): Handle correctly BBs
10450 which appears in the queue multiple times.
10452 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10453 Alan Hayward <alan.hayward@arm.com>
10454 David Sherwood <david.sherwood@arm.com>
10456 * tree-vectorizer.h (vec_lower_bound): New structure.
10457 (_loop_vec_info): Add check_nonzero and lower_bounds.
10458 (LOOP_VINFO_CHECK_NONZERO): New macro.
10459 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
10460 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
10461 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
10462 fields. Make seg_len the distance travelled, not including the
10464 (dr_direction_indicator): Declare.
10465 (dr_zero_step_indicator): Likewise.
10466 (dr_known_forward_stride_p): Likewise.
10467 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
10469 (runtime_alias_check_p): Allow runtime alias checks with
10471 (operator ==): Compare access_size and align.
10472 (prune_runtime_alias_test_list): Rework for new distinction between
10473 the access_size and seg_len.
10474 (create_intersect_range_checks_index): Likewise. Cope with polynomial
10476 (get_segment_min_max): New function.
10477 (create_intersect_range_checks): Use it.
10478 (dr_step_indicator): New function.
10479 (dr_direction_indicator): Likewise.
10480 (dr_zero_step_indicator): Likewise.
10481 (dr_known_forward_stride_p): Likewise.
10482 * tree-loop-distribution.c (data_ref_segment_size): Return
10483 DR_STEP * (niters - 1).
10484 (compute_alias_check_pairs): Update call to the dr_with_seg_len
10486 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
10487 (vect_preserves_scalar_order_p): New function, split out from...
10488 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
10489 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
10490 (vect_vfa_access_size): New function.
10491 (vect_vfa_align): Likewise.
10492 (vect_compile_time_alias): Take access_size_a and access_b arguments.
10493 (dump_lower_bound): New function.
10494 (vect_check_lower_bound): Likewise.
10495 (vect_small_gap_p): Likewise.
10496 (vectorizable_with_step_bound_p): Likewise.
10497 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
10498 depencies if the vectorization factor is 1. Convert the checks
10499 for nonzero steps into checks on the bounds of DR_STEP. Try using
10500 a bunds check for variable steps if the minimum required step is
10501 relatively small. Update calls to the dr_with_seg_len
10502 constructor and to vect_compile_time_alias.
10503 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
10505 (vect_loop_versioning): Call it.
10506 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
10508 (vect_estimate_min_profitable_iters): Account for any bounds checks.
10510 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10511 Alan Hayward <alan.hayward@arm.com>
10512 David Sherwood <david.sherwood@arm.com>
10514 * doc/sourcebuild.texi (vect_scatter_store): Document.
10515 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
10517 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
10519 * genopinit.c (main): Add supports_vec_scatter_store and
10520 supports_vec_scatter_store_cached to target_optabs.
10521 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
10522 IFN_MASK_SCATTER_STORE.
10523 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
10525 * internal-fn.h (internal_store_fn_p): Declare.
10526 (internal_fn_stored_value_index): Likewise.
10527 * internal-fn.c (scatter_store_direct): New macro.
10528 (expand_scatter_store_optab_fn): New function.
10529 (direct_scatter_store_optab_supported_p): New macro.
10530 (internal_store_fn_p): New function.
10531 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
10532 IFN_MASK_SCATTER_STORE.
10533 (internal_fn_mask_index): Likewise.
10534 (internal_fn_stored_value_index): New function.
10535 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
10536 for scatter stores.
10537 * optabs-query.h (supports_vec_scatter_store_p): Declare.
10538 * optabs-query.c (supports_vec_scatter_store_p): New function.
10539 * tree-vectorizer.h (vect_get_store_rhs): Declare.
10540 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
10541 true for scatter stores.
10542 (vect_gather_scatter_fn_p): Handle scatter stores too.
10543 (vect_check_gather_scatter): Consider using scatter stores if
10544 supports_vec_scatter_store_p.
10545 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
10546 scatter stores too.
10547 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10548 internal_fn_stored_value_index.
10549 (check_load_store_masking): Handle scatter stores too.
10550 (vect_get_store_rhs): Make public.
10551 (vectorizable_call): Use internal_store_fn_p.
10552 (vectorizable_store): Handle scatter store internal functions.
10553 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
10554 when deciding whether the end of the group has been reached.
10555 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
10556 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
10557 (mask_scatter_store<mode>): New insns.
10559 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10560 Alan Hayward <alan.hayward@arm.com>
10561 David Sherwood <david.sherwood@arm.com>
10563 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
10564 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
10565 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
10567 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
10568 Use vect_truncate_gather_scatter_offset if we can't treat the
10569 operation as a normal gather load or scatter store.
10570 (get_group_load_store_type): Take the gather_scatter_info
10571 as argument. Try using a gather load or scatter store for
10572 single-element groups.
10573 (get_load_store_type): Update calls to get_group_load_store_type
10574 and vect_use_strided_gather_scatters_p.
10576 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10577 Alan Hayward <alan.hayward@arm.com>
10578 David Sherwood <david.sherwood@arm.com>
10580 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
10581 optional tree argument.
10582 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
10584 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
10585 but continue to use the current value as a fallback.
10586 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
10587 to compare the updates.
10588 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
10589 (get_load_store_type): Use it when handling a strided access.
10590 (vect_get_strided_load_store_ops): New function.
10591 (vect_get_data_ptr_increment): Likewise.
10592 (vectorizable_load): Handle strided gather loads. Always pass
10593 a step to vect_create_data_ref_ptr and bump_vector_ptr.
10595 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10596 Alan Hayward <alan.hayward@arm.com>
10597 David Sherwood <david.sherwood@arm.com>
10599 * doc/md.texi (gather_load@var{m}): Document.
10600 (mask_gather_load@var{m}): Likewise.
10601 * genopinit.c (main): Add supports_vec_gather_load and
10602 supports_vec_gather_load_cached to target_optabs.
10603 * optabs-tree.c (init_tree_optimization_optabs): Use
10604 ggc_cleared_alloc to allocate target_optabs.
10605 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
10606 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
10608 * internal-fn.h (internal_load_fn_p): Declare.
10609 (internal_gather_scatter_fn_p): Likewise.
10610 (internal_fn_mask_index): Likewise.
10611 (internal_gather_scatter_fn_supported_p): Likewise.
10612 * internal-fn.c (gather_load_direct): New macro.
10613 (expand_gather_load_optab_fn): New function.
10614 (direct_gather_load_optab_supported_p): New macro.
10615 (direct_internal_fn_optab): New function.
10616 (internal_load_fn_p): Likewise.
10617 (internal_gather_scatter_fn_p): Likewise.
10618 (internal_fn_mask_index): Likewise.
10619 (internal_gather_scatter_fn_supported_p): Likewise.
10620 * optabs-query.c (supports_at_least_one_mode_p): New function.
10621 (supports_vec_gather_load_p): Likewise.
10622 * optabs-query.h (supports_vec_gather_load_p): Declare.
10623 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
10624 and memory_type field.
10625 (NUM_PATTERNS): Bump to 15.
10626 * tree-vect-data-refs.c: Include internal-fn.h.
10627 (vect_gather_scatter_fn_p): New function.
10628 (vect_describe_gather_scatter_call): Likewise.
10629 (vect_check_gather_scatter): Try using internal functions for
10630 gather loads. Recognize existing calls to a gather load function.
10631 (vect_analyze_data_refs): Consider using gather loads if
10632 supports_vec_gather_load_p.
10633 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
10634 (vect_get_gather_scatter_offset_type): Likewise.
10635 (vect_convert_mask_for_vectype): Likewise.
10636 (vect_add_conversion_to_patterm): Likewise.
10637 (vect_try_gather_scatter_pattern): Likewise.
10638 (vect_recog_gather_scatter_pattern): New pattern recognizer.
10639 (vect_vect_recog_func_ptrs): Add it.
10640 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10641 internal_fn_mask_index and internal_gather_scatter_fn_p.
10642 (check_load_store_masking): Take the gather_scatter_info as an
10643 argument and handle gather loads.
10644 (vect_get_gather_scatter_ops): New function.
10645 (vectorizable_call): Check internal_load_fn_p.
10646 (vectorizable_load): Likewise. Handle gather load internal
10648 (vectorizable_store): Update call to check_load_store_masking.
10649 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
10650 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
10651 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
10652 (aarch64_gather_scale_operand_d): New predicates.
10653 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
10654 (mask_gather_load<mode>): New insns.
10656 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10657 Alan Hayward <alan.hayward@arm.com>
10658 David Sherwood <david.sherwood@arm.com>
10660 * optabs.def (fold_left_plus_optab): New optab.
10661 * doc/md.texi (fold_left_plus_@var{m}): Document.
10662 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
10663 * internal-fn.c (fold_left_direct): Define.
10664 (expand_fold_left_optab_fn): Likewise.
10665 (direct_fold_left_optab_supported_p): Likewise.
10666 * fold-const-call.c (fold_const_fold_left): New function.
10667 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
10668 * tree-parloops.c (valid_reduction_p): New function.
10669 (gather_scalar_reductions): Use it.
10670 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
10671 (vect_finish_replace_stmt): Declare.
10672 * tree-vect-loop.c (fold_left_reduction_fn): New function.
10673 (needs_fold_left_reduction_p): New function, split out from...
10674 (vect_is_simple_reduction): ...here. Accept reductions that
10675 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
10676 (vect_force_simple_reduction): Also store the reduction type in
10677 the assignment's STMT_VINFO_REDUC_TYPE.
10678 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
10679 (merge_with_identity): New function.
10680 (vect_expand_fold_left): Likewise.
10681 (vectorize_fold_left_reduction): Likewise.
10682 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
10683 scalar phi in place for it. Check for target support and reject
10684 cases that would reassociate the operation. Defer the transform
10685 phase to vectorize_fold_left_reduction.
10686 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
10687 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
10688 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
10690 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10692 * tree-if-conv.c (predicate_mem_writes): Remove redundant
10693 call to ifc_temp_var.
10695 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10696 Alan Hayward <alan.hayward@arm.com>
10697 David Sherwood <david.sherwood@arm.com>
10699 * target.def (legitimize_address_displacement): Take the original
10700 offset as a poly_int.
10701 * targhooks.h (default_legitimize_address_displacement): Update
10703 * targhooks.c (default_legitimize_address_displacement): Likewise.
10704 * doc/tm.texi: Regenerate.
10705 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
10706 as an argument, moving assert of ad->disp == ad->disp_term to...
10707 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
10708 Try calling targetm.legitimize_address_displacement before expanding
10709 the address rather than afterwards, and adjust for the new interface.
10710 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
10711 Match the new hook interface. Handle SVE addresses.
10712 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
10713 new hook interface.
10715 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10717 * Makefile.in (OBJS): Add early-remat.o.
10718 * target.def (select_early_remat_modes): New hook.
10719 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
10720 * doc/tm.texi: Regenerate.
10721 * targhooks.h (default_select_early_remat_modes): Declare.
10722 * targhooks.c (default_select_early_remat_modes): New function.
10723 * timevar.def (TV_EARLY_REMAT): New timevar.
10724 * passes.def (pass_early_remat): New pass.
10725 * tree-pass.h (make_pass_early_remat): Declare.
10726 * early-remat.c: New file.
10727 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
10729 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
10731 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10732 Alan Hayward <alan.hayward@arm.com>
10733 David Sherwood <david.sherwood@arm.com>
10735 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
10736 vfm1 with a bound_epilog parameter.
10737 (vect_do_peeling): Update calls accordingly, and move the prologue
10738 call earlier in the function. Treat the base bound_epilog as 0 for
10739 fully-masked loops and retain vf - 1 for other loops. Add 1 to
10740 this base when peeling for gaps.
10741 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
10742 with fully-masked loops.
10743 (vect_estimate_min_profitable_iters): Handle the single peeled
10744 iteration in that case.
10746 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10747 Alan Hayward <alan.hayward@arm.com>
10748 David Sherwood <david.sherwood@arm.com>
10750 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
10751 single-element interleaving even if the size is not a power of 2.
10752 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
10753 accesses for single-element interleaving if the group size is
10756 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10757 Alan Hayward <alan.hayward@arm.com>
10758 David Sherwood <david.sherwood@arm.com>
10760 * doc/md.texi (fold_extract_last_@var{m}): Document.
10761 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
10762 * optabs.def (fold_extract_last_optab): New optab.
10763 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
10764 * internal-fn.c (fold_extract_direct): New macro.
10765 (expand_fold_extract_optab_fn): Likewise.
10766 (direct_fold_extract_optab_supported_p): Likewise.
10767 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
10768 * tree-vect-loop.c (vect_model_reduction_cost): Handle
10769 EXTRACT_LAST_REDUCTION.
10770 (get_initial_def_for_reduction): Do not create an initial vector
10771 for EXTRACT_LAST_REDUCTION reductions.
10772 (vectorizable_reduction): Leave the scalar phi in place for
10773 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
10774 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
10775 epilogue code for EXTRACT_LAST_REDUCTION and defer the
10776 transform phase to vectorizable_condition.
10777 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
10779 (vect_finish_stmt_generation): ...here.
10780 (vect_finish_replace_stmt): New function.
10781 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
10782 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
10784 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
10786 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10787 Alan Hayward <alan.hayward@arm.com>
10788 David Sherwood <david.sherwood@arm.com>
10790 * doc/md.texi (extract_last_@var{m}): Document.
10791 * optabs.def (extract_last_optab): New optab.
10792 * internal-fn.def (EXTRACT_LAST): New internal function.
10793 * internal-fn.c (cond_unary_direct): New macro.
10794 (expand_cond_unary_optab_fn): Likewise.
10795 (direct_cond_unary_optab_supported_p): Likewise.
10796 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
10797 loops using EXTRACT_LAST.
10798 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
10799 (extract_last_<mode>): ...this optab.
10800 (vec_extract<mode><Vel>): Update accordingly.
10802 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10803 Alan Hayward <alan.hayward@arm.com>
10804 David Sherwood <david.sherwood@arm.com>
10806 * target.def (empty_mask_is_expensive): New hook.
10807 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
10808 * doc/tm.texi: Regenerate.
10809 * targhooks.h (default_empty_mask_is_expensive): Declare.
10810 * targhooks.c (default_empty_mask_is_expensive): New function.
10811 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
10812 if the target says that empty masks are expensive.
10813 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
10815 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
10817 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10818 Alan Hayward <alan.hayward@arm.com>
10819 David Sherwood <david.sherwood@arm.com>
10821 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
10822 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
10823 (vect_use_loop_mask_for_alignment_p): New function.
10824 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
10825 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
10826 niters_skip argument. Make sure that the first niters_skip elements
10827 of the first iteration are inactive.
10828 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
10829 Update call to vect_set_loop_masks_directly.
10830 (get_misalign_in_elems): New function, split out from...
10831 (vect_gen_prolog_loop_niters): ...here.
10832 (vect_update_init_of_dr): Take a code argument that specifies whether
10833 the adjustment should be added or subtracted.
10834 (vect_update_init_of_drs): Likewise.
10835 (vect_prepare_for_masked_peels): New function.
10836 (vect_do_peeling): Skip prologue peeling if we're using a mask
10837 instead. Update call to vect_update_inits_of_drs.
10838 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10840 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
10841 alignment. Do not include the number of peeled iterations in
10842 the minimum threshold in that case.
10843 (vectorizable_induction): Adjust the start value down by
10844 LOOP_VINFO_MASK_SKIP_NITERS iterations.
10845 (vect_transform_loop): Call vect_prepare_for_masked_peels.
10846 Take the number of skipped iterations into account when calculating
10848 * tree-vect-stmts.c (vect_gen_while_not): New function.
10850 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10851 Alan Hayward <alan.hayward@arm.com>
10852 David Sherwood <david.sherwood@arm.com>
10854 * doc/sourcebuild.texi (vect_fully_masked): Document.
10855 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
10856 default value to 0.
10857 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
10859 (vect_analyze_loop_2): ...here. Don't check the vectorization
10860 factor against the number of loop iterations if the loop is
10863 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10864 Alan Hayward <alan.hayward@arm.com>
10865 David Sherwood <david.sherwood@arm.com>
10867 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
10868 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
10869 (dump_groups): Update accordingly.
10870 (iv_use::mem_type): New member variable.
10871 (address_p): New function.
10872 (record_use): Add a mem_type argument and initialize the new
10874 (record_group_use): Add a mem_type argument. Use address_p.
10875 Remove obsolete null checks of base_object. Update call to record_use.
10876 (find_interesting_uses_op): Update call to record_group_use.
10877 (find_interesting_uses_cond): Likewise.
10878 (find_interesting_uses_address): Likewise.
10879 (get_mem_type_for_internal_fn): New function.
10880 (find_address_like_use): Likewise.
10881 (find_interesting_uses_stmt): Try find_address_like_use before
10882 calling find_interesting_uses_op.
10883 (addr_offset_valid_p): Use the iv mem_type field as the type
10884 of the addressed memory.
10885 (add_autoinc_candidates): Likewise.
10886 (get_address_cost): Likewise.
10887 (split_small_address_groups_p): Use address_p.
10888 (split_address_groups): Likewise.
10889 (add_iv_candidate_for_use): Likewise.
