1 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
3 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
5 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
7 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
8 it down to simple_object_copy_lto_debug_sections.
9 (run_gcc): Determine incremental LTO link time and configure
10 lto1 into non-wpa mode, disable renaming of debug sections.
12 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
14 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
15 descriptions of various incorrectly documented functions.
17 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
20 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
21 address check not strict.
23 2018-05-30 Richard Biener <rguenther@suse.de>
25 PR tree-optimization/85964
26 * tracer.c (better_p): Drop initialized count check, we only
27 call the function with initialized counts now.
28 (find_best_successor): Do find a best edge if one
29 has uninitialized count.
30 (find_best_predecessor): Likewise. Do BB frequency check only
31 if count is initialized.
33 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
35 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
36 (aarch64_ldrstr_offset_compare): New.
37 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
39 (aarch64_gen_adjusted_ldpstp): Likewise.
41 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
43 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
44 Check for subset of GENERAL_REGS and FP_REGS.
45 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
48 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
50 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
51 and wi::to_poly_offset. Add the current offset and then check
52 whether the sum fits, rather than using an unchecked addition of
53 a checked term. Check for a shwi rather than a uhwi.
54 * expr.c (get_bit_range): Use tree_to_poly_uint64.
55 (store_constructor): Use poly_int_tree_p.
56 (expand_expr_real_1): Likewise.
57 * function.c (assign_temp): Likewise.
58 * fold-const.c (const_binop): Use poly_int_tree_p and
60 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
62 * ipa-icf-gimple.c (func_checker::compare_operand): Use
63 to_poly_offset for MEM offsets.
64 * ipa-icf.c (sem_variable::equals): Likewise.
65 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
66 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
67 wi::to_poly_offset for BIT_FIELD_REF offsets.
68 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
70 * var-tracking.c (emit_note_insn_var_location): Use
73 2018-05-29 Jim Wilson <jimw@sifive.com>
75 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
77 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
80 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
81 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
83 (sse4_1_round<mode>2): Use nonimmediate_operand
84 for operand 1 predicate.
86 2018-05-29 Martin Sebor <msebor@redhat.com>
87 Richard Biener <rguenther@suse.de>
90 * calls.c (get_size_range): Call determine_value_range instead
92 * tree-vrp.h (determine_value_range): Declared new function.
93 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
95 2018-05-29 Richard Biener <rguenther@suse.de>
97 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
98 sure to use non-pattern stmts for get_earlier_stmt arguments.
99 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
100 called on pattern stmts.
101 (get_later_stmt): Likewise.
103 2018-05-29 Martin Liska <mliska@suse.cz>
105 PR gcov-profile/85759
106 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
109 2018-05-29 Jakub Jelinek <jakub@redhat.com>
111 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
113 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
117 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
118 VEC_PACK_FLOAT_EXPR): New tree codes.
119 * tree-pretty-print.c (op_code_prio): Handle
120 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
121 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
122 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
123 * tree-inline.c (estimate_operator_cost): Likewise.
124 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
125 * fold-const.c (const_binop): Likewise.
126 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
127 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
128 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
129 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
130 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
131 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
132 * expr.c (expand_expr_real_2): Likewise.
133 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
134 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
135 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
137 * optabs.c (expand_widen_pattern_expr): For
138 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
139 sign from result type rather than operand's type.
140 (expand_binop_directly): For vec_packu_float_optab and
141 vec_packs_float_optab allow result type to be different from operand's
143 * optabs-tree.c (optab_for_tree_code): Handle
144 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
145 VEC_PACK_FLOAT_EXPR. Formatting fixes.
146 * tree-vect-generic.c (expand_vector_operations_1): Handle
147 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
149 * tree-vect-stmts.c (supportable_widening_operation): Handle
151 (supportable_narrowing_operation): Handle FLOAT_EXPR.
152 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
153 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
154 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
155 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
157 (vec_pack<floatprefix>_float_<mode>): New expander.
158 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
160 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
161 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
162 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
163 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
164 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
166 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
167 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
168 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
169 VEC_PACK_FLOAT_EXPR): Document.
171 2018-05-29 Richard Biener <rguenther@suse.de>
173 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
175 (stmt_vec_info_vec): Make pointer.
176 (init_stmt_vec_info_vec): Remove.
177 (free_stmt_vec_info_vec): Likewise.
178 (set_stmt_vec_info_vec): New function.
179 (free_stmt_vec_infos): Likewise.
180 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
181 (set_vinfo_for_stmt): Likewise.
182 (get_earlier_stmt): Likewise.
183 (get_later_stmt): Likewise.
184 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
185 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
186 (vec_info::~vec_info): Free stmt_vec_infos.
187 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
188 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
189 (pass_slp_vectorize::execute): Likewise.
190 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
191 (free_stmt_vec_info_vec): Likewise.
192 (set_stmt_vec_info_vec): New function.
193 (free_stmt_vec_infos): Likewise.
194 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
195 the global stmt_vec_info_vec.
196 * tree-parloops.c (gather_scalar_reductions): Use
197 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
200 2018-05-29 Richard Biener <rguenther@suse.de>
202 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
204 2018-05-29 Martin Liska <mliska@suse.cz>
205 David Malcolm <dmalcolm@redhat.com>
207 * vec.c (test_reverse): New.
208 (vec_c_tests): Add new test.
209 * vec.h (vl_ptr>::reverse): New function.
211 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
213 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
215 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
218 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
220 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
222 2018-05-28 Richard Biener <rguenther@suse.de>
224 PR tree-optimization/85933
225 * tree-vect-data-refs.c (vect_record_base_alignments): Only
226 look at stmts marked as vectorizable.
228 2018-05-28 Richard Biener <rguenther@suse.de>
230 PR tree-optimization/85934
231 * tree-vect-generic.c (expand_vector_operations_1): Hoist
232 vector boolean check before scalar optimization.
234 2018-05-28 Jakub Jelinek <jakub@redhat.com>
236 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
239 2018-05-28 Mark Wielaard <mark@klomp.org>
241 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
242 if it is an expression containing a minus sign.
244 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
246 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
248 2018-05-27 Paul Koning <ni1d@arrl.net>
250 * config/pdp11/pdp11.md (truncsihi2): Remove.
252 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
253 Chung-Ju Wu <jasonwucj@gmail.com>
255 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
257 (unaligned_store_dw): Ditto.
258 * config/nds32/nds32-memory-manipulation.c
259 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
260 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
261 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
262 (emit_setmem_word_loop): Rename to ...
263 (emit_setmem_doubleword_loop): ... this.
264 (nds32_gen_dup_4_byte_to_word_value): New function.
265 (nds32_gen_dup_8_byte_to_double_word_value): New function.
266 (nds32_expand_setmem_loop): Refine implementation.
267 (nds32_expand_setmem_loop_v3m): Ditto.
268 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
271 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
273 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
275 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
277 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
278 (nds32_init_machine_status): Initialize machine->attr_naked_p and
279 machine->attr_no_prologue_p.
280 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
281 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
282 (nds32_expand_epilogue): Consider attr_naked_p.
283 (nds32_expand_epilogue_v3pop): Likewise.
284 (nds32_can_use_return_insn): Likewise.
285 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
286 attr_no_prologue_p fields.
287 * config/nds32/nds32.opt (mret-in-naked-func): New option.
289 2018-05-27 Jakub Jelinek <jakub@redhat.com>
292 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
295 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
297 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
299 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
301 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
303 (*<floatsuffix>floatv2div2sf2): Rename to ...
304 (*float<floatunssuffix>v2div2sf2): ... this.
305 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
306 (float<floatunssuffix>v2div2sf2_mask): ... this.
307 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
308 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
309 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
311 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
313 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
315 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
317 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
319 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
321 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
322 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
323 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
324 gen_ufix_truncv8dfv8si2.
325 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
326 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
327 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
328 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
329 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
330 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
331 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
332 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
334 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
338 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
340 2018-05-25 Jim Wilson <jimw@sifive.com>
342 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
343 * config/riscv/riscv.c (struct machine_function): Add
344 interrupt_handler_p and attribute_checked_p fields.
345 (riscv_attribute_table): Add interrupt.
346 (riscv_interrupt_type_p): New.
347 (riscv_save_reg_p): Save extra regs for interrupt handler.
348 (riscv_use_save_libcall): Return false for interrupt handler.
349 (riscv_first_stack_step): Add forward declaration.
350 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
351 for interrupt handler with large frame. Use it for saved reg list.
352 (riscv_expand_prologue): Move flag_stack_usage_info support to
353 eliminate duplication.
354 (riscv_expand_epilogue): Generate mret for interrupt handler.
355 (riscv_epilogue_uses): New.
356 (riscv_can_use_return_insn): Return false for interrupt handler.
357 (riscv_function_ok_for_sibcall): Likewise.
358 (riscv_set_current_function): Add interrupt handler support.
359 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
360 * config/riscv/riscv.md (UNSPECV_MRET): New.
362 (riscv_frflags, riscv_fsflags): Use tab after opcode.
364 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
366 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
368 PR tree-optimization/85712
369 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
370 this candidate has already been replaced in-situ by a copy.
372 2018-05-25 Jason Merrill <jason@redhat.com>
374 PR c++/80485 - inline function non-zero address.
375 * symtab.c (nonzero_address): Check DECL_COMDAT.
377 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
380 * config/alpha/alpha.md (ashlsi3): New insn pattern.
381 (*ashlsi_se): Rename from *ashldi_se. Define as sign
382 extension of SImode operation. Use const123_operand predicate.
384 (*saddl_se_1): Ditto.
386 (*ssubl_se_1): Ditto.
387 * config/alpha/predicates.md (const123_operand): New predicate.
388 * config/alpha/constraints.md (P): Use IN_RANGE.
390 2018-05-25 Richard Biener <rguenther@suse.de>
392 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
394 (ref_maybe_used_by_stmt_p): Likewise.
395 (stmt_may_clobber_ref_p): Likewise.
396 (stmt_may_clobber_ref_p_1): Likewise.
397 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
399 (ref_maybe_used_by_stmt_p): Likewise.
400 (stmt_may_clobber_ref_p): Likewise.
401 (stmt_may_clobber_ref_p_1): Likewise.
402 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
403 the alias oracle to disambiguate DRs with stmts DR analysis
405 (vect_analyze_data_refs): Do not give up on not analyzable
406 DRs for BB vectorization. Remove code truncating the dataref
409 2018-05-25 Jakub Jelinek <jakub@redhat.com>
412 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
413 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
414 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
416 2018-05-25 Richard Biener <rguenther@suse.de>
418 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
419 function, combining stmt data ref gathering and fatal analysis
421 (vect_analyze_data_refs): Remove now redudnant code and simplify.
422 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
423 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
424 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
425 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
427 2018-05-25 Bin Cheng <bin.cheng@arm.com>
429 PR tree-optimization/85720
430 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
431 SCC if all partitions are builtins.
432 (version_loop_by_alias_check): New parameter. Generate cancelable
433 runtime alias check if all partitions are builtins.
434 (distribute_loop): Update call to above function.
436 2018-05-25 Bin Cheng <bin.cheng@arm.com>
438 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
439 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
440 (parm_default_def_partition_arg): Ditto.
441 (set_parm_default_def_partition): Ditto.
442 (get_parm_default_def_partitions): Ditto and make it static.
443 (get_undefined_value_partitions): Ditto and make it static.
444 (remove_ssa_form): Refactor call to init_var_map here.
445 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
446 computation for loop region.
447 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
448 (register_default_def): Delete.
449 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
450 (parm_default_def_partition_arg): Ditto.
451 (set_parm_default_def_partition): Ditto.
452 (get_parm_default_def_partitions): Ditto and make it static.
453 (get_undefined_value_partitions): Ditto and make it static.
454 (coalesce_with_default, coalesce_with_default): Update comment.
455 (create_coalesce_list_for_region): New func factored out from
456 create_outofssa_var_map.
457 (populate_coalesce_list_for_outofssa): New func factored out from
458 create_outofssa_var_map and coalesce_ssa_name.
459 (create_outofssa_var_map): Delete.
460 (coalesce_ssa_name): Refactor to support live range computation.
461 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
462 (get_parm_default_def_partitions): Delete.
463 (get_undefined_value_partitions): Ditto.
464 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
465 computation for loop region.
466 (new_tree_live_info, loe_visit_block): Ditto.
467 (live_worklist, set_var_live_on_entry): Ditto.
468 (calculate_live_on_exit, verify_live_on_entry): Ditto.
469 * tree-ssa-live.h (struct _var_map): New fields.
470 (init_var_map): Change decl.
471 (region_contains_p): New.
473 2018-05-25 Bin Cheng <bin.cheng@arm.com>
475 * tree-ssa-live.h (live_merge_and_clear): Delete.
477 2018-05-25 Richard Biener <rguenther@suse.de>
480 * tree-dump.c (dequeue_and_dump): Remove access to removed
481 operand 2 of a SWITCH_EXPR.
483 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
485 * doc/sourcebuild.texi (vect_double_cond_arith): Include
486 multiplication and division.
487 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
488 (cond_udiv@var{m}, cond_umod@var{m}): Document.
489 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
490 (cond_udiv_optab, cond_umod_optab): New optabs.
491 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
492 (IFN_COND_RDIV): New internal functions.
493 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
494 TRUNC_MOD_EXPR and RDIV_EXPR.
495 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
496 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
498 (SVE_INT_BINARY): Include mult.
499 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
500 (optab, sve_int_op): Handle mult.
501 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
503 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
504 for SVE_INT_BINARY_SD.
506 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
508 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
509 (optab, sve_int_op): Handle div and udiv.
510 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
511 for SVE_INT_BINARY_SD.
512 (*<optab><mode>3): New insn for the same.
514 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
516 * tree-vect-patterns.c: Include predict.h.
517 (vect_recog_divmod_pattern): Restrict check for division support
518 to when optimizing for size.
520 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
522 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
523 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
524 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
525 (gimple_match_op::set_op): Likewise.
526 (gimple_resimplify4): Declare.
527 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
528 (expr::gen_transform): Likewise.
529 (decision_tree::gen): Generate a simplification routine for 4 operands.
530 * gimple-match-head.c (gimple_simplify): Add an overload for
531 4 operands. In the top-level function, handle up to 4 call
532 arguments and call gimple_resimplify4.
533 (gimple_resimplify4): New function.
534 (build_call_internal): Pass a fourth operand.
535 (maybe_push_to_seq): Likewise.
536 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
537 Fold VEC_COND_EXPRs of an operation and a default value into
538 an IFN_COND_* function if possible.
539 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
541 (SVE_COND_FP_BINARY): Include them.
542 (optab, sve_fp_op): Handle them.
543 (SVE_INT_BINARY_REV): New code iterator.
544 (SVE_COND_FP_BINARY_REV): New int iterator.
545 (commutative): New int attribute.
546 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
548 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
550 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
551 (*cond_<optab><mode>): New patterns for reversed operands.
553 2018-05-25 Richard Biener <rguenther@suse.de>
555 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
556 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
557 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
558 (STMT_VINFO_GROUPED_ACCESS): Adjust.
559 * tree-vect-data-refs.c (everywhere): Adjust users.
560 * tree-vect-loop.c (everywhere): Likewise.
561 * tree-vect-slp.c (everywhere): Likewise.
562 * tree-vect-stmts.c (everywhere): Likewise.
563 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
565 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
567 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
569 (gcc_cv_as_section_exclude): ... this.
570 Try Solaris as #exclude syntax.
571 * configure: Regenerate.
572 * config.in: Regenerate.
573 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
575 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
576 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
578 * varasm.c (default_elf_asm_named_section): Don't check if
579 HAVE_GAS_SECTION_EXCLUDE is defined.
581 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
583 * doc/md.texi: Update the documentation of the cond_* optabs
584 to mention the new final operand. Fix GET_MODE_NUNITS call.
585 Describe the scalar case too.
586 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
587 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
589 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
590 (get_conditional_internal_fn): Update comment.
591 * tree-vect-loop.c (vectorizable_reduction): Pass the original
592 accumulator value as a final argument to conditional functions.
593 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
594 a define_expand and add an "else" operand. Assert for now that
595 the else operand is equal to operand 2. Use SVE_INT_BINARY and
596 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
597 (*cond_<optab><mode>): New patterns.
598 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
599 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
600 (UNSPEC_COND_EOR): Delete.
601 (optab): Remove associated mappings.
602 (SVE_INT_BINARY): New code iterator.
603 (sve_int_op): Remove int attribute and add "minus" to the code
605 (SVE_COND_INT_OP): Delete.
606 (SVE_COND_FP_OP): Rename to...
607 (SVE_COND_FP_BINARY): ...this.
609 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
611 * optabs.c (can_reuse_operands_p): New function.
612 (maybe_legitimize_operands): Try to reuse the results for
615 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
617 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
618 Add {q} suffix to insn mnemonic.
620 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
622 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
623 (msp430_warn_func_return): New.
625 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
627 * fold-const.c (tree_nonzero_bits): New function.
628 * fold-const.h (tree_nonzero_bits): Likewise.
629 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
630 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
632 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
636 * varasm.c (assemble_alias): Check ifunc_resolver only on
639 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
642 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
643 when memory input operand is handled.
645 2018-05-24 Luis Machado <luis.machado@linaro.org>
647 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
649 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
651 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
653 * match.pd: Delay FMA folds until after vectorization.
655 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
658 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
659 address check not strict.
661 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
663 * gimple-match.h (gimple_match_op): New class.
664 (mprts_hook): Replace parameters with a gimple_match_op *.
665 (maybe_build_generic_op): Likewise.
666 (gimple_simplified_result_is_gimple_val): Replace parameters with
667 a const gimple_match_op *.
668 (gimple_simplify): Replace code_helper * and tree * parameters with
669 a gimple_match_op * parameter.
670 (gimple_resimplify1): Replace code_helper *, tree and tree *
671 parameters with a gimple_match_op * parameter.
672 (gimple_resimplify2): Likewise.
673 (gimple_resimplify3): Likewise.
674 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
675 parameters with a gimple_match_op * parameter.
676 * gimple-match-head.c (gimple_simplify): Change prototypes of
677 auto-generated functions to take a gimple_match_op * instead of
678 separate code_helper * and tree * parameters. Make the same
679 change in the top-level overload and update calls to the
680 gimple_resimplify routines. Update calls to the auto-generated
681 functions and to maybe_push_res_to_seq in the publicly-facing
682 operation-specific gimple_simplify overloads.
683 (gimple_match_op::MAX_NUM_OPS): Define.
684 (gimple_resimplify1): Replace rcode and ops with a single res_op
685 parameter. Update call to gimple_simplify.
686 (gimple_resimplify2): Likewise.
687 (gimple_resimplify3): Likewise.
688 (mprts_hook): Replace parameters with a gimple_match_op *.
689 (maybe_build_generic_op): Likewise.
690 (build_call_internal): Replace type, nargs and ops with
692 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
693 with a single gimple_match_op *. Update calls to mprts_hook,
694 build_call_internal and gimple_simplified_result_is_gimple_val.
695 Factor out code that is common to the tree_code and combined_fn cases.
696 * genmatch.c (expr::gen_transform): Replace tem_code and
697 tem_ops with a gimple_match_op called tem_op. Update calls
698 to the gimple_resimplify functions and maybe_push_res_to_seq.
699 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
700 res_ops. Update call to the gimple_resimplify functions.
701 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
702 (decision_tree::gen): Make the functions take a gimple_match_op *
703 called res_op instead of separate res_code and res_ops parameters.
704 Update call accordingly.
705 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
706 and ops with a single res_op parameter. Update calls to
707 maybe_build_generic_op and maybe_push_res_to_seq.
708 (fold_stmt_1): Update calls to gimple_simplify and
709 replace_stmt_with_simplification.
710 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
711 and gimple_simplified_result_is_gimple_val.
712 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
714 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
715 with a gimple_match_op *.
716 (vn_nary_build_or_lookup): Likewise. Update call to
717 vn_nary_build_or_lookup_1.
718 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
719 gimple_match_op *. Update calls to the gimple_resimplify routines
720 and to gimple_simplified_result_is_gimple_val.
721 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
722 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
723 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
724 (visit_nary_op): Likewise.
725 (visit_reference_op_load): Likewise.
727 2018-05-23 Luis Machado <luis.machado@linaro.org>
729 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
730 modifier for printing the step amount.
732 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
735 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
738 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
740 * doc/sourcebuild.texi (Endianness): New subsubsection.
742 2018-05-23 Luis Machado <luis.machado@linaro.org>
744 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
745 <prefetch_dynamic_strides>: New const bool field.
746 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
747 prefetch_dynamic_strides.
748 (exynosm1_prefetch_tune): Likewise.
749 (thunderxt88_prefetch_tune): Likewise.
750 (thunderx_prefetch_tune): Likewise.
751 (thunderx2t99_prefetch_tune): Likewise.
752 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
754 (aarch64_override_options_internal): Update to set
755 PARAM_PREFETCH_DYNAMIC_STRIDES.
756 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
757 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
758 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
759 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
760 prefetch-dynamic-strides setting.
762 2018-05-23 Luis Machado <luis.machado@linaro.org>
764 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
765 <minimum_stride>: New const int field.
766 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
767 minimum_stride field defaulting to -1.
768 (exynosm1_prefetch_tune): Likewise.
769 (thunderxt88_prefetch_tune): Likewise.
770 (thunderx_prefetch_tune): Likewise.
771 (thunderx2t99_prefetch_tune): Likewise.
772 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
773 <default_opt_level>: Set to 3.
774 (aarch64_override_options_internal): Update to set
775 PARAM_PREFETCH_MINIMUM_STRIDE.
776 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
777 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
778 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
779 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
780 stride is constant and is below the minimum stride threshold.
782 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
784 * config/arm/arm-cpus.in (mode26): Delete.
785 (armv4): Delete mode26 reference.
786 * config/arm/arm.c (arm_configure_build_target): Delete use of
789 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
791 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
793 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
794 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
796 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
797 (floatunsdidf2): Ditto.
799 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
801 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
802 (fixuns_trunc<mode>si2_avx512f): Ditto.
803 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
804 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
805 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
807 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
809 PR rtl-optimization/79985
810 * df-scan.c (df_insn_refs_collect): Remove special case for
811 global registers and asm statements.
813 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
815 * extend.texi (Global Register Variables): Rewrite the bullet list.
816 Note that the register is available for allocation. Note that access
817 via inline asm must use constraints. Add note about async-signal
818 handlers. Remove paragraph about automagic register selection.
820 2018-05-23 Richard Biener <rguenther@suse.de>
822 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
823 of fixed offset from memset VN.
825 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
827 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
829 (alloc_cand_and_find_basis): Initialize first_interp field.
830 (slsr_process_mul): Modify first_interp field.
831 (slsr_process_add): Likewise.
832 (slsr_process_cast): Modify first_interp field for each new
834 (slsr_process_copy): Likewise.
835 (dump_candidate): Dump first_interp field.
836 (replace_mult_candidate): Process all interpretations, not just
838 (replace_rhs_if_not_dup): Likewise.
839 (replace_one_candidate): Likewise.
841 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
843 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
845 (aarch64_needs_frame_chain): New function.
846 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
848 2018-05-23 Sudakshina Das <sudi.das@arm.com>
851 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
852 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
853 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
854 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
855 as true for strict-align.
856 (aarch64_can_inline_p): Perform checks even when callee has no
857 attributes to check for strict alignment.
858 * doc/extend.texi (AArch64 Function Attributes): Document
860 * doc/invoke.texi: (AArch64 Options): Likewise.
862 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
864 PR tree-optimization/85853
865 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
866 the handling of the root of the node to...
867 (vect_slp_analyze_node_operations_1): ...this new function,
868 and run the whole thing with the child nodes' def types
869 set according to their SLP node's def type.
871 2018-05-23 Richard Biener <rguenther@suse.de>
874 * tree-data-ref.c (create_runtime_alias_checks): Defer
875 and ignore overflow warnings.
877 2018-05-23 Yury Gribov <tetra2005@gmail.com>
879 PR tree-optimization/85822
880 * tree-vrp.c (is_masked_range_test): Fix handling of negative
883 2018-05-23 Richard Biener <rguenther@suse.de>
885 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
886 memset constants via native_interpret_expr.
888 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
891 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
893 (cgraph_node::create_alias): Likewise.
894 (cgraph_node::get_availability): Check ifunc_resolver instead
895 of looking up ifunc attribute.
896 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
897 * varasm.c (do_assemble_alias): Likewise.
898 (assemble_alias): Likewise.
899 (default_binds_local_p_3): Likewise.
900 * cgraph.h (cgraph_node): Add ifunc_resolver.
901 (cgraph_node::only_called_directly_or_aliased_p): Return false
903 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
905 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
906 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
907 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
908 instead of looking up ifunc attribute.
910 2018-05-22 Luis Machado <luis.machado@linaro.org>
912 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
914 2018-05-22 Martin Sebor <msebor@redhat.com>
917 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
918 only when expasion succeeds.
919 (expand_builtin_strcmp): Same.
920 (expand_builtin_strncmp): Same.
922 2018-05-22 Martin Sebor <msebor@redhat.com>
924 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
926 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
927 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
929 * config/aarch64/aarch64-ldpstp.md: Replace uses of
930 aarch64_mem_pair_operand with memory_operand and delete operand swapping
932 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
933 Add check for legitimate_address.
934 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
935 (aarch64_swap_ldrstr_operands): New.
936 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
939 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
940 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
942 * config/aarch64/aarch64.md: New patterns to generate stp
944 (store_pair_sw, store_pair_dw): New patterns to generate stp for
945 single words and double words.
946 (load_pair_sw, load_pair_dw): Likewise.
947 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
949 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
951 * config/aarch64/aarch64-ldpstp.md: Modify peephole
952 for different mode ldpstp and add peephole for merged zero stores.
954 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
956 (aarch64_gen_store_pair): Rename calls to match new patterns.
957 (aarch64_gen_load_pair): Rename calls to match new patterns.
958 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
959 (load_pair<DREG:mode><DREG2:mode>): ... This.
960 (store_pair<mode>): Rename to...
961 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
962 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
964 (V_INT_EQUIV): Handle SImode.
965 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
968 2018-05-22 Martin Sebor <msebor@redhat.com>
971 * calls.c (maybe_warn_nonstring_arg): Use string length to set
972 or ajust the presumed bound on an operation to avoid unnecessary
975 2018-05-22 Martin Sebor <msebor@redhat.com>
977 PR tree-optimization/85826
978 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
979 assuming that a DECL necesarily has a constant size.
981 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
984 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
986 2018-05-22 Richard Biener <rguenther@suse.de>
988 PR tree-optimization/85834
989 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
990 non-constant and non-zero memset arguments.
992 2018-05-22 Martin Liska <mliska@suse.cz>
995 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
997 2018-05-22 Richard Biener <rguenther@suse.de>
999 PR tree-optimization/85863
1000 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
1001 comparisons when vectype is specified.
1002 (vectorizable_condition): Do not specify vectype for
1003 vect_is_simple_cond when SLP vectorizing.
1005 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
1008 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
1009 define __ibm128 as long double.
1010 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
1011 as a distinct type when IEEE 128-bit support is enabled.
1012 (init_float128_ieee): Fix up conversions between IFmode and IEEE
1013 128-bit types to use the correct functions.
1014 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
1015 convert between 128-bit floating point types that have different
1016 modes but the same representation, instead of using gen_lowpart to
1018 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
1020 (IFKF_reg): New attributes to give the register constraints for
1022 (extend<mode>tf2_internal): New insns to mark an explicit
1023 conversion between 128-bit floating point types that have a
1024 different mode but share the same representation.
1026 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
1028 PR tree-optimization/85814
1029 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
1030 a null return from get_strinfo when unsharing the next
1031 strinfo in the chain.
1033 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1036 * varasm.c (weak_finish): Clean up weak_decls.
1038 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1040 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
1041 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
1042 UNSPEC_UADALP values.
1043 * config/aarch64/iterators.md (ABAL): New int iterator.
1046 (sur): Add mappings for the above.
1047 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
1049 (aarch64_<sur>abal<mode>_4): Likewise.
1050 (aarch64_<sur>adalp<mode>_3): Likewise.
1051 (<sur>sadv16qi): New define_expand.
1053 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
1055 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
1056 (*movdf_internal): Ditto.
1057 (*rcpsf2_sse): Ditto.
1058 (*rsqrtsf2_sse): Ditto.
1059 (*sqrt<mode>2_sse): Ditto.
1061 2018-05-21 Tamar Christina <tamar.christina@arm.com>
1063 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
1065 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
1066 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
1067 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
1068 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
1070 * config/aarch64/arm_neon.h: Likewise.
1071 * config/aarch64/iterators.md (VQ_I): New.
1073 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
1075 * config.gcc: Add arc/t-multilib-linux to tmake_file for
1077 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
1080 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1082 * config/nds32/constraints.md (S): New constraint.
1083 * config/nds32/nds32.md (call_internal): Use constraint S.
1084 (call_value_internal): Likewise.
1085 (sibcall_internal): Likewise.
1086 (sibcall_value_internal): Likewise.
1088 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1089 Chung-Ju Wu <jasonwucj@gmail.com>
1091 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
1094 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1095 Chung-Ju Wu <jasonwucj@gmail.com>
1097 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
1098 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
1099 (nds32_rtx_costs_impl): Simplify.
1100 (nds32_address_cost_impl): Simplify.
1101 (nds32_init_rtx_costs): New function.
1102 (nds32_rtx_costs_speed_prefer): Likewise.
1103 (nds32_rtx_costs_size_prefer): Likewise.
1104 (nds32_address_cost_speed_prefer): Likewise.
1105 (nds32_address_cost_speed_fwprop): Likewise.
1106 (nds32_address_cost_size_prefer): Likewise.
1107 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
1108 * config/nds32/nds32.c (nds32_option_override): Use
1109 nds32_init_rtx_costs function.
1111 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1113 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
1114 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
1115 (TARGET_PIPELINE_N8): Likewise.
1116 (TARGET_PIPELINE_N10): Likewise.
1117 (TARGET_PIPELINE_N13): Likewise.
1118 (TARGET_PIPELINE_GRAYWOLF): Likewise.