10890 (autoinc_possible_for_pair): Likewise.
10891 (rewrite_groups): Likewise.
10892 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
10893 (determine_group_iv_cost): Update after split of USE_ADDRESS.
10894 (get_alias_ptr_type_for_ptr_address): New function.
10895 (rewrite_use_address): Rewrite address uses in calls that were
10896 identified by find_address_like_use.
10898 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10899 Alan Hayward <alan.hayward@arm.com>
10900 David Sherwood <david.sherwood@arm.com>
10902 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
10904 * gimple-expr.h (is_gimple_addressable: Likewise.
10905 * gimple-expr.c (is_gimple_address): Likewise.
10906 * internal-fn.c (expand_call_mem_ref): New function.
10907 (expand_mask_load_optab_fn): Use it.
10908 (expand_mask_store_optab_fn): Likewise.
10910 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10911 Alan Hayward <alan.hayward@arm.com>
10912 David Sherwood <david.sherwood@arm.com>
10914 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
10915 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
10916 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
10917 (cond_umax@var{mode}): Document.
10918 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
10919 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
10920 (cond_umin_optab, cond_umax_optab): New optabs.
10921 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
10922 (COND_IOR, COND_XOR): New internal functions.
10923 * internal-fn.h (get_conditional_internal_fn): Declare.
10924 * internal-fn.c (cond_binary_direct): New macro.
10925 (expand_cond_binary_optab_fn): Likewise.
10926 (direct_cond_binary_optab_supported_p): Likewise.
10927 (get_conditional_internal_fn): New function.
10928 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
10929 Cope with reduction statements that are vectorized as calls rather
10931 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
10932 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
10933 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
10934 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
10935 (UNSPEC_COND_EOR): New unspecs.
10936 (optab): Add mappings for them.
10937 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
10938 (sve_int_op, sve_fp_op): New int attributes.
10940 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10941 Alan Hayward <alan.hayward@arm.com>
10942 David Sherwood <david.sherwood@arm.com>
10944 * optabs.def (while_ult_optab): New optab.
10945 * doc/md.texi (while_ult@var{m}@var{n}): Document.
10946 * internal-fn.def (WHILE_ULT): New internal function.
10947 * internal-fn.h (direct_internal_fn_supported_p): New override
10948 that takes two types as argument.
10949 * internal-fn.c (while_direct): New macro.
10950 (expand_while_optab_fn): New function.
10951 (convert_optab_supported_p): Likewise.
10952 (direct_while_optab_supported_p): New macro.
10953 * wide-int.h (wi::udiv_ceil): New function.
10954 * tree-vectorizer.h (rgroup_masks): New structure.
10955 (vec_loop_masks): New typedef.
10956 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
10957 and fully_masked_p.
10958 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
10959 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
10960 (vect_max_vf): New function.
10961 (slpeel_make_loop_iterate_ntimes): Delete.
10962 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
10963 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
10964 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
10965 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
10966 internal-fn.h, stor-layout.h and optabs-query.h.
10967 (vect_set_loop_mask): New function.
10968 (add_preheader_seq): Likewise.
10969 (add_header_seq): Likewise.
10970 (interleave_supported_p): Likewise.
10971 (vect_maybe_permute_loop_masks): Likewise.
10972 (vect_set_loop_masks_directly): Likewise.
10973 (vect_set_loop_condition_masked): Likewise.
10974 (vect_set_loop_condition_unmasked): New function, split out from
10975 slpeel_make_loop_iterate_ntimes.
10976 (slpeel_make_loop_iterate_ntimes): Rename to..
10977 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
10978 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
10979 (vect_do_peeling): Update call accordingly.
10980 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
10982 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10983 mask_compare_type, can_fully_mask_p and fully_masked_p.
10984 (release_vec_loop_masks): New function.
10985 (_loop_vec_info): Use it to free the loop masks.
10986 (can_produce_all_loop_masks_p): New function.
10987 (vect_get_max_nscalars_per_iter): Likewise.
10988 (vect_verify_full_masking): Likewise.
10989 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
10990 retries, and free the mask rgroups before retrying. Check loop-wide
10991 reasons for disallowing fully-masked loops. Make the final decision
10992 about whether use a fully-masked loop or not.
10993 (vect_estimate_min_profitable_iters): Do not assume that peeling
10994 for the number of iterations will be needed for fully-masked loops.
10995 (vectorizable_reduction): Disable fully-masked loops.
10996 (vectorizable_live_operation): Likewise.
10997 (vect_halve_mask_nunits): New function.
10998 (vect_double_mask_nunits): Likewise.
10999 (vect_record_loop_mask): Likewise.
11000 (vect_get_loop_mask): Likewise.
11001 (vect_transform_loop): Handle the case in which the final loop
11002 iteration might handle a partial vector. Call vect_set_loop_condition
11003 instead of slpeel_make_loop_iterate_ntimes.
11004 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
11005 (check_load_store_masking): New function.
11006 (prepare_load_store_mask): Likewise.
11007 (vectorizable_store): Handle fully-masked loops.
11008 (vectorizable_load): Likewise.
11009 (supportable_widening_operation): Use vect_halve_mask_nunits for
11011 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
11012 (vect_gen_while): New function.
11013 * config/aarch64/aarch64.md (umax<mode>3): New expander.
11014 (aarch64_uqdec<mode>): New insn.
11016 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11017 Alan Hayward <alan.hayward@arm.com>
11018 David Sherwood <david.sherwood@arm.com>
11020 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
11021 (reduc_xor_scal_optab): New optabs.
11022 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
11023 (reduc_xor_scal_@var{m}): Document.
11024 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
11025 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
11026 internal functions.
11027 * fold-const-call.c (fold_const_call): Handle them.
11028 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
11029 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
11030 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
11031 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
11032 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
11033 (UNSPEC_XORV): New unspecs.
11034 (optab): Add entries for them.
11035 (BITWISEV): New int iterator.
11036 (bit_reduc_op): New int attributes.
11038 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11039 Alan Hayward <alan.hayward@arm.com>
11040 David Sherwood <david.sherwood@arm.com>
11042 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
11043 * internal-fn.def (VEC_SHL_INSERT): New internal function.
11044 * optabs.def (vec_shl_insert_optab): New optab.
11045 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
11046 (duplicate_and_interleave): Likewise.
11047 * tree-vect-loop.c: Include internal-fn.h.
11048 (neutral_op_for_slp_reduction): New function, split out from
11049 get_initial_defs_for_reduction.
11050 (get_initial_def_for_reduction): Handle option 2 for variable-length
11051 vectors by loading the neutral value into a vector and then shifting
11052 the initial value into element 0.
11053 (get_initial_defs_for_reduction): Replace the code argument with
11054 the neutral value calculated by neutral_op_for_slp_reduction.
11055 Use gimple_build_vector for constant-length vectors.
11056 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
11057 but the first group_size elements have a neutral value.
11058 Use duplicate_and_interleave otherwise.
11059 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
11060 Update call to get_initial_defs_for_reduction. Handle SLP
11061 reductions for variable-length vectors by creating one vector
11062 result for each scalar result, with the elements associated
11063 with other scalar results stubbed out with the neutral value.
11064 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
11065 Require IFN_VEC_SHL_INSERT for double reductions on
11066 variable-length vectors, or SLP reductions that have
11067 a neutral value. Require can_duplicate_and_interleave_p
11068 support for variable-length unchained SLP reductions if there
11069 is no neutral value, such as for MIN/MAX reductions. Also require
11070 the number of vector elements to be a multiple of the number of
11071 SLP statements when doing variable-length unchained SLP reductions.
11072 Update call to vect_create_epilog_for_reduction.
11073 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
11074 and remove initial values.
11075 (duplicate_and_interleave): Make public.
11076 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
11077 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
11079 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11080 Alan Hayward <alan.hayward@arm.com>
11081 David Sherwood <david.sherwood@arm.com>
11083 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
11084 (can_duplicate_and_interleave_p): New function.
11085 (vect_get_and_check_slp_defs): Take the vector of statements
11086 rather than just the current one. Remove excess parentheses.
11087 Restriction rejectinon of vect_constant_def and vect_external_def
11088 for variable-length vectors to boolean types, or types for which
11089 can_duplicate_and_interleave_p is false.
11090 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
11091 (duplicate_and_interleave): New function.
11092 (vect_get_constant_vectors): Use gimple_build_vector for
11093 constant-length vectors and suitable variable-length constant
11094 vectors. Use duplicate_and_interleave for other variable-length
11095 vectors. Don't defer the update when inserting new statements.
11097 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11098 Alan Hayward <alan.hayward@arm.com>
11099 David Sherwood <david.sherwood@arm.com>
11101 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
11102 min_profitable_iters doesn't go negative.
11104 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11105 Alan Hayward <alan.hayward@arm.com>
11106 David Sherwood <david.sherwood@arm.com>
11108 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
11109 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
11110 * optabs.def (vec_mask_load_lanes_optab): New optab.
11111 (vec_mask_store_lanes_optab): Likewise.
11112 * internal-fn.def (MASK_LOAD_LANES): New internal function.
11113 (MASK_STORE_LANES): Likewise.
11114 * internal-fn.c (mask_load_lanes_direct): New macro.
11115 (mask_store_lanes_direct): Likewise.
11116 (expand_mask_load_optab_fn): Handle masked operations.
11117 (expand_mask_load_lanes_optab_fn): New macro.
11118 (expand_mask_store_optab_fn): Handle masked operations.
11119 (expand_mask_store_lanes_optab_fn): New macro.
11120 (direct_mask_load_lanes_optab_supported_p): Likewise.
11121 (direct_mask_store_lanes_optab_supported_p): Likewise.
11122 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
11124 (vect_load_lanes_supported): Likewise.
11125 * tree-vect-data-refs.c (strip_conversion): New function.
11126 (can_group_stmts_p): Likewise.
11127 (vect_analyze_data_ref_accesses): Use it instead of checking
11128 for a pair of assignments.
11129 (vect_store_lanes_supported): Take a masked_p parameter.
11130 (vect_load_lanes_supported): Likewise.
11131 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
11132 vect_store_lanes_supported and vect_load_lanes_supported.
11133 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11134 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
11135 parameter. Don't allow gaps for masked accesses.
11136 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
11137 and vect_load_lanes_supported.
11138 (get_load_store_type): Take a masked_p parameter and update
11139 call to get_group_load_store_type.
11140 (vectorizable_store): Update call to get_load_store_type.
11141 Handle IFN_MASK_STORE_LANES.
11142 (vectorizable_load): Update call to get_load_store_type.
11143 Handle IFN_MASK_LOAD_LANES.
11145 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11146 Alan Hayward <alan.hayward@arm.com>
11147 David Sherwood <david.sherwood@arm.com>
11149 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
11151 * config/aarch64/aarch64-protos.h
11152 (aarch64_sve_struct_memory_operand_p): Declare.
11153 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
11154 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
11155 (VPRED, vpred): Handle SVE structure modes.
11156 * config/aarch64/constraints.md (Utx): New constraint.
11157 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
11158 (aarch64_sve_struct_nonimmediate_operand): New predicates.
11159 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
11160 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
11161 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
11162 structure modes. Split into pieces after RA.
11163 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
11164 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
11166 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
11167 SVE structure modes.
11168 (aarch64_classify_address): Likewise.
11169 (sizetochar): Move earlier in file.
11170 (aarch64_print_operand): Handle SVE register lists.
11171 (aarch64_array_mode): New function.
11172 (aarch64_sve_struct_memory_operand_p): Likewise.
11173 (TARGET_ARRAY_MODE): Redefine.
11175 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11176 Alan Hayward <alan.hayward@arm.com>
11177 David Sherwood <david.sherwood@arm.com>
11179 * target.def (array_mode): New target hook.
11180 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
11181 * doc/tm.texi: Regenerate.
11182 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
11183 * hooks.c (hook_optmode_mode_uhwi_none): New function.
11184 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
11185 targetm.array_mode.
11186 * stor-layout.c (mode_for_array): Likewise. Support polynomial
11189 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11190 Alan Hayward <alan.hayward@arm.com>
11191 David Sherwood <david.sherwood@arm.com>
11193 * fold-const.c (fold_binary_loc): Check the argument types
11194 rather than the result type when testing for a vector operation.
11196 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11198 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
11199 * doc/tm.texi: Regenerate.
11201 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11202 Alan Hayward <alan.hayward@arm.com>
11203 David Sherwood <david.sherwood@arm.com>
11205 * doc/invoke.texi (-msve-vector-bits=): Document new option.
11206 (sve): Document new AArch64 extension.
11207 * doc/md.texi (w): Extend the description of the AArch64
11208 constraint to include SVE vectors.
11209 (Upl, Upa): Document new AArch64 predicate constraints.
11210 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
11212 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
11213 (msve-vector-bits=): New option.
11214 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
11215 SVE when these are disabled.
11216 (sve): New extension.
11217 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
11218 modes. Adjust their number of units based on aarch64_sve_vg.
11219 (MAX_BITSIZE_MODE_ANY_MODE): Define.
11220 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
11221 aarch64_addr_query_type.
11222 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
11223 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
11224 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
11225 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
11226 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
11227 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
11228 (aarch64_simd_imm_zero_p): Delete.
11229 (aarch64_check_zero_based_sve_index_immediate): Declare.
11230 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11231 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11232 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11233 (aarch64_sve_float_mul_immediate_p): Likewise.
11234 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11235 rather than an rtx.
11236 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
11237 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
11238 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
11239 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
11240 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
11241 (aarch64_regmode_natural_size): Likewise.
11242 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
11243 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
11245 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
11246 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
11247 for VG and the SVE predicate registers.
11248 (V_ALIASES): Add a "z"-prefixed alias.
11249 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
11250 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
11251 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
11252 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
11253 (REG_CLASS_NAMES): Add entries for them.
11254 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
11255 and the predicate registers.
11256 (aarch64_sve_vg): Declare.
11257 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
11258 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
11259 (REGMODE_NATURAL_SIZE): Define.
11260 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
11262 * config/aarch64/aarch64.c: Include cfgrtl.h.
11263 (simd_immediate_info): Add a constructor for series vectors,
11264 and an associated step field.
11265 (aarch64_sve_vg): New variable.
11266 (aarch64_dbx_register_number): Handle VG and the predicate registers.
11267 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
11268 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
11269 (VEC_ANY_DATA, VEC_STRUCT): New constants.
11270 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
11271 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
11272 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
11273 (aarch64_get_mask_mode): New functions.
11274 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
11275 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
11276 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
11277 predicate modes and predicate registers. Explicitly restrict
11278 GPRs to modes of 16 bytes or smaller. Only allow FP registers
11279 to store a vector mode if it is recognized by
11280 aarch64_classify_vector_mode.
11281 (aarch64_regmode_natural_size): New function.
11282 (aarch64_hard_regno_caller_save_mode): Return the original mode
11284 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
11285 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
11286 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
11287 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
11289 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
11290 does not overlap dest if the function is frame-related. Handle
11292 (aarch64_split_add_offset): New function.
11293 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
11294 them aarch64_add_offset.
11295 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
11296 and update call to aarch64_sub_sp.
11297 (aarch64_add_cfa_expression): New function.
11298 (aarch64_expand_prologue): Pass extra temporary registers to the
11299 functions above. Handle the case in which we need to emit new
11300 DW_CFA_expressions for registers that were originally saved
11301 relative to the stack pointer, but now have to be expressed
11302 relative to the frame pointer.
11303 (aarch64_output_mi_thunk): Pass extra temporary registers to the
11305 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
11306 IP0 and IP1 values for SVE frames.
11307 (aarch64_expand_vec_series): New function.
11308 (aarch64_expand_sve_widened_duplicate): Likewise.
11309 (aarch64_expand_sve_const_vector): Likewise.
11310 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
11311 Handle SVE constants. Use emit_move_insn to move a force_const_mem
11312 into the register, rather than emitting a SET directly.
11313 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
11314 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
11315 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
11316 (offset_9bit_signed_scaled_p): New functions.
11317 (aarch64_replicate_bitmask_imm): New function.
11318 (aarch64_bitmask_imm): Use it.
11319 (aarch64_cannot_force_const_mem): Reject expressions involving
11320 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
11321 (aarch64_classify_index): Handle SVE indices, by requiring
11322 a plain register index with a scale that matches the element size.
11323 (aarch64_classify_address): Handle SVE addresses. Assert that
11324 the mode of the address is VOIDmode or an integer mode.
11325 Update call to aarch64_classify_symbol.
11326 (aarch64_classify_symbolic_expression): Update call to
11327 aarch64_classify_symbol.
11328 (aarch64_const_vec_all_in_range_p): New function.
11329 (aarch64_print_vector_float_operand): Likewise.
11330 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
11331 "vN" for FP registers with SVE modes. Handle (const ...) vectors
11332 and the FP immediates 1.0 and 0.5.
11333 (aarch64_print_address_internal): Handle SVE addresses.
11334 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
11335 (aarch64_regno_regclass): Handle predicate registers.
11336 (aarch64_secondary_reload): Handle big-endian reloads of SVE
11338 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
11339 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
11340 (aarch64_convert_sve_vector_bits): New function.
11341 (aarch64_override_options): Use it to handle -msve-vector-bits=.
11342 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11343 rather than an rtx.
11344 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
11345 Handle SVE vector and predicate modes. Accept VL-based constants
11346 that need only one temporary register, and VL offsets that require
11347 no temporary registers.
11348 (aarch64_conditional_register_usage): Mark the predicate registers
11349 as fixed if SVE isn't available.
11350 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
11351 Return true for SVE vector and predicate modes.
11352 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
11353 rather than an unsigned int. Handle SVE modes.
11354 (aarch64_preferred_simd_mode): Update call accordingly. Handle
11356 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
11358 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11359 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11360 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11361 (aarch64_sve_float_mul_immediate_p): New functions.
11362 (aarch64_sve_valid_immediate): New function.
11363 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
11364 Explicitly reject structure modes. Check for INDEX constants.
11365 Handle PTRUE and PFALSE constants.
11366 (aarch64_check_zero_based_sve_index_immediate): New function.
11367 (aarch64_simd_imm_zero_p): Delete.
11368 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
11369 vector modes. Accept constants in the range of CNT[BHWD].
11370 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
11371 ask for an Advanced SIMD mode.
11372 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
11373 (aarch64_simd_vector_alignment): Handle SVE predicates.
11374 (aarch64_vectorize_preferred_vector_alignment): New function.
11375 (aarch64_simd_vector_alignment_reachable): Use it instead of
11377 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
11378 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
11380 (MAX_VECT_LEN): Delete.
11381 (expand_vec_perm_d): Add a vec_flags field.
11382 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
11383 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
11384 (aarch64_evpc_ext): Don't apply a big-endian lane correction
11386 (aarch64_evpc_rev): Rename to...
11387 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
11388 (aarch64_evpc_rev_global): New function.
11389 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
11390 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
11392 (aarch64_evpc_sve_tbl): New function.
11393 (aarch64_expand_vec_perm_const_1): Update after rename of
11394 aarch64_evpc_rev. Handle SVE permutes too, trying
11395 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
11396 than aarch64_evpc_tbl.
11397 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
11398 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
11399 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
11400 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
11401 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
11402 (aarch64_expand_sve_vcond): New functions.
11403 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
11404 of aarch64_vector_mode_p.
11405 (aarch64_dwarf_poly_indeterminate_value): New function.
11406 (aarch64_compute_pressure_classes): Likewise.
11407 (aarch64_can_change_mode_class): Likewise.
11408 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
11409 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
11410 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
11411 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
11412 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
11413 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
11414 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
11415 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
11417 (Dn, Dl, Dr): Accept const as well as const_vector.
11418 (Dz): Likewise. Compare against CONST0_RTX.
11419 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
11420 of "vector" where appropriate.
11421 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
11422 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
11423 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
11424 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
11425 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
11426 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
11427 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
11428 (v_int_equiv): Extend to SVE modes.
11429 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
11431 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
11432 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
11433 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
11434 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
11435 (SVE_COND_FP_CMP): New int iterators.
11436 (perm_hilo): Handle the new unpack unspecs.
11437 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
11439 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
11440 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
11441 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
11442 (aarch64_equality_operator, aarch64_constant_vector_operand)
11443 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
11444 (aarch64_sve_nonimmediate_operand): Likewise.
11445 (aarch64_sve_general_operand): Likewise.
11446 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
11447 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
11448 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
11449 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
11450 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
11451 (aarch64_sve_float_arith_immediate): Likewise.
11452 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
11453 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
11454 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
11455 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
11456 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
11457 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
11458 (aarch64_sve_float_arith_operand): Likewise.
11459 (aarch64_sve_float_arith_with_sub_operand): Likewise.
11460 (aarch64_sve_float_mul_operand): Likewise.
11461 (aarch64_sve_vec_perm_operand): Likewise.
11462 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
11463 (aarch64_mov_operand): Accept const_poly_int and const_vector.
11464 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
11465 as well as const_vector.
11466 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
11467 in file. Use CONST0_RTX and CONSTM1_RTX.
11468 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
11469 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
11470 Use aarch64_simd_imm_zero.
11471 * config/aarch64/aarch64-sve.md: New file.
11472 * config/aarch64/aarch64.md: Include it.
11473 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
11474 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
11475 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
11476 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
11477 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
11478 (sve): New attribute.
11479 (enabled): Disable instructions with the sve attribute unless
11481 (movqi, movhi): Pass CONST_POLY_INT operaneds through
11482 aarch64_expand_mov_immediate.
11483 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
11484 CNT[BHSD] immediates.
11485 (movti): Split CONST_POLY_INT moves into two halves.
11486 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
11487 Split additions that need a temporary here if the destination
11488 is the stack pointer.
11489 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
11490 (*add<mode>3_poly_1): New instruction.
11491 (set_clobber_cc): New expander.
11493 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11495 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
11496 parameter and use it instead of GET_MODE_SIZE (innermode). Use
11497 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
11498 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
11499 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
11500 Change innermode from fixed_mode_size to machine_mode.
11501 (simplify_subreg): Update call accordingly. Handle a constant-sized
11502 subreg of a variable-length CONST_VECTOR.
11504 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11505 Alan Hayward <alan.hayward@arm.com>
11506 David Sherwood <david.sherwood@arm.com>
11508 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
11509 (add_offset_to_base): New function, split out from...
11510 (create_mem_ref): ...here. When handling a scale other than 1,
11511 check first whether the address is valid without the offset.
11512 Add it into the base if so, leaving the index and scale as-is.
11514 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11517 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
11518 fold_for_warn before checking if arg2 is INTEGER_CST.
11520 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
11522 * config/rs6000/predicates.md (load_multiple_operation): Delete.
11523 (store_multiple_operation): Delete.
11524 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
11525 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
11526 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
11527 guarded by TARGET_STRING.
11528 (rs6000_output_load_multiple): Delete.
11529 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
11530 OPTION_MASK_STRING / TARGET_STRING handling.
11531 (print_operand) <'N', 'O'>: Add comment that these are unused now.
11532 (const rs6000_opt_masks) <"string">: Change mask to 0.
11533 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
11534 (MASK_STRING): Delete.
11535 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
11537 (load_multiple): Delete.
11544 (store_multiple): Delete.
11551 (movmemsi_8reg): Delete.
11552 (corresponding unnamed define_insn): Delete.
11553 (movmemsi_6reg): Delete.
11554 (corresponding unnamed define_insn): Delete.
11555 (movmemsi_4reg): Delete.
11556 (corresponding unnamed define_insn): Delete.
11557 (movmemsi_2reg): Delete.
11558 (corresponding unnamed define_insn): Delete.
11559 (movmemsi_1reg): Delete.
11560 (corresponding unnamed define_insn): Delete.
11561 * config/rs6000/rs6000.opt (mno-string): New.
11562 (mstring): Replace by deprecation warning stub.
11563 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
11565 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11567 * regrename.c (regrename_do_replace): If replacing the same
11568 reg multiple times, try to reuse last created gen_raw_REG.
11571 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
11572 main to workaround a bug in GDB.
11574 2018-01-12 Tom de Vries <tom@codesourcery.com>
11577 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
11579 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
11581 PR rtl-optimization/80481
11582 * ira-color.c (get_cap_member): New function.
11583 (allocnos_conflict_by_live_ranges_p): Use it.
11584 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
11585 (setup_slot_coalesced_allocno_live_ranges): Ditto.
11587 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
11590 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
11591 (*saddl_se_1): Ditto.
11592 (*ssubsi_1): Ditto.
11593 (*ssubl_se_1): Ditto.
11595 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11597 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
11598 rather than wi::to_widest for DR_INITs.
11599 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
11600 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
11601 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
11603 (vect_analyze_group_access_1): Note that here.
11605 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11607 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
11608 polynomial type sizes.
11610 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11612 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
11613 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
11614 (gimple_add_tmp_var): Likewise.
11616 2018-01-12 Martin Liska <mliska@suse.cz>
11618 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
11619 (gimple_alloc_sizes): Likewise.
11620 (dump_gimple_statistics): Use PRIu64 in printf format.
11621 * gimple.h: Change uint64_t to int.
11623 2018-01-12 Martin Liska <mliska@suse.cz>
11625 * tree-core.h: Use uint64_t instead of int.
11626 * tree.c (tree_node_counts): Likewise.
11627 (tree_node_sizes): Likewise.
11628 (dump_tree_statistics): Use PRIu64 in printf format.
11630 2018-01-12 Martin Liska <mliska@suse.cz>
11632 * Makefile.in: As qsort_chk is implemented in vec.c, add
11633 vec.o to linkage of gencfn-macros.
11634 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
11635 passing the info to record_node_allocation_statistics.
11636 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
11638 * ggc-common.c (struct ggc_usage): Add operator== and use
11639 it in operator< and compare function.
11640 * mem-stats.h (struct mem_usage): Likewise.
11641 * vec.c (struct vec_usage): Remove operator< and compare
11642 function. Can be simply inherited.
11644 2018-01-12 Martin Jambor <mjambor@suse.cz>
11647 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
11648 * tree-ssa-math-opts.c: Include domwalk.h.
11649 (convert_mult_to_fma_1): New function.
11650 (fma_transformation_info): New type.
11651 (fma_deferring_state): Likewise.
11652 (cancel_fma_deferring): New function.
11653 (result_of_phi): Likewise.
11654 (last_fma_candidate_feeds_initial_phi): Likewise.
11655 (convert_mult_to_fma): Added deferring logic, split actual
11656 transformation to convert_mult_to_fma_1.
11657 (math_opts_dom_walker): New type.
11658 (math_opts_dom_walker::after_dom_children): New method, body moved
11659 here from pass_optimize_widening_mul::execute, added deferring logic
11661 (pass_optimize_widening_mul::execute): Moved most of code to
11662 math_opts_dom_walker::after_dom_children.
11663 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
11664 * config/i386/i386.c (ix86_option_override_internal): Added
11665 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
11667 2018-01-12 Richard Biener <rguenther@suse.de>
11670 * dwarf2out.c (gen_variable_die): Do not reset old_die for
11671 inline instance vars.
11673 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
11676 * config/rx/rx.c (rx_is_restricted_memory_address):
11677 Handle SUBREG case.
11679 2018-01-12 Richard Biener <rguenther@suse.de>
11681 PR tree-optimization/80846
11682 * target.def (split_reduction): New target hook.
11683 * targhooks.c (default_split_reduction): New function.
11684 * targhooks.h (default_split_reduction): Declare.
11685 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
11686 target requests first reduce vectors by combining low and high
11688 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
11689 (get_vectype_for_scalar_type_and_size): Export.
11690 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
11691 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
11692 * doc/tm.texi: Regenerate.
11693 * config/i386/i386.c (ix86_split_reduction): Implement
11694 TARGET_VECTORIZE_SPLIT_REDUCTION.
11696 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11699 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
11700 in PIC mode except for TARGET_VXWORKS_RTP.
11701 * config/sparc/sparc.c: Include cfgrtl.h.
11702 (TARGET_INIT_PIC_REG): Define.
11703 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
11704 (sparc_pic_register_p): New predicate.
11705 (sparc_legitimate_address_p): Use it.
11706 (sparc_legitimize_pic_address): Likewise.
11707 (sparc_delegitimize_address): Likewise.
11708 (sparc_mode_dependent_address_p): Likewise.
11709 (gen_load_pcrel_sym): Remove 4th parameter.
11710 (load_got_register): Adjust call to above. Remove obsolete stuff.
11711 (sparc_expand_prologue): Do not call load_got_register here.
11712 (sparc_flat_expand_prologue): Likewise.
11713 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
11714 (sparc_use_pseudo_pic_reg): New function.
11715 (sparc_init_pic_reg): Likewise.
11716 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11717 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
11719 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
11721 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
11722 Add item for branch_cost.
11724 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11726 PR rtl-optimization/83565
11727 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
11728 not extend the result to a larger mode for rotate operations.
11729 (num_sign_bit_copies1): Likewise.
11731 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11734 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
11736 Use values-Xc.o for -pedantic.
11737 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
11739 2018-01-12 Martin Liska <mliska@suse.cz>
11742 * ipa-devirt.c (final_warning_record::grow_type_warnings):
11744 (possible_polymorphic_call_targets): Use it.
11745 (ipa_devirt): Likewise.
11747 2018-01-12 Martin Liska <mliska@suse.cz>
11749 * profile-count.h (enum profile_quality): Use 0 as invalid
11750 enum value of profile_quality.
11752 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
11754 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
11755 -mext-string options.
11757 2018-01-12 Richard Biener <rguenther@suse.de>
11759 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
11760 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
11761 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
11763 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
11765 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
11767 * configure.ac (--with-long-double-format): Add support for the
11768 configuration option to change the default long double format on
11770 * config.gcc (powerpc*-linux*-*): Likewise.
11771 * configure: Regenerate.
11772 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
11773 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
11774 used without modification.
11776 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11778 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
11779 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
11780 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
11781 MISC_BUILTIN_SPEC_BARRIER.
11782 (rs6000_init_builtins): Likewise.
11783 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
11785 (speculation_barrier): New define_insn.
11786 * doc/extend.texi: Document __builtin_speculation_barrier.
11788 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11791 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
11792 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
11793 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
11795 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
11796 integral modes instead of "ss" and "sd".
11797 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
11798 vectors with 32-bit and 64-bit elements.
11799 (vecdupssescalarmodesuffix): New mode attribute.
11800 (vec_dup<mode>): Use it.
11802 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
11805 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
11806 frame if argument is passed on stack.
11808 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11811 * ree.c (combine_reaching_defs): Optimize also
11812 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
11813 reg2=any_extend(exp); reg1=reg2;, formatting fix.
11815 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11817 PR middle-end/83189
11818 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
11820 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11822 PR middle-end/83718
11823 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
11824 after they are computed.
11826 2018-01-11 Bin Cheng <bin.cheng@arm.com>
11828 PR tree-optimization/83695
11829 * gimple-loop-linterchange.cc
11830 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
11831 reset cached scev information after interchange.
11832 (pass_linterchange::execute): Remove call to scev_reset_htab.
11834 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11836 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
11837 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
11838 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
11839 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
11840 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
11841 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
11842 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
11843 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
11844 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
11845 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
11846 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
11847 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
11848 (V_lane_reg): Likewise.
11849 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
11851 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
11852 (vfmal_lane_low<mode>_intrinsic,
11853 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
11854 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
11855 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
11856 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
11857 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
11858 vfmsl_lane_high<mode>_intrinsic): New define_insns.
11860 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11862 * config/arm/arm-cpus.in (fp16fml): New feature.
11863 (ALL_SIMD): Add fp16fml.
11864 (armv8.2-a): Add fp16fml as an option.
11865 (armv8.3-a): Likewise.
11866 (armv8.4-a): Add fp16fml as part of fp16.
11867 * config/arm/arm.h (TARGET_FP16FML): Define.
11868 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
11870 * config/arm/arm-modes.def (V2HF): Define.
11871 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
11872 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
11873 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
11874 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
11875 vfmsl_low, vfmsl_high): New set of builtins.
11876 * config/arm/iterators.md (PLUSMINUS): New code iterator.
11877 (vfml_op): New code attribute.
11878 (VFMLHALVES): New int iterator.
11879 (VFML, VFMLSEL): New mode attributes.
11880 (V_reg): Define mapping for V2HF.
11881 (V_hi, V_lo): New mode attributes.
11882 (VF_constraint): Likewise.
11883 (vfml_half, vfml_half_selector): New int attributes.
11884 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
11886 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
11887 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
11889 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
11890 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
11891 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
11892 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
11894 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
11895 Document new effective target and option set.
11897 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11899 * config/arm/arm-cpus.in (armv8_4): New feature.
11900 (ARMv8_4a): New fgroup.
11901 (armv8.4-a): New arch.
11902 * config/arm/arm-tables.opt: Regenerate.
11903 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
11904 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
11905 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
11906 Add matching rules for -march=armv8.4-a and extensions.
11907 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
11909 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
11912 * config/rx/rx.md (BW): New mode attribute.
11913 (sync_lock_test_and_setsi): Add mode suffix to insn output.
11915 2018-01-11 Richard Biener <rguenther@suse.de>
11917 PR tree-optimization/83435
11918 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
11919 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
11920 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
11922 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11923 Alan Hayward <alan.hayward@arm.com>
11924 David Sherwood <david.sherwood@arm.com>
11926 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
11928 (aarch64_classify_address): Initialize it. Track polynomial offsets.
11929 (aarch64_print_address_internal): Use it to check for a zero offset.
11931 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11932 Alan Hayward <alan.hayward@arm.com>
11933 David Sherwood <david.sherwood@arm.com>
11935 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
11936 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
11937 Return a poly_int64 rather than a HOST_WIDE_INT.
11938 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
11939 rather than a HOST_WIDE_INT.
11940 * config/aarch64/aarch64.h (aarch64_frame): Protect with
11941 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
11942 hard_fp_offset, frame_size, initial_adjust, callee_offset and
11943 final_offset from HOST_WIDE_INT to poly_int64.
11944 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
11945 to_constant when getting the number of units in an Advanced SIMD
11947 (aarch64_builtin_vectorized_function): Check for a constant number
11949 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
11951 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
11952 attribute instead of GET_MODE_NUNITS.