1120 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
1122 * config/nds32/nds32-fpu.md: Update copyright year.
1124 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1126 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
1128 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1130 * config/nds32/nds32.c
1131 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
1132 * config/nds32/nds32.opt (minline-asm-r15): New option.
1134 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1136 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
1138 * config/nds32/nds32.md (abssi2): New pattern.
1140 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
1142 * config/i386/i386.md (rex64namesuffix): New mode attribute.
1143 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
1144 Merge insn pattern from sse_cvtsi2ss<round_name> and
1145 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
1146 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
1147 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
1148 using SWI48 mode iterator.
1149 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
1150 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
1151 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
1152 pattern from sse_cvttss2si<round_saeonly_name>
1153 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
1154 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
1155 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
1156 using SWI48 mode iterator.
1157 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
1158 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
1159 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
1160 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
1161 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
1162 using SWI48 mode iterator.
1163 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
1164 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
1165 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
1166 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
1167 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
1168 SWI48 mode iterator.
1169 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
1170 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
1171 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
1172 pattern from sse_cvttsd2si<round_saeonly_name>
1173 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
1175 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1177 * config/nds32/nds32-md-auxiliary.c
1178 (nds32_valid_smw_lwm_base_p): Refine.
1179 (nds32_output_smw_single_word): Refine.
1180 (nds32_output_smw_double_word): New.
1181 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
1183 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1185 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
1186 (nds32_output_stack_pop): Refine.
1187 (nds32_expand_unaligned_load): Refine.
1188 (nds32_expand_unaligned_store): Refine.
1190 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1191 Chung-Ju Wu <jasonwucj@gmail.com>
1193 * config/nds32/constants.md: Add TP_REGNUM constant.
1194 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
1195 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
1197 * config/nds32/nds32-doubleword.md: Consider flag_pic.
1198 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
1199 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
1200 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
1201 and PIC code generation.
1202 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
1204 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
1206 * config/nds32/nds32.md: Support TLS and PIC.
1207 * config/nds32/nds32.c: Support TLS and PIC.
1208 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
1209 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
1212 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1214 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
1215 mode with E_ prefix.
1217 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1218 Chung-Ju Wu <jasonwucj@gmail.com>
1220 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
1221 * config/nds32/nds32-md-auxiliary.c
1222 (symbolic_reference_mentioned_p): New.
1223 (nds32_legitimize_ict_address): New.
1224 (nds32_expand_ict_move): New.
1225 (nds32_indirect_call_referenced_p): New.
1226 (nds32_symbol_binds_local_p): Delete.
1227 (nds32_long_call_p): Modify.
1228 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
1229 * config/nds32/nds32-protos.h
1230 (symbolic_reference_mentioned_p): Declare.
1231 (nds32_legitimize_ict_address): Declare.
1232 (nds32_expand_ict_move): Declare.
1233 (nds32_indirect_call_referenced_p): Declare.
1234 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
1235 (nds32_relax_group): Use nds32_ict_const_p as condition.
1236 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
1237 (nds32_asm_file_start): Output ict_model directive in asm code.
1238 (nds32_legitimate_address_p): Consider indirect call.
1239 (nds32_print_operand): Consider indirect call.
1240 (nds32_print_operand_address): Consider indirect call.
1241 (nds32_insert_attributes): Handle "indirect_call" attribute.
1242 (TARGET_LEGITIMATE_ADDRESS_P): Define.
1243 (TARGET_LEGITIMATE_CONSTANT_P): Define.
1244 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
1245 (TARGET_DELEGITIMIZE_ADDRESS): Define.
1246 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
1247 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
1248 (TARGET_ICT_MODEL_SMALL): Define.
1249 (TARGET_ICT_MODEL_LARGE): Define.
1250 * config/nds32/nds32.md (movsi): Consider ict model.
1251 (call, call_value): Consider ict model.
1252 (sibcall, sibcall_value): Consider ict model.
1253 * config/nds32/nds32.opt (mict-model): New option.
1254 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
1257 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
1258 Monk Chiang <sh.chiang04@gmail.com>
1259 Jim Wilson <jimw@sifive.com>
1261 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
1262 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
1263 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
1264 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
1265 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
1266 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
1267 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
1268 compute save_libcall_adjustment properly.
1269 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
1270 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
1271 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
1272 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
1273 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
1274 (ABI_SPEC): Handle mabi=ilp32e.
1275 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
1276 (RVE): Add RVE mask.
1277 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
1278 <-march>: Add rv32e as an example.
1280 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
1283 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
1284 (intra_create_variable_infos): Handle C++ constructors.
1286 2018-05-18 Martin Liska <mliska@suse.cz>
1288 * passes.def: Remove a redundant pass.
1290 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
1293 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
1295 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1297 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
1299 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
1301 (armv2, armv2a, armv3, armv3m): Delete architectures.
1302 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
1303 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
1304 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
1306 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
1307 (*mulsidi3adddi): Likewise.
1308 (mulsidi3): Likewise.
1309 (*mulsidi3_nov6): Likewise.
1310 (umulsidi3): Likewise.
1311 (umulsidi3_nov6): Likewise.
1312 (umaddsidi4): Likewise.
1313 (*umulsidi3adddi): Likewise.
1314 (smulsi3_highpart): Likewise.
1315 (*smulsi3_highpart_nov6): Likewise.
1316 (umulsi3_highpart): Likewise.
1317 (*umulsi3_highpart_nov6): Likewise.
1318 * config/arm/arm.h (arm_arch3m): Delete.
1319 * config/arm/arm.c (arm_arch3m): Delete.
1320 (arm_option_override_internal): Update armv3-related comment.
1321 (arm_configure_build_target): Delete use of isa_bit_mode32.
1322 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
1323 (arm_rtx_costs_internal): Delete check of arm_arch3m.
1324 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
1326 (mulusa3): Likewise.
1327 * config/arm/arm-protos.h (arm_arch3m): Delete.
1328 * config/arm/arm-tables.opt: Regenerate.
1329 * config/arm/arm-tune.md: Likewise.
1330 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
1331 deleted architectures.
1333 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1335 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
1336 (armv5t, armv5te): New features.
1337 (ARMv5, ARMv5e): Delete fgroups.
1338 (ARMv5t, ARMv5te): Adjust for above changes.
1340 (armv5, armv5e): Delete arches.
1341 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
1343 (*call_reg_arm): Likewise.
1344 (*call_value_reg_armv5): Likewise.
1345 (*call_value_reg_arm): Likewise.
1346 (*call_symbol): Likewise.
1347 (*call_value_symbol): Likewise.
1348 (*sibcall_insn): Likewise.
1349 (*sibcall_value_insn): Likewise.
1351 (prefetch): Likewise.
1352 (define_split and define_peephole2 dependent on arm_arch5):
1354 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
1356 (TARGET_ARM_QBIT): Likewise.
1357 (TARGET_DSP_MULTIPLY): Likewise.
1358 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
1359 (arm_arch5, arm_arch5e): Delete.
1360 (arm_arch5t, arm_arch5te): Declare.
1361 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
1362 (arm_arch5t): Declare.
1363 (arm_option_reconfigure_globals): Update for the above.
1364 (arm_options_perform_arch_sanity_checks): Update comment, replace
1365 use of arm_arch5 with arm_arch5t.
1366 (use_return_insn): Likewise.
1367 (arm_emit_call_insn): Likewise.
1368 (output_return_instruction): Likewise.
1369 (arm_final_prescan_insn): Likewise.
1370 (arm_coproc_builtin_available): Likewise.
1371 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
1372 arm_arch5e with arm_arch5t and arm_arch5te.
1373 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
1374 (arm_arch5t, arm_arch5te): Declare.
1375 * config/arm/arm-tables.opt: Regenerate.
1376 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
1377 * config/arm/t-multilib: Likewise.
1378 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
1379 instead of arm_arch5.
1380 (*call_reg_thumb1): Likewise.
1381 (*call_value_reg_thumb1_v5): Likewise.
1382 (*call_value_reg_thumb1): Likewise.
1383 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
1385 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
1387 2018-05-18 Martin Liska <mliska@suse.cz>
1389 PR gcov-profile/84846
1390 * doc/gcov.texi: Document -t option of gcov tool.
1392 2018-05-18 Martin Liska <mliska@suse.cz>
1394 PR gcov-profile/84846
1395 * gcov.c (print_usage): Add new -t option.
1396 (process_args): Handle the option.
1397 (generate_results): Use stdout as output when requested by
1400 2018-05-18 Martin Liska <mliska@suse.cz>
1402 PR gcov-profile/84846
1403 * coverage.c (coverage_init): Write PWD to .gcno file.
1404 * doc/gcov.texi: Document how working directory is printed.
1405 * gcov-dump.c (dump_gcov_file): Print PWD.
1406 * gcov.c (output_intermediate_file): Likewise.
1407 (read_graph_file): Read PWD string.
1408 (output_lines): Print PWD.
1410 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1413 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
1414 for retval and return false if all args to phi are zero.
1416 2018-05-18 Richard Biener <rguenther@suse.de>
1418 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
1420 (evrp_dom_walker::before_dom_children): Call it.
1422 2018-05-18 Richard Biener <rguenther@suse.de>
1424 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
1425 results when processing array refs with variable index.
1427 2018-05-18 Toon Moene <toon@moene.org>
1429 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
1430 directly after that of -floop-interchange. Indicate that both
1431 options are enabled by default when specifying -O3.
1433 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1435 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
1436 iterator. Delete separate integer-mode vec_set<mode> expander.
1437 (aarch64_simd_vec_setv2di): Delete.
1438 (vec_setv2di): Delete.
1439 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
1440 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
1441 the "w, r" alternative.
1443 2018-05-18 Martin Liska <mliska@suse.cz>
1445 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
1446 * tree-pass.h (make_pass_lower_switch_O0): New function.
1447 * tree-switch-conversion.c (node_has_low_bound): Remove.
1448 (node_has_high_bound): Likewise.
1449 (node_is_bounded): Likewise.
1450 (class pass_lower_switch): Make it a template type and create
1452 (pass_lower_switch::execute): Add template argument.
1453 (make_pass_lower_switch): New function.
1454 (make_pass_lower_switch_O0): New function.
1455 (do_jump_if_equal): Remove.
1456 (emit_case_nodes): Simplify to just handle all 3 cases and leave
1457 all the hard work to tree optimization passes.
1459 2018-05-18 Martin Liska <mliska@suse.cz>
1461 * dbgcnt.c (limit_low): Renamed from limit.
1462 (limit_high): New variable.
1463 (dbg_cnt_is_enabled): Check for upper limit.
1464 (dbg_cnt): Adjust dumping.
1465 (dbg_cnt_set_limit_by_index): Add new argument for high
1467 (dbg_cnt_set_limit_by_name): Likewise.
1468 (dbg_cnt_process_single_pair): Parse new format.
1469 (dbg_cnt_process_opt): Use strtok.
1470 (dbg_cnt_list_all_counters): Remove 'value' and add
1472 * doc/invoke.texi: Document changes.
1474 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
1476 * doc/sourcebuild.texi (scalar_all_fma): Document.
1477 * tree.def (FMA_EXPR): Delete.
1478 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
1479 * internal-fn.c (ternary_direct): New macro.
1480 (expand_ternary_optab_fn): Likewise.
1481 (direct_ternary_optab_supported_p): Likewise.
1482 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
1483 * builtins.c (fold_builtin_fma): Delete.
1484 (fold_builtin_3): Don't call it.
1485 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
1486 * expr.c (expand_expr_real_2): Likewise.
1487 * fold-const.c (operand_equal_p): Likewise.
1488 (fold_ternary_loc): Likewise.
1489 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
1490 * gimple.c (DEFTREECODE): Likewise.
1491 * gimplify.c (gimplify_expr): Likewise.
1492 * optabs-tree.c (optab_for_tree_code): Likewise.
1493 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
1494 * tree-eh.c (operation_could_trap_p): Likewise.
1495 (stmt_could_throw_1_p): Likewise.
1496 * tree-inline.c (estimate_operator_cost): Likewise.
1497 * tree-pretty-print.c (dump_generic_node): Likewise.
1498 (op_code_prio): Likewise.
1499 * tree-ssa-loop-im.c (stmt_cost): Likewise.
1500 * tree-ssa-operands.c (get_expr_operands): Likewise.
1501 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
1502 * fold-const-call.h (fold_fma): Delete.
1503 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
1504 CFN_FNMA and CFN_FNMS.
1506 * genmatch.c (combined_fn): New enum.
1507 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
1508 (commutative_op): New function.
1509 (commutate): Use it. Handle more than 2 operands.
1510 (dt_operand::gen_gimple_expr): Use commutative_op.
1511 (parser::parse_expr): Allow :c to be used with non-binary
1512 operators if the commutative operand is known.
1513 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
1514 CFN_FMS, CFN_FNMA and CFN_FNMS.
1515 (backprop::process_assign_use): Remove FMA_EXPR handling.
1516 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
1517 (gen_hsa_fma): New function.
1518 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
1519 IFN_FNMA and IFN_FNMS.
1520 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
1521 * gimple-fold.h (follow_all_ssa_edges): Declare.
1522 * gimple-fold.c (follow_all_ssa_edges): New function.
1523 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
1524 gimple_build interface and use follow_all_ssa_edges to fold the result.
1525 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
1526 instead of checking for optabs directly.
1527 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
1528 rather than FMA_EXPRs.
1529 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
1530 call to IFN_FMA instead of an FMA_EXPR.
1532 2018-05-17 Jim Wilson <jimw@sifive.com>
1534 * expr.c (do_tablejump): When converting index to Pmode, if we have a
1535 sign extended promoted subreg, and the range does not have the sign bit
1536 set, then do a sign extend.
1538 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
1539 test, check for sign extended subreg and/or constant operands, and
1540 do a sign extend in that case.
1542 2018-05-17 Steve Ellcey <sellcey@cavium.com>
1544 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
1545 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
1547 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
1548 Change logics_shift_reg to logics_shift_imm.
1549 (thunderx2t99_fp_loadpair_basic): Delete.
1550 (thunderx2t99_fp_storepair_basic): Delete.
1551 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
1552 (thunderx2t99_asimd_polynomial): Delete.
1553 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
1554 and neon_fp_mul_d_scalar_q.
1555 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
1556 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
1557 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
1558 (thunderx2t99_asimd_lut): Add missing tbl types.
1559 (thunderx2t99_asimd_ext): Delete.
1560 (thunderx2t99_asimd_load1_1_mult): Delete.
1561 (thunderx2t99_asimd_load1_2_mult): Delete.
1562 (thunderx2t99_asimd_load1_ldp): New.
1563 (thunderx2t99_asimd_load1): New.
1564 (thunderx2t99_asimd_load2): Add missing *load2* types.
1565 (thunderx2t99_asimd_load3): New.
1566 (thunderx2t99_asimd_load4): New.
1567 (thunderx2t99_asimd_store1_1_mult): Delete.
1568 (thunderx2t99_asimd_store1_2_mult): Delete.
1569 (thunderx2t99_asimd_store2_mult): Delete.
1570 (thunderx2t99_asimd_store2_onelane): Delete.
1571 (thunderx2t99_asimd_store_stp): New.
1572 (thunderx2t99_asimd_store1): New.
1573 (thunderx2t99_asimd_store2): New.
1574 (thunderx2t99_asimd_store3): New.
1575 (thunderx2t99_asimd_store4): New.
1577 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
1579 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
1580 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
1582 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
1583 Segher Boessenkool <segher@kernel.crashing.org>
1586 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
1589 2018-05-17 Richard Biener <rguenther@suse.de>
1591 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
1592 for pruning loop and prune defs feeding only already visited PHIs.
1594 2018-05-17 Richard Biener <rguenther@suse.de>
1596 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
1598 2018-05-17 Bin Cheng <bin.cheng@arm.com>
1599 Richard Biener <rguenther@suse.de>
1601 PR tree-optimization/85793
1602 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
1603 for VMAT_ELEMENTWISE.
1605 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1607 * internal-fn.h (lookup_internal_fn): Declare
1608 * internal-fn.c (lookup_internal_fn): New function.
1609 * gimple.c (gimple_build_call_from_tree): Handle calls to
1611 * gimple-pretty-print.c (dump_gimple_call): Print "." before
1612 internal function names.
1613 * tree-pretty-print.c (dump_generic_node): Likewise.
1614 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
1616 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1618 * gimple-fold.h (gimple_build): Make the function forms take
1619 combined_fn rather than built_in_function.
1620 (gimple_simplify): Likewise.
1621 * gimple-match-head.c (gimple_simplify): Likewise.
1622 * gimple-fold.c (gimple_build): Likewise.
1623 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
1624 rather than gimple_build_call_internal.
1625 (get_initial_defs_for_reduction): Likewise.
1626 (vect_create_epilog_for_reduction): Likewise.
1627 (vectorizable_live_operation): Likewise.
1629 2018-05-17 Martin Liska <mliska@suse.cz>
1631 * gimple-ssa-sprintf.c (format_directive): Do not use
1632 space in between 'G_' and '('.
1634 2018-05-17 Jakub Jelinek <jakub@redhat.com>
1637 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
1638 even if the mask is not all ones.
1641 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
1643 (ix86_gimple_fold_builtin): Likewise.
1646 * config/i386/i386.c: Include tree-vector-builder.h.
1647 (ix86_vector_shift_count): New function.
1648 (ix86_fold_builtin): Fold shift builtins by scalar count.
1649 (ix86_gimple_fold_builtin): Likewise.
1651 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
1652 _mm512_setzero): New intrinsics.
1654 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
1655 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1657 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
1658 code generation for cases where splatting a value is not useful.
1659 * simplify-rtx.c (simplify_ternary_operation): Simplify
1660 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
1661 mode to a vec_concat.
1663 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
1665 * config.gcc: Support "goldmont-plus".
1666 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1668 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1669 PROCESSOR_GOLDMONT_PLUS.
1670 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
1671 (processor_target_table): Add "goldmont-plus".
1672 (PTA_GOLDMONT_PLUS): Define.
1673 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
1674 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
1675 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
1676 (fold_builtin_cpu): Add "goldmont-plus".
1677 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
1678 (ix86_option_override_internal): Add "goldmont-plus".
1679 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
1680 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
1681 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
1682 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
1684 2018-05-17 Richard Biener <rguenther@suse.de>
1686 PR tree-optimization/85757
1687 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
1688 remove defs that only feed that PHI from further processing.
1690 2018-05-16 Jim Wilson <jimw@sifive.com>
1692 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
1694 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
1696 2018-05-16 Mark Wielaard <mark@klomp.org>
1698 * dwarf2out.c (count_index_strings): New function.
1699 (output_indirect_strings): Call count_index_strings and generate
1700 header for dwarf_version >= 5.
1702 2018-05-16 Mark Wielaard <mark@klomp.org>
1704 * dwarf2out.c (dwarf_FORM): New function.
1705 (set_indirect_string): Use dwarf_FORM.
1706 (reset_indirect_string): Likewise.
1707 (size_of_die): Likewise.
1708 (value_format): Likewise.
1709 (output_die): Likewise.
1710 (add_skeleton_AT_string): Likewise.
1711 (output_macinfo_op): Likewise.
1712 (index_string): Likewise.
1713 (output_index_string_offset): Likewise.
1714 (output_index_string): Likewise.
1715 (count_index_strings): Likewise.
1717 2018-05-16 Carl Love <cel@us.ibm.com>
1719 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1720 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
1722 2018-05-16 Martin Jambor <mjambor@suse.cz>
1724 * ipa-prop.c (ipa_free_all_edge_args): Remove.
1725 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
1727 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
1729 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
1730 (fnma<mode>4): Likewise.
1731 (fms<mode>4): Likewise.
1732 (fnms<mode>4): Likewise.
1733 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
1734 (aarch64_fnma<mode>4): Likewise.
1735 (aarch64_fms<mode>4): Likewise.
1736 (aarch64_fnms<mode>4): Likewise.
1737 (aarch64_fnmadd<mode>4): Likewise.
1739 2018-05-16 Jason Merrill <jason@redhat.com>
1741 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
1743 2018-05-16 Richard Biener <rguenther@suse.de>
1745 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
1746 (dump_stmt_cost): Declare.
1747 (add_stmt_cost): Dump cost we add.
1748 (add_stmt_costs): New function.
1749 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
1751 (vect_analyze_stmt): Adjust prototype.
1752 (vectorizable_condition): Likewise.
1753 (vectorizable_live_operation): Likewise.
1754 (vectorizable_reduction): Likewise.
1755 (vectorizable_induction): Likewise.
1756 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
1757 cost vector to pass to vectorizable_ and record afterwards.
1758 (vect_model_reduction_cost): Take cost vector argument and adjust.
1759 (vect_model_induction_cost): Likewise.
1760 (vectorizable_reduction): Likewise.
1761 (vectorizable_induction): Likewise.
1762 (vectorizable_live_operation): Likewise.
1763 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
1764 SLP_TREE_NUMBER_OF_VEC_STMTS.
1765 (vect_analyze_slp_cost_1): Remove.
1766 (vect_analyze_slp_cost): Likewise.
1767 (vect_slp_analyze_node_operations): Take visited args and
1768 a target cost vector. Avoid processing already visited stmt sets.
1769 (vect_slp_analyze_operations): Use a local cost vector to gather
1770 costs and register those of non-discarded instances.
1771 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
1772 (vect_schedule_slp_instance): Remove copying of
1773 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
1775 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
1776 adding cost. Record cost entry location.
1777 (vect_prologue_cost_for_slp_op): Function to compute cost of
1778 a constant or invariant generated for SLP vect in the prologue,
1779 split out from vect_analyze_slp_cost_1.
1780 (vect_model_simple_cost): Make static. Adjust for SLP costing.
1781 (vect_model_promotion_demotion_cost): Likewise.
1782 (vect_model_store_cost): Likewise, make static.
1783 (vect_model_load_cost): Likewise.
1784 (vectorizable_bswap): Add cost vector arg and adjust.
1785 (vectorizable_call): Likewise.
1786 (vectorizable_simd_clone_call): Likewise.
1787 (vectorizable_conversion): Likewise.
1788 (vectorizable_assignment): Likewise.
1789 (vectorizable_shift): Likewise.
1790 (vectorizable_operation): Likewise.
1791 (vectorizable_store): Likewise.
1792 (vectorizable_load): Likewise.
1793 (vectorizable_condition): Likewise.
1794 (vectorizable_comparison): Likewise.
1795 (can_vectorize_live_stmts): Likewise.
1796 (vect_analyze_stmt): Likewise.
1797 (vect_transform_stmt): Adjust calls to vectorizable_*.
1798 * tree-vectorizer.c: Include gimple-pretty-print.h.
1799 (dump_stmt_cost): New function.
1801 2018-05-16 Richard Biener <rguenther@suse.de>
1803 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
1804 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
1805 * tree-ssa-dse.c: Include tree-ssa-loop.h.
1806 (check_name): New callback.
1807 (dse_classify_store): Track cycles via a visited bitmap of PHI
1808 defs and simplify handling of in-loop and across loop dead stores
1809 and properly fail for loop-variant refs. Handle byte-tracking with
1810 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
1813 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
1815 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
1816 (vect_get_mask_type_for_stmt): Likewise.
1817 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
1819 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
1820 to determine the statement's vector type and the vector type that
1821 should be used for calculating nunits. Deal with cases in which
1822 the type has to be deferred.
1823 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
1824 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
1825 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
1826 (vect_determine_vf_for_stmt): New functions, split out from...
1827 (vect_determine_vectorization_factor): ...here.
1828 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
1829 (vect_get_mask_type_for_stmt): New functions, split out from
1830 vect_determine_vectorization_factor.
1832 2018-05-16 Richard Biener <rguenther@suse.de>
1834 * tree-cfg.c (verify_gimple_assign_ternary): Properly
1835 verify the [VEC_]COND_EXPR embedded comparison.
1837 2018-05-15 Martin Sebor <msebor@redhat.com>
1839 PR tree-optimization/85753
1840 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
1841 RECORD_TYPE in addition to ARRAY_TYPE.
1843 2018-05-15 Martin Sebor <msebor@redhat.com>
1846 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
1848 2018-05-15 Richard Biener <rguenther@suse.de>
1850 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
1851 add by_clobber_p one. Change algorithm to collect all defs
1852 representing uses we need to walk and try reducing them to
1853 a single one before failing.
1854 (dse_dom_walker::dse_optimize_stmt): Adjust.
1856 2018-05-13 Mark Wielaard <mark@klomp.org>
1858 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
1859 (size_of_loc_descr): Likewise.
1860 (output_loc_operands): Likewise.
1861 (output_loc_operands_raw): Likewise.
1862 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
1863 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
1864 (hash_loc_operands): Likewise.
1865 (compare_loc_operands): Likewise.
1867 2018-05-14 Mark Wielaard <mark@klomp.org>
1869 * dwarf2out.c (count_index_addrs): New function.
1870 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
1872 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1874 PR tree-optimization/83648
1875 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
1876 return value as malloc candidate.
1878 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1881 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
1882 as true in call to suggest_attribute.
1884 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
1886 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
1887 -mreadonly-in-sdata.
1889 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1891 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
1893 (aarch64_crypto_aesd_fused): Likewise.
1895 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
1897 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
1898 (movsi_aarch64): Likewise.
1899 (load_pairsi): Likewise.
1900 (load_pairdi): Likewise.
1901 (store_pairsi): Likewise.
1902 (store_pairdi): Likewise.
1903 (load_pairsf): Likewise.
1904 (load_pairdf): Likewise.
1905 (store_pairsf): Likewise.
1906 (store_pairdf): Likewise.
1907 (zero_extend): Likewise.
1908 (trunc): Swap alternatives.
1909 (fcvt_target): Add '?' to prefer w over r.
1911 2018-05-14 Jakub Jelinek <jakub@redhat.com>
1914 * config/i386/i386.md: Disallow non-commutative arithmetics in
1915 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
1916 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
1917 in the peephole2 before it.
1919 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
1921 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
1922 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
1923 (ix86_handle_option): Handle -mcldemote.
1924 * config.gcc: New header.
1925 * config/i386/cldemoteintrin.h: New file.
1926 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
1927 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1929 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1930 OPTION_MASK_ISA_CLDEMOTE.
1931 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
1932 (ix86_valid_target_attribute_inner_p): Ditto.
1933 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
1934 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
1935 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
1936 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
1937 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
1939 * config/i386/i386.opt: Add -mcldemote.
1940 * config/i386/x86intrin.h: New header.
1941 * doc/invoke.texi: Add -mcldemote.
1943 2018-05-14 Richard Biener <rguenther@suse.de>
1945 * doc/match-and-simplify.texi: Adjust :s documentation.
1947 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
1949 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
1950 intended memcpy size.
1951 (REORDER_45): Likewise.
1953 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
1955 * sort.cc: New file.
1956 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
1957 * vec.c (qsort_chk): Use gcc_qsort.
1958 * Makefile.in (OBJS-libcommon): Add sort.o.
1959 (build/sort.o): New target. Use it...
1960 (BUILD_RTL): ... here, and...
1961 (build/gencfn-macros): ... here, and...
1962 (build/genmatch): ... here.
1964 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1965 Chung-Ju Wu <jasonwucj@gmail.com>
1967 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
1968 * config/nds32/nds32-graywolf.md: New file.
1969 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
1970 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
1972 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
1973 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
1974 * config/nds32/nds32.md (pipeline_model): Add graywolf.
1975 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
1976 * config/nds32/pipelines.md: Include n15 settings.
1978 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1979 Chung-Ju Wu <jasonwucj@gmail.com>
1981 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
1982 * config/nds32/nds32-n13.md: New file.
1983 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
1984 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
1986 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
1987 * config/nds32/nds32.md (pipeline_model): Add n13.
1988 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
1989 * config/nds32/pipelines.md: Include n13 settings.
1991 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1992 Chung-Ju Wu <jasonwucj@gmail.com>
1994 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
1995 * config/nds32/nds32-n10.md: New file.
1996 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
1997 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
1999 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
2000 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
2001 * config/nds32/nds32.md (pipeline_model): Add n10.
2002 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
2003 * config/nds32/pipelines.md: Include n10 settings.
2005 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
2006 Kito Cheng <kito.cheng@gmail.com>
2007 Chung-Ju Wu <jasonwucj@gmail.com>
2009 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
2010 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2011 Add enum values for DSP extension instructions.
2012 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
2014 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
2015 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
2017 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
2018 * config/nds32/nds32-dspext.md: New file for DSP implementation.
2019 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
2020 * config/nds32/nds32-intrinsic.md: Likewise.
2021 * config/nds32/nds32_intrinsic.h: Likewise.
2022 * config/nds32/nds32-md-auxiliary.c: Likewise.
2023 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
2024 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
2025 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
2026 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
2027 * config/nds32/nds32-protos.h: New declarations for DSP extension.
2028 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
2029 TYPE_DMAC in switch statement.
2030 * config/nds32/nds32.c: New checking and implementation for DSP
2031 extension instructions.
2032 * config/nds32/nds32.h: Likewise.
2033 * config/nds32/nds32.md: Likewise.
2034 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
2035 * config/nds32/predicates.md: Implement new predicates for DSP
2038 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
2040 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
2041 Reformat alternatives and attributes so it is easier to identify
2042 which constraints/attributes go with which instruction.
2043 (mov<mode>_hardfloat32, FMOVE64): Likewise.
2044 (mov<mode>_softfloat32, FMOVE64): Likewise.
2045 (mov<mode>_hardfloat64, FMOVE64): Likewise.
2046 (mov<mode>_softfloat64, FMOVE64): Likewise.
2048 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
2050 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2052 (Basic PowerPC Built-in Functions): The new name of the
2053 subsection previously known as "PowerPC Built-in Functions".
2054 (Basic PowerPC Built-in Functions Available on all Configurations):
2056 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
2057 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
2058 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
2059 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
2061 2018-05-11 Martin Jambor <mjambor@suse.cz>
2064 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
2067 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2070 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
2072 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
2074 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
2075 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
2076 (ix86_handle_option): Handle -mwaitpkg.
2077 * config.gcc: New header.
2078 * config/i386/cpuid.h (bit_WAITPKG): New bit.
2079 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
2080 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
2082 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2083 OPTION_MASK_ISA_WAITPKG.
2084 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
2085 (ix86_option_override_internal): Add PTA_WAITPKG.
2086 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
2087 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
2088 IX86_BUILTIN_TPAUSE.
2089 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
2090 __builtin_ia32_umwait and __builtin_ia32_tpause.
2091 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
2092 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
2093 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
2094 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
2095 UNSPECV_TPAUSE): New.
2096 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
2097 * config/i386/i386.opt: Add -mwaitpkg.
2098 * config/i386/waitpkgintrin.h: New file.
2099 * config/i386/x86intrin.h: New header.
2100 * doc/invoke.texi: Add -mwaitpkg.
2102 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2105 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
2107 (cortex-m0): Use armv6s-m isa.
2108 (cortex-m0plus): Likewise.
2109 (cortex-m1): Likewise.
2110 (cortex-m0.small-multiply): Likewise.
2111 (cortex-m0plus.small-multiply): Likewise.
2112 (cortex-m1.small-multiply): Likewise.
2114 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
2115 Jakub Jelinek <jakub@redhat.com>
2117 PR tree-optimization/85692
2118 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
2119 source permute as well.
2121 2018-05-11 Martin Liska <mliska@suse.cz>
2124 * doc/extend.texi: Document LLVM style format for no_sanitize
2127 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
2129 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
2130 mode_supports_vsx_dform_quad to mode_supports_dq_form.
2131 (mode_supports_vsx_dform_quad): Likewise.
2132 (mode_supports_vmx_dform): Move these functions to be next to the
2133 other mode_supports functions.
2134 (mode_supports_dq_form): Likewise.
2135 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
2136 mode_supports_dq_form.
2137 (reg_offset_addressing_ok_p): Likewise.
2138 (offsettable_ok_by_alignment): Likewise.
2139 (rs6000_legitimate_offset_address_p): Likewise.
2140 (legitimate_lo_sum_address_p): Likewise.
2141 (rs6000_legitimize_address): Likewise.
2142 (rs6000_legitimize_reload_address): Likewise.
2143 (rs6000_secondary_reload_inner): Likewise.
2144 (rs6000_preferred_reload_class): Likewise.
2145 (rs6000_output_move_128bit): Likewise.
2147 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2149 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
2150 Generate SImode target register for null target.
2151 <case IX86_BUILTIN_XGETBV>: Ditto.
2152 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
2153 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
2155 2018-05-10 Carl Love <cel@us.ibm.com>
2157 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2158 dcbtt and dcbtstt if operands[2] is 0.
2160 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2163 * config/i386/sse.md (usadv64qi): New expander.
2165 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
2167 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
2168 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
2169 -maltivec=be support.
2170 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
2171 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
2172 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
2173 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
2174 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
2175 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
2176 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
2177 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
2178 altivec_vsumsws): Adjust.
2179 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
2180 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
2181 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
2183 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
2184 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
2185 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
2186 (altivec_lve<VI_char>x): Delete expand.
2187 (*altivec_lve<VI_char>x_internal): Rename to...
2188 (altivec_lve<VI_char>x): ... this.
2189 (altivec_lvxl_<mode>): Delete expand.
2190 (*altivec_lvxl_<mode>_internal): Rename to ...
2191 (altivec_lvxl_<mode>): ... this.
2192 (altivec_stvxl_<mode>): Delete expand.
2193 (*altivec_stvxl_<mode>_internal): Rename to ...
2194 (altivec_stvxl_<mode>): ... this.
2195 (altivec_stve<VI_char>x): Delete expand.
2196 (*altivec_stve<VI_char>x_internal): Rename to ...
2197 (altivec_stve<VI_char>x): ... this.
2198 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
2199 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
2200 reduc_plus_scal_<mode>): Adjust.
2201 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
2203 (rs6000_cpu_cpp_builtins): Adjust.
2204 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
2205 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
2206 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
2207 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
2208 -maltivec=be support.
2209 (rs6000_split_vec_extract_var): Adjust.
2210 (rs6000_split_v4si_init): Adjust.
2211 (swap_selector_for_mode): Delete.
2212 (altivec_expand_lvx_be, altivec_expand_stvx_be,
2213 altivec_expand_stvex_be): Delete.
2214 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
2215 -maltivec=be support.
2216 (rs6000_gimple_fold_builtin): Ditto.
2217 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
2219 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
2220 (TARGET_DIRECT_MOVE_64BIT): Adjust.
2221 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
2222 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
2223 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
2224 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
2225 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
2226 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
2227 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
2228 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
2229 anonymous split): Adjust.
2230 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
2231 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
2233 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
2235 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
2236 when --with-gxx-include-dir is also specified.
2237 * configure: Regenerate.
2239 2018-05-09 Jim Wilson <jimw@sifive.com>
2242 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
2243 * config/riscv/t-withmultilib: New.
2244 * config/riscv/withmultilib.h: New.
2245 * doc/install.texi: Document RISC-V --with-multilib-list support.
2247 2018-05-09 Richard Biener <rguenther@suse.de>
2249 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
2251 (vect_bb_vectorization_profitable_p): Adjust. Compute
2252 actual scalar cost using the cost vector and the add_stmt_cost
2255 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2257 PR rtl-optimization/85645
2258 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
2259 in the REG_CFA_REGISTER note for LR, don't leave it empty.
2261 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2263 PR rtl-optimization/85645
2264 * shrink-wrap.c (spread_components): Return a boolean saying if
2265 anything was changed.
2266 (try_shrink_wrapping_separate): Iterate spread_components until
2267 nothing changes anymore.
2269 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2271 PR rtl-optimization/85645
2272 * regrename.c (build_def_use): Also kill the chains that include the
2273 destination of a REG_CFA_REGISTER note.
2275 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2277 PR rtl-optimization/85645
2278 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
2279 insn that has a REG_CFA_REGISTER note.
2281 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
2283 * cfgexpand.c (expand_clobber): New function.
2284 (expand_gimple_stmt_1): Use it.
2285 * tree-vect-stmts.c (vect_clobber_variable): New function,
2287 (vectorizable_simd_clone_call): ...here.
2288 (vectorizable_store): Emit a clobber either side of an
2289 IFN_STORE_LANES sequence.
2290 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
2292 2018-05-09 Tom de Vries <tom@codesourcery.com>
2295 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
2296 (define_insn "trap_if_false"): Add exit after trap.
2298 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
2300 PR rtl-optimization/85638
2301 * bb-reorder.c: Include common/common-target.h.
2302 (create_forwarder_block): New function extracted from...
2303 (fix_up_crossing_landing_pad): ...here. Rename into...
2304 (dw2_fix_up_crossing_landing_pad): ...this.
2305 (sjlj_fix_up_crossing_landing_pad): New function.
2306 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
2307 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
2308 from both partitions and exit the loop after one iteration.
2310 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2313 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2315 (Basic PowerPC Built-in Functions): The new name of the
2316 subsection previously known as "PowerPC Built-in Functions".
2317 (Basic PowerPC Built-in Functions Available on all Configurations):
2319 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2321 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2323 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2325 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2328 2018-05-08 Jim Wilson <jimw@sifive.com>
2330 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
2331 (LD_EMUL_SUFFIX): New.
2332 (LINK_SPEC): Use it.
2334 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2336 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2338 (Basic PowerPC Built-in Functions): The new name of the
2339 subsection previously known as "PowerPC Built-in Functions".
2340 (Basic PowerPC Built-in Functions Available on all Configurations):
2342 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2344 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2346 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2348 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2351 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2354 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
2355 after cmpelim optimization.
2357 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
2359 * config.gcc: Support "goldmont".
2360 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
2361 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2363 * config/i386/i386.c (m_GOLDMONT): Define.
2364 (processor_target_table): Add "goldmont".
2365 (PTA_GOLDMONT): Define.
2366 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
2367 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
2368 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
2369 (fold_builtin_cpu): Add "goldmont".
2370 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
2371 (ix86_option_override_internal): Add "goldmont".
2372 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
2373 (processor_type): Add PROCESSOR_GOLDMONT.
2374 * config/i386/i386.md: Add CPU "glm".
2375 * config/i386/glm.md: New file.
2376 * config/i386/x86-tune.def: Add m_GOLDMONT.
2377 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
2379 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2382 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
2384 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
2385 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
2386 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
2389 * config/i386/i386.c (ix86_fold_builtin): Handle
2390 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
2393 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
2394 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
2396 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
2399 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
2400 (check_arch): Likewise.
2401 (check_fpu): Return the result rather than printing it.
2402 (end arch): Fix operator precedence.
2403 (end cpu): Likewise.
2404 (END): Print the result from check_fpu.
2406 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2407 Alan Hayward <alan.hayward@arm.com>
2408 David Sherwood <david.sherwood@arm.com>
2410 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
2411 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
2412 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
2413 (*fcmuo<mode>_and): New patterns.
2415 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2417 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
2418 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
2419 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
2420 (cmp_op, sve_imm_con): New code attributes.
2421 (SVE_COND_INT_CMP, imm_con): Delete.
2422 (cmp_op): Remove above unspecs from int attribute.
2423 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
2425 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
2426 comparison-specific unspecs.
2427 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
2428 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
2429 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
2430 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
2431 (*vec_fcm<cmp_op><mode>): Rename to...
2432 (*fcm<cmp_op><mode>): ...this and adjust likewise.
2433 (*vec_fcmuo<mode>): Rename to...
2434 (*fcmuo<mode>): ...this and adjust likewise.
2435 (*pred_fcm<cmp_op><mode>): New pattern.
2436 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
2437 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
2439 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
2441 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
2442 (aarch64_emit_sve_predicated_cond): New function.
2443 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
2444 (aarch64_emit_unspec_cond_or): Replace with...
2445 (aarch64_emit_sve_or_conds): ...this new function. Use
2446 aarch64_emit_sve_ptrue_op for the individual comparisons and
2447 aarch64_emit_binop to OR them together.
2448 (aarch64_emit_inverted_unspec_cond): Replace with...
2449 (aarch64_emit_sve_inverted_cond): ...this new function. Use
2450 aarch64_emit_sve_ptrue_op for the comparison and
2451 aarch64_emit_unop to invert the result.
2452 (aarch64_expand_sve_vec_cmp_float): Update after the above
2453 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
2455 2018-05-07 Nathan Sidwell <nathan@acm.org>
2457 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
2458 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
2459 (Backwards Compatibility): Likewise.
2461 2018-05-07 Luis Machado <luis.machado@linaro.org>
2465 2018-05-07 Luis Machado <luis.machado@linaro.org>
2467 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2468 <prefetch_dynamic_strides>: New const bool field.
2469 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2470 prefetch_dynamic_strides.
2471 (exynosm1_prefetch_tune): Likewise.
2472 (thunderxt88_prefetch_tune): Likewise.
2473 (thunderx_prefetch_tune): Likewise.
2474 (thunderx2t99_prefetch_tune): Likewise.
2475 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2476 (aarch64_override_options_internal): Update to set
2477 PARAM_PREFETCH_DYNAMIC_STRIDES.
2478 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2479 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2480 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2481 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2482 prefetch-dynamic-strides setting.
2484 2018-05-07 Luis Machado <luis.machado@linaro.org>
2486 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2487 <minimum_stride>: New const int field.
2488 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2489 minimum_stride field.
2490 (exynosm1_prefetch_tune): Likewise.
2491 (thunderxt88_prefetch_tune): Likewise.
2492 (thunderx_prefetch_tune): Likewise.
2493 (thunderx2t99_prefetch_tune): Likewise.
2494 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2495 (aarch64_override_options_internal): Update to set
2496 PARAM_PREFETCH_MINIMUM_STRIDE.
2497 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2498 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2499 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2500 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2501 stride is constant and is below the minimum stride threshold.
2503 2018-05-07 Luis Machado <luis.machado@linaro.org>
2505 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
2508 2018-05-07 Luis Machado <luis.machado@linaro.org>
2510 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2511 <prefetch_dynamic_strides>: New const bool field.
2512 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2513 prefetch_dynamic_strides.
2514 (exynosm1_prefetch_tune): Likewise.
2515 (thunderxt88_prefetch_tune): Likewise.
2516 (thunderx_prefetch_tune): Likewise.
2517 (thunderx2t99_prefetch_tune): Likewise.
2518 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2519 (aarch64_override_options_internal): Update to set
2520 PARAM_PREFETCH_DYNAMIC_STRIDES.
2521 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2522 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2523 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2524 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2525 prefetch-dynamic-strides setting.
2527 2018-05-07 Luis Machado <luis.machado@linaro.org>
2529 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2530 <minimum_stride>: New const int field.
2531 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2532 minimum_stride field.
2533 (exynosm1_prefetch_tune): Likewise.
2534 (thunderxt88_prefetch_tune): Likewise.
2535 (thunderx_prefetch_tune): Likewise.
2536 (thunderx2t99_prefetch_tune): Likewise.
2537 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2538 (aarch64_override_options_internal): Update to set
2539 PARAM_PREFETCH_MINIMUM_STRIDE.
2540 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2541 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2542 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2543 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2544 stride is constant and is below the minimum stride threshold.
2546 2018-05-06 Jakub Jelinek <jakub@redhat.com>
2549 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
2550 the type is addressable. Don't force op into register if it has
2553 2018-05-05 Roland McGrath <mcgrathr@google.com>
2556 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
2557 any section for which we don't know a specific type it should have,
2558 regardless of name. Previously this was done only for the exact
2559 names ".init_array", ".fini_array", and ".preinit_array".
2560 (default_elf_asm_named_section): Add comment about
2561 relationship with default_section_type_flags and SECTION_NOTYPE.
2562 (get_section): Don't consider it a type conflict if one side has
2563 SECTION_NOTYPE and the other doesn't, as long as neither has the
2564 SECTION_BSS et al used in the default_section_type_flags logic.
2566 2018-05-05 Tom de Vries <tom@codesourcery.com>
2569 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
2570 (workaround_barsyncs): New function.
2571 (nvptx_reorg): Use workaround_barsyncs.
2572 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2573 (define_expand "nvptx_membar_cta"): New define_expand.
2574 (define_insn "*nvptx_membar_cta"): New insn.
2576 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
2578 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
2579 To improve optimization opportunities.
2580 * builtin-types.def: The new needed builtin types for the above.
2582 2018-05-04 Richard Biener <rguenther@suse.de>
2584 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
2585 * gimple-ssa-store-merging.c
2586 (imm_store_chain_info::output_merged_store): Remove redundant create,
2587 release split_store vector contents on failure.
2588 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
2589 scalar stmt vector on cache hit.
2591 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
2593 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
2595 * config.gcc (powerpc-xilinx-eabi*): Remove.
2596 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
2598 (fusion_addis_mem_combo_load): Ditto.
2599 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
2601 (rs6000_cpu_cpp_builtins): Ditto.
2602 * config/rs6000/rs6000-linux.c
2603 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
2604 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
2605 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
2607 (rs6000_setup_reg_addr_masks): Ditto.
2608 (rs6000_init_hard_regno_mode_ok): Ditto.
2609 (rs6000_option_override_internal): Ditto.
2610 (legitimate_lo_sum_address_p): Ditto.
2611 (rs6000_legitimize_address): Ditto.
2612 (rs6000_legitimize_reload_address): Ditto.
2613 (rs6000_legitimate_address_p): Ditto.
2614 (abi_v4_pass_in_fpr): Ditto.
2615 (setup_incoming_varargs): Ditto.
2616 (rs6000_gimplify_va_arg): Ditto.
2617 (rs6000_split_multireg_move): Ditto.
2618 (rs6000_savres_strategy): Ditto.
2619 (rs6000_emit_prologue_components): Ditto.
2620 (rs6000_emit_epilogue_components): Ditto.
2621 (rs6000_emit_prologue): Ditto.
2622 (rs6000_emit_epilogue): Ditto.
2623 (rs6000_elf_file_end): Ditto.
2624 (rs6000_function_value): Ditto.
2625 (rs6000_libcall_value): Ditto.
2626 * config/rs6000/rs6000.h: Ditto.
2627 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
2628 (TARGET_MINMAX): ... this. New.
2629 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
2630 * config/rs6000/rs6000.md: Remove Xilinx FP support.
2631 (*movsi_internal1_single): Delete.
2632 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
2633 mfpu=, mxilinx-fpu): Delete.
2634 * config/rs6000/singlefp.h: Delete.
2635 * config/rs6000/sysv4.h: Remove Xilinx FP support.
2636 * config/rs6000/t-rs6000: Ditto.
2637 * config/rs6000/t-xilinx: Delete.
2638 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
2639 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
2640 (VStype_simple): Delete.
2641 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
2642 * config/rs6000/xfpu.h: Delete.
2643 * config/rs6000/xfpu.md: Delete.
2644 * config/rs6000/xilinx.h: Delete.
2645 * config/rs6000/xilinx.opt: Delete.
2646 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
2647 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
2649 2018-05-04 Tom de Vries <tom@codesourcery.com>
2652 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
2655 2018-05-04 Richard Biener <rguenther@suse.de>
2658 * tree-complex.c (update_complex_assignment): We are always in SSA form.
2659 (expand_complex_div_wide): Likewise.
2660 (expand_complex_operations_1): Likewise.
2661 (expand_complex_libcall): Preserve EH info of the original stmt.
2662 (tree_lower_complex): Handle removed blocks.
2663 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
2664 on complex multiplication and division libcall builtins.
2666 2018-05-04 Richard Biener <rguenther@suse.de>
2669 * fold-const.c (negate_expr_p): Restrict negation of operand
2670 zero of a division to when we know that can happen without
2672 (fold_negate_expr_1): Likewise.
2674 2018-05-04 Jakub Jelinek <jakub@redhat.com>
2677 * real.h (real_nextafter): Declare.
2678 * real.c (real_nextafter): New function.
2679 * fold-const-call.c (fold_const_nextafter): New function.
2680 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
2681 CASE_CFN_NEXTTOWARD.
2682 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
2683 even when arg1_mode is different from arg0_mode.
2685 2018-05-03 Nathan Sidwell <nathan@acm.org>
2687 * doc/extend.texi (Deprecated Features): Remove
2689 (Backwards Compatibility): Likewise.
2690 * doc/invoke.texi (C++ Language Options): Likewise.
2691 (C++ Dialect Options): Likewise.
2693 2018-05-03 Jakub Jelinek <jakub@redhat.com>
2696 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
2697 _mm512_mask_mullox_epi64): New intrinsics.
2699 2018-05-03 Tom de Vries <tom@codesourcery.com>
2702 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2703 dump files): Add offload-tree.
2705 2018-05-03 Richard Biener <rguenther@suse.de>
2707 PR tree-optimization/85615
2708 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
2709 to loops not nested in BBs loop father to avoid creating multi-entry
2712 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2714 PR tree-optimization/70291
2715 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
2716 arguments. Change return type to tree. Emit libcall as a new
2717 statement rather than replacing existing one when inplace_p is true.
2718 (expand_complex_multiplication_components): New function.
2719 (expand_complex_multiplication): Expand floating-point complex
2720 multiplication using the above.
2721 (expand_complex_division): Rename inner_type parameter to type.
2722 Update expand_complex_libcall call-site.
2723 (expand_complex_operations_1): Update expand_complex_multiplication
2724 and expand_complex_division call-sites.
2726 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2729 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2730 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2731 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
2732 the highest significant bit of the shift count mask is clear. In
2733 check whether and[sq]i3 is needed verify that all significant bits
2734 of the shift count other than the highest are set.
2736 2018-05-02 Tom de Vries <tom@codesourcery.com>
2739 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
2740 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
2741 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
2742 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
2743 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
2744 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
2745 __builtin_goacc_parlevel_size.
2747 2018-05-02 Richard Biener <rguenther@suse.de>
2749 PR tree-optimization/85597
2750 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
2751 do not use split vect_get_vec_defs call but call vect_get_slp_defs
2754 2018-05-02 Tom de Vries <tom@codesourcery.com>
2757 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2758 dump files): Add ltrans-tree.
2760 2018-05-02 Tom de Vries <tom@codesourcery.com>
2763 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2764 dump files): Add wpa-ipa.
2766 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
2768 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
2769 powerpc*-*-linux*paired* target.
2770 * config/rs6000/750cl.h: Delete.
2771 * config/rs6000/paired.h: Delete.
2772 * config/rs6000/paired.md: Delete.
2773 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
2775 * config/rs6000/rs6000-builtin.def: Remove paired float support.
2776 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
2777 comment. Remove paired float support.
2778 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
2779 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
2781 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
2782 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
2784 * config/rs6000/rs6000.c: Remove paired float support.
2785 (paired_expand_vector_init, paired_expand_vector_move,
2786 paired_emit_vector_compare, paired_emit_vector_cond_expr,
2787 (paired_expand_lv_builtin, paired_expand_stv_builtin,
2788 paired_expand_builtin, paired_expand_predicate_builtin,
2789 paired_init_builtins): Delete.
2790 * config/rs6000/rs6000.h: Remove paired float support.
2791 * config/rs6000/rs6000.md: Remove paired float support.
2792 (move_from_CR_ov_bit): Delete.
2793 * config/rs6000/rs6000.opt (mpaired): Delete.
2794 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
2795 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
2797 2018-05-02 Richard Biener <rguenther@suse.de>
2800 * gimplify.c (gimplify_save_expr): When in SSA form allow
2801 SAVE_EXPRs to compute to SSA vars.
2803 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2806 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2807 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2808 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
2809 clobber operands[2], instead use a new pseudo. Formatting fixes.
2811 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
2813 PR tree-optimization/85586
2814 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
2815 exit early for statements in the same group if the accesses are
2818 2018-05-02 Tom de Vries <tom@codesourcery.com>
2821 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
2824 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
2826 PR tree-optimization/85143
2827 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
2829 2018-05-01 Tom de Vries <tom@codesourcery.com>
2832 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
2833 not found" error message.
2835 2018-05-01 Tom de Vries <tom@codesourcery.com>
2838 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
2839 * vec.c (test_ordered_remove_if): New function.
2840 (vec_c_tests): Call test_ordered_remove_if.
2841 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
2842 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
2843 * tree-vect-patterns.c (vect_pattern_recog_1): Use
2844 VEC_ORDERED_REMOVE_IF.
2846 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2848 PR tree-optimization/82665
2849 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
2850 pointer subtraction where arguments come from a memchr call.
2852 2018-05-01 Jakub Jelinek <jakub@redhat.com>
2854 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
2855 --push-state --as-needed and --pop-state instead of --as-needed and
2856 --no-as-needed if ld supports it.
2857 * configure: Regenerated.
2860 * doc/install.texi2html: Replace _002d with - and _002a with * in
2861 generated html files using sed.
2863 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2866 * gcc-rich-location.c (blank_line_before_p): New function.
2867 (use_new_line): New function.
2868 (gcc_rich_location::add_fixit_insert_formatted): New function.
2869 * gcc-rich-location.h
2870 (gcc_rich_location::add_fixit_insert_formatted): New function.
2872 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2874 * selftest.c (assert_streq): Rename "expected" and "actual" to
2875 "val1" and "val2". Extend NULL-handling to cover both inputs
2876 symmetrically, while still requiring both to be non-NULL for a pass.
2877 * selftest.h (assert_streq): Rename "expected" and "actual" to
2879 (ASSERT_EQ): Likewise.
2880 (ASSERT_EQ_AT): Likewise.
2881 (ASSERT_KNOWN_EQ): Likewise.
2882 (ASSERT_KNOWN_EQ_AT): Likewise.
2883 (ASSERT_NE): Likewise.
2884 (ASSERT_MAYBE_NE): Likewise.
2885 (ASSERT_MAYBE_NE_AT): Likewise.
2886 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
2887 the assertion to pass.
2888 (ASSERT_STREQ_AT): Likewise.
2890 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2892 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
2893 interaction with -pie.
2895 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2897 * selftest.h: Fix alphabetization of per-source-file selftest
2900 2018-04-30 Jason Merrill <jason@redhat.com>
2902 PR c++/61982 - dead stores to destroyed objects.
2903 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
2906 2018-04-30 Jason Merrill <jason@redhat.com>
2908 * tree.c (build_clobber): New.
2909 * tree.h: Declare it.
2910 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
2912 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2914 * diagnostic-show-locus.c (layout::layout): Update for
2915 location_get_source_line returning a char_span.
2916 (struct char_span): Move to input.h.
2917 (struct correction): Update for fields in char_span becoming
2919 (struct source_line): Update for location_get_source_line
2920 returning a char_span.
2921 (layout::print_line): Likewise.
2922 * edit-context.c (edited_file::print_content): Likewise.
2923 (edited_file::print_diff_hunk): Likewise.
2924 (edited_file::print_run_of_changed_lines): Likewise.
2925 (edited_file::get_num_lines): Likewise.
2926 (edited_line::edited_line): Likewise.
2927 * final.c (asm_show_source): Likewise.
2928 * input.c (location_get_source_line): Convert return type
2929 from const char * to char_span, losing the final "line_len"
2931 (dump_location_info): Update for the above.
2932 (get_substring_ranges_for_loc): Likewise. Use a char_span
2933 when handling the literal within the line.
2934 (test_reading_source_line): Update for location_get_source_line
2935 returning a char_span.
2936 * input.h (class char_span): Move here from
2937 diagnostic-show-locus.c, converting from a struct to a class.
2938 Make data members private.
2939 (char_span::operator bool): New.
2940 (char_span::length): New.
2941 (char_span::get_buffer): New.
2942 (char_span::operator[]): New.
2943 (char_span::subspan): Make const.
2944 (char_span::xstrdup): New.
2945 (location_get_source_line): Convert return type from const char *
2946 to char_span, losing the final "line_size" param.
2948 2018-04-30 Jan Hubicka <jh@suse.cz>
2950 * lto-wrapper.c (ltrans_priorities): New static var.
2951 (cmp_priority): New.
2952 (run_gcc): Read priorities and if doing parallel build order
2953 the Makefile by them.
2955 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2957 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
2959 2018-04-30 Richard Biener <rguenther@suse.de>
2961 * tree-cfg.c (verify_address): Remove base argument, add
2962 flag whether to check TREE_ADDRESSABLE and do that.
2963 (verify_expr): Remove.
2964 (verify_types_in_gimple_reference): Add pieces from verify_expr.
2965 (verify_gimple_assign_single): Likewise.
2966 (verify_gimple_switch): Likewise.
2967 (verify_expr_location_1): Dereference tp once. Add (disabled)
2968 piece from verify_expr.
2969 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
2971 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2973 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
2975 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2977 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
2978 (small_data_pattern): Likewise.
2979 (arc_rewrite_small_data): Likewise.
2980 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
2981 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
2982 (get_symbol_alignment): New function.
2983 (legitimate_small_data_address_p): Likewise.
2984 (legitimate_scaled_address): Update, call
2985 legitimate_small_data_address_p.
2986 (output_sdata): New static variable.
2987 (arc_print_operand): Update how we handle small data operands.
2988 (arc_print_operand_address): Likewise.
2989 (arc_legitimate_address_p): Update, use
2990 legitimate_small_data_address_p.
2991 (arc_rewrite_small_data_p): Remove.
2992 (arc_rewrite_small_data_1): Likewise.
2993 (arc_rewrite_small_data): Likewise.
2994 (small_data_pattern): Likewise.
2995 (compact_sda_memory_operand): Update to use
2996 legitimate_small_data_address_p and get_symbol_alignment.
2997 (prepare_move_operands): Don't rewite sdata pattern.
2998 (prepare_extend_operands): Remove.
2999 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
3001 (zero_extendqisi2): Likewise.
3002 (zero_extendhisi2): Likewise.
3003 (extendqihi2): Likewise.
3004 (extendqisi2): Likewise.
3005 (extendhisi2): Likewise.
3009 * config/arc/constraints.md (Usd): Change it to memory constraint.
3011 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3013 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
3014 as source of std instructions.
3015 * config/arc/arc.md (movsi_insn): Update pattern predicate to
3016 allow 6-bit constants as source for store instructions.
3017 (movdi_insn): Update instruction pattern to allow 6-bit constants
3018 as source for store instructions.
3020 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3022 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
3024 2018-04-30 Nathan Sidwell <nathan@acm.org>
3025 Sandra Loosemore <sandra@codesourcery.com>
3027 * dumpfile.c (dump_open): Allow '-' for stdout.
3028 * doc/invoke.texi (Developer Options): Document dump filename
3029 determination early. Document stdin/stdout selection.
3031 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
3033 Microblaze Target: PIC data text relative
3035 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
3036 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
3038 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
3039 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
3040 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
3041 New addressing mode for data-text relative position indepenedent code.
3042 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
3043 'ADDRESS_SYMBOLIC_TXT_REL'.
3044 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
3045 (microblaze_legitimate_pic_operand): Exclude function calls from
3046 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
3047 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
3049 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3050 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3051 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
3052 for 'address + offset'.
3053 (microblaze_expand_prologue): Add new function prologue call for
3055 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
3056 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
3057 table in case of TARGET_PIC_DATA_TEXT_REL.
3058 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
3059 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
3060 Add new macros 'UNSPEC_TEXT',
3061 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
3062 + exclude function calls from 'UNSPEC_PLT' in case of data text
3064 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3065 new target hook for generating address diff vector tables in case of
3067 * doc/tm.texi : Regenerate.
3068 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
3069 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
3070 of addr diff vector generation.
3071 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3072 target hook definition.
3073 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
3074 Add default function for generate_pic_addr_diff_vec -> flag_pic.
3075 * doc/invoke.texi (Add new pic option): Add new microblaze pic
3076 option for data text relative.
3078 2018-04-30 Richard Biener <rguenther@suse.de>
3080 * tree-chrec.h (evolution_function_is_constant_p): Remove
3082 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
3084 2018-04-30 Richard Biener <rguenther@suse.de>
3087 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
3089 2018-04-30 Richard Biener <rguenther@suse.de>
3091 PR tree-optimization/28364
3092 PR tree-optimization/85275
3093 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
3094 copying first exit test.
3096 2018-04-28 Mark Wielaard <mark@klomp.org>
3098 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
3100 (dwarf_AT): Handle DW_AT_addr_base.