11953 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
11954 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
11955 GET_MODE_SIZE for fixed-size registers.
11956 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
11957 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
11958 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
11959 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
11960 (aarch64_print_operand, aarch64_print_address_internal)
11961 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
11962 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
11963 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
11964 Handle polynomial GET_MODE_SIZE.
11965 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
11966 wider than SImode without modification.
11967 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
11968 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
11969 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
11970 passing and returning SVE modes.
11971 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
11972 rather than GEN_INT.
11973 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
11974 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
11975 (aarch64_allocate_and_probe_stack_space): Likewise.
11976 (aarch64_layout_frame): Cope with polynomial offsets.
11977 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
11978 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
11979 polynomial offsets.
11980 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
11981 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
11982 poly_int64 rather than a HOST_WIDE_INT.
11983 (aarch64_get_separate_components, aarch64_process_components)
11984 (aarch64_expand_prologue, aarch64_expand_epilogue)
11985 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
11986 (aarch64_anchor_offset): New function, split out from...
11987 (aarch64_legitimize_address): ...here.
11988 (aarch64_builtin_vectorization_cost): Handle polynomial
11989 TYPE_VECTOR_SUBPARTS.
11990 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
11992 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
11993 number of elements from the PARALLEL rather than the mode.
11994 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
11995 rather than GET_MODE_BITSIZE.
11996 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
11997 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
11998 (aarch64_expand_vec_perm_const_1): Handle polynomial
11999 d->perm.length () and d->perm elements.
12000 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
12001 Apply to_constant to d->perm elements.
12002 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
12003 polynomial CONST_VECTOR_NUNITS.
12004 (aarch64_move_pointer): Take amount as a poly_int64 rather
12006 (aarch64_progress_pointer): Avoid temporary variable.
12007 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
12008 the mode attribute instead of GET_MODE.
12010 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12011 Alan Hayward <alan.hayward@arm.com>
12012 David Sherwood <david.sherwood@arm.com>
12014 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
12015 x exists before using it.
12016 (aarch64_add_constant_internal): Rename to...
12017 (aarch64_add_offset_1): ...this. Replace regnum with separate
12018 src and dest rtxes. Handle the case in which they're different,
12019 including when the offset is zero. Replace scratchreg with an rtx.
12020 Use 2 additions if there is no spare register into which we can
12021 move a 16-bit constant.
12022 (aarch64_add_constant): Delete.
12023 (aarch64_add_offset): Replace reg with separate src and dest
12024 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
12025 Use aarch64_add_offset_1.
12026 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
12027 an rtx rather than an int. Take the delta as a poly_int64
12028 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
12029 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
12030 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
12031 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
12032 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
12033 and aarch64_add_sp.
12034 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
12035 aarch64_add_constant.
12037 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12039 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
12040 Use scalar_float_mode.
12042 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12044 * config/aarch64/aarch64-simd.md
12045 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
12046 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
12047 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
12048 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
12049 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
12050 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
12051 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
12052 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
12053 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
12054 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
12056 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12059 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
12060 targ_options->x_arm_arch_string is non NULL.
12062 2018-01-11 Tamar Christina <tamar.christina@arm.com>
12064 * config/aarch64/aarch64.h
12065 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
12067 2018-01-11 Sudakshina Das <sudi.das@arm.com>
12070 * expmed.c (emit_store_flag_force): Swap if const op0
12071 and change VOIDmode to mode of op0.
12073 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12075 PR rtl-optimization/83761
12076 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
12077 than bytes to mode_for_size.
12079 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12081 PR middle-end/83189
12082 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
12083 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
12086 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12088 PR middle-end/83575
12089 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
12090 when in layout mode.
12091 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
12092 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
12095 2018-01-10 Michael Collison <michael.collison@arm.com>
12097 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
12098 * config/aarch64/aarch64-option-extension.def: Add
12099 AARCH64_OPT_EXTENSION of 'fp16fml'.
12100 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12101 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
12102 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
12103 * config/aarch64/constraints.md (Ui7): New constraint.
12104 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
12105 (VFMLA_SEL_W): Ditto.
12108 (VFMLA16_LOW): New int iterator.
12109 (VFMLA16_HIGH): Ditto.
12110 (UNSPEC_FMLAL): New unspec.
12111 (UNSPEC_FMLSL): Ditto.
12112 (UNSPEC_FMLAL2): Ditto.
12113 (UNSPEC_FMLSL2): Ditto.
12114 (f16mac): New code attribute.
12115 * config/aarch64/aarch64-simd-builtins.def
12116 (aarch64_fmlal_lowv2sf): Ditto.
12117 (aarch64_fmlsl_lowv2sf): Ditto.
12118 (aarch64_fmlalq_lowv4sf): Ditto.
12119 (aarch64_fmlslq_lowv4sf): Ditto.
12120 (aarch64_fmlal_highv2sf): Ditto.
12121 (aarch64_fmlsl_highv2sf): Ditto.
12122 (aarch64_fmlalq_highv4sf): Ditto.
12123 (aarch64_fmlslq_highv4sf): Ditto.
12124 (aarch64_fmlal_lane_lowv2sf): Ditto.
12125 (aarch64_fmlsl_lane_lowv2sf): Ditto.
12126 (aarch64_fmlal_laneq_lowv2sf): Ditto.
12127 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
12128 (aarch64_fmlalq_lane_lowv4sf): Ditto.
12129 (aarch64_fmlsl_lane_lowv4sf): Ditto.
12130 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
12131 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
12132 (aarch64_fmlal_lane_highv2sf): Ditto.
12133 (aarch64_fmlsl_lane_highv2sf): Ditto.
12134 (aarch64_fmlal_laneq_highv2sf): Ditto.
12135 (aarch64_fmlsl_laneq_highv2sf): Ditto.
12136 (aarch64_fmlalq_lane_highv4sf): Ditto.
12137 (aarch64_fmlsl_lane_highv4sf): Ditto.
12138 (aarch64_fmlalq_laneq_highv4sf): Ditto.
12139 (aarch64_fmlsl_laneq_highv4sf): Ditto.
12140 * config/aarch64/aarch64-simd.md:
12141 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
12142 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12143 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
12144 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12145 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
12146 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
12147 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
12148 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
12149 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
12150 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
12151 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
12152 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
12153 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
12154 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
12155 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
12156 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
12157 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
12158 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
12159 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
12160 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
12161 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
12162 (vfmlsl_low_u32): Ditto.
12163 (vfmlalq_low_u32): Ditto.
12164 (vfmlslq_low_u32): Ditto.
12165 (vfmlal_high_u32): Ditto.
12166 (vfmlsl_high_u32): Ditto.
12167 (vfmlalq_high_u32): Ditto.
12168 (vfmlslq_high_u32): Ditto.
12169 (vfmlal_lane_low_u32): Ditto.
12170 (vfmlsl_lane_low_u32): Ditto.
12171 (vfmlal_laneq_low_u32): Ditto.
12172 (vfmlsl_laneq_low_u32): Ditto.
12173 (vfmlalq_lane_low_u32): Ditto.
12174 (vfmlslq_lane_low_u32): Ditto.
12175 (vfmlalq_laneq_low_u32): Ditto.
12176 (vfmlslq_laneq_low_u32): Ditto.
12177 (vfmlal_lane_high_u32): Ditto.
12178 (vfmlsl_lane_high_u32): Ditto.
12179 (vfmlal_laneq_high_u32): Ditto.
12180 (vfmlsl_laneq_high_u32): Ditto.
12181 (vfmlalq_lane_high_u32): Ditto.
12182 (vfmlslq_lane_high_u32): Ditto.
12183 (vfmlalq_laneq_high_u32): Ditto.
12184 (vfmlslq_laneq_high_u32): Ditto.
12185 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
12186 (AARCH64_FL_FOR_ARCH8_4): New.
12187 (AARCH64_ISA_F16FML): New ISA flag.
12188 (TARGET_F16FML): New feature flag for fp16fml.
12189 (doc/invoke.texi): Document new fp16fml option.
12191 2018-01-10 Michael Collison <michael.collison@arm.com>
12193 * config/aarch64/aarch64-builtins.c:
12194 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
12195 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12196 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
12197 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
12198 (AARCH64_ISA_SHA3): New ISA flag.
12199 (TARGET_SHA3): New feature flag for sha3.
12200 * config/aarch64/iterators.md (sha512_op): New int attribute.
12201 (CRYPTO_SHA512): New int iterator.
12202 (UNSPEC_SHA512H): New unspec.
12203 (UNSPEC_SHA512H2): Ditto.
12204 (UNSPEC_SHA512SU0): Ditto.
12205 (UNSPEC_SHA512SU1): Ditto.
12206 * config/aarch64/aarch64-simd-builtins.def
12207 (aarch64_crypto_sha512hqv2di): New builtin.
12208 (aarch64_crypto_sha512h2qv2di): Ditto.
12209 (aarch64_crypto_sha512su0qv2di): Ditto.
12210 (aarch64_crypto_sha512su1qv2di): Ditto.
12211 (aarch64_eor3qv8hi): Ditto.
12212 (aarch64_rax1qv2di): Ditto.
12213 (aarch64_xarqv2di): Ditto.
12214 (aarch64_bcaxqv8hi): Ditto.
12215 * config/aarch64/aarch64-simd.md:
12216 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
12217 (aarch64_crypto_sha512su0qv2di): Ditto.
12218 (aarch64_crypto_sha512su1qv2di): Ditto.
12219 (aarch64_eor3qv8hi): Ditto.
12220 (aarch64_rax1qv2di): Ditto.
12221 (aarch64_xarqv2di): Ditto.
12222 (aarch64_bcaxqv8hi): Ditto.
12223 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
12224 (vsha512h2q_u64): Ditto.
12225 (vsha512su0q_u64): Ditto.
12226 (vsha512su1q_u64): Ditto.
12227 (veor3q_u16): Ditto.
12228 (vrax1q_u64): Ditto.
12229 (vxarq_u64): Ditto.
12230 (vbcaxq_u16): Ditto.
12231 * config/arm/types.md (crypto_sha512): New type attribute.
12232 (crypto_sha3): Ditto.
12233 (doc/invoke.texi): Document new sha3 option.
12235 2018-01-10 Michael Collison <michael.collison@arm.com>
12237 * config/aarch64/aarch64-builtins.c:
12238 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
12239 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12240 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
12241 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
12242 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
12243 (AARCH64_ISA_SM4): New ISA flag.
12244 (TARGET_SM4): New feature flag for sm4.
12245 * config/aarch64/aarch64-simd-builtins.def
12246 (aarch64_sm3ss1qv4si): Ditto.
12247 (aarch64_sm3tt1aq4si): Ditto.
12248 (aarch64_sm3tt1bq4si): Ditto.
12249 (aarch64_sm3tt2aq4si): Ditto.
12250 (aarch64_sm3tt2bq4si): Ditto.
12251 (aarch64_sm3partw1qv4si): Ditto.
12252 (aarch64_sm3partw2qv4si): Ditto.
12253 (aarch64_sm4eqv4si): Ditto.
12254 (aarch64_sm4ekeyqv4si): Ditto.
12255 * config/aarch64/aarch64-simd.md:
12256 (aarch64_sm3ss1qv4si): Ditto.
12257 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
12258 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
12259 (aarch64_sm4eqv4si): Ditto.
12260 (aarch64_sm4ekeyqv4si): Ditto.
12261 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
12262 (sm3part_op): Ditto.
12263 (CRYPTO_SM3TT): Ditto.
12264 (CRYPTO_SM3PART): Ditto.
12265 (UNSPEC_SM3SS1): New unspec.
12266 (UNSPEC_SM3TT1A): Ditto.
12267 (UNSPEC_SM3TT1B): Ditto.
12268 (UNSPEC_SM3TT2A): Ditto.
12269 (UNSPEC_SM3TT2B): Ditto.
12270 (UNSPEC_SM3PARTW1): Ditto.
12271 (UNSPEC_SM3PARTW2): Ditto.
12272 (UNSPEC_SM4E): Ditto.
12273 (UNSPEC_SM4EKEY): Ditto.
12274 * config/aarch64/constraints.md (Ui2): New constraint.
12275 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
12276 * config/arm/types.md (crypto_sm3): New type attribute.
12277 (crypto_sm4): Ditto.
12278 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
12279 (vsm3tt1aq_u32): Ditto.
12280 (vsm3tt1bq_u32): Ditto.
12281 (vsm3tt2aq_u32): Ditto.
12282 (vsm3tt2bq_u32): Ditto.
12283 (vsm3partw1q_u32): Ditto.
12284 (vsm3partw2q_u32): Ditto.
12285 (vsm4eq_u32): Ditto.
12286 (vsm4ekeyq_u32): Ditto.
12287 (doc/invoke.texi): Document new sm4 option.
12289 2018-01-10 Michael Collison <michael.collison@arm.com>
12291 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
12292 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
12293 (AARCH64_FL_FOR_ARCH8_4): New.
12294 (AARCH64_FL_V8_4): New flag.
12295 (doc/invoke.texi): Document new armv8.4-a option.
12297 2018-01-10 Michael Collison <michael.collison@arm.com>
12299 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12300 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
12301 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
12302 * config/aarch64/aarch64-option-extension.def: Add
12303 AARCH64_OPT_EXTENSION of 'sha2'.
12304 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
12305 (crypto): Disable sha2 and aes if crypto disabled.
12306 (crypto): Enable aes and sha2 if enabled.
12307 (simd): Disable sha2 and aes if simd disabled.
12308 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
12310 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
12311 (TARGET_SHA2): New feature flag for sha2.
12312 (TARGET_AES): New feature flag for aes.
12313 * config/aarch64/aarch64-simd.md:
12314 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
12315 conditional on TARGET_AES.
12316 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
12317 (aarch64_crypto_sha1hsi): Make pattern conditional
12319 (aarch64_crypto_sha1hv4si): Ditto.
12320 (aarch64_be_crypto_sha1hv4si): Ditto.
12321 (aarch64_crypto_sha1su1v4si): Ditto.
12322 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
12323 (aarch64_crypto_sha1su0v4si): Ditto.
12324 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
12325 (aarch64_crypto_sha256su0v4si): Ditto.
12326 (aarch64_crypto_sha256su1v4si): Ditto.
12327 (doc/invoke.texi): Document new aes and sha2 options.
12329 2018-01-10 Martin Sebor <msebor@redhat.com>
12331 PR tree-optimization/83781
12332 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
12335 2018-01-11 Martin Sebor <msebor@gmail.com>
12336 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12338 PR tree-optimization/83501
12339 PR tree-optimization/81703
12341 * tree-ssa-strlen.c (get_string_cst): Rename...
12342 (get_string_len): ...to this. Handle global constants.
12343 (handle_char_store): Adjust.
12345 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
12346 Jim Wilson <jimw@sifive.com>
12348 * config/riscv/riscv-protos.h (riscv_output_return): New.
12349 * config/riscv/riscv.c (struct machine_function): New naked_p field.
12350 (riscv_attribute_table, riscv_output_return),
12351 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
12352 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
12353 (riscv_compute_frame_info): Only compute frame->mask if not a naked
12355 (riscv_expand_prologue): Add early return for naked function.
12356 (riscv_expand_epilogue): Likewise.
12357 (riscv_function_ok_for_sibcall): Return false for naked function.
12358 (riscv_set_current_function): New.
12359 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
12360 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
12361 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
12362 * doc/extend.texi (RISC-V Function Attributes): New.
12364 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
12366 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
12367 check for 128-bit long double before checking TCmode.
12368 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
12369 128-bit long doubles before checking TFmode or TCmode.
12370 (FLOAT128_IBM_P): Likewise.
12372 2018-01-10 Martin Sebor <msebor@redhat.com>
12374 PR tree-optimization/83671
12375 * builtins.c (c_strlen): Unconditionally return zero for the empty
12377 Use -Warray-bounds for warnings.
12378 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
12379 for non-constant array indices with COMPONENT_REF, arrays of
12380 arrays, and pointers to arrays.
12381 (gimple_fold_builtin_strlen): Determine and set length range for
12382 non-constant character arrays.
12384 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
12386 PR middle-end/81897
12387 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
12390 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
12392 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
12394 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12397 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
12398 VECTOR_MEM_ALTIVEC_OR_VSX_P.
12399 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
12400 indexed_or_indirect_operand predicate.
12401 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
12402 (*vsx_le_perm_load_v8hi): Likewise.
12403 (*vsx_le_perm_load_v16qi): Likewise.
12404 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
12405 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
12406 (*vsx_le_perm_store_v8hi): Likewise.
12407 (*vsx_le_perm_store_v16qi): Likewise.
12408 (eight unnamed splitters): Likewise.
12410 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12412 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
12413 * config/rs6000/emmintrin.h: Likewise.
12414 * config/rs6000/mmintrin.h: Likewise.
12415 * config/rs6000/xmmintrin.h: Likewise.
12417 2018-01-10 David Malcolm <dmalcolm@redhat.com>
12420 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
12422 * tree.c (tree_nop_conversion): Return true for location wrapper
12424 (maybe_wrap_with_location): New function.
12425 (selftest::check_strip_nops): New function.
12426 (selftest::test_location_wrappers): New function.
12427 (selftest::tree_c_tests): Call it.