3101 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
3103 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
3106 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
3107 (*ashl<dwi>3_doubleword_mask_1): Ditto.
3108 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
3109 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
3111 2018-04-28 Richard Biener <rguenther@suse.de>
3113 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
3114 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
3115 to reflect use. Only add interesting stmts.
3117 2018-04-27 Martin Jambor <mjambor@suse.cz>
3120 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
3121 the jump function allows for passing through aggregate values.
3123 2018-04-27 David Malcolm <dmalcolm@redhat.com>
3125 * input.h (in_system_header_at): Convert from macro to inline
3127 (from_macro_expansion_at): Likewise.
3128 (from_macro_definition_at): Likewise.
3130 2018-04-27 Jeff Law <law@redhat.com>
3132 * config.gcc: Mark tile* targets as deprecated/obsolete.
3134 2018-04-27 Richard Biener <rguenther@suse.de>
3136 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
3139 2018-04-27 Richard Biener <rguenther@suse.de>
3141 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
3143 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
3145 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
3146 with Yd constraint. Set "preferred_for_speed" attribute from
3147 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
3149 (*movdi_internal): Ditto.
3150 (movti_interunit splitters): Remove
3151 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
3152 (movdi_interunit splitters): Ditto.
3153 * config/i386/constraints.md (Ye): Remove.
3154 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
3156 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3159 * config/aarch64/constraints.md (Usg): Limit to 31.
3162 2018-04-27 Jakub Jelinek <jakub@redhat.com>
3164 PR tree-optimization/85529
3165 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
3166 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
3167 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
3168 zero extension or masking of the MSB bit.
3169 (optimize_range_tests): Add FIRST_BB argument, pass it through
3170 to optimize_range_tests_var_bound.
3171 (maybe_optimize_range_tests, reassociate_bb): Adjust
3172 optimize_range_tests callers.
3174 2018-04-26 Richard Biener <rguenther@suse.de>
3175 Jakub Jelinek <jakub@redhat.com>
3177 * cgraph.h (symbol_table): Just declare debug method here.
3178 * symtab.c (symbol_table::debug): Define.
3180 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
3182 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
3184 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
3186 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
3187 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
3188 (*movdi_internal): Substitute Yi and Yj constraint with x
3189 and Ym and Yn constraint with y constraint. Update "isa"
3190 attribute and set "preferred_for_speed" attribute from
3191 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
3192 (*movsi_internal): Ditto.
3193 (*movdf_internal): Ditto.
3194 (*movsf_internal): Ditto.
3195 (*zero_extendsidi2): Ditto.
3196 * config/i386/sse.md (vec_set<mode>_0): Ditto.
3197 (sse2_loadld): Ditto.
3198 (*vec_extract<ssevecmodelower>_0): Ditto.
3199 (*vec_extractv4si_0_zext_sse4): Ditto.
3200 (vec_concatv2di): Ditto.
3201 (*vec_dup<mode>): Ditto.
3202 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
3203 * config/i386/constraints.md (Yi): Remove.
3208 2018-04-26 Nathan Sidwell <nathan@acm.org>
3210 * dumpfile.c (dump_open): New.
3211 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
3212 (dump_finish): Detect stdio/stderr by value not name.
3214 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
3216 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
3218 2018-04-26 Tom de Vries <tom@codesourcery.com>
3221 * config/nvptx/nvptx.c (verify_neutering_jumps)
3222 (verify_neutering_labels): New function
3223 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
3225 2018-04-26 Tom de Vries <tom@codesourcery.com>
3228 * config/nvptx/nvptx.c (needs_neutering_p): New function.
3229 (nvptx_single): Use needs_neutering_p to skip over insns that do not
3232 2018-04-26 Richard Biener <rguenther@suse.de>
3233 Tom de Vries <tom@codesourcery.com>
3236 * lto-streamer-out.c (output_function): Fixup loops if required to match
3237 discovery done in the reader.
3239 2018-04-26 Richard Biener <rguenther@suse.de>
3241 PR tree-optimization/85116
3242 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
3243 have a loop exit from the single latch predecessor. Remove
3244 case of header with just condition.
3245 (ch_base::copy_headers): Exclude infinite loops from any
3247 (pass_ch::execute): Record exits.
3249 2018-04-26 Richard Biener <rguenther@suse.de>
3251 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
3252 prologue cost vector and pass it to vect_get_load_cost.
3253 (vect_get_peeling_costs_all_drs): Likewise.
3254 (vect_peeling_hash_get_lowest_cost): Likewise.
3255 (vect_enhance_data_refs_alignment): Likewise.
3257 2018-04-26 Richard Biener <rguenther@suse.de>
3260 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
3261 checking of integer<->pointer conversions.
3262 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
3263 sign-/zero-extending pointer types.
3264 (expand_omp_for_static_chunk): Likewise.
3266 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
3267 Jean Lee <xiaoyur347@gmail.com>
3269 * config/mips/mips.c (mips_asan_shadow_offset): New function.
3270 (TARGET_ASAN_SHADOW_OFFSET): Define.
3271 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
3272 true for -fsanitize=address.
3274 2018-04-25 Mark Wielaard <mark@klomp.org>
3276 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
3279 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3281 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
3282 than "alu", remove explicit "memory" and "imm_disp" attributes.
3283 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
3286 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
3287 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
3288 gen_lowpart_no_emit.
3290 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
3293 * config/i386/i386.c (ix86_expand_builtin): Change memory
3294 operand to XI, extend p0 to Pmode.
3295 * config/i386/i386.md: Change unspec volatile and operand
3296 1 mode to XI, change operand 0 mode to P.
3298 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3300 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
3301 GET_MODE_MASK before any checking.
3302 (nds32_can_use_bset_p): Likewise.
3303 (nds32_can_use_btgl_p): Likewise.
3305 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3307 * config/nds32/nds32-doubleword.md: New define_split pattern for
3308 illegal register number.
3310 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3312 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
3314 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3316 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
3318 2018-04-25 Richard Biener <rguenther@suse.de>
3320 * lto-streamer.h (LTO_major_version): Bump to 8.
3322 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3324 * BASE-VER: Set to 9.0.0.
3326 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
3328 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
3329 in __abskf2 and __powikf2.
3331 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3334 * config/aarch64/constraints.md (Usg, Usj): New constraints.
3335 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
3336 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
3337 Use the above on operand 2. Reindent.
3338 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
3340 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3343 * common/config/i386/i386-common.c (ix86_handle_option): Don't
3345 * config/i386/i386.opt (mcet): Removed.
3346 * doc/install.texi: Remove -mcet documentation.
3347 * doc/invoke.texi: Likewise.
3349 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3352 * doc/install.texi: Remove -mcet from bootstrap-cet.
3354 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3357 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
3358 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
3362 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
3363 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
3364 containing a CONST_VECTOR.
3366 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
3368 * doc/install.texi: Update newlib dependency for nvptx.
3370 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3373 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
3374 instead of INTVAL when shifting x left.
3376 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
3378 PR tree-optimization/85478
3379 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
3380 vect_grouped_store_supported for single element vectors.
3382 2018-04-24 Richard Biener <rguenther@suse.de>
3385 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
3386 load cost increase to the case of non-constant step.
3388 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3391 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
3392 destination if any_malformed_asm.
3394 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
3397 * expr.c (store_field): In the bitfield case, if the value comes from
3398 a function call and is returned in registers by means of a PARALLEL,
3399 do not change the mode of the temporary unless BLKmode and VOIDmode.
3401 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
3403 PR rtl-optimization/85423
3404 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
3405 dependencies to debug insns when the previous insn is non-debug.
3407 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3409 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
3410 enums into a single definition.
3411 (fls): Fix predicates and printing.
3414 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3416 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
3417 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
3418 and short u6 immediate.
3419 (check_if_valid_sleep_operand): Remove.
3420 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
3422 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3424 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
3425 flag_always_save_lp condition.
3426 * config/nds32/nds32.opt (malways-save-lp): New option.
3428 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3430 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
3431 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
3432 * config/nds32/nds32.h
3433 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
3434 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
3436 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3438 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
3439 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
3441 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3442 Chung-Ju Wu <jasonwucj@gmail.com>
3444 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
3446 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
3447 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
3449 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3451 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
3453 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3455 * config/nds32/nds32-protos.h (nds32_data_alignment,
3456 nds32_local_alignment): Declare.
3457 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
3458 nds32_local_alignment): New functions.
3459 (TARGET_CONSTANT_ALIGNMENT): Define.
3460 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
3462 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3464 * config/nds32/nds32.c
3465 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
3466 (TARGET_MODES_TIEABLE_P): Likewise.
3468 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3470 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
3473 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
3474 Chung-Ju Wu <jasonwucj@gmail.com>
3476 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3477 for unaligned access.
3478 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3480 * config/nds32/nds32-intrinsic.md: Likewise.
3481 * config/nds32/nds32_intrinsic.h: Likewise.
3482 * config/nds32/nds32.h (nds32_builtins): Likewise.
3483 * config/nds32/nds32.opt (munaligned-access): New option.
3484 * config/nds32/nds32.c (nds32_asm_file_start): Display
3485 flag_unaligned_access status.
3487 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3489 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
3490 -mno-relax is present.
3491 * config/riscv/linux.h (LINK_SPEC): Ditto.
3493 2018-04-20 Martin Sebor <msebor@redhat.com>
3496 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
3498 (gimple_fold_builtin_stxcpy_chk): Same.
3499 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
3501 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
3504 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
3505 __powikf2 when long double is IEEE 128-bit.
3507 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3509 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
3510 step to make sure stack always aligned.
3512 2018-04-20 Carl Love <cel@us.ibm.com>
3515 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
3516 size check for arg0.
3518 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
3519 Tom de Vries <tom@codesourcery.com>
3522 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
3523 Emit insns for calls too.
3524 (nvptx_find_par): Always look for worker-level predecessor insn.
3525 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
3527 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
3528 (nvptx_process_pars): Propagate frames for calls.
3530 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
3533 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
3535 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
3536 (ix86_handle_option): Don't handle OPT_mibt.
3537 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
3539 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
3541 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
3542 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
3543 (ix86_target_macros): Define __CET__ with flag_cf_protection
3544 for -fcf-protection.
3545 * config/i386/i386.c (isa2_opts): Remove -mibt.
3546 * config/i386/i386.h (TARGET_IBT): Removed.
3547 (TARGET_IBT_P): Likewise.
3548 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
3549 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
3550 * config/i386/i386.opt (mcet): Update help message.
3553 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
3554 -mcet as an alias for -mshstk.
3556 2018-04-20 Richard Biener <rguenther@suse.de>
3559 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
3560 complexity by forcing a single use of the multiply operand.
3562 2018-04-20 Martin Jambor <mjambor@suse.cz>
3565 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
3566 recursion dependency to only apply to non-clones.
3568 2018-04-20 Martin Jambor <mjambor@suse.cz>
3571 * ipa-cp.c (create_specialized_node): Check that clones of
3572 self-recursive edges exist during IPA-CP.
3574 2018-04-19 Toon Moene <toon@moene.org>
3576 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
3579 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3581 PR tree-optimization/85467
3582 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
3583 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
3584 VECTOR_CST element to type.
3586 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3589 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
3590 * config/i386/i386.md (builtin_setjmp_setup): Removed.
3591 (builtin_longjmp): Likewise.
3592 (save_stack_nonlocal): New pattern.
3593 (restore_stack_nonlocal): Likewise.
3595 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3598 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3599 Replace ASM_OUTPUT_LABEL with fprintf.
3601 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3604 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3605 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
3606 * config/i386/i386-c.c (ix86_target_macros_internal): Also
3607 define __IBT__ and __SHSTK__ for -fcf-protection.
3608 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
3610 (ix86_trampoline_init): Likewise.
3611 (x86_output_mi_thunk): Likewise.
3612 (ix86_notrack_prefixed_insn_p): Likewise.
3613 (ix86_option_override_internal): Don't disallow -fcf-protection.
3614 * config/i386/i386.md (rdssp<mode>): Also enable for
3616 (incssp<mode>): Likewise.
3617 (nop_endbr): Likewise.
3618 * config/i386/i386.opt (mcet): Change help message to built-in
3622 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
3623 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
3624 enable CET built-in functions.
3626 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
3628 * common/config/i386/i386-common.c
3629 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
3630 OPTION_MASK_ISA_MOVDIRI_UNSET,
3631 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
3632 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
3633 * config.gcc (movdirintrin.h): New header.
3634 * config/i386/cpuid.h (bit_MOVDIRI,
3635 bit_MOVDIR64B): New bits.
3636 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
3638 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
3639 (VOID, PVOID, PCVOID)): New function types.
3640 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
3641 __builtin_ia32_directstoreu_u64,
3642 __builtin_ia32_movdir64b): New builtins.
3643 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
3644 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
3646 (ix86_valid_target_attribute_inner_p): Ditto.
3647 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
3648 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
3649 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
3650 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
3651 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
3652 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
3653 (movdiri<mode>, movdir64b_<mode>): New.
3654 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
3655 * config/i386/immintrin.h: Include movdirintrin.h.
3656 * config/i386/movdirintrin.h: New file.
3657 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
3659 2018-04-19 Richard Biener <rguenther@suse.de>
3662 * cfg.c (clear_bb_flags): When loop state says we have
3663 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
3665 2018-04-19 Richard Biener <rguenther@suse.de>
3667 PR tree-optimization/84737
3668 * tree-vect-data-refs.c (vect_copy_ref_info): New function
3669 copying restrict info.
3670 (vect_setup_realignment): Use it.
3671 * tree-vectorizer.h (vect_copy_ref_info): Declare.
3672 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
3673 the first DR to all generated stores.
3674 (vectorizable_load): Likewise for loads.
3676 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3678 PR tree-optimization/85446
3679 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
3680 the integral and pointer types to have the same precision.
3682 * doc/install.texi: Document --disable-cet being the default and
3685 2018-04-18 Martin Liska <mliska@suse.cz>
3687 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
3690 2018-04-18 Martin Liska <mliska@suse.cz>
3693 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3696 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3697 arguments if they are comparable.
3699 2018-04-18 Martin Liska <mliska@suse.cz>
3702 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3705 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3708 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
3711 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
3712 ENDBR after calling __morestack.
3714 2018-04-18 David Malcolm <dmalcolm@redhat.com>
3717 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
3718 by using gcc_base_ver to generate a gcc_driver_version, and use
3719 it when generating GCC_DRIVER_NAME.
3720 * configure: Regenerate.
3722 2018-04-18 Jakub Jelinek <jakub@redhat.com>
3725 * config.gcc: Obsolete powerpc*-*-*spe*.
3727 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3730 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
3731 (stabstr_D): Change type of unum from unsigned int to
3732 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
3735 2018-04-17 Jim Wilson <jimw@sifive.com>
3738 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
3739 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
3740 Set arg_pointer_offset after using pretend_args_size.
3742 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3744 PR rtl-optimization/85431
3745 * dse.c (record_store): Ignore zero width stores.
3748 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
3749 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
3750 __builtin_stack_restore rather than after it.
3751 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
3752 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
3753 argument instead of virtual_dynamic_stack_rtx.
3755 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
3757 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
3759 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
3760 Add note to error message to explain internal mapping of overloaded
3761 built-in function name to non-overloaded built-in function name.
3762 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
3765 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3768 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
3769 where the inputs overlap with the output.
3771 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3773 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
3774 (=v, v) alternative and explicit "memory" attribute.
3775 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
3776 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3778 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
3779 "sselog1" type instead of "sselog".
3780 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
3781 "sselog". Remove explicit "memory" attribute.
3782 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
3783 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3785 (vec_extract_hi_v32hi): Merge all alternatives into one, use
3786 "sselog1" type instead of "sselog". Remove explicit "memory"
3788 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
3789 use "sselog1" type instead of "sselog". Remove explicit "memory"
3791 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
3792 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3794 (vec_extract_hi_v64qi): Merge all alternatives into one, use
3795 "sselog1" type instead of "sselog". Remove explicit "memory"
3797 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
3798 use "sselog1" type instead of "sselog". Remove explicit "memory"
3802 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
3805 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
3808 2018-04-17 Martin Jambor <mjambor@suse.cz>
3811 * ipa-cp.c (create_specialized_node): Call
3812 expand_all_artificial_thunks if necessary.
3814 2018-04-17 Martin Liska <mliska@suse.cz>
3817 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
3818 in message, remote space in between '_G' and '('.
3820 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3823 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
3824 avx512f_vmcmp<mode>3<round_saeonly_name>,
3825 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
3826 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
3827 avx512f_rndscale<mode><round_saeonly_name>,
3828 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
3829 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
3830 Use %<iptr>2 instead of %2 for -masm=intel.
3831 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
3832 avx512f_vcvttss2usi<round_saeonly_name>,
3833 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
3835 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
3836 avx512f_vcvttsd2usi<round_saeonly_name>,
3837 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
3838 Use %q1 instead of %1 for -masm=intel.
3839 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
3840 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
3841 of %3 for -masm=intel.
3842 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
3844 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
3846 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
3848 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
3849 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
3851 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
3853 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
3854 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
3855 %g1 and one with %0 and %1.
3856 (avx512er_vmrcp28<mode><round_saeonly_name>,
3857 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
3859 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
3860 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
3861 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
3862 of %0 and %{%4%} for -masm=intel.
3863 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
3864 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
3865 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
3866 order of %0 and %{%5%}%{z%} for -masm=intel.
3868 2018-04-17 Jan Hubicka <jh@suse.cz>
3871 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
3873 2018-04-17 Martin Liska <mliska@suse.cz>
3876 * multiple_target.c (create_dispatcher_calls): Set apostrophes
3877 for target_clone error message. Make default implementation
3878 clone to be a local declaration.
3879 (separate_attrs): Add new argument and check for an empty
3881 (expand_target_clones): Handle it.
3882 (ipa_target_clone): Make redirection just for target_clones
3885 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
3886 Tom de Vries <tom@codesourcery.com>
3889 * omp-expand.c (expand_oacc_for): Add dummy false branch for
3890 tiled basic blocks without omp continue statements.
3892 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
3895 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
3896 vec_extract expression as having side effects to make sure it gets
3899 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
3902 * config/i386/i386.c (get_builtin_code_for_version): Check
3905 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
3908 * gcc/config.gcc: Support "skylake".
3909 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
3911 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
3912 (processor_target_table): Add "skylake".
3913 (ix86_option_override_internal): Add "skylake".
3914 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
3915 PROCESSOR_CANNONLAKE.
3916 (get_builtin_code_for_version): Fix priority for
3917 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
3918 PROCESSOR_SKYLAKE-AVX512.
3919 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
3920 (processor_type): Add PROCESSOR_SKYLAKE.
3922 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
3923 Jason Merrill <jason@redhat.com>
3926 * convert.c (convert_to_integer_1): Use direct recursion for
3927 enumeral types and types with a precision less than the number
3928 of bits in their mode.
3930 2018-04-16 Julia Koval <julia.koval@intel.com>
3933 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
3934 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
3936 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
3939 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
3940 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
3941 and -mno-direct-move.
3943 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
3946 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
3947 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
3948 Ensure negative shifts result in {0}.
3950 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
3952 PR rtl-optimization/79916
3953 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
3954 regs (if any) to define how to gnerate SD moves when LRA is in
3957 2018-04-13 Jakub Jelinek <jakub@redhat.com>
3959 PR rtl-optimization/85393
3960 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
3961 * except.c (expand_dw2_landing_pad_for_region): Make static.
3962 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
3963 a label and unconditional jump to old_bb, rather than
3964 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
3967 PR rtl-optimization/85376
3968 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
3969 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
3970 instead of a specific value.
3972 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3973 Bin Cheng <bin.cheng@arm.com>
3975 PR tree-optimization/82965
3976 PR tree-optimization/83991
3977 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
3978 by_profile_only parameter.
3979 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
3980 information if the loop was predicted to iterate too many times.
3981 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
3983 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3986 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
3989 2018-04-13 Martin Liska <mliska@suse.cz>
3990 Jakub Jelinek <jakub@redhat.com>
3993 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
3994 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
3995 * builtins.c (expand_builtin_memory_copy_args): Use
3996 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
3997 handle dest_addr == pc_rtx.
3999 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
4002 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
4003 asked to not generate direct moves.
4004 (fix_trunc<mode>si2_stfiwx): Similar.
4005 (fix_trunc<mode>si2_internal): Similar.
4007 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4010 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
4011 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
4012 lookup if dest in some wider mode is known to be const0_rtx and
4013 if so, record permanent equivalence for it to be ZERO_EXTEND of
4014 the narrower mode destination.
4016 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4018 * lto-streamer-out.c (output_function): Revert 259346.
4019 * omp-expand.c (expand_oacc_for): Likewise.
4021 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
4023 PR rtl-optimization/85354
4024 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
4025 * sel-sched.c (sel_global_init): ... here.
4027 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
4030 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
4031 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
4032 mode for PE-COFF targets.
4033 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
4034 (i386_pe_asm_lto_end): Likewise.
4035 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
4036 (TARGET_ASM_LTO_END): Likewise.
4037 * config/i386/winnt.c (saved_debug_info_level): New static variable.
4038 (i386_pe_asm_lto_start): New function.
4039 (i386_pe_asm_lto_end): Likewise.
4041 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4042 Richard Biener <rguenther@suse.de>
4045 * lto-streamer-out.c (output_function): Fix CFG loop state before
4047 * omp-expand.c (expand_oacc_for): Handle calls to internal
4048 functions like regular functions.
4050 2018-04-12 Richard Biener <rguenther@suse.de>
4053 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
4054 for the early LTO debug to properly generate references to it
4055 during DIE emission. Do not re-use that for the skeleton for
4057 (dwarf2out_early_finish): Likewise.
4059 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4062 * config/i386/sse.md
4063 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
4064 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
4065 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
4066 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
4067 and output is a reg, avoid creating invalid lowpart subreg, but
4068 instead split into a 512-bit move. Don't split if not AVX512VL,
4069 input is xmm16+ reg and output is a mem.
4070 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
4071 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
4072 xmm16+ reg and output is a mem.
4074 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4076 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
4077 also for flag_dwarf2_cfi_asm.
4079 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4081 PR rtl-optimization/85342
4082 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
4083 a bool scalar var inside of the loop instead. Don't try to update
4084 recog_data.operand after failed apply_change_group.
4086 2018-04-12 Tom de Vries <tom@codesourcery.com>
4089 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
4090 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
4091 array with flexible array member as array without given dimension.
4092 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
4093 argument for undefined param to true.
4095 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
4098 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4099 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
4100 from PowerPC section.
4101 * config/rs6000/sysv4.opt (mcall-): Improve help text.
4102 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
4103 help text that is too long.
4104 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
4105 help text that is too long.
4106 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
4107 help text that is too long.
4109 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
4111 * config/alpha/alpha.md (stack_probe_internal): Rename
4112 from "probe_stack". Update all callers.
4114 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4116 PR rtl-optimization/84566
4117 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
4118 sched_macro_fuse_insns.
4120 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4123 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
4124 (compute_block_dependences): ... from here.
4126 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4128 PR tree-optimization/85331
4129 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
4130 from int to HOST_WIDE_INT.
4132 2018-04-11 Martin Jambor <mjambor@suse.cz>
4135 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
4136 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
4137 not the same as the source val.
4138 (cgraph_edge_brings_value_p): New parameter.
4139 (gather_edges_for_value): Pass destination value to
4140 cgraph_edge_brings_value_p.
4141 (perhaps_add_new_callers): Likewise.
4142 (get_info_about_necessary_edges): Likewise and exclude values brought
4143 only by self-recursive edges.
4144 (create_specialized_node): Redirect only clones of self-calling edges.
4145 (+self_recursive_pass_through_p): New function.
4146 (find_more_scalar_values_for_callers_subset): Use it.
4147 (find_aggregate_values_for_callers_subset): Likewise.
4148 (known_aggs_to_agg_replacement_list): Removed.
4149 (decide_whether_version_node): Re-calculate known constants for all
4150 remaining context clones.
4152 2018-04-11 Richard Biener <rguenther@suse.de>
4155 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
4156 from early DWARF output.
4157 (dwarf2out_early_finish): Output line info unconditionally into
4158 early DWARF and add reference to it.
4160 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4163 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
4164 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
4165 other than V2DFmode using iptr mode attribute.
4166 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
4168 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4170 PR rtl-optimization/84659
4171 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
4173 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4176 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
4178 (output_loc_list): Pass address of a dummy size variable even in the
4179 locview handling loop.
4180 (index_location_lists): Add comment on why skip_loc_list_entry can't
4183 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
4186 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
4189 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
4192 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4193 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
4194 and -mstring-compare-inline-limit.
4196 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4199 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
4200 for stack clash protection in a register whenever we need it to be in
4203 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4205 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
4206 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
4208 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4211 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
4213 (mlong-double-): Ditto.
4214 * config/rs6000/sysv4.opt (msdata=): Ditto.
4215 (mtls-size=): Ditto.
4217 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4219 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4220 erroneous entries for
4221 "vector int vec_ldl (int, long int *)", and
4222 "vector unsigned int vec_ldl (int, unsigned long int *)".
4223 Add comments and entries for
4224 "vector bool char vec_ldl (int, bool char *)",
4225 "vector bool short vec_ldl (int, bool short *)",
4226 "vector bool int vec_ldl (int, bool int *)",
4227 "vector bool long long vec_ldl (int, bool long long *)",
4228 "vector pixel vec_ldl (int, pixel *)",
4229 "vector long long vec_ldl (int, long long *)",
4230 "vector unsigned long long vec_ldl (int, unsigned long long *)".
4231 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
4232 type tree bool_long_long_type_node and correct definition of
4233 bool_V2DI_type_node to make reference to this new type tree.
4234 (rs6000_mangle_type): Replace erroneous reference to
4235 bool_long_type_node with bool_long_long_type_node.
4236 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
4237 comments to emphasize sign distinctions for char and int types and
4238 replace RS6000_BTI_bool_long constant with
4239 RS6000_BTI_bool_long_long constant. Also add comment to restrict
4240 use of RS6000_BTI_pixel.
4241 (bool_long_type_node): Remove this macro definition.
4242 (bool_long_long_type_node): New macro definition
4244 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4246 PR rtl-optimization/85300
4247 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
4248 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
4249 simplify_unary_operation fails.
4251 2018-04-10 Martin Liska <mliska@suse.cz>
4253 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
4254 cgraph_edge and ipa_ref.
4256 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4260 * config/i386/sse.md
4261 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
4262 computation of the VEC_MERGE selector from mask.
4263 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
4264 Fix decoding of the VEC_MERGE selector into mask.
4266 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4268 PR tree-optimization/85286
4269 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
4271 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4273 * final.c (final_1): Set insn_last_address as well as
4274 insn_current_address.
4276 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4279 * explow.c (emit_stack_probe): Call validize_mem on memory location
4280 before passing it to gen_probe_stack. Create address operand and
4281 legitimize it for the probe_stack_address case.
4283 2018-04-09 Jan Hubicka <jh@suse.cz>
4286 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
4287 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
4288 * tree.c (free_lang_data_in_type): Fix handling of binfos;
4290 (free_lang_data): Rebuild type inheritance graph.
4292 2018-04-09 Martin Sebor <msebor@redhat.com>
4294 * invoke.texi (-finline-small-functions): Mention other optimization
4296 (-findirect-inlining, -fpartial-inlining): Same.
4297 (-finline-functions-called-once): Same.
4298 (-freorder-blocks-and-partition): Same.
4300 2018-04-09 Jan Hubicka <jh@suse.cz>
4303 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
4304 jumps; choose last target that matches the criteria (i.e.
4305 no partition changes for non-crossing jumps).
4306 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
4307 support for redirecting crossing jumps to non-crossing.
4309 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
4311 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
4312 also for naked functions.
4314 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
4316 * config/arc/arc.md (add_shift): New pattern.
4317 (add_shift2): Likewise.
4318 (sub_shift): Likewise.
4319 (sub_shift_cmp0_noout): Likewise.
4320 (compare_si_ashiftsi): Likewise.
4321 (xbfu_cmp0_noout): New combine pattern.
4322 (xbfu_cmp0"): Likewise.
4323 (movsi_set_cc_insn): Place the predicable variant first.
4324 (commutative_binary_cmp0_noout): Remove clobber.
4325 (commutative_binary_cmp0): New pattern.
4326 (noncommutative_binary_cmp0): Likewise.
4327 (noncommutative_binary_cmp0_noout): Likewise.
4328 (noncommutative_binary_comparison_result_used): Removed.
4329 (rsub_cmp0): New pattern.
4330 (rsub_cmp0_noout): Likewise.
4331 (extzvsi): Changed, keep only meaningful variants.
4332 (SQH, SEZ): New iterators.
4333 (SQH_postfix): New mode attribute.
4334 (SEZ_prefix): New code attribute.
4335 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
4336 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
4337 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
4339 (noncommutative_operator): Check the availability of barrel
4342 2018-04-09 Richard Biener <rguenther@suse.de>
4344 PR tree-optimization/85284
4345 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
4346 Only use the niter constraining form of simple_iv when the exit
4349 2018-04-09 Tom de Vries <tom@codesourcery.com>
4352 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4353 (define_expand "*memory_barrier"): New define_expand.
4354 (define_insn "memory_barrier"): New insn.
4356 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4358 PR rtl-optimization/80463
4359 PR rtl-optimization/83972
4360 PR rtl-optimization/83480
4362 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
4363 correct producer for the insn.
4364 (tidy_control_flow): Fixup seqnos in case of debug insns.
4366 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4368 PR rtl-optimization/83913
4370 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
4371 different sched-times when merging exprs.
4373 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4375 PR rtl-optimization/83962
4377 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
4378 tidy_fallthru_edge and tidy_control_flow.
4380 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4382 PR rtl-optimization/83530
4384 * sel-sched.c (force_next_insn): New global variable.
4385 (remove_insn_for_debug): When force_next_insn is true, also leave only
4386 next insn in the ready list.
4387 (sel_sched_region): When the region wasn't scheduled, make another pass
4388 over it with force_next_insn set to 1.
4390 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4392 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
4394 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4395 for interrupt control.
4396 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4397 functions for interrupt control.
4398 * config/nds32/nds32-intrinsic.md: Likewise.
4399 * config/nds32/nds32_intrinsic.h: Likewise.
4400 * config/nds32/nds32.h (nds32_builtins): Likewise.
4402 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4404 * config/nds32/nds32.c (nds32_init_machine_status,
4405 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
4406 strict_aligned_p field.
4407 (nds32_expand_to_rtl_hook): New function.
4408 (TARGET_EXPAND_TO_RTL_HOOK): Define.
4409 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
4411 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4412 Chung-Ju Wu <jasonwucj@gmail.com>
4414 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
4415 * config/nds32/nds32-n7.md: New file.
4416 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
4417 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
4419 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
4420 * config/nds32/nds32.md (pipeline_model): Add n7.
4421 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
4422 * config/nds32/pipelines.md: Include n7 settings.
4424 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4425 Chung-Ju Wu <jasonwucj@gmail.com>
4427 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
4428 * config/nds32/nds32-e8.md: New file.
4429 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
4430 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
4432 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
4433 * config/nds32/nds32.md (pipeline_model): Add e8.
4434 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
4435 * config/nds32/pipelines.md: Include e8 settings.
4437 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4438 Chung-Ju Wu <jasonwucj@gmail.com>
4440 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
4441 * config/nds32/nds32-n8.md: New file.
4442 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
4443 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
4445 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
4446 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
4447 * config/nds32/nds32.md (pipeline_model): Add n8.
4448 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
4449 * config/nds32/pipelines.md: Include n8 settings.
4451 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4452 Chung-Ju Wu <jasonwucj@gmail.com>
4454 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
4455 * config/nds32/nds32-n9-2r1w.md: New file.
4456 * config/nds32/nds32-n9-3r2w.md: New file.
4457 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
4458 nds32_register_ports): New or modify for cpu n9.
4459 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
4461 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
4462 * config/nds32/nds32-utils.c: New file.
4463 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
4464 TARGET_MUL_SLOW): Define.
4465 * config/nds32/nds32.md (pipeline_model): New attribute.
4466 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
4467 New options that support cpu n9.
4468 * config/nds32/pipelines.md: Include n9 settings.
4469 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
4471 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4473 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
4474 information if necessary.
4475 (output_cond_branch_compare_zero): Likewise.
4476 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
4477 (nds32_target_alignment): Refine for alignment.
4478 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
4479 (FUNCTION_BOUNDARY): Modify.
4480 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
4482 * config/nds32/nds32.opt (malways-align, malign-functions): New.
4484 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4486 * config/nds32/constants.md (unspec_volatile_element): Add values for
4487 TLB operation and data prefetch.
4488 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4489 functions for TLB operation and data prefetch.
4490 * config/nds32/nds32-intrinsic.md: Likewise.
4491 * config/nds32/nds32_intrinsic.h: Likewise.
4492 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
4493 (nds32_print_operand): Likewise.
4494 * config/nds32/nds32.h (nds32_builtins): Likewise.
4496 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
4497 Andrew Pinski <pinsika@gcc.gnu.org>
4500 * match.pd: Use constant_boolean_node of correct type instead of
4501 boolean_true_node or boolean_false_node for simplifying
4502 pointer comparisons to zero.
4504 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4506 PR tree-optimization/80021
4507 * tree.c (verify_type_variant): Make error call in verify_variant_match
4508 translatable and remove final full stop.
4510 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4512 * config/nds32/constants.md (unspec_volatile_element): Add
4513 UNSPEC_VOLATILE_EH_RETURN.
4514 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
4515 nds32_output_stack_pop): Support dwarf exception handling process.
4516 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
4517 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
4518 exception handling process.
4519 (nds32_compute_stack_frame): Likewise.
4520 (nds32_return_addr_rtx): Likewise.
4521 (nds32_initial_elimination_offset): Likewise.
4522 (nds32_expand_prologue): Likewise.
4523 (nds32_expand_epilogue): Likewise.
4524 (nds32_dynamic_chain_address): New function.
4525 * config/nds32/nds32.h (machine_function): Add fields for dwarf
4527 (DYNAMIC_CHAIN_ADDRESS): Define.
4528 (EH_RETURN_DATA_REGNO): Define.
4529 (EH_RETURN_STACKADJ_RTX): Define.
4530 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
4531 patterns for dwarf exception handling.
4533 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4535 * config/nds32/nds32.h: Clean up obsolete macros.
4537 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4539 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4540 Add enum values for particular instructions.
4541 * config/nds32/nds32-intrinsic.c: Implementation of expanding
4542 particular intrinsic functions.
4543 * config/nds32/nds32-intrinsic.md: Likewise.
4544 * config/nds32/nds32_intrinsic.h: Likewise.
4545 * config/nds32/nds32.h (nds32_builtins): Likewise.
4546 * config/nds32/nds32.md (type): Add pbsad and pbsada.
4547 (btst, ave): New patterns for particular instructions.
4549 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4551 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4552 Add enum values for atomic load/store and memory sync.
4553 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
4555 * config/nds32/nds32-intrinsic.md: Likewise.
4556 * config/nds32/nds32_intrinsic.h: Likewise.
4557 * config/nds32/nds32.h (nds32_builtins): Likewise.
4559 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4561 PR tree-optimization/85257
4562 * fold-const.c (native_encode_vector): If not all elts could fit
4563 and off is -1, return 0 rather than offset.
4564 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
4565 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
4566 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
4567 adjust buffer in native_interpret_expr call.
4569 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4571 * config/nds32/constants.md (unspec_volatile_element): Add cache
4572 control enum values.
4573 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
4574 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
4575 * config/nds32/nds32.c (nds32_cctl_names): New.
4576 (nds32_print_operand): Handle cache control register names.
4577 * config/nds32/nds32.h (nds32_builtins): New enum values.
4578 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
4580 * config/nds32/nds32.md (type): Add mmu.
4581 * config/nds32/pipelines.md (simple_insn): Add mmu.
4583 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4585 * config/nds32/nds32.md (type): Remove call.
4586 * config/nds32/pipelines.md (simple_insn): Likewise.
4588 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4590 * config/nds32/constants.md (unspec_volatile_element): Add
4591 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
4592 UNSPEC_VOLATILE_FMFCFG.
4593 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
4594 description for fmfcfg and fmfcsr.
4595 (bdesc_1arg): Add fmtcsr.
4596 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
4597 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
4598 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
4599 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
4600 unspec_fmfcfg): New patterns.
4601 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
4602 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
4603 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
4604 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
4605 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
4606 __nds32__fmfcfg): Define.
4608 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4610 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
4611 intrinsic register names.
4612 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
4613 intrinsic register enum values and macros.
4615 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4617 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
4618 for load/store addressing form.
4619 (nds32_print_operand_address): Likewise.
4621 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
4624 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
4625 based on LABEL_REF. Remove useless assertion.
4626 (pic_address_needs_scratch): Fix formatting.
4627 (sparc_legitimize_pic_address): Minor tweaks.
4628 (sparc_delegitimize_address): Adjust assertion accordingly.
4629 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
4630 into symbolic_operand.
4631 (movsi_high_pic_label_ref): Likewise.
4632 (movsi_lo_sum_pic_label_ref): Likewise.
4633 (movdi_pic_label_ref): Likewise.
4634 (movdi_high_pic_label_ref): Likewise.
4635 (movdi_lo_sum_pic_label_ref): Likewise.
4637 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
4639 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
4640 custom LIB_SPEC setup.
4642 2018-04-06 Ruslan Bukin <br@bsdpad.com>
4643 Kito Cheng <kito.cheng@gmail.com>
4645 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
4646 * config/riscv/freebsd.h: New.
4648 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4650 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
4651 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
4654 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4655 Kito Cheng <kito.cheng@gmail.com>
4657 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
4658 nds32_output_call, nds32_symbol_binds_local_p): New functions.
4659 * config/nds32/nds32-protos.h (nds32_output_call,
4660 nds32_output_return): Declare.
4661 * config/nds32/nds32.md: Refine all the call and return patterns.
4663 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4666 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
4667 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
4669 PR rtl-optimization/84872
4670 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
4671 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
4674 2018-04-06 Tamar Christina <tamar.christina@arm.com>
4676 * expr.c (copy_blkmode_to_reg): Revert 254862.
4677 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
4679 2018-04-06 Richard Biener <rguenther@suse.de>
4682 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
4683 after seeing a component reference with an adjacent field. Treat
4684 refs to arrays at struct end of external decls similar to
4685 refs to unconstrained commons.
4687 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4690 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
4691 look through SAVE_EXPRs with non-side-effects argument. Adjust
4693 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
4696 2018-04-06 Richard Biener <rguenther@suse.de>
4699 * alias.c (find_base_term): New wrapper around find_base_term
4700 unwinding CSELIB_VAL_PTR changes.
4701 (find_base_term): Do not restore CSELIB_VAL_PTR during the
4704 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4706 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
4708 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
4709 constant definitions.
4710 ("nop"): lr 0,0 -> nopr r0
4711 ("nop_lr0", "nop_lr1"): New insn definitions.
4713 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4715 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
4716 NDS32_V3PUSH_AVAILABLE_P macro.
4718 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
4719 Chung-Ju Wu <jasonwucj@gmail.com>
4721 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
4722 (nds32*-*-*): Add float and fpu_config into supported_defaults.
4723 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
4724 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
4725 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
4726 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
4727 * config/nds32/constraints.md: New constraints and checking for hard
4728 float configuration.
4729 * config/nds32/iterators.md: New mode iterator and attribute for hard
4730 float configuration.
4731 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
4733 * config/nds32/nds32-fpu.md: New file.
4734 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
4735 deal with hard float code generation.
4736 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
4738 (abi_type, float_reg_number): New enum type.
4739 * config/nds32/nds32-predicates.c: New predicates for hard float.
4740 * config/nds32/nds32-protos.h: Declare functions for hard float.
4741 * config/nds32/nds32.c: Implementation for hard float configuration.
4742 * config/nds32/nds32.h: Definitions for hard float configuration.
4743 * config/nds32/nds32.md: Include hard float machine description and
4744 modify patterns for hard float configuration.
4745 * config/nds32/nds32.opt: New options for hard float configuration.
4746 * config/nds32/predicates.md: New predicates for hard float
4749 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4751 * common/config/nds32/nds32-common.c
4752 (nds32_option_optimization_table): Enable -mreleax-hint by default.
4754 2018-04-05 Jakub Jelinek <jakub@redhat.com>
4757 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
4758 CONSTRUCTOR_ELT (ctor, ...)->value.
4760 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
4763 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
4765 2018-04-05 Tom de Vries <tom@codesourcery.com>
4768 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
4771 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4772 Kito Cheng <kito.cheng@gmail.com>
4774 * config/nds32/constraints.md (U33): Fine-tune checking condition.
4775 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
4776 * config/nds32/nds32.h (nds32_16bit_address_type): Add
4777 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
4779 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4780 Kito Cheng <kito.cheng@gmail.com>
4782 * config/nds32/constraints.md (Ufe): New memory constraint.
4783 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
4784 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
4785 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
4787 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
4788 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
4790 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4792 * config/nds32/nds32.md: Use optimize_size in the condition for
4793 alu-shift instructions.
4795 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4797 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
4799 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4801 * config/nds32/nds32.md (negsi2): Refine pattern.
4803 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4804 Chung-Ju Wu <jasonwucj@gmail.com>
4806 * config/nds32/iterators.md (shift_rotate): New code iterator.
4807 (shift): New code attribute.
4808 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
4809 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
4810 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
4811 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
4812 bit-wise operations.
4813 (andsi3, *andsi3): Ditto.
4814 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
4815 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
4816 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
4817 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
4818 nds32_ior_operand, nds32_xor_operand): New predicates.
4820 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4822 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
4823 (addsi3, subsi3): ... this.
4825 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4827 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
4829 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4831 * config/nds32/nds32.md: Adjust indention.
4833 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4835 * config/nds32/nds32.md (feature): New attribute.
4837 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4839 * config/nds32/nds32.md (subtype): New attribute.
4841 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
4844 * config/arm/arm-builtins.c (arm_expand_builtin): Change
4845 expansion to perform a bitwise AND of the argument followed by a
4846 boolean negation of the result.
4848 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
4850 PR rtl-optimization/84878
4851 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
4852 the basic block. Assert the use reference is not artificial and that
4853 it has an associated insn.
4855 2018-04-04 Michael Matz <matz@suse.de>
4857 * builtins.c (compute_objsize): Pass correct operand
4858 to array_at_struct_end_p.
4860 2018-04-04 Richard Biener <rguenther@suse.de>
4863 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
4864 from contexts for DINFO_LEVEL_TERSE and below.
4866 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4868 * config/nds32/nds32-doubleword.md (move_<mode>): Require
4869 resiter_operand condition.
4870 * config/nds32/nds32.md (*move<mode>): Ditto.
4872 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4873 Monk Chiang <sh.chiang04@gmail.com>
4875 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
4877 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4879 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
4881 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4882 Kito Cheng <kito.cheng@gmail.com>
4884 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
4885 nds32_cond_code_str, output_cond_branch,
4886 output_cond_branch_compare_zero, nds32_expand_cbranch,
4887 nds32_expand_cstore, nds32_expand_movcc,
4888 nds32_output_cbranchsi4_equality_zero,
4889 nds32_output_cbranchsi4_equality_reg,
4890 nds32_output_cbranchsi4_equality_reg_or_const_int,
4891 nds32_output_cbranchsi4_greater_less_zero: New functions.
4892 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
4893 nds32_expand_cstore, nds32_expand_movcc,
4894 nds32_output_cbranchsi4_equality_zero,
4895 nds32_output_cbranchsi4_equality_reg,
4896 nds32_output_cbranchsi4_equality_reg_or_const_int,
4897 nds32_output_cbranchsi4_greater_less_zero): Declare.
4898 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
4899 nds32_rimm11s_operand): New predicates.
4900 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
4901 * config/nds32/nds32.md: Rewrite all the branch and conditional move
4904 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4906 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
4907 * config/nds32/nds32.md: Ditto.
4908 * config/nds32/pipelines.md: Ditto.
4910 2018-04-04 Richard Biener <rguenther@suse.de>
4912 PR tree-optimization/85168
4913 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
4914 propagating abnormals.
4916 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4918 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
4920 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4921 Kito Cheng <kito.cheng@gmail.com>
4923 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
4924 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
4925 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
4926 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
4927 * config/nds32/nds32.md (sibcall_internal): New.
4928 (sibcall_register): Remove.
4929 (sibcall_immediate): Remove.
4930 (sibcall_value_internal): New.
4931 (sibcall_value_register): Remove.
4932 (sibcall_value_immediate): Remove.
4933 * config/nds32/predicates.md (nds32_general_register_operand): New.
4934 (nds32_call_address_operand): New.
4936 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4938 PR rtl-optimization/85167
4939 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
4940 bb_defs if *split_p, instead preinitialize it to NULL.
4942 PR tree-optimization/85156
4943 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
4944 evaluating the argument multiple times.
4946 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
4948 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
4950 (_mm_cvtpd_ps): Likewise.
4951 (_mm_cvttpd_epi32): Likewise.
4952 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
4953 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
4954 vector, pixel, and bool following altivec.h include.
4956 2018-04-03 Martin Sebor <msebor@redhat.com>
4958 * doc/extend.texi (Common Function Attributes): Clarify.
4959 (const attribute): Likewise.
4960 (pure attribute): Likewise.
4962 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4965 * config/i386/i386.c (ix86_expand_vector_set): Use
4966 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
4968 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
4970 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
4971 instructions when changing rounding bits to preserve precision bits
4972 in the x87 control word.
4974 2018-04-03 Martin Liska <mliska@suse.cz>
4976 PR tree-optimization/82491
4977 * rtl.h (strip_offset_and_add): Replace += suboffset with
4978 poly_uint64 () + suboffset.
4980 2018-03-29 Martin Liska <mliska@suse.cz>
4981 Martin Jambor <mjambor@suse.cz>
4984 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
4985 param_type is not an integral or pointer type.
4987 2018-04-03 Richard Biener <rguenther@suse.de>
4989 * sese.h (recompute_all_dominators): Remove.
4991 2018-04-02 Martin Sebor <msebor@redhat.com>
4993 * doc/invoke.texi (-Wrestrict): Fix typos.
4995 2018-04-02 Jim Wilson <jimw@sifive.com>
4997 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
4998 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
4999 (<optab>di3, <optab>si3_extend): Likewise.
5000 (<optab>si3_mask, <optab>si3_mask_1): New.
5001 (<optab>di3_mask, <optab>di3_mask_1): New.
5002 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
5003 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
5004 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
5006 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
5008 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
5011 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
5013 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
5014 (nds32_canonicalize_comparison): New function.
5016 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5017 Kito Cheng <kito.cheng@gmail.com>
5018 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5020 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
5021 * config/nds32/constants.md (unspec_volatile_element): Add
5022 UNSPEC_VOLATILE_RELAX_GROUP.
5023 * config/nds32/nds32-relax-opt.c: New file.
5024 * config/nds32/nds32-predicates.c
5025 (nds32_symbol_load_store_p): New function.
5026 * config/nds32/nds32-protos.h
5027 (nds32_symbol_load_store_p): Declare function.
5028 (make_pass_nds32_relax_opt): Declare new rtl pass function.
5029 * config/nds32/nds32.c
5030 (nds32_register_pass): New function to register pass.
5031 (nds32_register_passes): New function to register passes.
5032 * config/nds32/nds32.md (relax_group): New pattern.
5033 * config/nds32/nds32.opt (mrelax-hint): New option.
5034 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
5036 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
5038 * config/nds32/t-nds32: Modify files dependency.
5040 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5042 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
5043 (PROFILE_HOOK): Define its implementation.
5045 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5047 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
5048 type and 32-bit size.
5050 2018-04-01 Jakub Jelinek <jakub@redhat.com>
5053 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
5054 (V_128_256): New mode iterator.
5055 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
5056 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
5057 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
5059 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
5060 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
5062 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
5065 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
5066 NaN inputs correctly.
5068 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
5071 * config/rs6000/vsx.md (??r): New mode attribute.
5072 (*vsx_mov<mode>_64bit): Use it.
5073 (*vsx_mov<mode>_32bit): Likewise.
5075 2018-03-30 Martin Sebor <msebor@redhat.com>
5077 PR tree-optimization/84818
5078 * builtins.c (check_access): Use warning_n.
5080 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5083 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
5085 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
5088 2018-03-30 Julia Koval <julia.koval@intel.com>
5091 * x86-tune.def (movx, partial_reg_dependency): Enable for
5094 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
5097 * lra-constraints.c (process_alt_operands): Move setting
5098 this_alternative_matches below.
5100 2018-03-29 Martin Liska <mliska@suse.cz>
5103 * doc/invoke.texi: Document how LTO works with debug info.
5104 Describe auto-load support of binutils. Mention 'x86-64'
5105 as valid option value of -march option.
5107 2018-03-29 Jakub Jelinek <jakub@redhat.com>
5109 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
5112 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
5113 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
5114 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
5117 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5120 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
5121 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
5122 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
5123 for RS6000_BTM_POWERPC64.
5124 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
5125 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
5126 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
5131 2018-03-28 Carl Love <cel@us.ibm.com>
5134 2017-09-27 Carl Love <cel@us.ibm.com>
5136 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
5137 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
5138 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
5141 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5143 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
5144 instead of __vector bool.
5145 (_mm_max_pu8): Likewise.
5146 (_mm_min_pi16): Likewise.
5148 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5151 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
5152 (DIVWEUO): Likewise.
5154 (DIVDEUO): Likewise.
5155 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
5156 DIVWEUO and DIVDEUO.
5157 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
5158 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
5159 (div_extend): Likewise.
5160 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
5162 (__builtin_divweuo): Likewise.
5163 (__builtin_divdeo): Likewise.
5164 (__builtin_divdeuo): Likewise.
5166 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5169 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
5170 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
5172 PR tree-optimization/82004
5173 * gimple-match-head.c (optimize_pow_to_exp): New function.
5174 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
5175 Don't fold to exp if optimize_pow_to_exp is false.
5177 2018-03-28 Martin Liska <mliska@suse.cz>
5180 * calls.c (initialize_argument_information): Fix trailing space.
5181 * common.opt: Fix typo and provide better explanation for
5182 -fsanitize-coverage option.
5183 * config/i386/i386.opt: Fix typo.
5185 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5186 Martin Liska <mliska@suse.cz>
5189 * gimplify.c (asan_poison_variable): Don't do the check for
5190 gimplify_omp_ctxp here.
5191 (gimplify_decl_expr): Do it here.
5192 (gimplify_target_expr): Likewise.
5194 2018-03-28 Martin Liska <mliska@suse.cz>
5197 * config/i386/i386.c (ix86_function_arg_advance): Do not call
5198 chkp_type_bounds_count if MPX is not enabled.
5200 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
5202 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
5204 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
5207 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
5208 function to create the function decl for complex long double
5209 multiply and divide for -mabi=ieeelongdouble.
5210 (init_float128_ieee): Call it.
5212 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
5215 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
5216 -fcf-protection=branch -mibt.
5217 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
5219 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5222 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
5224 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
5227 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
5228 extern array declarations.
5230 2018-03-27 Richard Biener <rguenther@suse.de>
5233 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
5234 explicit single_use checks.
5236 2018-03-27 Richard Biener <rguenther@suse.de>
5238 PR tree-optimization/85082
5239 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
5242 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5244 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
5245 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
5246 Turn on fasynchronous-unwind-tables and funwind-tables.
5248 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
5251 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
5252 (*bmi_blsr_<mode>_ccz): Ditto.
5254 2018-03-26 Tom de Vries <tom@codesourcery.com>
5256 PR tree-optimization/85063
5257 * omp-general.c (offloading_function_p): New function. Factor out
5259 * omp-offload.c (pass_omp_target_link::gate): ... here.
5260 * omp-general.h (offloading_function_p): Declare.
5261 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
5262 with attribute omp declare target for offloading functions.
5264 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
5266 PR tree-optimization/84005
5267 * tree-data-ref.h (get_base_for_alignment): Declare.
5268 * tree-data-ref.c (get_base_for_alignment_1): New function.
5269 (get_base_for_alignment): Likewise.
5270 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
5271 get_base_for_alignment to find a suitable base object, instead
5272 of always using drb->base_address.
5274 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5277 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
5278 known size by default.
5280 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
5283 * lra-constraints.c (process_alt_operands): Don't match BLKmode
5284 and non BLKmode operands.
5286 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5289 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
5290 Clean up attributes.
5292 2018-03-23 Richard Biener <rguenther@suse.de>
5295 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
5296 we are going to emit early debug for LTO.
5298 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5301 * function.c (match_asm_constraints_1): Don't optimize if input
5302 doesn't satisfy general_operand predicate for output's mode.
5305 * alias.c (write_dependence_p): Don't require for x_canonicalized
5306 non-VOIDmode if x has VOIDmode.
5309 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
5310 just don't try to optimize it rather than assert it never happens.
5312 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5314 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
5315 macro expansions for definition of ST_INTERNAL_<mode> and
5316 LD_INTERNAL_<mode> builtins.
5317 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
5319 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
5321 (altivec_expand_st_builtin): Likewise.
5322 (altivec_expand_builtin): Remove calls to deleted functions.
5323 (rs6000_address_for_altivec): Delete this function.
5324 * config/rs6000/vector.md: Remove expands for
5325 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
5327 2018-03-22 Sudakshina Das <sudi.das@arm.com>
5330 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
5331 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
5332 re-computing once computed.
5333 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
5334 (arm_init_machine_status): Initialize
5335 machine->static_chain_stack_bytes.
5337 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5340 * doc/extend.texi: Add four new prototypes for vec_ld.
5341 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
5342 definitions for more logical presentation.
5343 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
5344 entries for V1TI variants of __builtin_altivec_ld builtin.
5345 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
5346 handling of V1TI variant of LVX icode pattern.
5347 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
5348 (rs6000_gimple_fold_builtin): Likewise.
5349 (altivec_init_builtins): Add code to define
5350 __builtin_altivec_lvx_v1ti function.
5352 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5355 * function.c (match_asm_constraints_1): Don't do the optimization
5356 if input isn't a REG, SUBREG, MEM or constant.
5358 2018-03-22 Tom de Vries <tom@codesourcery.com>
5360 PR tree-optimization/84956
5361 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
5362 bb_has_abnormal_pred.
5364 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5367 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
5368 DECL_INITIAL (decl) to decl at the end.
5369 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
5372 2018-03-21 Joseph Myers <joseph@codesourcery.com>
5374 * doc/extend.texi (__builtin_tgmath): Document when complex
5375 integer types are treated as _Complex _Float64.
5377 2018-03-21 Tom de Vries <tom@codesourcery.com>
5379 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
5381 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5383 PR tree-optimization/84960
5384 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
5385 if it is ENTRY block, move them into single succ of ENTRY in that case.
5387 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
5389 PR tree-optimization/84811
5390 * poly-int.h (poly_span_traits): Remove the T3 parameter and
5391 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
5392 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
5393 (known_subrange_p): Update accordingly. Cast each value involved
5394 in the size comparison, rather than casting the result of the
5397 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5399 PR tree-optimization/84982
5400 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
5401 by flipping the least significant bit rather than all bits from
5402 bitpos to bitpos + bitsize - 1.
5404 2018-03-21 Nathan Sidwell <nathan@acm.org>
5406 * doc/extend.texi (Deprecated Features): Remove mention of
5407 long-deleted deprecations.
5409 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5412 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
5413 * configure: Regenerate.
5415 2018-03-21 Tom de Vries <tom@codesourcery.com>
5417 PR tree-optimization/83126
5418 * tree-parloops.c (num_phis): New function.
5419 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
5421 2018-03-21 Nathan Sidwell <nathan@acm.org>
5423 * doc/extend.texi (Deprecated Features): Update deprecated flags,
5424 mention anon-struct/union members and trailing attributes.
5426 2018-03-21 Bin Cheng <bin.cheng@arm.com>
5428 PR tree-optimization/84969
5429 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
5430 builtin memset partitions if they set different rhs values.
5432 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5434 PR rtl-optimization/84989
5435 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
5436 VEC_DUPLICATE with scalar result mode.
5438 2018-03-21 Martin Liska <mliska@suse.cz>
5441 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
5442 not intended return statement.
5444 2018-03-21 Martin Liska <mliska@suse.cz>
5447 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
5448 (chkp_find_bound_slots_1): Limit number of iterations.
5450 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
5453 * Minor grammar fixes for x86 options.
5455 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5458 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
5459 holding REG_CFA_RESTORE notes, instead turn them into a USE.
5461 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
5464 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
5465 (altivec_lvx_<mode>_1op): Likewise.
5466 (altivec_stvx_<mode>_2op): Likewise.
5467 (altivec_stvx_<mode>_1op): Likewise.
5468 (altivec_lvx_<VM2:mode>): New define_expand.
5469 (altivec_stvx_<VM2:mode>): Likewise.
5470 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
5471 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5472 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
5473 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5474 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
5475 (rs6000_gen_lvx): Likewise.
5476 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
5477 (altivec_expand_stv_builtin): Likewise.
5478 (altivec_expand_builtin): Likewise.
5479 * config/rs6000/vector.md: Likewise.
5481 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5484 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
5487 2018-03-20 Richard Biener <rguenther@suse.de>
5490 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
5491 sign-conversions as zero, fall back to standard scalar_stmt
5494 2018-03-20 Martin Liska <mliska@suse.cz>
5497 * predict.c (rebuild_frequencies): Handle case when we have
5498 PROFILE_ABSENT, but flag_guess_branch_prob is false.
5500 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5503 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
5504 flag_section_anchors.
5505 * varasm.c (use_blocks_for_decl_p): Remove hack for
5506 dw2_force_const_mem.
5509 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
5511 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
5512 be created, use lowpart_subreg of operands[0] rather than operands[0]
5514 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
5515 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
5516 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
5517 and n constraint instead of aarch64_shift_imm_di and Usd.
5518 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
5519 (*aarch64_<optab>_reg_minus<mode>3): ... this.
5521 2018-03-20 Sudakshina Das <sudi.das@arm.com>
5524 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
5525 to favor GPR over NEON registers.
5526 (<shift>di3_neon): Likewise.
5528 2018-03-20 Tom de Vries <tom@codesourcery.com>
5531 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
5532 (nvptx_process_pars): Emit bar.sync asap and alap.
5534 2018-03-20 Tom de Vries <tom@codesourcery.com>
5537 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
5538 seen_label if seen_label is already set.
5540 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5543 * config/i386/i386.c (fold_builtin_cpu): For features above 31
5544 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
5545 Use 1U instead of 1. Formatting fixes.
5548 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
5549 instead of TREE_TYPE (s1) for the return value.
5551 2018-03-19 Jakub Jelinek <jakub@redhat.com>
5553 PR tree-optimization/84946
5554 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
5555 bitsize + bitsize in poly_uint64 rather than poly_int64.
5558 * dwarf2asm.c: Include fold-const.c.
5559 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
5560 of decl rather than decl itself.
5562 PR rtl-optimization/84643
5563 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
5565 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
5568 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
5569 calling assemble_variable.
5571 2018-03-19 Sudakshina Das <sudi.das@arm.com>
5574 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
5575 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
5577 2018-03-19 Jim Wilson <jimw@sifive.com>
5580 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
5581 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
5582 (riscv_first_stack_step): Likewise.
5583 (riscv_option_override): Use STACK_BOUNDARY instead of
5585 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
5587 (BIGGEST_ALIGNMENT): Set to 128.
5588 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
5589 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
5592 2018-03-19 Richard Biener <rguenther@suse.de>
5594 PR tree-optimization/84933
5595 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
5596 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
5598 2018-03-19 Richard Biener <rguenther@suse.de>
5600 PR tree-optimization/84859
5601 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
5602 (cond_if_else_store_replacement): Perform sinking operation on
5603 single-store BBs regardless of MAX_STORES_TO_SINK setting.
5604 Generalize what a BB with a single eligible store is.
5606 2018-03-19 Richard Biener <rguenther@suse.de>
5608 PR tree-optimization/84929
5609 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
5610 chrec_is_positive against non-chrec arg.