12428 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
12429 (maybe_wrap_with_location): New decl.
12430 (EXPR_LOCATION_WRAPPER_P): New macro.
12431 (location_wrapper_p): New inline function.
12432 (tree_strip_any_location_wrapper): New inline function.
12434 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
12437 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
12438 stack_realign_offset for the largest alignment of stack slot
12440 (ix86_find_max_used_stack_alignment): New function.
12441 (ix86_finalize_stack_frame_flags): Use it. Set
12442 max_used_stack_alignment if we don't realign stack.
12443 * config/i386/i386.h (machine_function): Add
12444 max_used_stack_alignment.
12446 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
12448 * config/arm/arm.opt (-mbranch-cost): New option.
12449 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
12452 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
12455 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
12456 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
12458 2018-01-10 Richard Biener <rguenther@suse.de>
12461 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
12462 early out so it also covers the case where we have a non-NULL
12465 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12467 PR tree-optimization/83753
12468 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
12469 for non-strided grouped accesses if the number of elements is 1.
12471 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12474 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
12475 * i386.h (TARGET_USE_GATHER): Define.
12476 * x86-tune.def (X86_TUNE_USE_GATHER): New.
12478 2018-01-10 Martin Liska <mliska@suse.cz>
12481 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
12482 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
12484 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
12485 CLEANUP_NO_PARTITIONING is not set.
12487 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12489 * doc/rtl.texi: Remove documentation of (const ...) wrappers
12490 for vectors, as a partial revert of r254296.
12491 * rtl.h (const_vec_p): Delete.
12492 (const_vec_duplicate_p): Don't test for vector CONSTs.
12493 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
12494 * expmed.c (make_tree): Likewise.
12497 * common.md (E, F): Use CONSTANT_P instead of checking for
12499 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
12500 checking for CONST_VECTOR.
12502 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12504 PR middle-end/83575
12505 * predict.c (force_edge_cold): Handle in more sane way edges
12506 with no prediction.
12508 2018-01-09 Carl Love <cel@us.ibm.com>
12510 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
12512 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
12513 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
12514 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
12515 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
12516 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
12517 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
12518 * config/rs6000/rs6000-protos.h: Add extern defition for
12519 rs6000_generate_float2_double_code.
12520 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
12522 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
12523 (float2_v2df): Add define_expand.
12525 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
12528 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
12529 op_mode in the force_to_mode call.
12531 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12533 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
12534 instead of checking each element individually.
12535 (aarch64_evpc_uzp): Likewise.
12536 (aarch64_evpc_zip): Likewise.
12537 (aarch64_evpc_ext): Likewise.
12538 (aarch64_evpc_rev): Likewise.
12539 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
12540 instead of checking each element individually. Return true without
12542 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
12543 whether all selected elements come from the same input, instead of
12544 checking each element individually. Remove calls to gen_rtx_REG,
12545 start_sequence and end_sequence and instead assert that no rtl is
12548 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12550 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
12551 order of HIGH and CONST checks.
12553 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12555 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
12556 if the destination isn't an SSA_NAME.
12558 2018-01-09 Richard Biener <rguenther@suse.de>
12560 PR tree-optimization/83668
12561 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
12563 (canonicalize_loop_form): ... here, renamed from ...
12564 (canonicalize_loop_closed_ssa_form): ... this and amended to
12565 swap successor edges for loop exit blocks to make us use
12566 the RPO order we need for initial schedule generation.
12568 2018-01-09 Joseph Myers <joseph@codesourcery.com>
12570 PR tree-optimization/64811
12571 * match.pd: When optimizing comparisons with Inf, avoid
12572 introducing or losing exceptions from comparisons with NaN.
12574 2018-01-09 Martin Liska <mliska@suse.cz>
12577 * asan.c (shadow_mem_size): Add gcc_assert.
12579 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
12581 Don't save registers in main().
12584 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
12585 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
12586 * config/avr/avr.c (avr_set_current_function): Don't error if
12587 naked, OS_task or OS_main are specified at the same time.
12588 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
12590 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
12592 * common/config/avr/avr-common.c (avr_option_optimization_table):
12593 Switch on -mmain-is-OS_task for optimizing compilations.
12595 2018-01-09 Richard Biener <rguenther@suse.de>
12597 PR tree-optimization/83572
12598 * graphite.c: Include cfganal.h.
12599 (graphite_transform_loops): Connect infinite loops to exit
12600 and remove fake edges at the end.
12602 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12604 * ipa-inline.c (edge_badness): Revert accidental checkin.
12606 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12609 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
12610 symbols; not inline clones.
12612 2018-01-09 Jakub Jelinek <jakub@redhat.com>
12615 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
12616 hard registers. Formatting fixes.
12618 PR preprocessor/83722
12619 * gcc.c (try_generate_repro): Pass
12620 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
12621 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
12624 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
12625 Kito Cheng <kito.cheng@gmail.com>
12627 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
12628 (riscv_leaf_function_p): Delete.
12629 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
12631 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12633 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
12635 (do_ifelse): New function.
12636 (do_isel): New function.
12637 (do_sub3): New function.
12638 (do_add3): New function.
12639 (do_load_mask_compare): New function.
12640 (do_overlap_load_compare): New function.
12641 (expand_compare_loop): New function.
12642 (expand_block_compare): Call expand_compare_loop() when appropriate.
12643 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
12644 option description.
12645 (-mblock-compare-inline-loop-limit): New option.
12647 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12650 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
12651 Reverse order of second and third operands in first alternative.
12652 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
12653 of first and second elements in UNSPEC_VPERMR vector.
12654 (altivec_expand_vec_perm_le): Likewise.
12656 2018-01-08 Jeff Law <law@redhat.com>
12658 PR rtl-optimizatin/81308
12659 * tree-switch-conversion.c (cfg_altered): New file scoped static.
12660 (process_switch): If group_case_labels makes a change, then set
12662 (pass_convert_switch::execute): If a switch is converted, then
12663 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
12665 PR rtl-optimization/81308
12666 * recog.c (split_all_insns): Conditionally cleanup the CFG after
12669 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
12671 PR target/83663 - Revert r255946
12672 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
12673 generation for cases where splatting a value is not useful.
12674 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
12675 across a vec_duplicate and a paradoxical subreg forming a vector
12676 mode to a vec_concat.
12678 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12680 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
12681 -march=armv8.3-a variants.
12682 * config/arm/t-multilib: Likewise.
12683 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
12685 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12687 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
12689 (cceq_ior_compare_complement): Give it a name so I can use it, and
12690 change boolean_or_operator predicate to boolean_operator so it can
12691 be used to generate a crand.
12692 (eqne): New code iterator.
12693 (bd/bd_neg): New code_attrs.
12694 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
12695 a single define_insn.
12696 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
12697 decrement (bdnzt/bdnzf/bdzt/bdzf).
12698 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
12699 with the new names of the branch decrement patterns, and added the
12700 names of the branch decrement conditional patterns.
12702 2018-01-08 Richard Biener <rguenther@suse.de>
12704 PR tree-optimization/83563
12705 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
12708 2018-01-08 Richard Biener <rguenther@suse.de>
12710 PR middle-end/83713
12711 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
12713 2018-01-08 Richard Biener <rguenther@suse.de>
12715 PR tree-optimization/83685
12716 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
12717 references to abnormals.
12719 2018-01-08 Richard Biener <rguenther@suse.de>
12722 * dwarf2out.c (output_indirect_strings): Handle empty
12723 skeleton_debug_str_hash.
12724 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
12726 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12728 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
12729 (emit_store_direct): Likewise.
12730 (arc_trampoline_adjust_address): Likewise.
12731 (arc_asm_trampoline_template): New function.
12732 (arc_initialize_trampoline): Use asm_trampoline_template.
12733 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
12734 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
12735 * config/arc/arc.md (flush_icache): Delete pattern.
12737 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12739 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
12740 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
12743 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12746 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
12747 by not USED_FOR_TARGET.
12748 (make_pass_resolve_sw_modes): Likewise.
12750 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12752 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
12755 2018-01-08 Richard Biener <rguenther@suse.de>
12757 PR middle-end/83580
12758 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
12760 2018-01-08 Richard Biener <rguenther@suse.de>
12762 PR middle-end/83517
12763 * match.pd ((t * 2) / 2) -> t): Add missing :c.
12765 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
12767 PR middle-end/81897
12768 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
12769 basic blocks with a small number of successors.
12770 (convert_control_dep_chain_into_preds): Improve handling of
12772 (dump_predicates): Split apart into...
12773 (dump_pred_chain): ...here...
12774 (dump_pred_info): ...and here.
12775 (can_one_predicate_be_invalidated_p): Add debugging printfs.
12776 (can_chain_union_be_invalidated_p): Improve check for invalidation
12778 (uninit_uses_cannot_happen): Avoid unnecessary if
12779 convert_control_dep_chain_into_preds yielded nothing.
12781 2018-01-06 Martin Sebor <msebor@redhat.com>
12783 PR tree-optimization/83640
12784 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
12785 subtracting negative offset from size.
12786 (builtin_access::overlap): Adjust offset bounds of the access to fall
12787 within the size of the object if possible.
12789 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
12791 PR rtl-optimization/83699
12792 * expmed.c (extract_bit_field_1): Restrict the vector usage of
12793 extract_bit_field_as_subreg to cases in which the extracted
12794 value is also a vector.
12796 * lra-constraints.c (process_alt_operands): Test for the equivalence
12797 substitutions when detecting a possible reload cycle.
12799 2018-01-06 Jakub Jelinek <jakub@redhat.com>
12802 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
12803 by default if flag_selective_schedling{,2}. Formatting fixes.
12805 PR rtl-optimization/83682
12806 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
12807 if it has non-VECTOR_MODE element mode.
12808 (vec_duplicate_p): Likewise.
12810 PR middle-end/83694
12811 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
12812 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
12814 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12817 * config/i386/i386-builtin.def
12818 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
12819 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
12820 Require also OPTION_MASK_ISA_AVX512F in addition to
12821 OPTION_MASK_ISA_GFNI.
12822 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
12823 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
12824 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
12825 to OPTION_MASK_ISA_GFNI.
12826 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
12827 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
12828 OPTION_MASK_ISA_AVX512BW.
12829 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
12830 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
12831 addition to OPTION_MASK_ISA_GFNI.
12832 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
12833 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
12834 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
12835 to OPTION_MASK_ISA_GFNI.
12836 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
12837 a requirement for all ISAs rather than any of them with a few
12839 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
12841 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
12842 bitmasks to be enabled with 3 exceptions, instead of requiring any
12843 enabled ISA with lots of exceptions.
12844 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
12845 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
12846 Change avx512bw in isa attribute to avx512f.
12847 * config/i386/sgxintrin.h: Add license boilerplate.
12848 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
12849 to __AVX512F__ and __AVX512VL to __AVX512VL__.
12850 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
12851 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
12853 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
12854 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
12855 temporarily sse2 rather than sse if not enabled already.
12858 * config/i386/sse.md (VI248_VLBW): Rename to ...
12859 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
12860 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
12861 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
12862 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
12863 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
12864 mode iterator instead of VI248_VLBW.
12866 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
12868 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
12869 (record_modified): Skip clobbers; add debug output.
12870 (param_change_prob): Use sreal frequencies.
12872 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12874 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
12875 punt for user-aligned variables.
12877 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12879 * tree-chrec.c (chrec_contains_symbols): Return true for
12882 2018-01-05 Sudakshina Das <sudi.das@arm.com>
12885 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
12886 of (x|y) == x for BICS pattern.
12888 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12890 PR tree-optimization/83605
12891 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
12892 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
12895 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
12897 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
12898 * config/epiphany/rtems.h: New file.
12900 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12901 Uros Bizjak <ubizjak@gmail.com>
12904 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
12905 QIreg_operand instead of register_operand predicate.
12906 * config/i386/i386.c (ix86_rop_should_change_byte_p,
12907 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
12908 comments instead of -fmitigate[-_]rop.
12910 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12913 * cgraphunit.c (symbol_table::compile): Switch to text_section
12914 before calling assembly_start debug hook.
12915 * run-rtl-passes.c (run_rtl_passes): Likewise.
12918 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12920 * tree-vrp.c (extract_range_from_binary_expr_1): Check
12921 range_int_cst_p rather than !symbolic_range_p before calling
12922 extract_range_from_multiplicative_op_1.
12924 2018-01-04 Jeff Law <law@redhat.com>
12926 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
12927 redundant test in assertion.
12929 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12931 * doc/rtl.texi: Document machine_mode wrapper classes.
12933 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12935 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
12936 using tree_to_uhwi.
12938 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12940 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
12941 the VEC_PERM_EXPR fold to fail.
12943 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12946 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
12947 to switched_sections.
12949 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12952 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
12953 test for d.testing.
12955 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
12958 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
12959 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
12961 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12964 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
12965 is BLKmode and bitpos not zero or mode change is needed.
12967 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12970 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
12973 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
12976 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
12977 instead of MULT rtx. Update all corresponding splitters.
12978 (*saddl_se): Ditto.
12979 (*ssub<modesuffix>): Ditto.
12980 (*ssubl_se): Ditto.
12981 (*cmp_sadd_di): Update split patterns.
12982 (*cmp_sadd_si): Ditto.
12983 (*cmp_sadd_sidi): Ditto.
12984 (*cmp_ssub_di): Ditto.
12985 (*cmp_ssub_si): Ditto.
12986 (*cmp_ssub_sidi): Ditto.
12987 * config/alpha/predicates.md (const23_operand): New predicate.
12988 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
12989 Look for ASHIFT, not MULT inner operand.
12990 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
12992 2018-01-04 Martin Liska <mliska@suse.cz>
12994 PR gcov-profile/83669
12995 * gcov.c (output_intermediate_file): Add version to intermediate
12997 * doc/gcov.texi: Document new field 'version' in intermediate
12998 file format. Fix location of '-k' option of gcov command.
13000 2018-01-04 Martin Liska <mliska@suse.cz>
13003 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
13005 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13007 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
13009 2018-01-03 Martin Sebor <msebor@redhat.com>
13011 PR tree-optimization/83655
13012 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
13013 checking calls with invalid arguments.
13015 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13017 * tree-vect-stmts.c (vect_get_store_rhs): New function.
13018 (vectorizable_mask_load_store): Delete.
13019 (vectorizable_call): Return false for masked loads and stores.
13020 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
13021 instead of gimple_assign_rhs1.
13022 (vectorizable_load): Handle IFN_MASK_LOAD.
13023 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
13025 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13027 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
13029 (vectorizable_mask_load_store): ...here.
13030 (vectorizable_load): ...and here.
13032 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13034 * tree-vect-stmts.c (vect_build_all_ones_mask)
13035 (vect_build_zero_merge_argument): New functions, split out from...
13036 (vectorizable_load): ...here.
13038 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13040 * tree-vect-stmts.c (vect_check_store_rhs): New function,
13042 (vectorizable_mask_load_store): ...here.
13043 (vectorizable_store): ...and here.
13045 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13047 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
13049 (vectorizable_mask_load_store): ...here.
13051 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13053 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
13054 (vect_model_store_cost): Take a vec_load_store_type instead of a
13056 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
13057 (vect_model_store_cost): Take a vec_load_store_type instead of a
13059 (vectorizable_mask_load_store): Update accordingly.
13060 (vectorizable_store): Likewise.
13061 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
13063 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13065 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
13066 IFN_MASK_LOAD calls here rather than...
13067 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
13069 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13070 Alan Hayward <alan.hayward@arm.com>
13071 David Sherwood <david.sherwood@arm.com>
13073 * expmed.c (extract_bit_field_1): For vector extracts,
13074 fall back to extract_bit_field_as_subreg if vec_extract
13077 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13078 Alan Hayward <alan.hayward@arm.com>
13079 David Sherwood <david.sherwood@arm.com>
13081 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
13082 they are variable or constant sized.
13083 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
13084 slots for constant-sized data.
13086 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13087 Alan Hayward <alan.hayward@arm.com>
13088 David Sherwood <david.sherwood@arm.com>
13090 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
13091 handling COND_EXPRs with boolean comparisons, try to find a better
13092 basis for the mask type than the boolean itself.
13094 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13096 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
13097 is calculated and how it can be overridden.
13098 * genmodes.c (max_bitsize_mode_any_mode): New variable.
13099 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
13101 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
13104 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13105 Alan Hayward <alan.hayward@arm.com>
13106 David Sherwood <david.sherwood@arm.com>
13108 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
13109 Remove the mode argument.
13110 (aarch64_simd_valid_immediate): Remove the mode and inverse
13112 * config/aarch64/iterators.md (bitsize): New iterator.
13113 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
13114 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
13115 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
13116 aarch64_simd_valid_immediate.
13117 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
13118 (aarch64_reg_or_bic_imm): Likewise.
13119 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
13120 with an insn_type enum and msl with a modifier_type enum.
13121 Replace element_width with a scalar_mode. Change the shift
13122 to unsigned int. Add constructors for scalar_float_mode and
13123 scalar_int_mode elements.
13124 (aarch64_vect_float_const_representable_p): Delete.
13125 (aarch64_can_const_movi_rtx_p)
13126 (aarch64_simd_scalar_immediate_valid_for_move)
13127 (aarch64_simd_make_constant): Update call to
13128 aarch64_simd_valid_immediate.