5612 2018-03-19 Tamar Christina <tamar.christina@arm.com>
5615 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
5617 2018-03-18 Martin Liska <mliska@suse.cz>
5619 PR rtl-optimization/84635
5620 * regrename.c (build_def_use): Use matches_mode only when
5623 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
5625 PR tree-optimization/84913
5626 * tree-vect-loop.c (vectorizable_reduction): Don't try to
5627 vectorize chains of COND_EXPRs.
5629 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5631 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
5633 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5635 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
5637 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5639 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
5641 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
5642 Kito Cheng <kito.cheng@gmail.com>
5644 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
5645 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
5646 (nds32_adjust_reg_alloc_order): New function.
5647 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
5649 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
5651 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
5652 nds32_print_operand, nds32_print_operand_address): Use
5653 HOST_WIDE_INT_PRINT_DEC instead.
5655 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
5657 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
5659 2018-03-17 Jakub Jelinek <jakub@redhat.com>
5662 * config/i386/i386.c (initial_ix86_tune_features,
5663 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
5665 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
5666 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
5667 rather than 1u << ix86_tune. Formatting fix.
5668 (ix86_option_override_internal): Change ix86_arch_mask from
5669 unsigned int to unsigned HOST_WIDE_INT, initialize to
5670 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
5671 (ix86_function_specific_restore): Likewise.
5673 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5676 * postreload.c (reload_combine_recognize_pattern): Perform
5677 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
5678 truncate_int_for_mode the result for the destination's mode.
5681 * hsa-gen.c (mem_type_for_type): Fix comment typo.
5682 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
5684 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5687 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
5690 * lra-assigns.c (lra_split_hard_reg_for): Don't use
5691 regno_allocno_class_array and sorted_pseudos.
5692 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
5693 insns where regno is used.
5695 2018-03-16 Martin Liska <mliska@suse.cz>
5698 * multiple_target.c (create_dispatcher_calls): Redirect
5699 reference in the symbol table.
5701 2018-03-16 Martin Liska <mliska@suse.cz>
5704 * multiple_target.c (create_dispatcher_calls): Redirect also
5707 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5711 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
5712 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
5713 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
5715 2018-03-16 Julia Koval <julia.koval@intel.com>
5717 * doc/invoke.texi (Skylake Server): Add CLWB.
5718 Cannonlake): Remove CLWB.
5720 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5722 PR tree-optimization/84841
5723 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
5725 (FLOAT_ONE_CONST_TYPE): Define.
5726 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
5727 (sort_by_operand_rank): Put entries with higher constant_type last
5728 rather than first to match comments.
5730 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
5732 * config/nios2/nios2.md (movsi_internal): Fix thinko in
5735 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5738 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
5739 check and use address of target always.
5741 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
5744 * config/i386/i386.c (indirect_thunk_needed): Update comments.
5745 (indirect_thunk_bnd_needed): Likewise.
5746 (indirect_thunks_used): Likewise.
5747 (indirect_thunks_bnd_used): Likewise.
5748 (indirect_return_needed): New.
5749 (indirect_return_bnd_needed): Likewise.
5750 (output_indirect_thunk_function): Add a bool argument for
5752 (output_indirect_thunk_function): Don't generate alias for
5753 function return thunk.
5754 (ix86_code_end): Call output_indirect_thunk_function to generate
5755 function return thunks.
5756 (ix86_output_function_return): Set indirect_return_bnd_needed
5757 and indirect_return_needed instead of indirect_thunk_bnd_needed
5758 and indirect_thunk_needed.
5760 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
5762 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
5763 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
5764 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
5766 2018-03-15 David Malcolm <dmalcolm@redhat.com>
5767 Paul Hua <paul.hua.gm@gmail.com>
5770 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
5772 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
5774 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
5775 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
5778 2018-03-15 Tamar Christina <tamar.christina@arm.com>
5781 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
5782 instead of GET_MODE_SIZE when comparing Units.
5784 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
5787 * varasm.c (hash_section): Return an unchangeble hash value
5788 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
5789 Return !aarch64_can_use_per_function_literal_pools_p ().
5791 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5794 * optabs.c (emit_conditional_move): Pass address of cmode's copy
5795 rather than address of cmode as last argument to prepare_cmp_insn.
5797 2018-03-15 Julia Koval <julia.koval@intel.com>
5799 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
5800 F_AVX512VNNI, F_AVX512BITALG): New.
5802 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
5805 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
5806 insn for floating-point loads and stores.
5808 2018-03-14 Carl Love <cel@us.ibm.com>
5810 * config/rs6000/rs6000-c.c: Add macro definitions for
5811 ALTIVEC_BUILTIN_VEC_PERMXOR.
5812 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
5813 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
5814 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
5815 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
5817 * config/doc/extend.texi: Add prototypes for vec_permxor.
5819 2018-03-14 David Malcolm <dmalcolm@redhat.com>
5822 * diagnostic-show-locus.c (class layout_point): Convert m_line
5823 from int to linenum_type.
5824 (line_span::comparator): Use linenum "compare" function when
5825 comparing line numbers.
5826 (test_line_span): New function.
5827 (layout_range::contains_point): Convert param "row" from int to
5829 (layout_range::intersects_line_p): Likewise.
5830 (layout::will_show_line_p): Likewise.
5831 (layout::print_source_line): Likewise.
5832 (layout::should_print_annotation_line_p): Likewise.
5833 (layout::print_annotation_line): Likewise.
5834 (layout::print_leading_fixits): Likewise.
5835 (layout::annotation_line_showed_range_p): Likewise.
5836 (struct line_corrections): Likewise for field m_row.
5837 (line_corrections::line_corrections): Likewise for param "row".
5838 (layout::print_trailing_fixits): Likewise.
5839 (layout::get_state_at_point): Likewise.
5840 (layout::get_x_bound_for_row): Likewise.
5841 (layout::print_line): Likewise.
5842 (diagnostic_show_locus): Likewise for locals "last_line" and
5844 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
5845 * input.c (selftest::test_linenum_comparisons): New function.
5846 (selftest::input_c_tests): Call it.
5847 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
5848 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
5849 * selftest.h (ASSERT_GT): New macro.
5850 (ASSERT_GT_AT): New macro.
5851 (ASSERT_LT): New macro.
5852 (ASSERT_LT_AT): New macro.
5854 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
5856 PR rtl-optimization/84780
5857 * combine.c (distribute_links): Don't make a link based on pc_rtx.
5859 2018-03-14 Martin Liska <mliska@suse.cz>
5861 * tree.c (record_node_allocation_statistics): Use
5862 get_stats_node_kind.
5863 (get_stats_node_kind): New function extracted from
5864 record_node_allocation_statistics.
5865 (free_node): Use get_stats_node_kind.
5867 2018-03-14 Richard Biener <rguenther@suse.de>
5869 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
5870 that the value-set of ANTIC_IN doesn't grow.
5873 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5875 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5876 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5878 2018-03-14 Julia Koval <julia.koval@intel.com>
5880 * config.gcc (icelake-client, icelake-server): New.
5882 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
5883 (initial_ix86_arch_features): Ditto.
5884 (PTA_SKYLAKE): Add SGX.
5885 (PTA_ICELAKE): Remove.
5886 (PTA_ICELAKE_CLIENT): New.
5887 (PTA_ICELAKE_SERVER): New.
5888 (ix86_option_override_internal): Split up icelake on icelake client and
5890 (get_builtin_code_for_version): Ditto.
5891 (fold_builtin_cpu): Ditto.
5892 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
5893 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
5894 * config/i386/i386.h (processor_type): Ditto.
5895 * doc/invoke.texi: Ditto.
5897 2018-03-14 Jakub Jelinek <jakub@redhat.com>
5900 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
5901 INTEGER_CST offset, add it together with bitpos / 8 and
5902 sign extend based on POINTER_SIZE.
5906 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
5909 * config/i386/constraints.md (Yc): New register constraint.
5910 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
5911 Use Yc constraint for alternative 2 of operand 0. Remove
5912 preferred_for_speed attribute.
5914 2018-03-14 Richard Biener <rguenther@suse.de>
5916 PR tree-optimization/84830
5917 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
5918 with the old one to avoid oscillations.
5920 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
5923 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
5925 (assign_by_spills): Return a flag of reload assignment failure.
5926 Do not process the reload assignment failures. Do not spill other
5927 reload pseudos if they has the same reg class. Update n if
5929 (lra_assign): Add a return arg. Set up from the result of
5930 assign_by_spills call.
5931 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5932 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5933 usage_insns if it is not NULL.
5934 (spill_hard_reg_in_range): New function.
5935 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5936 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5937 function prototypes.
5938 (lra_assign): Change prototype.
5939 * lra.c (lra): Add code to deal with fails by splitting hard reg
5942 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
5944 * config/riscv/riscv.opt (mrelax): New option.
5945 * config/riscv/riscv.c (riscv_file_start): Emit ".option
5946 "norelax" when riscv_mrelax is disabled.
5947 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
5949 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5952 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
5953 reassociation for int modes.
5955 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5957 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
5958 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
5960 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
5961 * config/aarch64/aarch64-sve.md
5962 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
5963 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
5964 (*extend<mode><Vwide>2): Rename to...
5965 (aarch64_sve_extend<mode><Vwide>2): ...this.
5966 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
5967 renaming the old pattern to...
5968 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
5970 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
5971 define_expand, renaming the old pattern to...
5972 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
5973 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
5974 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
5975 account when deciding which SVE instruction the optab should use.
5976 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
5978 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5980 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
5981 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
5982 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
5983 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
5984 (tlsdesc_small_<mode>): Turn a define_expand and use
5985 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
5986 (tlsdesc_small_advsimd_<mode>): ...this.
5987 (tlsdesc_small_sve_<mode>): New pattern.
5989 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5991 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
5992 (UNSPEC_UMUL_HIGHPART): New constants.
5993 (MUL_HIGHPART): New int iteraor.
5994 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
5995 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
5997 (*<su>mul<mode>3_highpart): New define_insn.
5999 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6002 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6005 2018-03-13 Martin Liska <mliska@suse.cz>
6008 * (sem_item_optimizer::sem_item_optimizer): Initialize new
6010 (sem_item_optimizer::~sem_item_optimizer): Release it.
6011 (sem_item_optimizer::merge_classes): Register variable aliases.
6012 (sem_item_optimizer::fixup_pt_set): New function.
6013 (sem_item_optimizer::fixup_points_to_sets): Likewise.
6014 * ipa-icf.h: Declare new variables and functions.
6016 2018-03-13 Jakub Jelinek <jakub@redhat.com>
6019 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
6020 integer_pow2p@2 and test integer_pow2p in condition.
6021 (A < 0 ? C : 0): Similarly for @1.
6024 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
6025 characters starting at p contain '\0' character, don't look beyond
6029 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
6030 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
6033 * reg-stack.c (change_stack): Change update_end var from int to
6034 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
6035 also call set_block_for_insn on the newly added insns and rescan.
6038 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
6039 on the last operand.
6042 * tree.c (stabilize_reference_1): Return save_expr (e) for
6043 STATEMENT_LIST even if it doesn't have side-effects.
6045 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
6047 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
6049 2018-03-12 Renlin Li <renlin.li@arm.com>
6051 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
6052 aarch64_output_scalar_simd_mov_immediate.
6054 2018-03-12 Martin Sebor <msebor@redhat.com>
6056 PR tree-optimization/83456
6057 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
6058 for perfectly overlapping calls to memcpy.
6059 (gimple_fold_builtin_memory_chk): Same.
6060 (gimple_fold_builtin_strcpy): Handle no-warning.
6061 (gimple_fold_builtin_stxcpy_chk): Same.
6062 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
6064 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6066 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
6067 parameter. Use it for SFmode.
6068 (rs6000_function_arg_advance_1): Adjust.
6069 (rs6000_function_arg): Adjust.
6070 (rs6000_gimplify_va_arg): Pass false for that new parameter.
6072 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6074 PR rtl-optimization/84169
6075 PR rtl-optimization/84780
6076 * combine.c (can_combine_p): Check for a 2-insn combination whether
6077 the destination register is used between the two insns, too.
6079 2018-03-12 Richard Biener <rguenther@suse.de>
6081 PR tree-optimization/84803
6082 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
6083 for refs DR analysis didn't process.
6085 2018-03-12 Richard Biener <rguenther@suse.de>
6087 PR tree-optimization/84777
6088 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
6089 force-vectorize loops ignore whether we are optimizing for size.
6091 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
6093 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
6094 (TARGET_MD_ASM_ADJUST): Define.
6096 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
6097 Kito Cheng <kito.cheng@gmail.com>
6098 Chung-Ju Wu <jasonwucj@gmail.com>
6100 * config/nds32/nds32.c (nds32_compute_stack_frame,
6101 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
6102 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
6103 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
6104 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
6105 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
6106 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
6107 * config/nds32/nds32.md (prologue, epilogue): Use macro
6108 NDS32_V3PUSH_AVAILABLE_P to do checking.
6110 2018-03-11 Jakub Jelinek <jakub@redhat.com>
6113 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
6114 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
6115 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
6116 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
6117 addition of most attributes on !orig_type_die or the attribute not
6118 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
6120 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6121 Chung-Ju Wu <jasonwucj@gmail.com>
6123 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
6125 * config/nds32/nds32.opt (mvh): New option.
6127 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6128 Chung-Ju Wu <jasonwucj@gmail.com>
6130 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
6132 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
6133 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
6136 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6137 Chung-Ju Wu <jasonwucj@gmail.com>
6139 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
6141 * config/nds32/nds32-multiple.md (strlensi): New pattern.
6142 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
6144 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
6145 Kito Cheng <kito.cheng@gmail.com>
6146 Chung-Ju Wu <jasonwucj@gmail.com>
6148 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
6149 UNSPEC_FFMISM and UNSPEC_FLMISM.
6150 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
6151 for ffb, ffmism and flmism.
6152 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
6153 (unspec_ffmism): Ditto.
6154 (unspec_flmism): Ditto.
6155 (nds32_expand_builtin_impl): Check if string extension is available.
6156 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
6157 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
6159 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
6162 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6165 * lra-assigns.c (assign_by_spills): Return a flag of reload
6166 assignment failure. Do not process the reload assignment
6167 failures. Do not spill other reload pseudos if they has the same
6169 (lra_assign): Add a return arg. Set up from the result of
6170 assign_by_spills call.
6171 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6172 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6173 usage_insns if it is not NULL.
6174 (spill_hard_reg_in_range): New function.
6175 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6176 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6177 function prototypes.
6178 (lra_assign): Change prototype.
6179 * lra.c (lra): Add code to deal with fails by splitting hard reg
6182 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
6185 * config/i386/i386.opt: Replace Enforcment with Enforcement.
6187 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
6190 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
6191 (dw_val_node): Add val_symbolic_view.
6192 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
6193 (symview_upper_bound): New.
6194 (new_line_info_table): Initialize symviews_since_reset.
6195 (dwarf2out_source_line): Count symviews_since_reset and set
6196 symview_upper_bound.
6197 (dw_val_equal_p): Handle symview.
6198 (add_AT_symview): New.
6199 (print_dw_val): Handle symview.
6200 (attr_checksum, attr_checksum_ordered): Likewise.
6201 (same_dw_val_p, size_of_die): Likewise.
6202 (value_format, output_die): Likewise.
6203 (add_high_low_attributes): Use add_AT_symview for entry_view.
6204 (dwarf2out_finish): Reset symview_upper_bound, clear
6207 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
6210 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
6211 Add strict argument and use it.
6212 (rs6000_split_multireg_move): Update for new strict argument.
6213 (mem_operand_gpr): Disallow all non-offsettable addresses.
6214 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
6216 2018-03-09 Jakub Jelinek <jakub@redhat.com>
6219 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
6220 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
6221 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
6224 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
6225 decl, use remap_type if we want to use the type.
6227 2018-03-09 Martin Sebor <msebor@redhat.com>
6229 PR tree-optimization/84526
6230 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6232 (builtin_access::generic_overlap): Be prepared to handle non-array
6235 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
6237 PR rtl-optimization/84682
6238 * lra-constraints.c (process_address_1): Check is_address flag
6239 for address constraints.
6240 (process_alt_operands): Likewise.
6241 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
6242 preprocess_constraints.
6243 * recog.h (preprocess_constraints): Add oploc parameter.
6245 * recog.c (preprocess_constraints): Test address_operand for
6246 CT_ADDRESS constraints.
6248 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6251 * lra-assigns.c (assign_by_spills): Return a flag of reload
6252 assignment failure. Do not process the reload assignment
6253 failures. Do not spill other reload pseudos if they has the same
6255 (lra_assign): Add a return arg. Set up from the result of
6256 assign_by_spills call.
6257 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6258 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6259 usage_insns if it is not NULL.
6260 (spill_hard_reg_in_range): New function.
6261 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6262 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6263 function prototypes.
6264 (lra_assign): Change prototype.
6265 * lra.c (lra): Add code to deal with fails by splitting hard reg
6268 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6271 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
6272 Accept complain bool parameter. Only emit errors if it is true.
6273 (arm_parse_cpu_option_name): Likewise.
6274 (arm_target_thumb_only): Adjust callers of the above.
6275 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
6276 prototype to take a default true bool parameter.
6277 (arm_parse_arch_option_name): Likewise.
6279 2018-03-09 David Malcolm <dmalcolm@redhat.com>
6280 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
6284 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
6285 * configure: Regenerate.
6286 * configure.ac ("linker --version-script option"): New.
6287 ("linker soname option"): New.
6289 2018-03-09 Richard Biener <rguenther@suse.de>
6291 PR tree-optimization/84775
6292 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
6293 immediate uses of predicate stmts and mark them modified.
6296 PR tree-optimization/84178
6297 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6299 (version_loop_for_if_conversion): Delay update_ssa call.
6300 (tree_if_conversion): Delay update_ssa until after predicate
6303 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
6306 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
6307 when the function accesses prior frames.
6309 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6312 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
6313 gen_llsym, otherwise call maybe_gen_llsym.
6316 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
6317 has ',' character inside of it.
6319 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6322 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
6323 as clobbering CC_REGNUM.
6325 2018-03-08 Richard Biener <rguenther@suse.de>
6328 * tree-scalar-evolution.c: Include tree-into-ssa.h.
6329 (follow_copies_to_constant): Do not follow SSA names registered
6332 2018-03-08 Richard Biener <rguenther@suse.de>
6334 PR tree-optimization/84178
6335 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6337 (version_loop_for_if_conversion): Delay update_ssa call.
6338 (tree_if_conversion): Delay update_ssa until after predicate
6341 2018-03-08 David Malcolm <dmalcolm@redhat.com>
6343 PR tree-optimization/84178
6344 * tree-if-conv.c (release_bb_predicate): Remove the
6345 the assertion that the stmts have NULL use_ops.
6346 Discard the statements, asserting that they haven't
6347 yet been added to a BB.
6349 2018-03-08 Richard Biener <rguenther@suse.de>
6351 PR tree-optimization/84746
6352 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
6353 (phi_translate): Pass in destination ANTIC_OUT set.
6354 (phi_translate_1): Likewise. For a simplified result lookup
6355 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
6356 (phi_translate_set): Adjust.
6357 (do_pre_regular_insertion): Likewise.
6358 (do_pre_partial_partial_insertion): Likewise.
6360 2018-03-08 Martin Liska <mliska@suse.cz>
6362 PR gcov-profile/84735
6363 * doc/gcov.texi: Document usage of profile files.
6364 * gcov-io.h: Document changes in the format.
6366 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
6370 * dwarf2out.c (struct dw_line_info_table): Update comments for
6372 (FORCE_RESET_NEXT_VIEW): New.
6373 (FORCE_RESETTING_VIEW_P): New.
6374 (RESETTING_VIEW_P): Check for -1 too.
6375 (ZERO_VIEW_P): Likewise.
6376 (new_line_info_table): Force-reset next view.
6377 (dwarf2out_begin_function): Likewise.
6378 (dwarf2out_source_line): Simplify zero_view_p initialization.
6379 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
6380 view directly. Omit view when omitting .loc at line 0.
6382 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6384 PR tree-optimization/84740
6385 * tree-switch-conversion.c (process_switch): Call build_constructors
6386 only if info.phi_count is non-zero.
6388 PR tree-optimization/84739
6389 * tree-tailcall.c (find_tail_calls): Check call arguments against
6390 DECL_ARGUMENTS (current_function_decl) rather than
6391 DECL_ARGUMENTS (func) when checking for tail recursion.
6393 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6395 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
6396 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
6397 Volker Reichelt's entry and add entries for people that perform
6398 GCC fuzzy testing and report numerous bugs.
6400 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
6403 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
6404 readonly data in sdata, if that is disabled.
6405 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
6406 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
6407 -mreadonly-in-sdata option.
6409 2018-03-07 Martin Sebor <msebor@redhat.com>
6411 PR tree-optimization/84468
6412 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
6413 basic block when looking for nul assignment.
6415 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
6418 * except.h (output_function_exception_table): Adjust prototype.
6419 * except.c (output_function_exception_table): Remove FNNAME parameter
6420 and add SECTION parameter. Ouput one part of the table at a time.
6421 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
6422 the first part of the exception table and emit unwind directives.
6423 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
6424 (i386_pe_seh_cold_init): Likewise.
6425 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
6426 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
6427 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
6428 (ix86_output_call_insn): Emit a nop in one more case for SEH.
6429 * config/i386/winnt.c: Include except.h.
6430 (struct seh_frame_state): Add reg_offset, after_prologue and
6431 in_cold_section fields.
6432 (i386_pe_seh_end_prologue): Set seh->after_prologue.
6433 (i386_pe_seh_cold_init): New function.
6434 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
6435 to seh->in_cold_section.
6436 (seh_emit_push): Record the offset of the push.
6437 (seh_emit_save): Record the offet of the save.
6438 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
6439 Test seh->after_prologue to disregard the epilogue.
6440 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
6441 (i386_pe_end_cold_function): New function.
6443 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6446 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
6447 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
6450 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
6452 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
6453 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
6456 * multiple_target.c: Include tree-inline.h and intl.h.
6457 (expand_target_clones): Diagnose and fail if node->definition and
6458 !tree_versionable_function_p (node->decl).
6460 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
6462 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
6464 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
6465 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
6466 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
6468 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6471 * combine.c (try_combine): Use reg_or_subregno instead of handling
6472 just paradoxical SUBREGs and REGs.
6474 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
6476 * config/arc/arc.c (arc_finalize_pic): Remove function.
6477 (arc_must_save_register): We use single base PIC register, remove
6478 checks to save/restore the PIC register.
6479 (arc_expand_prologue): Likewise.
6480 * config/arc/arc-protos.h (arc_set_default_type_attributes):
6482 (arc_verify_short): Likewise.
6483 (arc_attr_type): Likewise.
6484 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
6485 (walk_stores): Likewise.
6486 (arc_address_cost): Make it static.
6487 (arc_verify_short): Likewise.
6488 (branch_dest): Likewise.
6489 (arc_attr_type): Likewise.
6490 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
6491 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
6492 (arc_final_prescan_insn): Remove inserting the nops due to
6493 hardware hazards. It is done in reorg step.
6494 (insn_length_variant_t): Remove.
6495 (insn_length_parameters_t): Likewise.
6496 (arc_insn_length_parameters): Likewise.
6497 (arc_get_insn_variants): Likewise.
6498 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
6500 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6503 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
6506 PR tree-optimization/84687
6507 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
6509 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
6511 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6513 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
6514 Rename to ppc_speculation_barrier.
6515 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
6516 __builtin_ppc_speculation_barrier.
6518 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6521 * combine.c (combine_simplify_rtx): Don't try to simplify if
6522 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
6525 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
6527 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
6528 to 32 bytes when compiling for POWER9.
6530 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6533 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
6534 regparm >= 3 with no arg reg available also for calls with
6535 flag_force_indirect_call. Pass decl to ix86_function_regparm.
6538 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
6540 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
6542 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
6545 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
6547 2018-03-05 Richard Biener <rguenther@suse.de>
6549 PR tree-optimization/84486
6550 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
6551 When inserting a __builtin_assume_aligned call set the LHS
6552 SSA name alignment info accordingly.
6554 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
6556 PR tree-optimization/84114
6557 * config/aarch64/aarch64.c (aarch64_reassociation_width)
6558 Avoid reassociation of FLOAT_MODE addition.
6560 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
6562 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
6563 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
6564 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
6565 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
6566 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
6567 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
6568 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
6570 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
6571 __builtin_ia32_wbinvd): New builtins.
6572 (SPECIAL_ARGS2): New.
6573 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
6574 (SPECIAL_ARGS2): New.
6575 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
6576 (ix86_valid_target_attribute_inner_p): Ditto.
6577 (ix86_init_mmx_sse_builtins): Add special_args2.
6578 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
6579 TARGET_WBNOINVD_P): New.
6580 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
6581 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
6582 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
6583 * config/i386/immintrin.h (_wbinvd): New intrinsic.
6584 * config/i386/pconfigintrin.h: New file.
6585 * config/i386/wbnoinvdintrin.h: Ditto.
6586 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
6587 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
6589 2018-03-05 Richard Biener <rguenther@suse.de>
6591 PR tree-optimization/84670
6592 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6594 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6595 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6596 (compute_antic_aux): Only assert the number of values in ANTIC_IN
6597 doesn't grow if all successors (recursively) were visited at least
6600 2018-03-05 Richard Biener <rguenther@suse.de>
6602 PR tree-optimization/84650
6603 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
6604 if executed in the loop pipeline.
6606 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
6608 * doc/configfiles.texi (Configuration Files): Move info about
6609 conditionalizing $target-protos.h to...
6610 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
6611 differs from $target-protos.h.
6613 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
6614 Chung-Ju Wu <jasonwucj@gmail.com>
6616 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
6617 * config/nds32/nds32-multiple.md (setmemsi): Define.
6618 * config/nds32/nds32-memory-manipulation.c
6619 (nds32_gen_dup_4_byte_to_word_value): New.
6620 (emit_setmem_word_loop): New.
6621 (emit_setmem_byte_loop): New.
6622 (nds32_expand_setmem_loop): New.
6623 (nds32_expand_setmem_loop_v3m): New.
6624 (nds32_expand_setmem_unroll): New.
6625 (nds32_expand_setmem): New.
6627 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6628 Chung-Ju Wu <jasonwucj@gmail.com>
6630 * config/nds32/nds32-memory-manipulation.c
6631 (nds32_emit_load_store): New.
6632 (nds32_emit_post_inc_load_store): New.
6633 (nds32_emit_mem_move): New.
6634 (nds32_emit_mem_move_block): New.
6635 (nds32_expand_movmemsi_loop_unknown_size): New.
6636 (nds32_expand_movmemsi_loop_known_size): New.
6637 (nds32_expand_movmemsi_loop): New.
6638 (nds32_expand_movmemsi_unroll): New.
6639 (nds32_expand_movmemqi): Rename ...
6640 (nds32_expand_movmemsi): ... to this.
6641 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
6642 (movmemsi): ... to this.
6643 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
6644 (nds32_expand_movmemsi): ... to this.
6646 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6647 Monk Chiang <sh.chiang04@gmail.com>
6648 Chung-Ju Wu <jasonwucj@gmail.com>
6650 * config/nds32/nds32-protos.h
6651 (nds32_expand_load_multiple): New arguments.
6652 (nds32_expand_store_multiple): Ditto.
6653 (nds32_valid_multiple_load_store): Rename ...
6654 (nds32_valid_multiple_load_store_p): ... to this.
6655 * config/nds32/nds32-memory-manipulation.c
6656 (nds32_expand_load_multiple): Refine implementation.
6657 (nds32_expand_store_multiple): Ditto.
6658 * config/nds32/nds32-multiple.md
6659 (load_multiple): Update nds32_expand_load_multiple interface.
6660 (store_multiple): Update nds32_expand_store_multiple interface.
6661 * config/nds32/nds32-predicates.c
6662 (nds32_valid_multiple_load_store): Rename ...
6663 (nds32_valid_multiple_load_store_p): ... to this and refine
6665 * config/nds32/predicates.md
6666 (nds32_load_multiple_and_update_address_operation): New predicate.
6667 (nds32_store_multiple_and_update_address_operation): New predicate.
6669 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6670 Chung-Ju Wu <jasonwucj@gmail.com>
6672 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
6673 (combo): New attribute.
6674 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
6676 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
6678 * config/nds32/nds32.opt: Change -mcmodel= default value.
6680 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
6681 Monk Chiang <sh.chiang04@gmail.com>
6682 Chung-Ju Wu <jasonwucj@gmail.com>
6684 * config/nds32/constants.md (unspec_element): New enum.
6685 * config/nds32/constraints.md (Umw): New constraint.
6686 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
6687 * config/nds32/nds32-intrinsic.md: Likewise.
6688 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
6689 (nds32_valid_smw_lwm_base_p): New.
6690 (nds32_output_smw_single_word): New.
6691 (nds32_output_lmw_single_word): New.
6692 (nds32_expand_unaligned_load): New.
6693 (nds32_expand_unaligned_store): New.
6694 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
6695 (nds32_output_smw_single_word): Declare.
6696 (nds32_output_lmw_single_word): Declare.
6697 (nds32_expand_unaligned_load): Declare.
6698 (nds32_expand_unaligned_store): Declare.
6699 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
6700 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
6701 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
6702 NDS32_BUILTIN_UASTORE_DW.
6703 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
6706 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
6707 Kito Cheng <kito.cheng@gmail.com>
6708 Chung-Ju Wu <jasonwucj@gmail.com>
6710 * config/nds32/nds32-intrinsic.c
6711 (nds32_expand_builtin_null_ftype_reg): Delete.
6712 (nds32_expand_builtin_reg_ftype_imm): Ditto.
6713 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
6714 (nds32_read_argument): New.
6715 (nds32_legitimize_target): Ditto.
6716 (nds32_legitimize_argument): Ditto.
6717 (nds32_check_constant_argument): Ditto.
6718 (nds32_expand_unop_builtin): Ditto.
6719 (nds32_expand_unopimm_builtin): Ditto.
6720 (nds32_expand_binop_builtin): Ditto.