13129 (aarch64_advsimd_valid_immediate_hs): New function.
13130 (aarch64_advsimd_valid_immediate): Likewise.
13131 (aarch64_simd_valid_immediate): Remove mode and inverse
13132 arguments. Rewrite to use the above. Use const_vec_duplicate_p
13133 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
13134 and aarch64_float_const_representable_p on the result.
13135 (aarch64_output_simd_mov_immediate): Remove mode argument.
13136 Update call to aarch64_simd_valid_immediate and use of
13137 simd_immediate_info.
13138 (aarch64_output_scalar_simd_mov_immediate): Update call
13141 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13142 Alan Hayward <alan.hayward@arm.com>
13143 David Sherwood <david.sherwood@arm.com>
13145 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
13146 (mode_nunits): Likewise CONST_MODE_NUNITS.
13147 * machmode.def (ADJUST_NUNITS): Document.
13148 * genmodes.c (mode_data::need_nunits_adj): New field.
13149 (blank_mode): Update accordingly.
13150 (adj_nunits): New variable.
13151 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
13153 (emit_mode_size_inline): Set need_bytesize_adj for all modes
13154 listed in adj_nunits.
13155 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
13156 listed in adj_nunits. Don't emit case statements for such modes.
13157 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
13158 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
13159 nothing if adj_nunits is nonnull.
13160 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
13161 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
13162 (emit_mode_fbit): Update use of print_maybe_const_decl.
13163 (emit_move_size): Likewise. Treat the array as non-const
13165 (emit_mode_adjustments): Handle adj_nunits.
13167 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13169 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
13170 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
13171 (VECTOR_MODES): Use it.
13172 (make_vector_modes): Take the prefix as an argument.
13174 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13175 Alan Hayward <alan.hayward@arm.com>
13176 David Sherwood <david.sherwood@arm.com>
13178 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
13179 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
13180 for MODE_VECTOR_BOOL.
13181 * machmode.def (VECTOR_BOOL_MODE): Document.
13182 * genmodes.c (VECTOR_BOOL_MODE): New macro.
13183 (make_vector_bool_mode): New function.
13184 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
13186 * lto-streamer-in.c (lto_input_mode_table): Likewise.
13187 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
13189 * stor-layout.c (int_mode_for_mode): Likewise.
13190 * tree.c (build_vector_type_for_mode): Likewise.
13191 * varasm.c (output_constant_pool_2): Likewise.
13192 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
13193 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
13194 for MODE_VECTOR_BOOL.
13195 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
13196 of mode class checks.
13197 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
13198 instead of a list of mode class checks.
13199 (expand_vector_scalar_condition): Likewise.
13200 (type_for_widest_vector_mode): Handle BImode as an inner mode.
13202 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13203 Alan Hayward <alan.hayward@arm.com>
13204 David Sherwood <david.sherwood@arm.com>
13206 * machmode.h (mode_size): Change from unsigned short to
13208 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
13209 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13210 or if measurement_type is not polynomial.
13211 (fixed_size_mode::includes_p): Check for constant-sized modes.
13212 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
13213 return a poly_uint16 rather than an unsigned short.
13214 (emit_mode_size): Change the type of mode_size from unsigned short
13215 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
13216 (emit_mode_adjustments): Cope with polynomial vector sizes.
13217 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13219 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13221 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
13222 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
13223 * caller-save.c (setup_save_areas): Likewise.
13224 (replace_reg_with_saved_mem): Likewise.
13225 * calls.c (emit_library_call_value_1): Likewise.
13226 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
13227 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
13228 (gen_lowpart_for_combine): Likewise.
13229 * convert.c (convert_to_integer_1): Likewise.
13230 * cse.c (equiv_constant, cse_insn): Likewise.
13231 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
13232 (cselib_subst_to_values): Likewise.
13233 * dce.c (word_dce_process_block): Likewise.
13234 * df-problems.c (df_word_lr_mark_ref): Likewise.
13235 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
13236 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
13237 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
13238 (rtl_for_decl_location): Likewise.
13239 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
13240 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
13241 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
13242 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
13243 (expand_expr_real_1): Likewise.
13244 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
13245 (pad_below): Likewise.
13246 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13247 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
13248 * ira.c (get_subreg_tracking_sizes): Likewise.
13249 * ira-build.c (ira_create_allocno_objects): Likewise.
13250 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
13251 (ira_sort_regnos_for_alter_reg): Likewise.
13252 * ira-costs.c (record_operand_costs): Likewise.
13253 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
13254 (resolve_simple_move): Likewise.
13255 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
13256 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
13257 (lra_constraints): Likewise.
13258 (CONST_POOL_OK_P): Reject variable-sized modes.
13259 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
13260 (add_pseudo_to_slot, lra_spill): Likewise.
13261 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13262 * optabs-query.c (get_best_extraction_insn): Likewise.
13263 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13264 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
13265 (expand_mult_highpart, valid_multiword_target_p): Likewise.
13266 * recog.c (offsettable_address_addr_space_p): Likewise.
13267 * regcprop.c (maybe_mode_change): Likewise.
13268 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
13269 * regrename.c (build_def_use): Likewise.
13270 * regstat.c (dump_reg_info): Likewise.
13271 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
13272 (find_reloads, find_reloads_subreg_address): Likewise.
13273 * reload1.c (eliminate_regs_1): Likewise.
13274 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
13275 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
13276 (simplify_binary_operation_1, simplify_subreg): Likewise.
13277 * targhooks.c (default_function_arg_padding): Likewise.
13278 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
13279 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
13280 (verify_gimple_assign_ternary): Likewise.
13281 * tree-inline.c (estimate_move_cost): Likewise.
13282 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13283 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
13284 (get_address_cost_ainc): Likewise.
13285 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
13286 (vect_supportable_dr_alignment): Likewise.
13287 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13288 (vectorizable_reduction): Likewise.
13289 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
13290 (vectorizable_operation, vectorizable_load): Likewise.
13291 * tree.c (build_same_sized_truth_vector_type): Likewise.
13292 * valtrack.c (cleanup_auto_inc_dec): Likewise.
13293 * var-tracking.c (emit_note_insn_var_location): Likewise.
13294 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
13295 (ADDR_VEC_ALIGN): Likewise.
13297 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13298 Alan Hayward <alan.hayward@arm.com>
13299 David Sherwood <david.sherwood@arm.com>
13301 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
13303 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13304 or if measurement_type is polynomial.
13305 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
13306 * combine.c (make_extraction): Likewise.
13307 * dse.c (find_shift_sequence): Likewise.
13308 * dwarf2out.c (mem_loc_descriptor): Likewise.
13309 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
13310 (extract_bit_field, extract_low_bits): Likewise.
13311 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
13312 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
13313 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
13314 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
13315 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13316 * reload.c (find_reloads): Likewise.
13317 * reload1.c (alter_reg): Likewise.
13318 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
13319 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
13320 * tree-if-conv.c (predicate_mem_writes): Likewise.
13321 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
13322 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
13323 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
13324 * valtrack.c (dead_debug_insert_temp): Likewise.
13325 * varasm.c (mergeable_constant_section): Likewise.
13326 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
13328 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13329 Alan Hayward <alan.hayward@arm.com>
13330 David Sherwood <david.sherwood@arm.com>
13332 * expr.c (expand_assignment): Cope with polynomial mode sizes
13333 when assigning to a CONCAT.
13335 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13336 Alan Hayward <alan.hayward@arm.com>
13337 David Sherwood <david.sherwood@arm.com>
13339 * machmode.h (mode_precision): Change from unsigned short to
13341 (mode_to_precision): Return a poly_uint16 rather than an unsigned
13343 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
13344 or if measurement_type is not polynomial.
13345 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
13346 in which the mode is already known to be a scalar_int_mode.
13347 * genmodes.c (emit_mode_precision): Change the type of mode_precision
13348 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
13350 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13351 for GET_MODE_PRECISION.
13352 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13353 for GET_MODE_PRECISION.
13354 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
13356 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
13357 (expand_field_assignment, make_extraction): Likewise.
13358 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
13359 (get_last_value): Likewise.
13360 * convert.c (convert_to_integer_1): Likewise.
13361 * cse.c (cse_insn): Likewise.
13362 * expr.c (expand_expr_real_1): Likewise.
13363 * lra-constraints.c (simplify_operand_subreg): Likewise.
13364 * optabs-query.c (can_atomic_load_p): Likewise.
13365 * optabs.c (expand_atomic_load): Likewise.
13366 (expand_atomic_store): Likewise.
13367 * ree.c (combine_reaching_defs): Likewise.
13368 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
13369 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
13370 * tree.h (type_has_mode_precision_p): Likewise.
13371 * ubsan.c (instrument_si_overflow): Likewise.
13373 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13374 Alan Hayward <alan.hayward@arm.com>
13375 David Sherwood <david.sherwood@arm.com>
13377 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
13378 polynomial numbers of units.
13379 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
13380 (valid_vector_subparts_p): New function.
13381 (build_vector_type): Remove temporary shim and take the number
13382 of units as a poly_uint64 rather than an int.
13383 (build_opaque_vector_type): Take the number of units as a
13384 poly_uint64 rather than an int.
13385 * tree.c (build_vector_from_ctor): Handle polynomial
13386 TYPE_VECTOR_SUBPARTS.
13387 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
13388 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
13389 (build_vector_from_val): If the number of units is variable,
13390 use build_vec_duplicate_cst for constant operands and
13391 VEC_DUPLICATE_EXPR otherwise.
13392 (make_vector_type): Remove temporary is_constant ().
13393 (build_vector_type, build_opaque_vector_type): Take the number of
13394 units as a poly_uint64 rather than an int.
13395 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
13397 * cfgexpand.c (expand_debug_expr): Likewise.
13398 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
13399 (store_constructor, expand_expr_real_1): Likewise.
13400 (const_scalar_mask_from_tree): Likewise.
13401 * fold-const-call.c (fold_const_reduction): Likewise.
13402 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
13403 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
13404 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
13405 (fold_relational_const): Likewise.
13406 (native_interpret_vector): Likewise. Change the size from an
13407 int to an unsigned int.
13408 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
13409 TYPE_VECTOR_SUBPARTS.
13410 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
13411 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
13412 duplicating a non-constant operand into a variable-length vector.
13413 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
13414 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
13415 * ipa-icf.c (sem_variable::equals): Likewise.
13416 * match.pd: Likewise.
13417 * omp-simd-clone.c (simd_clone_subparts): Likewise.
13418 * print-tree.c (print_node): Likewise.
13419 * stor-layout.c (layout_type): Likewise.
13420 * targhooks.c (default_builtin_vectorization_cost): Likewise.
13421 * tree-cfg.c (verify_gimple_comparison): Likewise.
13422 (verify_gimple_assign_binary): Likewise.
13423 (verify_gimple_assign_ternary): Likewise.
13424 (verify_gimple_assign_single): Likewise.
13425 * tree-pretty-print.c (dump_generic_node): Likewise.
13426 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13427 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
13428 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
13429 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
13430 (vect_shift_permute_load_chain): Likewise.
13431 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
13432 (expand_vector_condition, optimize_vector_constructor): Likewise.
13433 (lower_vec_perm, get_compute_type): Likewise.
13434 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13435 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
13436 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
13437 (vect_recog_mask_conversion_pattern): Likewise.
13438 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
13439 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
13440 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13441 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
13442 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
13443 (vectorizable_shift, vectorizable_operation, vectorizable_store)
13444 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
13445 (supportable_widening_operation): Likewise.
13446 (supportable_narrowing_operation): Likewise.
13447 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
13449 * varasm.c (output_constant): Likewise.
13451 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13452 Alan Hayward <alan.hayward@arm.com>
13453 David Sherwood <david.sherwood@arm.com>
13455 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
13456 so that both the length == 3 and length != 3 cases set up their
13457 own permute vectors. Add comments explaining why we know the
13458 number of elements is constant.
13459 (vect_permute_load_chain): Likewise.
13461 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13462 Alan Hayward <alan.hayward@arm.com>
13463 David Sherwood <david.sherwood@arm.com>
13465 * machmode.h (mode_nunits): Change from unsigned char to
13467 (ONLY_FIXED_SIZE_MODES): New macro.
13468 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
13469 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
13470 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
13472 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
13473 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
13474 or if measurement_type is not polynomial.
13475 * genmodes.c (ZERO_COEFFS): New macro.
13476 (emit_mode_nunits_inline): Make mode_nunits_inline return a
13478 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
13479 Use ZERO_COEFFS when emitting initializers.
13480 * data-streamer.h (bp_pack_poly_value): New function.
13481 (bp_unpack_poly_value): Likewise.
13482 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13483 for GET_MODE_NUNITS.
13484 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13485 for GET_MODE_NUNITS.
13486 * tree.c (make_vector_type): Remove temporary shim and make
13487 the real function take the number of units as a poly_uint64
13488 rather than an int.
13489 (build_vector_type_for_mode): Handle polynomial nunits.
13490 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
13491 * emit-rtl.c (const_vec_series_p_1): Likewise.
13492 (gen_rtx_CONST_VECTOR): Likewise.
13493 * fold-const.c (test_vec_duplicate_folding): Likewise.
13494 * genrecog.c (validate_pattern): Likewise.
13495 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
13496 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13497 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
13498 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
13499 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
13500 * rtlanal.c (subreg_get_info): Likewise.
13501 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13502 (vect_grouped_load_supported): Likewise.
13503 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
13504 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
13505 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
13506 (simplify_const_unary_operation, simplify_binary_operation_1)
13507 (simplify_const_binary_operation, simplify_ternary_operation)
13508 (test_vector_ops_duplicate, test_vector_ops): Likewise.
13509 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
13510 instead of CONST_VECTOR_NUNITS.
13511 * varasm.c (output_constant_pool_2): Likewise.
13512 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
13513 explicit-encoded elements in the XVEC for variable-length vectors.
13515 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13517 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
13519 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13520 Alan Hayward <alan.hayward@arm.com>
13521 David Sherwood <david.sherwood@arm.com>
13523 * coretypes.h (fixed_size_mode): Declare.
13524 (fixed_size_mode_pod): New typedef.
13525 * builtins.h (target_builtins::x_apply_args_mode)
13526 (target_builtins::x_apply_result_mode): Change type to
13527 fixed_size_mode_pod.
13528 * builtins.c (apply_args_size, apply_result_size, result_vector)
13529 (expand_builtin_apply_args_1, expand_builtin_apply)
13530 (expand_builtin_return): Update accordingly.
13532 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13534 * cse.c (hash_rtx_cb): Hash only the encoded elements.
13535 * cselib.c (cselib_hash_rtx): Likewise.
13536 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
13537 CONST_VECTOR encoding.
13539 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13540 Jeff Law <law@redhat.com>
13543 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
13544 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
13545 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
13546 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
13549 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
13550 explicitly probe *sp in a noreturn function if there were any callee
13551 register saves or frame pointer is needed.
13553 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13556 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
13557 BLKmode for ternary, binary or unary expressions.
13560 * var-tracking.c (delete_vta_debug_insn): New inline function.
13561 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
13562 insns from get_insns () to NULL instead of each bb separately.
13563 Use delete_vta_debug_insn. No longer static.
13564 (vt_debug_insns_local, variable_tracking_main_1): Adjust
13565 delete_vta_debug_insns callers.
13566 * rtl.h (delete_vta_debug_insns): Declare.
13567 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
13568 instead of variable_tracking_main.
13570 2018-01-03 Martin Sebor <msebor@redhat.com>
13572 PR tree-optimization/83603
13573 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
13574 arguments past the endof the argument list in functions declared
13575 without a prototype.
13576 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
13577 Avoid checking when arguments are null.
13579 2018-01-03 Martin Sebor <msebor@redhat.com>
13582 * doc/extend.texi (attribute const): Fix a typo.
13583 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
13584 issuing -Wsuggest-attribute for void functions.
13586 2018-01-03 Martin Sebor <msebor@redhat.com>
13588 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
13589 offset_int::from instead of wide_int::to_shwi.
13590 (maybe_diag_overlap): Remove assertion.
13591 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
13592 * gimple-ssa-sprintf.c (format_directive): Same.
13593 (parse_directive): Same.
13594 (sprintf_dom_walker::compute_format_length): Same.
13595 (try_substitute_return_value): Same.
13597 2018-01-03 Jeff Law <law@redhat.com>
13599 PR middle-end/83654
13600 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
13601 non-constant residual for zero at runtime and avoid probing in
13602 that case. Reorganize code for trailing problem to mirror handling
13605 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13607 PR tree-optimization/83501
13608 * tree-ssa-strlen.c (get_string_cst): New.
13609 (handle_char_store): Call get_string_cst.
13611 2018-01-03 Martin Liska <mliska@suse.cz>
13613 PR tree-optimization/83593
13614 * tree-ssa-strlen.c: Include tree-cfg.h.
13615 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
13616 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
13617 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
13619 (strlen_dom_walker::before_dom_children): Call
13620 gimple_purge_dead_eh_edges. Dump tranformation with details
13622 (strlen_dom_walker::before_dom_children): Update call by adding
13623 new argument cleanup_eh.
13624 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
13626 2018-01-03 Martin Liska <mliska@suse.cz>
13629 * cif-code.def (VARIADIC_THUNK): New enum value.
13630 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
13633 2018-01-03 Jan Beulich <jbeulich@suse.com>
13635 * sse.md (mov<mode>_internal): Tighten condition for when to use
13636 vmovdqu<ssescalarsize> for TI and OI modes.
13638 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13640 Update copyright years.
13642 2018-01-03 Martin Liska <mliska@suse.cz>
13645 * ipa-visibility.c (function_and_variable_visibility): Skip
13646 functions with noipa attribure.
13648 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13650 * gcc.c (process_command): Update copyright notice dates.
13651 * gcov-dump.c (print_version): Ditto.
13652 * gcov.c (print_version): Ditto.
13653 * gcov-tool.c (print_version): Ditto.
13654 * gengtype.c (create_file): Ditto.
13655 * doc/cpp.texi: Bump @copying's copyright year.
13656 * doc/cppinternals.texi: Ditto.
13657 * doc/gcc.texi: Ditto.
13658 * doc/gccint.texi: Ditto.
13659 * doc/gcov.texi: Ditto.
13660 * doc/install.texi: Ditto.
13661 * doc/invoke.texi: Ditto.
13663 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13665 * vector-builder.h (vector_builder::m_full_nelts): Change from
13666 unsigned int to poly_uint64.
13667 (vector_builder::full_nelts): Update prototype accordingly.
13668 (vector_builder::new_vector): Likewise.
13669 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
13670 (vector_builder::operator ==): Likewise.
13671 (vector_builder::finalize): Likewise.
13672 * int-vector-builder.h (int_vector_builder::int_vector_builder):
13673 Take the number of elements as a poly_uint64 rather than an
13675 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
13676 from unsigned int to poly_uint64.
13677 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
13678 (vec_perm_indices::new_vector): Likewise.
13679 (vec_perm_indices::length): Likewise.
13680 (vec_perm_indices::nelts_per_input): Likewise.
13681 (vec_perm_indices::input_nelts): Likewise.
13682 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
13683 number of elements per input as a poly_uint64 rather than an
13684 unsigned int. Use the original encoding for variable-length
13685 vectors, rather than clamping each individual element.
13686 For the second and subsequent elements in each pattern,
13687 clamp the step and base before clamping their sum.
13688 (vec_perm_indices::series_p): Handle polynomial element counts.
13689 (vec_perm_indices::all_in_range_p): Likewise.
13690 (vec_perm_indices_to_tree): Likewise.
13691 (vec_perm_indices_to_rtx): Likewise.
13692 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
13693 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
13694 (tree_vector_builder::new_binary_operation): Handle polynomial
13695 element counts. Return false if we need to know the number
13696 of elements at compile time.
13697 * fold-const.c (fold_vec_perm): Punt if the number of elements
13698 isn't known at compile time.
13700 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13702 * vec-perm-indices.h (vec_perm_builder): Change element type
13703 from HOST_WIDE_INT to poly_int64.
13704 (vec_perm_indices::element_type): Update accordingly.
13705 (vec_perm_indices::clamp): Handle polynomial element_types.
13706 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13707 (vec_perm_indices::all_in_range_p): Likewise.
13708 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
13710 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
13711 polynomial vec_perm_indices element types.
13712 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
13713 * fold-const.c (fold_vec_perm): Likewise.
13714 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
13715 * tree-vect-generic.c (lower_vec_perm): Likewise.
13716 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13717 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
13718 element type to HOST_WIDE_INT.
13720 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13721 Alan Hayward <alan.hayward@arm.com>
13722 David Sherwood <david.sherwood@arm.com>
13724 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
13725 rather than an int. Use plus_constant.
13726 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
13727 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
13729 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13730 Alan Hayward <alan.hayward@arm.com>
13731 David Sherwood <david.sherwood@arm.com>
13733 * calls.c (emit_call_1, expand_call): Change struct_value_size from
13734 a HOST_WIDE_INT to a poly_int64.
13736 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13737 Alan Hayward <alan.hayward@arm.com>
13738 David Sherwood <david.sherwood@arm.com>
13740 * calls.c (load_register_parameters): Cope with polynomial
13741 mode sizes. Require a constant size for BLKmode parameters
13742 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
13743 forces a parameter to be padded at the lsb end in order to
13744 fill a complete number of words, require the parameter size
13745 to be ordered wrt UNITS_PER_WORD.
13747 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13748 Alan Hayward <alan.hayward@arm.com>
13749 David Sherwood <david.sherwood@arm.com>
13751 * reload1.c (spill_stack_slot_width): Change element type
13752 from unsigned int to poly_uint64_pod.
13753 (alter_reg): Treat mode sizes as polynomial.
13755 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13756 Alan Hayward <alan.hayward@arm.com>
13757 David Sherwood <david.sherwood@arm.com>
13759 * reload.c (complex_word_subreg_p): New function.
13760 (reload_inner_reg_of_subreg, push_reload): Use it.
13762 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13763 Alan Hayward <alan.hayward@arm.com>
13764 David Sherwood <david.sherwood@arm.com>
13766 * lra-constraints.c (process_alt_operands): Reject matched
13767 operands whose sizes aren't ordered.
13768 (match_reload): Refer to this check here.
13770 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13771 Alan Hayward <alan.hayward@arm.com>
13772 David Sherwood <david.sherwood@arm.com>
13774 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
13775 that the mode size is in the set {1, 2, 4, 8, 16}.
13777 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13778 Alan Hayward <alan.hayward@arm.com>
13779 David Sherwood <david.sherwood@arm.com>
13781 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
13782 Use plus_constant instead of gen_rtx_PLUS.
13784 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13785 Alan Hayward <alan.hayward@arm.com>
13786 David Sherwood <david.sherwood@arm.com>
13788 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
13789 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
13790 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
13791 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
13792 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
13793 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
13794 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
13795 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
13796 * config/i386/i386.c (ix86_push_rounding): ...this new function.
13797 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
13799 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
13800 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
13801 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
13802 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
13803 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
13804 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
13805 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
13806 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
13807 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
13808 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
13810 * expr.c (emit_move_resolve_push): Treat the input and result
13811 of PUSH_ROUNDING as a poly_int64.
13812 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
13813 (emit_push_insn): Likewise.
13814 * lra-eliminations.c (mark_not_eliminable): Likewise.
13815 * recog.c (push_operand): Likewise.
13816 * reload1.c (elimination_effects): Likewise.
13817 * rtlanal.c (nonzero_bits1): Likewise.
13818 * calls.c (store_one_arg): Likewise. Require the padding to be
13819 known at compile time.
13821 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13822 Alan Hayward <alan.hayward@arm.com>
13823 David Sherwood <david.sherwood@arm.com>
13825 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
13826 Use plus_constant instead of gen_rtx_PLUS.
13828 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13829 Alan Hayward <alan.hayward@arm.com>
13830 David Sherwood <david.sherwood@arm.com>
13832 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
13833 rather than an int.
13835 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13836 Alan Hayward <alan.hayward@arm.com>
13837 David Sherwood <david.sherwood@arm.com>
13839 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
13840 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
13841 via stack temporaries. Treat the mode size as polynomial too.
13843 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13844 Alan Hayward <alan.hayward@arm.com>
13845 David Sherwood <david.sherwood@arm.com>
13847 * expr.c (expand_expr_real_2): When handling conversions involving
13848 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
13849 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
13850 as a poly_uint64 too.
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 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
13858 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13859 Alan Hayward <alan.hayward@arm.com>
13860 David Sherwood <david.sherwood@arm.com>
13862 * combine.c (can_change_dest_mode): Handle polynomial
13863 REGMODE_NATURAL_SIZE.
13864 * expmed.c (store_bit_field_1): Likewise.
13865 * expr.c (store_constructor): Likewise.
13866 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
13867 and polynomial REGMODE_NATURAL_SIZE.
13868 (gen_lowpart_common): Likewise.
13869 * reginfo.c (record_subregs_of_mode): Likewise.
13870 * rtlanal.c (read_modify_subreg_p): Likewise.
13872 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13873 Alan Hayward <alan.hayward@arm.com>
13874 David Sherwood <david.sherwood@arm.com>
13876 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
13877 numbers of elements.
13879 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13880 Alan Hayward <alan.hayward@arm.com>
13881 David Sherwood <david.sherwood@arm.com>
13883 * match.pd: Cope with polynomial numbers of vector elements.
13885 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13886 Alan Hayward <alan.hayward@arm.com>
13887 David Sherwood <david.sherwood@arm.com>
13889 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
13890 in a POINTER_PLUS_EXPR.
13892 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13893 Alan Hayward <alan.hayward@arm.com>
13894 David Sherwood <david.sherwood@arm.com>
13896 * omp-simd-clone.c (simd_clone_subparts): New function.
13897 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
13898 (ipa_simd_modify_function_body): Likewise.
13900 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13901 Alan Hayward <alan.hayward@arm.com>
13902 David Sherwood <david.sherwood@arm.com>
13904 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
13905 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
13906 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
13907 (expand_vector_condition, vector_element): Likewise.
13908 (subparts_gt): New function.
13909 (get_compute_type): Use subparts_gt.
13910 (count_type_subparts): Delete.
13911 (expand_vector_operations_1): Use subparts_gt instead of
13912 count_type_subparts.
13914 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13915 Alan Hayward <alan.hayward@arm.com>
13916 David Sherwood <david.sherwood@arm.com>
13918 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
13919 (vect_compile_time_alias): ...this new function. Do the calculation
13920 on poly_ints rather than trees.
13921 (vect_prune_runtime_alias_test_list): Update call accordingly.
13923 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13924 Alan Hayward <alan.hayward@arm.com>
13925 David Sherwood <david.sherwood@arm.com>
13927 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
13929 (vect_schedule_slp_instance): Likewise.
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-vect-slp.c (vect_get_and_check_slp_defs): Reject
13936 constant and extern definitions for variable-length vectors.
13937 (vect_get_constant_vectors): Note that the number of units
13938 is known to be constant.
13940 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13941 Alan Hayward <alan.hayward@arm.com>
13942 David Sherwood <david.sherwood@arm.com>
13944 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
13945 of units as polynomial. Choose between WIDE and NARROW based
13948 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13949 Alan Hayward <alan.hayward@arm.com>
13950 David Sherwood <david.sherwood@arm.com>
13952 * tree-vect-stmts.c (simd_clone_subparts): New function.
13953 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
13955 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13956 Alan Hayward <alan.hayward@arm.com>
13957 David Sherwood <david.sherwood@arm.com>
13959 * tree-vect-stmts.c (vectorizable_call): Treat the number of
13960 vectors as polynomial. Use build_index_vector for
13961 IFN_GOMP_SIMD_LANE.
13963 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13964 Alan Hayward <alan.hayward@arm.com>
13965 David Sherwood <david.sherwood@arm.com>
13967 * tree-vect-stmts.c (get_load_store_type): Treat the number of
13968 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
13969 for variable-length vectors.
13970 (vectorizable_mask_load_store): Treat the number of units as
13971 polynomial, asserting that it is constant if the condition has
13972 already been enforced.
13973 (vectorizable_store, vectorizable_load): Likewise.
13975 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13976 Alan Hayward <alan.hayward@arm.com>
13977 David Sherwood <david.sherwood@arm.com>
13979 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
13980 of units as polynomial. Punt if we can't tell at compile time
13981 which vector contains the final result.
13983 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13984 Alan Hayward <alan.hayward@arm.com>
13985 David Sherwood <david.sherwood@arm.com>
13987 * tree-vect-loop.c (vectorizable_induction): Treat the number
13988 of units as polynomial. Punt on SLP inductions. Use an integer
13989 VEC_SERIES_EXPR for variable-length integer reductions. Use a
13990 cast of such a series for variable-length floating-point
13993 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13994 Alan Hayward <alan.hayward@arm.com>
13995 David Sherwood <david.sherwood@arm.com>
13997 * tree.h (build_index_vector): Declare.
13998 * tree.c (build_index_vector): New function.
13999 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
14000 of units as polynomial, forcibly converting it to a constant if
14001 vectorizable_reduction has already enforced the condition.
14002 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
14003 to create a {1,2,3,...} vector.
14004 (vectorizable_reduction): Treat the number of units as polynomial.
14005 Choose vectype_in based on the largest scalar element size rather
14006 than the smallest number of units. Enforce the restrictions
14009 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14010 Alan Hayward <alan.hayward@arm.com>
14011 David Sherwood <david.sherwood@arm.com>
14013 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
14014 number of units as polynomial.
14016 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14017 Alan Hayward <alan.hayward@arm.com>
14018 David Sherwood <david.sherwood@arm.com>
14020 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
14021 * target.def (autovectorize_vector_sizes): Return the vector sizes
14022 by pointer, using vector_sizes rather than a bitmask.
14023 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
14024 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
14025 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
14027 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
14028 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
14029 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
14030 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
14031 * omp-general.c (omp_max_vf): Likewise.
14032 * omp-low.c (omp_clause_aligned_alignment): Likewise.
14033 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
14034 * tree-vect-loop.c (vect_analyze_loop): Likewise.
14035 * tree-vect-slp.c (vect_slp_bb): Likewise.
14036 * doc/tm.texi: Regenerate.
14037 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
14039 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
14040 the vector size as a poly_uint64 rather than an unsigned int.
14041 (current_vector_size): Change from an unsigned int to a poly_uint64.
14042 (get_vectype_for_scalar_type): Update accordingly.
14043 * tree.h (build_truth_vector_type): Take the size and number of
14044 units as a poly_uint64 rather than an unsigned int.
14045 (build_vector_type): Add a temporary overload that takes
14046 the number of units as a poly_uint64 rather than an unsigned int.
14047 * tree.c (make_vector_type): Likewise.
14048 (build_truth_vector_type): Take the number of units as a poly_uint64
14049 rather than an unsigned int.
14051 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14052 Alan Hayward <alan.hayward@arm.com>
14053 David Sherwood <david.sherwood@arm.com>
14055 * target.def (get_mask_mode): Take the number of units and length
14056 as poly_uint64s rather than unsigned ints.
14057 * targhooks.h (default_get_mask_mode): Update accordingly.
14058 * targhooks.c (default_get_mask_mode): Likewise.
14059 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
14060 * doc/tm.texi: Regenerate.
14062 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14063 Alan Hayward <alan.hayward@arm.com>
14064 David Sherwood <david.sherwood@arm.com>
14066 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
14067 * omp-general.c (omp_max_vf): Likewise.
14068 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
14069 (expand_omp_simd): Handle polynomial safelen.
14070 * omp-low.c (omplow_simd_context): Add a default constructor.
14071 (omplow_simd_context::max_vf): Change from int to poly_uint64.
14072 (lower_rec_simd_input_clauses): Update accordingly.
14073 (lower_rec_input_clauses): Likewise.
14075 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14076 Alan Hayward <alan.hayward@arm.com>
14077 David Sherwood <david.sherwood@arm.com>
14079 * tree-vectorizer.h (vect_nunits_for_cost): New function.
14080 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
14081 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
14082 (vect_analyze_slp_cost): Likewise.
14083 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
14084 (vect_model_load_cost): Likewise.
14086 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14087 Alan Hayward <alan.hayward@arm.com>
14088 David Sherwood <david.sherwood@arm.com>
14090 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
14091 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
14092 from an unsigned int * to a poly_uint64_pod *.
14093 (calculate_unrolling_factor): New function.
14094 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
14096 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14097 Alan Hayward <alan.hayward@arm.com>
14098 David Sherwood <david.sherwood@arm.com>
14100 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
14101 from an unsigned int to a poly_uint64.
14102 (_loop_vec_info::slp_unrolling_factor): Likewise.
14103 (_loop_vec_info::vectorization_factor): Change from an int
14105 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
14106 (vect_get_num_vectors): New function.
14107 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
14108 (vect_get_num_copies): Use vect_get_num_vectors.
14109 (vect_analyze_data_ref_dependences): Change max_vf from an int *
14110 to an unsigned int *.
14111 (vect_analyze_data_refs): Change min_vf from an int * to a
14113 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14114 than an unsigned HOST_WIDE_INT.
14115 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
14116 (vect_analyze_data_ref_dependence): Change max_vf from an int *
14117 to an unsigned int *.
14118 (vect_analyze_data_ref_dependences): Likewise.
14119 (vect_compute_data_ref_alignment): Handle polynomial vf.
14120 (vect_enhance_data_refs_alignment): Likewise.
14121 (vect_prune_runtime_alias_test_list): Likewise.
14122 (vect_shift_permute_load_chain): Likewise.
14123 (vect_supportable_dr_alignment): Likewise.
14124 (dependence_distance_ge_vf): Take the vectorization factor as a
14125 poly_uint64 rather than an unsigned HOST_WIDE_INT.
14126 (vect_analyze_data_refs): Change min_vf from an int * to a
14128 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
14129 vfm1 as a poly_uint64 rather than an int. Make the same change
14130 for the returned bound_scalar.
14131 (vect_gen_vector_loop_niters): Handle polynomial vf.
14132 (vect_do_peeling): Likewise. Update call to
14133 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
14134 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
14136 * tree-vect-loop.c (vect_determine_vectorization_factor)
14137 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
14138 (vect_get_known_peeling_cost): Likewise.