6721 (nds32_builtin_decl_impl): Ditto.
6722 (builtin_description): Ditto.
6723 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
6724 (nds32_init_builtins_impl): Ditto.
6725 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
6726 (nds32_builtin_decl): New.
6727 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
6728 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
6730 2018-03-02 Jeff Law <law@redhat.com>
6732 * reorg.c (stop_search_p): Handle DEBUG_INSN.
6733 (redundant_insn, fill_simple_delay_slots): Likewise.
6734 (fill_slots_from_thread): Likewise.
6735 * resource.c (mark_referenced_resources): Likewise.
6736 (mark_set_resources, find_dead_or_set_registers): Likewise.
6738 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6740 * substring-locations.h (format_warning_va): Formatting fix for
6742 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
6744 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
6745 * substring-locations.c: Include intl.h.
6746 (format_warning_va): Turned into small wrapper around
6747 format_warning_n_va, renamed to ...
6748 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
6749 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
6751 (format_warning_at_substring_n): New function.
6752 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
6753 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
6754 format_warning_at_substring with just a shorter name instead of
6755 const function pointer.
6756 (fmtwarn_n): New function.
6757 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
6758 appropriate, get rid of all the fmtstr temporaries, move conditionals
6759 with G_() wrapped string literals directly into fmtwarn arguments,
6760 cast dir.len to (int), formatting fixes.
6762 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
6764 * doc/invoke.texi: Remove "Cilk Plus" references.
6766 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6767 Richard Biener <rguenther@suse.de>
6770 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
6771 for error or warning attributes if CALL_FROM_THUNK_P is set.
6774 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6777 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
6778 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
6781 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
6782 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
6784 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
6785 instead of -1U in last predictors element's probability member.
6787 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
6790 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
6791 arguments if they are comparable.
6793 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6795 PR tree-optimization/84634
6796 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
6797 masks and masked_loop_p with a single loop_masks, making sure it's
6798 null for bb vectorization.
6800 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6802 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
6803 (vect_analyze_data_ref_access): Use loop->safe_len rather than
6804 loop->force_vectorize to check whether there is no alias.
6806 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6809 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
6811 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
6813 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
6814 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
6815 instead of a loop around prev_real_insn.
6816 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
6820 * config/i386/i386.c (ix86_print_operand): Use conditional
6821 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
6824 2018-03-02 Richard Biener <rguenther@suse.de>
6826 PR tree-optimization/84427
6827 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
6828 (bitmap_set_subtract_values): Rewrite to handle multiple
6831 (prune_clobbered_mems): Likewise.
6832 (phi_translate): Take edge instead of pred/phiblock.
6833 (phi_translate_1): Likewise.
6834 (phi_translate_set): Likewise. Insert all translated
6835 exprs for a value into the set, keeping possibly multiple
6836 expressions per value.
6837 (compute_antic_aux): Adjust for phi_translate changes.
6838 When intersecting union the expressions and prune those
6839 not in the final value set, keeping possibly multiple
6840 expressions per value. Do not use value-insertion
6841 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
6842 all expressions. Add verification that the value-sets
6843 only shrink during iteration.
6844 (compute_partial_antic_aux): Adjust for the phi_translate changes.
6845 (do_pre_regular_insertion): Likewise.
6846 (do_pre_partial_partial_insertion): Likewise.
6848 2018-03-02 Richard Biener <rguenther@suse.de>
6851 * config/darwin.c (saved_debug_info_level): New static global.
6852 (darwin_asm_lto_start): Disable debug info generation for LTO out.
6853 (darwin_asm_lto_end): Restore debug info generation settings.
6855 2018-03-01 Martin Liska <mliska@suse.cz>
6858 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
6861 2018-03-01 Richard Biener <rguenther@suse.de>
6864 * dwarf2out.c (gen_variable_die): Properly handle late VLA
6865 type annotation with LTO when debug was disabled at compile-time.
6867 2018-03-01 Matthew Fortune <mfortune@gmail.com>
6869 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
6871 (mips_final_postscan_insn): Likewise.
6873 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
6875 PR rtl-optimization/84528
6876 * alias.c (init_alias_target): Add commentary.
6877 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
6878 a unique base value if the frame pointer is not eliminated
6879 to the stack pointer.
6881 2018-03-01 Tom de Vries <tom@codesourcery.com>
6883 PR rtl-optimization/83327
6884 * lra-int.h (hard_regs_spilled_into): Declare.
6885 * lra.c (hard_regs_spilled_into): Define.
6886 (init_reg_info): Init hard_regs_spilled_into.
6887 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
6888 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
6889 (process_bb_lives): Handle hard_regs_spilled_into.
6890 (lra_create_live_ranges_1): Before doing liveness propagation, clear
6891 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
6893 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
6895 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
6896 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
6897 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
6898 * config/rs6000/aix72.h: New file.
6900 2018-02-28 Jakub Jelinek <jakub@redhat.com>
6902 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
6903 instead of warning_at with conditional singular and plural messages
6907 * stor-layout.c (update_alignment_for_field): For
6908 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
6909 && !DECL_PACKED (field), do the alignment update, just use
6910 only desired_align instead of MAX (type_align, desired_align)
6912 (place_field): Don't do known_align < desired_align handling
6913 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
6914 is non-NULL, instead do it after rli->prev_field handling and
6915 only if not within a bitfield word. For DECL_PACKED (field)
6916 use type_align of BITS_PER_UNIT.
6918 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
6920 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
6921 superfluous parentheses and trailing spaces.
6923 2018-02-28 Richard Biener <rguenther@suse.de>
6925 PR tree-optimization/84584
6926 * graphite-scop-detection.c (scop_detection::add_scop): Discard
6927 SCoPs with fake exit edge.
6929 2018-02-28 Martin Liska <mliska@suse.cz>
6932 * timevar.c (timer::print): Fix format to properly print 100%
6935 2018-02-28 Richard Biener <rguenther@suse.de>
6938 * genmatch.c (capture_info::walk_match): Do not mark
6939 captured expressions without operands as expr_p given
6940 they act more like predicates and should be subject to
6941 "lost tail" side-effect preserving.
6943 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
6945 PR rtl-optimization/81611
6946 * auto-inc-dec.c (attempt_change): Move dead note from
6947 mem_insn if it's the next use of regno
6948 (find_address): Take address use of reg holding
6949 non-incremented value. Add parm to limit search to the named
6951 (merge_in_block): Attempt to use a mem insn that is the next
6952 use of the original regno.
6954 2018-02-27 Martin Sebor <msebor@redhat.com>
6957 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
6958 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
6960 2018-02-27 Martin Sebor <msebor@redhat.com>
6962 PR translation/84207
6963 * diagnostic-core.h (warning_n, error_n, inform_n): Change
6964 n argument to unsigned HOST_WIDE_INT.
6965 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
6966 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
6967 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
6968 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
6970 2018-02-27 Richard Biener <rguenther@suse.de>
6972 PR tree-optimization/84512
6973 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
6974 Do not use the estimate returned from record_stmt_cost for
6975 the scalar iteration cost but sum properly using add_stmt_cost.
6977 2018-02-27 Richard Biener <rguenther@suse.de>
6979 PR tree-optimization/84466
6980 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6981 Adjust last change to less strictly validate use operands.
6983 2018-02-27 Martin Liska <mliska@suse.cz>
6985 PR gcov-profile/84548
6986 * gcov.c (process_file): Allow partial overlap and consider it
6987 also as group functions.
6988 (output_lines): Properly calculate range of lines for a group.
6990 2018-02-27 Martin Liska <mliska@suse.cz>
6992 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
6993 'ggc' suffixes. Change first column width.
6994 (timer::print): Fix formatting of the column.
6996 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
6998 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
6999 preserve inline entry blocks for the sake of debug inline
7000 entry point markers alone.
7001 (remove_unused_locals): Suggest in comments a better place to
7002 force the preservation of inline entry blocks that are
7003 otherwise unused, but do not preserve them.
7005 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7007 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
7009 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7012 * config/i386/constraints.md (Bs): Replace
7013 ix86_indirect_branch_register with
7014 TARGET_INDIRECT_BRANCH_REGISTER.
7016 * config/i386/i386.md (indirect_jump): Likewise.
7017 (tablejump): Likewise.
7018 (*sibcall_memory): Likewise.
7019 (*sibcall_value_memory): Likewise.
7020 Peepholes of indirect call and jump via memory: Likewise.
7021 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
7022 (*sibcall_value_GOT_32): Likewise.
7023 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7024 (GOT_memory_operand): Likewise.
7025 (call_insn_operand): Likewise.
7026 (sibcall_insn_operand): Likewise.
7027 (GOT32_symbol_operand): Likewise.
7028 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
7030 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7032 PR rtl-optimization/83496
7033 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
7034 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
7035 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
7036 redundant insn, if any.
7037 (relax_delay_slots): Likewise.
7038 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
7040 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
7042 PR tree-optimization/83965
7043 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
7044 that grouped statements are part of a reduction chain. Return
7045 true if the statement is not marked as a reduction itself but
7047 (vect_recog_dot_prod_pattern): Don't check whether the statement
7048 is part of a group here.
7049 (vect_recog_sad_pattern): Likewise.
7050 (vect_recog_widen_sum_pattern): Likewise.
7052 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7055 * final.c (rest_of_clean_state): Also look for calls inside sequences.
7057 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7060 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
7062 (ix86_output_indirect_function_return): New prototype.
7063 (ix86_split_simple_return_pop_internal): Likewise.
7064 * config/i386/i386.c (indirect_return_via_cx): New.
7065 (indirect_return_via_cx_bnd): Likewise.
7066 (indirect_thunk_name): Handle return va CX_REG.
7067 (output_indirect_thunk_function): Create alias for
7068 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
7069 (ix86_output_indirect_jmp): Remove the bool argument.
7070 (ix86_output_indirect_function_return): New function.
7071 (ix86_split_simple_return_pop_internal): Likewise.
7072 * config/i386/i386.md (*indirect_jump): Don't pass false
7073 to ix86_output_indirect_jmp.
7074 (*tablejump_1): Likewise.
7075 (simple_return_pop_internal): Change it to define_insn_and_split.
7076 Call ix86_split_simple_return_pop_internal to split it for
7078 (simple_return_indirect_internal): Call
7079 ix86_output_indirect_function_return instead of
7080 ix86_output_indirect_jmp.
7082 2018-02-26 Jakub Jelinek <jakub@redhat.com>
7085 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
7086 memset and value initialization afterwards.
7088 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
7090 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
7092 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7095 * common/config/aarch64/aarch64-common.c
7096 (aarch_option_optimization_table[]): Switch
7097 off fomit-frame-pointer
7099 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7100 Chung-Ju Wu <jasonwucj@gmail.com>
7102 * config/nds32/nds32-multiple.md (load_multiple): Disallow
7104 (store_multiple): Ditto.
7106 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7108 * config.gcc: Add --with-cpu support for nds32 target.
7109 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
7110 * config/nds32/nds32.opt: Add -mcpu= option.
7112 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
7114 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
7115 isel=yes): Warn for these deprecated options.
7117 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
7119 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
7120 ISA_2_5_MASKS_EMBEDDED.
7122 2018-02-23 Jakub Jelinek <jakub@redhat.com>
7124 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
7125 p->max as pointers rather than using iterative_hash_expr.
7127 2018-02-23 Carl Love <cel@us.ibm.com>
7129 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
7130 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
7132 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
7133 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
7134 P8V_BUILTIN_VEC_VUNSIGNED2.
7136 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
7139 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
7140 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
7142 * lra-constraints.c (curr_insn_transform): Set up
7143 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
7145 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
7146 * lra-eliminations.c (spill_pseudos): Ditto.
7147 (process_insn_for_elimination): Ditto.
7148 * lra-lives.c (reg_early_clobber_p): Use the new macros.
7149 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
7150 LRA_NON_CLOBBERED_ALT.
7152 2018-02-22 Martin Sebor <msebor@redhat.com>
7154 PR tree-optimization/84480
7155 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
7156 to maybe_diag_stxncpy_trunc. Call it.
7157 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
7158 from gimple_fold_builtin_strcpy. Print inlining stack.
7159 (handle_builtin_stxncpy): Print inlining stack.
7160 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
7162 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
7165 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
7166 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
7167 and -fcheck-pointer-bounds are used together.
7168 (indirect_thunk_prefix): New enum.
7169 (indirect_thunk_need_prefix): New function.
7170 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
7171 "_nt" instead of "_bnd" for NOTRACK prefix.
7172 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
7173 (output_indirect_thunk_function): Likewise.
7175 (ix86_code_end): Update output_indirect_thunk_function calls.
7176 (ix86_output_indirect_branch_via_reg): Replace
7177 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
7178 (ix86_output_indirect_branch_via_push): Likewise.
7179 (ix86_output_function_return): Likewise.
7180 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
7181 incompatible with -fcf-protection=branch and
7182 -fcheck-pointer-bounds.
7184 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7187 * config/aarch64/aarch64.c (aarch64_print_address_internal):
7188 Change gcc_assert call to output_operand_lossage.
7190 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7192 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
7194 2018-02-22 DJ Delorie <dj@redhat.com>
7195 Sebastian Perta <sebastian.perta@renesas.com>
7196 Oleg Endo <olegendo@gcc.gnu.org>
7198 * config/rx/rx.c (rx_rtx_costs): New function.
7199 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
7201 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
7203 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
7205 2018-02-22 Martin Liska <mliska@suse.cz>
7208 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
7209 Add "native" as a possible value.
7211 2018-02-22 Martin Liska <mliska@suse.cz>
7214 * config/i386/i386.c (ix86_option_override_internal):
7215 Add "native" as a possible value for -march and -mtune.
7217 2018-02-22 Jakub Jelinek <jakub@redhat.com>
7220 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
7221 to all type variants.
7223 PR tree-optimization/84503
7224 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
7225 width as info->bitpos + info->bitsize - start.
7226 (merged_store_group::merge_overlapping): Simplify width computation.
7227 (check_no_overlap): New function.
7228 (imm_store_chain_info::try_coalesce_bswap): Compute expected
7229 start + width and last_order of the group, fail if check_no_overlap
7231 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
7232 to group if check_no_overlap fails.
7234 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7236 * config/rs6000/altivec.md: Delete contraint arguments to
7237 define_expand, define_split, and define_peephole2, and in
7238 define_insn_and_split if always unused.
7239 * config/rs6000/darwin.md: Ditto.
7240 * config/rs6000/dfp.md: Ditto.
7241 * config/rs6000/rs6000.md: Ditto.
7242 * config/rs6000/sync.md: Ditto.
7243 * config/rs6000/vector.md: Ditto.
7244 * config/rs6000/vsx.md: Ditto.
7246 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7248 * config/rs6000/altivec.md: Write output control strings as braced
7249 blocks instead of double-quoted strings.
7250 * config/rs6000/darwin.md: Ditto.
7251 * config/rs6000/rs6000.md: Ditto.
7252 * config/rs6000/vector.md: Ditto.
7253 * config/rs6000/vsx.md: Ditto.
7255 2018-02-21 Jason Merrill <jason@redhat.com>
7257 PR c++/84314 - ICE with templates and fastcall attribute.
7258 * attribs.c (build_type_attribute_qual_variant): Remove assert.
7260 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7262 * ipa-cp.c (determine_versionability): Fix comment typos.
7264 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7267 * ipa-cp.c (determine_versionability): Do not version functions caling
7270 2018-02-21 Martin Liska <mliska@suse.cz>
7273 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
7274 Add "native" as a possible value.
7275 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
7276 the macro when native cpu detection is available.
7278 2018-02-21 Martin Liska <mliska@suse.cz>
7281 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
7282 Add "native" as a possible value.
7283 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
7284 when native cpu detection is available.
7286 2018-02-21 Jakub Jelinek <jakub@redhat.com>
7287 Martin Sebor <msebor@redhat.com>
7289 PR tree-optimization/84478
7290 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
7292 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
7293 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
7294 support which is conservatively correct, for 2 only stay conservative
7295 for maxlen. Formatting and comment capitalization fixes. Add STRICT
7296 argument to the 2 argument get_range_strlen, adjust 6 arg
7297 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
7299 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
7300 (gimple_fold_builtin_strlen): Pass true as last argument to
7303 2018-02-20 Martin Sebor <msebor@redhat.com>
7306 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
7307 (builtin_memref::set_base_and_offset): Same. Handle inner references.
7308 (builtin_memref::builtin_memref): Factor out parts into
7309 set_base_and_offset and call it.
7311 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7314 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
7315 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
7316 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
7317 search at the associated MODE_INT.
7319 2018-02-20 Jeff Law <law@redhat.com>
7322 PR tree-optimization/81592
7324 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
7325 for range data rather than using global data.
7326 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
7327 range data rather than using global data.
7328 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
7329 pass it to children as needed.
7330 (struct directive::fmtresult): Similarly.
7331 (struct directive::set_width): Similarly.
7332 (struct directive::set_precision): Similarly.
7333 (format_integer, format_directive, parse_directive): Similarly.
7334 (format_none): Accept unnamed vr_values parameter.
7335 (format_percent, format_floating, format_character): Similarly.
7336 (format_string, format_plain): Similarly.
7337 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
7338 the EVRP range analyzer for range data rather than using global data.
7339 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
7340 gimple-ssa-evrp-analyze.h
7341 (class sprintf_dom_walker): Add after_dom_children member function.
7342 Add evrp_range_analyzer member.
7343 (sprintf_dom_walker::before_dom_children): Call into the EVRP
7344 range analyzer as needed.
7345 (sprintf_dom_walker::after_dom_children): New member function.
7346 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
7348 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
7349 (evrp_range_analyzer::pop_to_marker): Likewise.
7351 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7353 PR tree-optimization/84419
7354 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
7355 with the required type if its current type is compatible but
7358 2018-02-20 Jakub Jelinek <jakub@redhat.com>
7361 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
7362 after vectorization.
7364 2018-02-20 Martin Liska <mliska@suse.cz>
7367 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
7368 possible values if we don't have a hint.
7370 2018-02-20 Martin Liska <mliska@suse.cz>
7374 * final.c (shorten_branches): Build align_tab array with one
7376 * opts.c (finish_options): Add alignment option limit check.
7377 (MAX_CODE_ALIGN): Likewise.
7378 (MAX_CODE_ALIGN_VALUE): Likewise.
7379 * doc/invoke.texi: Document maximum allowed option value for
7380 all -falign-* options.
7382 2018-02-19 Jakub Jelinek <jakub@redhat.com>
7385 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
7386 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
7387 * var-tracking.c (emit_note_insn_var_location): Remove all references
7388 to NOTE_INSN_CALL_ARG_LOCATION.
7389 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
7390 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
7391 Use copy_rtx_if_shared.
7392 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
7393 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
7394 (dwarf2out_var_location): Remove handling of
7395 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
7397 * final.c (final_scan_insn): Remove all references to
7398 NOTE_INSN_CALL_ARG_LOCATION.
7399 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
7400 before dumping final insns.
7401 * except.c (emit_note_eh_region_end): Remove all references to
7402 NOTE_INSN_CALL_ARG_LOCATION.
7403 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
7404 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
7405 * config/arc/arc.c (hwloop_optimize): Likewise.
7406 * config/arm/arm.c (create_fix_barrier): Likewise.
7407 * config/s390/s390.c (s390_chunkify_start): Likewise.
7408 * config/sh/sh.c (find_barrier): Likewise.
7409 * config/i386/i386.c (rest_of_insert_endbranch,
7410 ix86_seh_fixup_eh_fallthru): Likewise.
7411 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
7412 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
7413 * config/frv/frv.c (frv_function_prologue): Likewise.
7414 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
7416 (note_outside_basic_block_p): Remove all references to
7417 NOTE_INSN_CALL_ARG_LOCATION.
7418 * gengtype.c (adjust_field_rtx_def): Likewise.
7419 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
7421 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
7422 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
7425 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
7428 PR tree-optimization/84452
7429 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
7430 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
7433 2018-02-19 Martin Liska <mliska@suse.cz>
7436 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
7438 2018-02-19 Martin Liska <mliska@suse.cz>
7439 Richard Sandiford <richard.sandiford@linaro.org>
7441 PR tree-optimization/82491
7442 * gimple-fold.c (get_base_constructor): Make earlier bail out
7445 2018-02-19 Carl Love <cel@us.ibm.com>
7447 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
7448 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
7450 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
7451 P8V_BUILTIN_VEC_NEG.
7453 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
7455 * config/rl78/rl78.md (movdf): New define expand.
7457 2018-02-19 Martin Liska <mliska@suse.cz>
7460 * doc/invoke.texi: Fix typo.
7461 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
7463 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
7465 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
7466 handle rs6000_single_float and rs6000_double_float specially for
7469 2018-02-16 Jeff Law <law@redhat.com>
7471 * config/rx/rx.c (add_pop_cfi_notes): New function.;
7474 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7477 * ipa-inline.c (inline_small_functions): Fix a typo.
7479 2018-02-16 Nathan Sidwell <nathan@acm.org>
7481 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
7483 2018-02-16 Carl Love <cel@us.ibm.com>
7485 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
7486 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
7487 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
7488 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
7489 expansion to P8V_BUILTIN_VEC_FLOAT2.
7491 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
7493 PR rtl-optimization/70023
7494 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
7495 src_regno into account.
7497 2018-02-16 Carl Love <cel@us.ibm.com>
7499 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
7500 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
7501 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
7502 * config/rs6000/rs6000.c: Remove case statements for
7503 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
7504 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
7505 and P9V_BUILTIN_VEC_VINSERT4B.
7506 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
7507 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
7508 * config/rs6000/vsx.md:
7509 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
7512 2018-02-16 Carl Love <cel@us.ibm.com>
7514 * config/rs6000/altivec.h: Add builtin names vec_extract4b
7516 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
7518 * config/rs6000/rs6000-c.c: Add the definitions for
7519 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
7520 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
7521 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
7522 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
7523 definition for insert4b and define insn *insert3b_internal.
7524 * doc/extend.texi: Add documentation for vec_extract4b.
7526 2018-02-16 Nathan Sidwell <nathan@acm.org>
7528 * doc/extend.texi (Backwards Compatibility): Mention friend
7529 injection. Note for-scope is deprecated.
7530 * doc/invoke.texi (-ffriend-injection): Deprecate.
7532 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
7534 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
7535 that moved to I2, also allow destinations that are a paradoxical
7536 subreg (instead of a normal reg).
7538 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
7541 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
7544 2018-02-16 Richard Biener <rguenther@suse.de>
7546 PR tree-optimization/84037
7547 PR tree-optimization/84016
7549 * config/i386/i386.c (ix86_builtin_vectorization_cost):
7550 Adjust vec_construct for the fact we need additional higher latency
7551 128bit inserts for AVX256 and AVX512 vector builds.
7552 (ix86_add_stmt_cost): Scale vector construction cost for
7555 2018-02-16 Richard Biener <rguenther@suse.de>
7557 PR tree-optimization/84417
7558 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
7559 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
7560 (non_rewritable_lvalue_p): Likewise, use poly-ints.
7562 2018-02-16 Martin Liska <mliska@suse.cz>
7565 * internal-fn.def (ASAN_CHECK): Set proper flags.
7566 (ASAN_MARK): Likewise.
7568 2018-02-16 Julia Koval <julia.koval@intel.com>
7570 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
7571 from PTA_CANNONLAKE.
7573 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7576 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
7577 Use ++iter rather than iter++ for std::list iterators.
7578 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
7579 defer deleting them until all nodes in the forest are processed. Do
7580 free even leaf nodes. Change to_process into auto_vec.
7583 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
7584 * vec.h (vec_default_construct): Use memset instead of placement new
7585 if BROKEN_VALUE_INITIALIZATION is defined.
7586 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
7587 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
7590 PR rtl-optimization/83723
7591 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
7592 * lra.c (lra_substitute_pseudo): Likewise. If true, use
7593 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
7595 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
7597 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
7599 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
7601 PR rtl-optimization/81443
7602 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
7603 from inner REGs to paradoxical SUBREGs.
7605 2018-02-16 Richard Biener <rguenther@suse.de>
7607 PR tree-optimization/84399
7608 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7609 For operands we can analyze at their definition make sure we can
7610 analyze them at each use as well.
7612 2018-02-16 Richard Biener <rguenther@suse.de>
7614 PR tree-optimization/84190
7615 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
7616 volatile accesses if the decl isn't volatile.
7618 2018-02-15 Jason Merrill <jason@redhat.com>
7620 PR c++/84314 - ICE with templates and fastcall attribute.
7621 * attribs.c (build_type_attribute_qual_variant): Don't clobber
7622 TYPE_CANONICAL on an existing type.
7624 2018-02-15 Jakub Jelinek <jakub@redhat.com>
7626 PR tree-optimization/84383
7627 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
7628 dstoff nor call operand_equal_p if dstbase is NULL.
7630 PR tree-optimization/84334
7631 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
7632 also a CONSTANT_CLASS_P, punt.
7634 2018-02-14 Jim Wilson <jimw@sifive.com>
7636 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
7637 first SMALL_OPERAND check. New local min_second_step. Move assert
7638 to where locals are set. Add TARGET_RVC support.
7639 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
7641 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
7643 * doc/invoke.texi: Correct -Wformat-overflow code sample.
7645 2018-02-14 Martin Sebor <msebor@redhat.com>
7647 PR tree-optimization/83698
7648 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
7649 arrays constrain the offset range to their bounds.
7650 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
7651 (builtin_access::overlap): Avoid setting the size of overlap if it's
7653 (maybe_diag_overlap): Also consider arrays when deciding what values
7654 of offsets to include in diagnostics.
7656 2018-02-14 Martin Sebor <msebor@redhat.com>
7659 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
7660 that correspond to the kind of a declaration.
7662 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
7665 * config/pa/pa.md: Load address of PIC label using the linkage table
7666 if the label is nonlocal.
7668 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
7670 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
7671 warning message if user requests -maltivec=be.
7672 * doc/invoke.texi: Document deprecation of -maltivec=be.
7674 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
7677 * config/rs6000/rs6000-c.c: Update definitions for
7678 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
7679 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
7681 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7684 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
7685 add _get_ssp intrinsics. Remove argument from
7686 __builtin_ia32_rdssp[d|q].
7687 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
7688 * config/i386/i386-builtin.def: Remove argument from
7689 __builtin_ia32_rdssp[d|q].
7690 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
7691 ix86_expand_special_args_builtin for _rdssp[d|q].
7692 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
7693 Clear register before usage.
7694 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
7695 Add documentation for new _get_ssp and _inc_ssp intrinsics.
7697 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
7699 PR tree-optimization/84357
7700 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
7701 operand 1 of an ARRAY_REF too.
7703 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
7706 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
7707 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7709 (set_of_reg): New struct.
7710 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
7711 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
7712 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7714 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
7715 Split into bitclr, bitset, bitinvert patterns if appropriate.
7716 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
7717 use rx_fuse_in_memory_bitop.
7718 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
7719 to named insn, correct maximum insn length.
7721 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
7724 * machmode.def: Define a complex mode for PARTIAL_INT.
7725 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
7727 * doc/rtl.texi: Document CSPImode.
7728 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
7730 (msp430_hard_regno_nregs_with_padding): Likewise.
7732 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
7735 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
7737 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
7739 PR rtl-optimization/84169
7740 * combine.c (try_combine): New variable split_i2i3. Set it to true if
7741 we generated a parallel as new i3 and we split that to new i2 and i3
7742 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
7743 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
7744 those to i2, not i1. Partially rewrite this scan code.
7746 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7749 * stor-layout.c (place_field): For variable length fields, adjust
7750 offset_align afterwards not just based on the field's alignment,
7751 but also on the size.
7754 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
7755 of exps and logs in the use_exp2 case.
7757 2018-02-13 Jeff Law <law@redhat.com>
7759 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
7762 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
7765 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
7769 * common.opt (gas-loc-support, gas-locview-support): New.
7770 (ginline-points, ginternal-reset-location-views): New.
7771 * doc/invoke.texi: Document them. Use @itemx where intended.
7772 (gvariable-location-views): Adjust.
7773 * target.def (reset_location_view): New.
7774 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
7775 (TARGET_RESET_LOCATION_VIEW): New.
7776 * doc/tm.texi: Rebuilt.
7777 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
7778 (dwarf2out_default_as_locview_support): New.
7779 (output_asm_line_debug_info): Use option variables.
7780 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
7781 (output_loc_list): Likewise.
7782 (add_high_low_attributes): Check option variables.
7783 Don't output entry view attribute in strict mode.
7784 (gen_inlined_subroutine_die): Check option variables.
7785 (dwarf2out_inline_entry): Likewise.
7786 (init_sections_and_labels): Likewise.
7787 (dwarf2out_early_finish): Likewise.
7788 (maybe_reset_location_view): New, from...
7789 (dwarf2out_var_location): ... here. Call it.
7790 * debug.h (dwarf2out_default_as_loc_support): Declare.
7791 (dwarf2out_default_as_locview_support): Declare.
7792 * hooks.c (hook_int_rtx_insn_0): New.
7793 * hooks.h (hook_int_rtx_insn_0): Declare.
7794 * toplev.c (process_options): Take -gas-loc-support and
7795 -gas-locview-support from dwarf2out. Enable
7796 -gvariable-location-views by default only with locview
7797 assembler support. Enable -ginternal-reset-location-views by
7798 default only if the target defines the corresponding hook.
7799 Enable -ginline-points by default if location views are
7800 enabled; force it disabled if statement frontiers are
7802 * tree-inline.c (expand_call_inline): Check option variables.
7803 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
7805 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
7807 PR tree-optimization/84321
7808 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
7809 handling. Also check whether the anti-range contains any values
7810 that satisfy the mask; switch to a VR_RANGE if not.
7812 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
7815 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
7817 2018-02-13 Martin Jambor <mjambor@suse.cz>
7820 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
7821 of call statements, also set location of a load to a temporary.
7823 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
7825 * config/rl78/rl78.c (add_vector_labels): New function.
7826 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
7827 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
7828 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
7829 which checks that no arguments are passed.
7830 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
7831 * doc/extend.texi: Documentation for the new attribute.