14139 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
14140 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
14141 (vect_transform_loop): Likewise. Use the lowest possible VF when
14142 updating the upper bounds of the loop.
14143 (vect_min_worthwhile_factor): Make static. Return an unsigned int
14144 rather than an int.
14145 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
14146 polynomial unroll factors.
14147 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
14148 (vect_make_slp_decision): Likewise.
14149 (vect_supported_load_permutation_p): Likewise, and polynomial
14151 (vect_analyze_slp_cost): Handle polynomial vf.
14152 (vect_slp_analyze_node_operations): Likewise.
14153 (vect_slp_analyze_bb_1): Likewise.
14154 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14155 than an unsigned HOST_WIDE_INT.
14156 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
14157 (vectorizable_load): Handle polynomial vf.
14158 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
14160 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
14162 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14163 Alan Hayward <alan.hayward@arm.com>
14164 David Sherwood <david.sherwood@arm.com>
14166 * match.pd: Handle bit operations involving three constants
14167 and try to fold one pair.
14169 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14171 * tree-vect-loop-manip.c: Include gimple-fold.h.
14172 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
14173 niters_maybe_zero parameters. Handle other cases besides a step of 1.
14174 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
14175 Add a path that uses a step of VF instead of 1, but disable it
14177 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
14178 and niters_no_overflow parameters. Update calls to
14179 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
14180 Create a new SSA name if the latter choses to use a ste other
14181 than zero, and return it via niters_vector_mult_vf_var.
14182 * tree-vect-loop.c (vect_transform_loop): Update calls to
14183 vect_do_peeling, vect_gen_vector_loop_niters and
14184 slpeel_make_loop_iterate_ntimes.
14185 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
14186 (vect_gen_vector_loop_niters): Update declarations after above changes.
14188 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
14190 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
14191 128-bit round to integer instructions.
14192 (ceil<mode>2): Likewise.
14193 (btrunc<mode>2): Likewise.
14194 (round<mode>2): Likewise.
14196 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14198 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
14199 unaligned VSX load/store on P8/P9.
14200 (expand_block_clear): Allow the use of unaligned VSX
14201 load/store on P8/P9.
14203 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14205 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
14207 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
14208 swap associated with both a load and a store.
14210 2018-01-02 Andrew Waterman <andrew@sifive.com>
14212 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
14213 * config/riscv/riscv.md (clear_cache): Use it.
14215 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
14217 * web.c: Remove out-of-date comment.
14219 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14221 * expr.c (fixup_args_size_notes): Check that any existing
14222 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
14223 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
14224 (emit_single_push_insn): ...here.
14226 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14228 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
14229 (const_vector_encoded_nelts): New function.
14230 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
14231 (const_vector_int_elt, const_vector_elt): Declare.
14232 * emit-rtl.c (const_vector_int_elt_1): New function.
14233 (const_vector_elt): Likewise.
14234 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
14235 of CONST_VECTOR_ELT.
14237 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14239 * expr.c: Include rtx-vector-builder.h.
14240 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
14241 directly on the tree encoding.
14242 (const_vector_from_tree): Likewise.
14243 * optabs.c: Include rtx-vector-builder.h.
14244 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
14245 sequence of "u" values.
14246 * vec-perm-indices.c: Include rtx-vector-builder.h.
14247 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
14248 directly on the vec_perm_indices encoding.
14250 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14252 * doc/rtl.texi (const_vector): Describe new encoding scheme.
14253 * Makefile.in (OBJS): Add rtx-vector-builder.o.
14254 * rtx-vector-builder.h: New file.
14255 * rtx-vector-builder.c: Likewise.
14256 * rtl.h (rtx_def::u2): Add a const_vector field.
14257 (CONST_VECTOR_NPATTERNS): New macro.
14258 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
14259 (CONST_VECTOR_DUPLICATE_P): Likewise.
14260 (CONST_VECTOR_STEPPED_P): Likewise.
14261 (CONST_VECTOR_ENCODED_ELT): Likewise.
14262 (const_vec_duplicate_p): Check for a duplicated vector encoding.
14263 (unwrap_const_vec_duplicate): Likewise.
14264 (const_vec_series_p): Check for a non-duplicated vector encoding.
14265 Say that the function only returns true for integer vectors.
14266 * emit-rtl.c: Include rtx-vector-builder.h.
14267 (gen_const_vec_duplicate_1): Delete.
14268 (gen_const_vector): Call gen_const_vec_duplicate instead of
14269 gen_const_vec_duplicate_1.
14270 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
14271 (gen_const_vec_duplicate): Use rtx_vector_builder.
14272 (gen_const_vec_series): Likewise.
14273 (gen_rtx_CONST_VECTOR): Likewise.
14274 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
14275 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14276 Build a new vector rather than modifying a CONST_VECTOR in-place.
14277 (handle_special_swappables): Update call accordingly.
14278 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
14279 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14280 Build a new vector rather than modifying a CONST_VECTOR in-place.
14281 (handle_special_swappables): Update call accordingly.
14283 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14285 * simplify-rtx.c (simplify_const_binary_operation): Use
14286 CONST_VECTOR_ELT instead of XVECEXP.
14288 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14290 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
14291 the selector elements to be different from the data elements
14292 if the selector is a VECTOR_CST.
14293 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
14294 ssizetype for the selector.
14296 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14298 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
14299 before testing each element individually.
14300 * tree-vect-generic.c (lower_vec_perm): Likewise.
14302 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14304 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
14305 * selftest-run-tests.c (selftest::run_tests): Call it.
14306 * vector-builder.h (vector_builder::operator ==): New function.
14307 (vector_builder::operator !=): Likewise.
14308 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
14309 (vec_perm_indices::all_from_input_p): New function.
14310 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14311 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
14312 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
14313 instead of reading the VECTOR_CST directly. Detect whether both
14314 vector inputs are the same before constructing the vec_perm_indices,
14315 and update the number of inputs argument accordingly. Use the
14316 utility functions added above. Only construct sel2 if we need to.
14318 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14320 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
14321 the broadcast of the low byte.
14322 (expand_mult_highpart): Use an explicit encoding for the permutes.
14323 * optabs-query.c (can_mult_highpart_p): Likewise.
14324 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
14325 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14326 (vectorizable_bswap): Likewise.
14327 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
14328 explicit encoding for the power-of-2 permutes.
14329 (vect_permute_store_chain): Likewise.
14330 (vect_grouped_load_supported): Likewise.
14331 (vect_permute_load_chain): Likewise.
14333 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14335 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
14336 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
14337 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
14338 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14339 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
14340 (vect_gen_perm_mask_any): Likewise.
14342 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14344 * int-vector-builder.h: New file.
14345 * vec-perm-indices.h: Include int-vector-builder.h.
14346 (vec_perm_indices): Redefine as an int_vector_builder.
14347 (auto_vec_perm_indices): Delete.
14348 (vec_perm_builder): Redefine as a stand-alone class.
14349 (vec_perm_indices::vec_perm_indices): New function.
14350 (vec_perm_indices::clamp): Likewise.
14351 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
14352 (vec_perm_indices::new_vector): New function.
14353 (vec_perm_indices::new_expanded_vector): Update for new
14354 vec_perm_indices class.
14355 (vec_perm_indices::rotate_inputs): New function.
14356 (vec_perm_indices::all_in_range_p): Operate directly on the
14357 encoded form, without computing elided elements.
14358 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
14359 encoding. Update for new vec_perm_indices class.
14360 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
14361 the given vec_perm_builder.
14362 (expand_vec_perm_var): Update vec_perm_builder constructor.
14363 (expand_mult_highpart): Use vec_perm_builder instead of
14364 auto_vec_perm_indices.
14365 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
14366 vec_perm_indices instead of auto_vec_perm_indices. Use a single
14367 or double series encoding as appropriate.
14368 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
14369 vec_perm_indices instead of auto_vec_perm_indices.
14370 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14371 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14372 (vect_permute_store_chain): Likewise.
14373 (vect_grouped_load_supported): Likewise.
14374 (vect_permute_load_chain): Likewise.
14375 (vect_shift_permute_load_chain): Likewise.
14376 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14377 (vect_transform_slp_perm_load): Likewise.
14378 (vect_schedule_slp_instance): Likewise.
14379 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14380 (vectorizable_mask_load_store): Likewise.
14381 (vectorizable_bswap): Likewise.
14382 (vectorizable_store): Likewise.
14383 (vectorizable_load): Likewise.
14384 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
14385 vec_perm_indices instead of auto_vec_perm_indices. Use
14386 tree_to_vec_perm_builder to read the vector from a tree.
14387 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
14388 vec_perm_builder instead of a vec_perm_indices.
14389 (have_whole_vector_shift): Use vec_perm_builder and
14390 vec_perm_indices instead of auto_vec_perm_indices. Leave the
14391 truncation to calc_vec_perm_mask_for_shift.
14392 (vect_create_epilog_for_reduction): Likewise.
14393 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
14394 from auto_vec_perm_indices to vec_perm_indices.
14395 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14396 instead of changing individual elements.
14397 (aarch64_vectorize_vec_perm_const): Use new_vector to install
14398 the vector in d.perm.
14399 * config/arm/arm.c (expand_vec_perm_d::perm): Change
14400 from auto_vec_perm_indices to vec_perm_indices.
14401 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14402 instead of changing individual elements.
14403 (arm_vectorize_vec_perm_const): Use new_vector to install
14404 the vector in d.perm.
14405 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
14406 Update vec_perm_builder constructor.
14407 (rs6000_expand_interleave): Likewise.
14408 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
14409 (rs6000_expand_interleave): Likewise.
14411 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14413 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
14414 to qimode could truncate the indices.
14415 * optabs.c (expand_vec_perm_var): Likewise.
14417 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14419 * Makefile.in (OBJS): Add vec-perm-indices.o.
14420 * vec-perm-indices.h: New file.
14421 * vec-perm-indices.c: Likewise.
14422 * target.h (vec_perm_indices): Replace with a forward class
14424 (auto_vec_perm_indices): Move to vec-perm-indices.h.
14425 * optabs.h: Include vec-perm-indices.h.
14426 (expand_vec_perm): Delete.
14427 (selector_fits_mode_p, expand_vec_perm_var): Declare.
14428 (expand_vec_perm_const): Declare.
14429 * target.def (vec_perm_const_ok): Replace with...
14430 (vec_perm_const): ...this new hook.
14431 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
14432 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
14433 * doc/tm.texi: Regenerate.
14434 * optabs.def (vec_perm_const): Delete.
14435 * doc/md.texi (vec_perm_const): Likewise.
14436 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
14437 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
14438 expand_vec_perm for constant permutation vectors. Assert that
14439 the mode of variable permutation vectors is the integer equivalent
14440 of the mode that is being permuted.
14441 * optabs-query.h (selector_fits_mode_p): Declare.
14442 * optabs-query.c: Include vec-perm-indices.h.
14443 (selector_fits_mode_p): New function.
14444 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
14445 is defined, instead of checking whether the vec_perm_const_optab
14446 exists. Use targetm.vectorize.vec_perm_const instead of
14447 targetm.vectorize.vec_perm_const_ok. Check whether the indices
14448 fit in the vector mode before using a variable permute.
14449 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
14450 vec_perm_indices instead of an rtx.
14451 (expand_vec_perm): Replace with...
14452 (expand_vec_perm_const): ...this new function. Take the selector
14453 as a vec_perm_indices rather than an rtx. Also take the mode of
14454 the selector. Update call to shift_amt_for_vec_perm_mask.
14455 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
14456 Use vec_perm_indices::new_expanded_vector to expand the original
14457 selector into bytes. Check whether the indices fit in the vector
14458 mode before using a variable permute.
14459 (expand_vec_perm_var): Make global.
14460 (expand_mult_highpart): Use expand_vec_perm_const.
14461 * fold-const.c: Includes vec-perm-indices.h.
14462 * tree-ssa-forwprop.c: Likewise.
14463 * tree-vect-data-refs.c: Likewise.
14464 * tree-vect-generic.c: Likewise.
14465 * tree-vect-loop.c: Likewise.
14466 * tree-vect-slp.c: Likewise.
14467 * tree-vect-stmts.c: Likewise.
14468 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
14470 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
14471 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
14472 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
14473 (aarch64_vectorize_vec_perm_const): ...this new function.
14474 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14475 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14476 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
14477 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
14478 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14479 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14480 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
14482 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
14483 check for NEON modes.
14484 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
14485 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
14486 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
14487 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
14489 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
14490 the old VEC_PERM_CONST conditions.
14491 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
14492 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
14493 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
14494 (ia64_vectorize_vec_perm_const_ok): Merge into...
14495 (ia64_vectorize_vec_perm_const): ...this new function.
14496 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
14497 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
14498 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
14499 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
14500 * config/mips/mips.c (mips_expand_vec_perm_const)
14501 (mips_vectorize_vec_perm_const_ok): Merge into...
14502 (mips_vectorize_vec_perm_const): ...this new function.
14503 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
14504 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
14505 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
14506 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
14507 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
14508 (rs6000_expand_vec_perm_const): Delete.
14509 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
14511 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14512 (altivec_expand_vec_perm_const_le): Take each operand individually.
14513 Operate on constant selectors rather than rtxes.
14514 (altivec_expand_vec_perm_const): Likewise. Update call to
14515 altivec_expand_vec_perm_const_le.
14516 (rs6000_expand_vec_perm_const): Delete.
14517 (rs6000_vectorize_vec_perm_const_ok): Delete.
14518 (rs6000_vectorize_vec_perm_const): New function.
14519 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14520 an element count and rtx array.
14521 (rs6000_expand_extract_even): Update call accordingly.
14522 (rs6000_expand_interleave): Likewise.
14523 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
14524 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
14525 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
14526 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
14527 (rs6000_expand_vec_perm_const): Delete.
14528 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14529 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14530 (altivec_expand_vec_perm_const_le): Take each operand individually.
14531 Operate on constant selectors rather than rtxes.
14532 (altivec_expand_vec_perm_const): Likewise. Update call to
14533 altivec_expand_vec_perm_const_le.
14534 (rs6000_expand_vec_perm_const): Delete.
14535 (rs6000_vectorize_vec_perm_const_ok): Delete.
14536 (rs6000_vectorize_vec_perm_const): New function. Remove stray
14537 reference to the SPE evmerge intructions.
14538 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14539 an element count and rtx array.
14540 (rs6000_expand_extract_even): Update call accordingly.
14541 (rs6000_expand_interleave): Likewise.
14542 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
14543 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
14545 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14547 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14549 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
14550 vector mode and that that mode matches the mode of the data
14552 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
14553 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
14554 directly using expand_vec_perm_1 when forcing selectors into
14556 (expand_vec_perm_var): New function, split out from expand_vec_perm.
14558 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14560 * optabs-query.h (can_vec_perm_p): Delete.
14561 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
14562 * optabs-query.c (can_vec_perm_p): Split into...
14563 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
14564 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
14565 particular selector is valid.
14566 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14567 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14568 (vect_grouped_load_supported): Likewise.
14569 (vect_shift_permute_load_chain): Likewise.
14570 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14571 (vect_transform_slp_perm_load): Likewise.
14572 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14573 (vectorizable_bswap): Likewise.
14574 (vect_gen_perm_mask_checked): Likewise.
14575 * fold-const.c (fold_ternary_loc): Likewise. Don't take
14576 implementations of variable permutation vectors into account
14577 when deciding which selector to use.
14578 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
14579 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
14580 with a false third argument.
14581 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
14582 to test whether the constant selector is valid and can_vec_perm_var_p
14583 to test whether a variable selector is valid.
14585 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14587 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
14588 * optabs-query.c (can_vec_perm_p): Likewise.
14589 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
14590 instead of vec_perm_indices.
14591 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
14592 (vect_gen_perm_mask_checked): Likewise,
14593 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
14594 (vect_gen_perm_mask_checked): Likewise,
14596 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14598 * optabs-query.h (qimode_for_vec_perm): Declare.
14599 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
14600 (qimode_for_vec_perm): ...this new function.
14601 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
14603 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14605 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
14606 does not have a conditional at the top.
14608 2018-01-02 Richard Biener <rguenther@suse.de>
14610 * ipa-inline.c (big_speedup_p): Fix expression.
14612 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14615 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
14618 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14622 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
14623 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
14624 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
14625 cond_taken_branch_cost 3->4.
14627 2018-01-01 Jakub Jelinek <jakub@redhat.com>
14629 PR tree-optimization/83581
14630 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
14631 TODO_cleanup_cfg if any changes have been made.
14633 PR middle-end/83608
14634 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
14635 convert_modes if target mode has the right side, but different mode
14638 PR middle-end/83609
14639 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
14640 last argument when extracting from CONCAT. If either from_real or
14641 from_imag is NULL, use expansion through memory. If result is not
14642 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
14643 the parts directly to inner mode, if even that fails, use expansion
14646 PR middle-end/83623
14647 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
14648 check for bswap in mode rather than HImode and use that in expand_unop
14651 Copyright (C) 2018 Free Software Foundation, Inc.
14653 Copying and distribution of this file, with or without modification,
14654 are permitted in any medium without royalty provided the copyright
14655 notice and this notice are preserved.