7833 2018-02-13 Andreas Schwab <schwab@suse.de>
7835 * config/riscv/linux.h (CPP_SPEC): Define.
7837 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7840 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
7841 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
7842 OPTION_MASK_ISA_AES as first argument to def_builtin_const
7843 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
7844 instead of OPTION_MASK_ISA_PCLMUL as first argument to
7845 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
7846 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
7847 temporarily for AES and PCLMUL builtins.
7849 PR tree-optimization/84339
7850 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
7851 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
7855 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
7856 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
7857 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
7859 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
7861 * omp-simd-clone.h: New file.
7862 * omp-simd-clone.c: Include omp-simd-clone.h.
7863 (expand_simd_clones): No longer static.
7864 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
7865 cgraph.h and omp-simd-clone.h.
7866 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
7867 (vect_recog_widen_shift_pattern): Formatting fix.
7868 (vect_pattern_recog_1): Don't check optab for calls.
7871 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
7872 operands[2] into a REG before using gen_lowpart on it.
7874 2018-02-12 Jeff Law <law@redhat.com>
7877 * config/sh/sh.c (find_barrier): Consider a sibling call
7880 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
7881 successfully back substituting a reg.
7883 2018-02-12 Richard Biener <rguenther@suse.de>
7885 PR tree-optimization/84037
7886 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
7887 parameter, move visited init to caller.
7888 (vect_slp_analyze_operations): Separate cost from validity
7889 check, initialize visited once for all instances.
7890 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
7892 * tree-vect-stmts.c (vect_model_simple_cost): Make early
7894 (vect_model_promotion_demotion_cost): Likewise.
7895 (vectorizable_bswap): Guard cost modeling with !slp_node
7896 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
7898 (vectorizable_call): Likewise.
7899 (vectorizable_conversion): Likewise.
7900 (vectorizable_assignment): Likewise.
7901 (vectorizable_shift): Likewise.
7902 (vectorizable_operation): Likewise.
7903 (vectorizable_store): Likewise.
7904 (vectorizable_load): Likewise.
7905 (vectorizable_condition): Likewise.
7906 (vectorizable_comparison): Likewise.
7908 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
7911 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
7912 (ASAN_MARK): Fix fnspec to account for return value, change pointer
7913 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
7915 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7918 * params.def (inline-min-speedup): Increase from 8 to 15.
7919 (max-inline-insns-auto): Decrease from 40 to 30.
7920 * ipa-split.c (consider_split): Add some buffer for function to
7921 be considered inlining candidate.
7922 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
7925 2018-02-12 Richard Biener <rguenther@suse.de>
7927 PR tree-optimization/84037
7928 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
7929 matched stmts if we cannot swap the non-matched ones.
7931 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7933 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
7934 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
7935 _mm_maskz_scalef_round_ss): New intrinsics.
7936 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
7937 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
7938 __builtin_ia32_scalefss_round): Remove.
7939 (__builtin_ia32_scalefsd_mask_round,
7940 __builtin_ia32_scalefss_mask_round): New intrinsics.
7941 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
7942 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
7943 ((match_operand:VF_128 2 "<round_nimm_predicate>"
7944 "<round_constraint>")): Changed to ...
7945 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
7946 "<round_scalar_constraint>")): ... this.
7947 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
7948 %0, %1, %2<round_op3>}"): Changed to ...
7949 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
7950 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
7951 %2<round_scalar_mask_op3>}"): ... this.
7952 * config/i386/subst.md (round_scalar_nimm_predicate): New.
7954 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7956 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
7957 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
7958 (_mm_maskz_sqrt_round_ss): New intrinsics.
7959 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
7960 (__builtin_ia32_sqrtsd_mask_round)
7961 (__builtin_ia32_sqrtss_mask_round): New builtins.
7962 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
7963 (__builtin_ia32_sqrtss_round): Remove.
7964 (__builtin_ia32_sqrtsd_mask_round)
7965 (__builtin_ia32_sqrtss_mask_round): New builtins.
7966 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
7967 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
7968 ((match_operand:VF_128 1 "vector_operand"
7969 "xBm,<round_constraint>")): Changed to ...
7970 ((match_operand:VF_128 1 "vector_operand"
7971 "xBm,<round_scalar_constraint>")): ... this.
7972 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
7973 %0, %2, %<iptr>1<round_op3>}): Changed to ...
7974 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
7975 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
7976 %<iptr>1<round_scalar_mask_op3>}): ... this.
7977 ((set_attr "prefix" "<round_prefix>")): Changed to ...
7978 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
7980 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
7983 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
7984 Cast vec_cmpeq result to correct type.
7985 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
7986 Cast vec_cmpgt result to correct type.
7988 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
7990 * final.c (final_scan_insn_1): Renamed from...
7991 (final_scan_insn): ... this. New wrapper, to recover
7992 seen from the outermost call in recursive ones.
7993 * config/sparc/sparc.c (output_return): Drop seen from call.
7994 (output_sibcall): Likewise.
7995 * config/visium/visium.c (output_branch): Likewise.
7997 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
7999 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
8002 2018-02-10 Alan Modra <amodra@gmail.com>
8005 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
8006 Specify LR as an input.
8008 2018-02-10 Jakub Jelinek <jakub@redhat.com>
8011 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
8012 remove_member_access_dummy_vars): New functions.
8013 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
8014 lower_omp_1, execute_lower_omp): Use them.
8016 PR rtl-optimization/84308
8017 * shrink-wrap.c (spread_components): Release todo vector.
8019 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
8021 PR rtl-optimization/57193
8022 * ira-color.c (struct allocno_color_data): Add member
8023 conflict_allocno_hard_prefs.
8024 (update_conflict_allocno_hard_prefs): New.
8025 (bucket_allocno_compare_func): Add a preference based on
8026 conflict_allocno_hard_prefs.
8027 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
8028 (color_allocnos): Remove a dead code. Initiate
8029 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
8031 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8034 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
8035 constraint from =wa to wa. Avoid a subreg on the output operand,
8036 instead use a pseudo and subreg it in a move.
8037 (p9_xxbrd_<mode>): Changed to ...
8038 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
8039 (p9_xxbrd_v2df): New expander.
8040 (p9_xxbrw_<mode>): Changed to ...
8041 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
8042 (p9_xxbrw_v4sf): New expander.
8044 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8046 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
8048 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
8051 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
8052 multiply in 32-bit mode.
8053 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
8054 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
8057 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8059 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
8060 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
8061 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
8062 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
8064 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
8067 * dwarf2out.c (is_trivial_indirect_ref): New function.
8068 (dwarf2out_late_global_decl): Do not generate a location
8069 attribute for variables that have a non-trivial DECL_VALUE_EXPR
8070 and that are not defined in the current unit.
8072 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8074 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
8075 instead of a libcall for UNORDERED.
8077 2018-02-09 Tamar Christina <tamar.christina@arm.com>
8080 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
8081 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
8083 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8086 * config/s390/s390.c (s390_set_current_function): Invoke
8087 s390_indirect_branch_settings also if fndecl didn't change.
8089 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8091 * config/rs6000/rs6000.md (blockage): Set length to zero.
8093 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8095 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
8097 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8100 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
8101 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
8105 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
8106 PARALLEL incoming that failed vt_get_decl_and_offset check.
8109 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
8110 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
8112 (get_variable_section): For decls in named .bss* sections pass true as
8113 second argument to bss_initializer_p.
8115 2018-02-09 Marek Polacek <polacek@redhat.com>
8116 Jakub Jelinek <jakub@redhat.com>
8119 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
8120 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
8121 Sync some changes from cxx_fold_indirect_ref.
8123 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8125 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
8127 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
8128 (BLOCK_INLINE_ENTRY_LABEL): New.
8129 (dwarf2out_var_location): Disregard inline entry markers.
8130 (inline_entry_data): New struct.
8131 (inline_entry_data_hasher): New hashtable type.
8132 (inline_entry_data_hasher::hash): New.
8133 (inline_entry_data_hasher::equal): New.
8134 (inline_entry_data_table): New variable.
8135 (add_high_low_attributes): Add DW_AT_entry_pc and
8136 DW_AT_GNU_entry_view attributes if a pending entry is found
8137 in inline_entry_data_table. Add old entry_pc attribute only
8138 if debug nonbinding markers are disabled.
8139 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
8140 markers are enabled.
8141 (block_within_block_p, dwarf2out_inline_entry): New.
8142 (dwarf2out_finish): Check that no entries remained in
8143 inline_entry_data_table.
8144 * final.c (reemit_insn_block_notes): Handle inline entry notes.
8145 (final_scan_insn, notice_source_line): Likewise.
8146 (rest_of_clean_state): Skip inline entry markers.
8147 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
8149 * gimple.c (gimple_build_debug_inline_entry): New.
8150 * gimple.h (enum gimple_debug_subcode): Add
8151 GIMPLE_DEBUG_INLINE_ENTRY.
8152 (gimple_build_debug_inline_entry): Declare.
8153 (gimple_debug_inline_entry_p): New.
8154 (gimple_debug_nonbind_marker_p): Adjust.
8155 * insn-notes.def (INLINE_ENTRY): New.
8156 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
8157 inline entry marker notes.
8158 (print_insn): Likewise.
8159 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
8160 (INSN_DEBUG_MARKER_KIND): Likewise.
8161 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
8162 * tree-inline.c (expand_call_inline): Build and insert
8163 debug_inline_entry stmt.
8164 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
8165 inline entry blocks early, if nonbind markers are enabled.
8166 (dump_scope_block): Dump fragment info.
8167 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
8168 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
8169 (gimple_build_debug_inline_entry): New.
8170 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
8171 Enable/disable inline entry points too.
8172 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
8173 (DEBUG_INSN): Describe inline entry markers.
8175 * common.opt (gvariable-location-views): New.
8176 (gvariable-location-views=incompat5): New.
8177 * config.in: Rebuilt.
8178 * configure: Rebuilt.
8179 * configure.ac: Test assembler for view support.
8180 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
8181 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
8182 * dwarf2out.c (var_loc_view): New typedef.
8183 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
8184 (dwarf2out_locviews_in_attribute): New.
8185 (dwarf2out_locviews_in_loclist): New.
8186 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
8187 (enum dw_line_info_opcode): Add LI_adv_address.
8188 (struct dw_line_info_table): Add view.
8189 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
8190 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
8191 (zero_view_p): New variable.
8192 (ZERO_VIEW_P): New macro.
8193 (output_asm_line_debug_info): New.
8194 (struct var_loc_node): Add view.
8195 (add_AT_view_list, AT_loc_list): New.
8196 (add_var_loc_to_decl): Add view param. Test it against last.
8197 (new_loc_list): Add view params. Record them.
8198 (AT_loc_list_ptr): Handle loc and view lists.
8199 (view_list_to_loc_list_val_node): New.
8200 (print_dw_val): Handle dw_val_class_view_list.
8201 (size_of_die): Likewise.
8202 (value_format): Likewise.
8203 (loc_list_has_views): New.
8204 (gen_llsym): Set vl_symbol too.
8205 (maybe_gen_llsym, skip_loc_list_entry): New.
8206 (dwarf2out_maybe_output_loclist_view_pair): New.
8207 (output_loc_list): Output view list or entries too.
8208 (output_view_list_offset): New.
8209 (output_die): Handle dw_val_class_view_list.
8210 (output_dwarf_version): New.
8211 (output_compilation_unit_header): Use it.
8212 (output_skeleton_debug_sections): Likewise.
8213 (output_rnglists, output_line_info): Likewise.
8214 (output_pubnames, output_aranges): Update version comments.
8215 (output_one_line_info_table): Output view numbers in asm comments.
8216 (dw_loc_list): Determine current endview, pass it to new_loc_list.
8217 Call maybe_gen_llsym.
8218 (loc_list_from_tree_1): Adjust.
8219 (add_AT_location_description): Create view list attribute if
8220 needed, check it's absent otherwise.
8221 (convert_cfa_to_fb_loc_list): Adjust.
8222 (maybe_emit_file): Call output_asm_line_debug_info for test.
8223 (dwarf2out_var_location): Reset views as needed. Precompute
8224 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
8225 attribute. Set view.
8226 (new_line_info_table): Reset next view.
8227 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
8228 (dwarf2out_source_line): Likewise. Output view resets and labels to
8229 the assembler, or select appropriate line info opcodes.
8230 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
8231 (optimize_string_length): Catch it. Adjust.
8232 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
8233 dw_val_class_view_list, and remove it if no longer needed.
8234 (hash_loc_list): Hash view numbers.
8235 (loc_list_hasher::equal): Compare them.
8236 (optimize_location_lists): Check whether a view list symbol is
8237 needed, and whether the locview attribute is present, and
8238 whether they match. Remove the locview attribute if no longer
8240 (index_location_lists): Call skip_loc_list_entry for test.
8241 (dwarf2out_finish): Call output_asm_line_debug_info for test.
8242 Use output_dwarf_version.
8243 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
8244 (struct dw_val_node): Add val_view_list.
8245 * final.c (SEEN_NEXT_VIEW): New.
8246 (set_next_view_needed): New.
8247 (clear_next_view_needed): New.
8248 (maybe_output_next_view): New.
8249 (final_start_function): Rename to...
8250 (final_start_function_1): ... this. Take pointer to FIRST,
8251 add SEEN parameter. Emit param bindings in the initial view.
8252 (final_start_function): Reintroduce SEEN-less interface.
8253 (final): Rename to...
8254 (final_1): ... this. Take SEEN parameter. Output final pending
8255 next view at the end.
8256 (final): Reintroduce seen-less interface.
8257 (final_scan_insn): Output pending next view before switching
8258 sections or ending a block. Mark the next view as needed when
8259 outputting variable locations. Notify debug backend of section
8260 changes, and of location view changes.
8261 (rest_of_handle_final): Adjust.
8262 * toplev.c (process_options): Autodetect value for debug variable
8263 location views option. Warn on incompat5 without -gdwarf-5.
8264 * doc/invoke.texi (gvariable-location-views): New.
8265 (gvariable-location-views=incompat5): New.
8266 (gno-variable-location-views): New.
8268 2018-02-08 David Malcolm <dmalcolm@redhat.com>
8270 PR tree-optimization/84136
8271 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
8272 that the result of find_edge is non-NULL.
8274 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8277 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
8278 storing integer register in SImode. Fix cost of 256 and 512
8279 byte aligned SSE register store.
8281 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8283 * config/i386/i386.c (ix86_multiplication_cost): Fix
8284 multiplication cost for TARGET_AVX512DQ.
8286 2018-02-08 Marek Polacek <polacek@redhat.com>
8288 PR tree-optimization/84238
8289 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
8292 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8294 PR tree-optimization/84265
8295 * tree-vect-stmts.c (vectorizable_store): Don't treat
8296 VMAT_CONTIGUOUS accesses as grouped.
8297 (vectorizable_load): Likewise.
8299 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8301 PR tree-optimization/81635
8302 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
8303 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
8304 (test_round_for_mask): New functions.
8305 (wide_int_cc_tests): Call test_round_for_mask.
8306 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
8307 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
8308 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
8309 range returned by get_range_info.
8311 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8314 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
8315 * symtab.c: Include builtins.h
8316 (symtab_node::output_to_lto_symbol_table_p): Move here
8317 from lto-streamer-out.c:output_symbol_p.
8318 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
8319 (output_symbol_p): Move all logic to symtab.c
8320 (produce_symtab): Update.
8322 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8324 * config/s390/s390-opts.h (enum indirect_branch): Define.
8325 * config/s390/s390-protos.h (s390_return_addr_from_memory)
8326 (s390_indirect_branch_via_thunk)
8327 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
8328 (enum s390_indirect_branch_type): Define.
8329 * config/s390/s390.c (struct s390_frame_layout, struct
8330 machine_function): Remove.
8331 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
8332 (indirect_branch_table_label_no, indirect_branch_table_name):
8334 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
8335 (enum s390_indirect_branch_option): Define.
8336 (s390_return_addr_from_memory): New function.
8337 (s390_handle_string_attribute): New function.
8338 (s390_attribute_table): Add new attribute handler.
8339 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
8340 (s390_indirect_branch_via_thunk): New function.
8341 (s390_indirect_branch_via_inline_thunk): New function.
8342 (s390_function_ok_for_sibcall): When jumping via thunk disallow
8343 sibling call optimization for non z10 compiles.
8344 (s390_emit_call): Force indirect branch target to be a single
8345 register. Add r1 clobber for non-z10 compiles.
8346 (s390_emit_epilogue): Emit return jump via return_use expander.
8347 (s390_reorg): Handle JUMP_INSNs as execute targets.
8348 (s390_option_override_internal): Perform validity checks for the
8349 new command line options.
8350 (s390_indirect_branch_attrvalue): New function.
8351 (s390_indirect_branch_settings): New function.
8352 (s390_set_current_function): Invoke s390_indirect_branch_settings.
8353 (s390_output_indirect_thunk_function): New function.
8354 (s390_code_end): Implement target hook.
8355 (s390_case_values_threshold): Implement target hook.
8356 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
8358 * config/s390/s390.h (struct s390_frame_layout)
8359 (struct machine_function): Move here from s390.c.
8360 (TARGET_INDIRECT_BRANCH_NOBP_RET)
8361 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
8362 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
8363 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
8364 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
8365 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
8366 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
8367 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
8368 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
8369 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
8370 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
8371 (mnemonic attribute): Add values which aren't recognized
8373 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
8374 pattern for branch conversion. Fix mnemonic attribute.
8375 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
8376 indirect branch via thunk if requested.
8377 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
8378 ("*indirect_jump"): Disable for branch conversion using out of
8380 ("indirect_jump_via_thunk<mode>_z10")
8381 ("indirect_jump_via_thunk<mode>")
8382 ("indirect_jump_via_inlinethunk<mode>_z10")
8383 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
8384 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
8385 ("casesi_jump_via_inlinethunk<mode>_z10")
8386 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
8387 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
8388 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
8389 ("*indirect2_jump"): Disable for branch conversion.
8390 ("casesi_jump"): Turn into expander and expand patterns for branch
8392 ("return_use"): New expander.
8393 ("*return"): Emit return via thunk and rename it to ...
8394 ("*return<mode>"): ... this one.
8395 * config/s390/s390.opt: Add new options and and enum for the
8398 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8400 * lra-constraints.c (match_reload): Unconditionally use
8401 gen_lowpart_SUBREG, rather than selecting between that
8402 and equivalent gen_rtx_SUBREG code.
8404 2018-02-08 Richard Biener <rguenther@suse.de>
8406 PR tree-optimization/84233
8407 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
8408 changed flag instead of boguously re-using phi_inserted.
8410 2018-02-08 Martin Jambor <mjambor@suse.cz>
8412 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
8413 static local variables.
8415 2018-02-08 Richard Biener <rguenther@suse.de>
8417 PR tree-optimization/84278
8418 * tree-vect-stmts.c (vectorizable_store): When looking for
8419 smaller vector types to perform grouped strided loads/stores
8420 make sure the mode is supported by the target.
8421 (vectorizable_load): Likewise.
8423 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8425 * config/aarch64/aarch64.c (aarch64_components_for_bb):
8426 Increase LDP/STP opportunities by adding adjacent callee-saves.
8428 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8430 PR rtl-optimization/84068
8431 PR rtl-optimization/83459
8432 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
8434 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
8436 PR tree-optimization/84224
8437 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
8438 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
8441 2018-02-07 Iain Sandoe <iain@codesourcery.com>
8444 * config/rs6000/altivec.md (*restore_world): Remove LR use.
8445 * config/rs6000/predicates.md (restore_world_operation): Adjust op
8446 count, remove one USE.
8448 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
8450 * doc/install.texi (Configuration): Document the
8451 --with-long-double-format={ibm,ieee} PowerPC configuration
8455 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
8456 Convert from define_expand to be define_insn_and_split. Rework
8457 float/double/_Float128 conversions to QI/HI/SImode to work with
8458 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
8459 conversions to QI/HImode types did a store and then a load to
8460 truncate the value. For conversions to VSX registers, don't split
8461 the insn, instead emit the code directly. Use the code iterator
8462 any_fix to combine signed and unsigned conversions.
8463 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
8464 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
8465 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8466 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
8467 (fix_<mode>di2_hw): Likewise.
8468 (fixuns_<mode>di2_hw): Likewise.
8469 (fix_<mode>si2_hw): Likewise.
8470 (fixuns_<mode>si2_hw): Likewise.
8471 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
8472 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8473 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
8474 fix<uns>_trunc<SFDF:mode>si2_p8.
8475 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
8477 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
8478 (fix<uns>_<mode>_mem): Likewise.
8479 (fctiw<u>z_<mode>_mem): Likewise.
8480 (fix<uns>_<mode>_mem): Likewise.
8481 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
8482 the register allocator from doing a direct move to the GPRs to do
8483 a store, and instead use the ISA 3.0 store byte/half-word from
8484 vector register instruction. For IEEE 128-bit floating point,
8485 also optimize stores of 32-bit ints.
8486 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
8488 2018-02-07 Alan Hayward <alan.hayward@arm.com>
8490 * genextract.c (push_pathstr_operand): New function to support
8492 (walk_rtx): Call push_pathstr_operand.
8493 (print_path): Support [a-zA-Z].
8495 2018-02-07 Richard Biener <rguenther@suse.de>
8497 PR tree-optimization/84037
8498 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
8499 (cse_and_gimplify_to_preheader): Declare.
8500 (vect_get_place_in_interleaving_chain): Likewise.
8501 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8503 (_loop_vec_info::~_loop_vec_info): Delete it.
8504 (cse_and_gimplify_to_preheader): New function.
8505 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
8506 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
8507 (vectorizable_load): Likewise. For grouped stores always base
8508 the IV on the first element.
8509 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
8510 condition before gimplifying.
8512 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8514 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
8515 *DIV_EXPR and *MOD_EXPR.
8517 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
8520 * config/i386/i386.c (ix86_option_override_internal): Mask out
8521 the CF_SET bit when checking -fcf-protection.
8523 2018-02-07 Tom de Vries <tom@codesourcery.com>
8526 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
8529 2018-02-07 Richard Biener <rguenther@suse.de>
8531 PR tree-optimization/84204
8532 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
8535 PR tree-optimization/84205
8536 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
8537 special-case isl_ast_op_zdiv_r.
8539 PR tree-optimization/84223
8540 * graphite-scop-detection.c (gather_bbs::before_dom_children):
8541 Only add conditions from within the region.
8542 (gather_bbs::after_dom_children): Adjust.
8544 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
8547 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
8548 * config/avr/avr.md: Only post-reload split REG-REG moves if
8549 either register is GENERAL_REG_P.
8551 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8553 PR tree-optimization/84235
8554 * tree-ssa-scopedtables.c
8555 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
8556 if the subtraction is performed in floating point type where NaNs are
8557 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
8558 build 1. Formatting fix.
8560 2018-02-06 Jakub Jelinek <jakub@redhat.com>
8563 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
8564 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
8565 and skip it regardless of bb boundaries. Use CALL_P macro,
8566 don't test INSN_P (insn) together with CALL_P or JUMP_P check
8567 unnecessarily, formatting fix.
8569 2018-02-06 Michael Collison <michael.collison@arm.com>
8571 * config/arm/thumb2.md:
8572 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
8573 (*thumb_mov_notscc): Ditto.
8575 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
8578 * config/rs6000/rs6000.md (su code attribute): Use "u" for
8579 unsigned_fix, not "s".
8581 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8583 * configure.ac (gcc_fn_eh_frame_ro): New function.
8584 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
8585 correct .eh_frame permissions.
8586 * configure: Regenerate.
8588 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
8590 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
8593 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8595 * config/rs6000/rs6000.c (rs6000_option_override_internal):
8596 Display warning message for -mno-speculate-indirect-jumps.
8598 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
8600 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
8602 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
8604 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
8606 PR tree-optimization/84225
8607 * tree-eh.c (find_trapping_overflow): Only call
8608 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
8610 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8613 * config/i386/i386.c: Reimplement the check of possible options
8614 -mibt/-mshstk conbination. Change error messages.
8615 * doc/invoke.texi: Fix a typo: remove extra '='.
8617 2018-02-06 Marek Polacek <polacek@redhat.com>
8619 PR tree-optimization/84228
8620 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
8622 2018-02-06 Tamar Christina <tamar.christina@arm.com>
8625 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
8626 emitted arch directives.
8627 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
8628 __ARM_FEATURE_COPROC before changing architectures.
8630 2018-02-06 Richard Biener <rguenther@suse.de>
8632 * config/i386/i386.c (print_reg): Fix typo.
8633 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
8635 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
8637 * configure: Regenerate.
8639 2018-02-05 Martin Sebor <msebor@redhat.com>
8641 PR tree-optimization/83369
8642 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
8645 2018-02-05 Martin Liska <mliska@suse.cz>
8647 * doc/invoke.texi: Cherry-pick upstream r323995.
8649 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
8651 * ira.c (ira_init_register_move_cost): Adjust comment.
8653 2018-02-05 Martin Liska <mliska@suse.cz>
8655 PR gcov-profile/84137
8656 * doc/gcov.texi: Fix typo in documentation.
8658 2018-02-05 Martin Liska <mliska@suse.cz>
8660 PR gcov-profile/83879
8661 * doc/gcov.texi: Document necessity of --dynamic-list-data when
8662 using dlopen functionality.
8664 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
8666 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
8667 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
8668 _mm_maskz_range_ss, _mm_mask_range_round_ss,
8669 _mm_maskz_range_round_ss): New intrinsics.
8670 (__builtin_ia32_rangesd128_round)
8671 (__builtin_ia32_rangess128_round): Remove.
8672 (__builtin_ia32_rangesd128_mask_round,
8673 __builtin_ia32_rangess128_mask_round): New builtins.
8674 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
8675 __builtin_ia32_rangess128_round): Remove.
8676 (__builtin_ia32_rangesd128_mask_round,
8677 __builtin_ia32_rangess128_mask_round): New builtins.
8678 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
8679 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
8680 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
8681 "<round_saeonly_constraint>")): Changed to ...
8682 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
8683 "<round_saeonly_scalar_constraint>")): ... this.
8684 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
8685 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
8686 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
8687 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
8688 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
8690 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
8692 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
8694 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
8695 Remove all values except native, 8540 and 8548.
8697 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
8699 * config/i386/i386.c (ix86_output_function_return): Pass
8700 INVALID_REGNUM, instead of -1, as invalid register number to
8701 indirect_thunk_name and output_indirect_thunk.
8703 2018-02-02 Julia Koval <julia.koval@intel.com>
8705 * config.gcc: Add -march=icelake.
8706 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
8707 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
8708 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
8709 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
8710 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
8711 (processor_target_table): Add icelake.
8712 (ix86_option_override_internal): Handle new PTAs.
8713 (get_builtin_code_for_version): Handle icelake.
8714 (M_INTEL_COREI7_ICELAKE): New.
8715 (fold_builtin_cpu): Handle icelake.
8716 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
8717 * doc/invoke.texi: Add -march=icelake.
8719 2018-02-02 Julia Koval <julia.koval@intel.com>
8721 * config/i386/i386.c (ix86_option_override_internal): Change flags type
8722 to wide_int_bitmask.
8723 * wide-int-bitmask.h: New.
8725 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8728 * config/i386/i386.md: Replace Pmode with word_mode in
8729 builtin_setjmp_setup and builtin_longjmp to support x32.
8731 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
8735 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
8737 (rs6000_supported_cpu_names): New static variable.
8738 (linux_cpu_translation_table): Likewise.
8739 (elf_platform) <cpu>: Define new static variable and use it.
8740 Translate kernel AT_PLATFORM name to canonical name if needed.
8741 Error if platform name is unknown.
8743 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
8746 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
8748 2018-02-01 Jeff Law <law@redhat.com>
8751 * config/i386/i386.c (release_scratch_register_on_entry): Add new
8752 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
8753 the scratch if RELEASE_VIA_POP is false.
8754 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
8755 If we have to save a temporary register, decrement SIZE appropriately.
8756 Pass new arguments to release_scratch_register_on_entry.
8757 (ix86_adjust_stack_and_probe): Likewise.
8758 (ix86_emit_probe_stack_range): Pass new arguments to
8759 release_scratch_register_on_entry.
8761 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
8763 PR rtl-optimization/84157
8764 * combine.c (change_zero_ext): Use REG_P predicate in
8765 front of HARD_REGISTER_P predicate.
8767 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
8769 * config/avr/avr.c (avr_option_override): Move disabling of
8770 -fdelete-null-pointer-checks to...
8771 * common/config/avr/avr-common.c (avr_option_optimization_table):
8774 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8776 PR tree-optimization/81635
8777 * tree-data-ref.c (split_constant_offset_1): For types that
8778 wrap on overflow, try to use range info to prove that wrapping
8781 2018-02-01 Renlin Li <renlin.li@arm.com>
8784 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
8786 (aarch64_register_move_cost): Likewise.
8787 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
8789 (REG_CLASS_NAMES): Likewise.
8790 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
8791 TAILCALL_ADDR_REGS. Remove IP registers.
8792 * config/aarch64/aarch64.md (Ucs): Update register constraint.
8794 2018-02-01 Richard Biener <rguenther@suse.de>
8796 * domwalk.h (dom_walker::dom_walker): Add additional constructor
8797 for specifying RPO order and allow NULL for that.
8798 * domwalk.c (dom_walker::dom_walker): Likewise.
8799 (dom_walker::walk): Handle NULL RPO order.
8800 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
8802 (rewrite_update_dom_walker): Likewise.
8803 (mark_def_dom_walker): Likewise.
8805 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8807 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
8808 (aarch64_maybe_expand_sve_subreg_move): Declare.
8809 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
8810 * config/aarch64/predicates.md (aarch64_any_register_operand): New
8812 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
8813 that are semantically a reverse operation.
8814 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
8815 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
8816 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
8818 (aarch64_can_change_mode_class): For big-endian, forbid changes
8819 between two SVE modes if they have different element sizes.
8821 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8823 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
8824 the TImode handling for big-endian targets.
8826 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8828 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
8829 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
8831 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
8832 Remove BSWAP handing for big-endian targets and use the form of
8833 LD1RQ appropariate for the mode.
8835 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8837 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
8838 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
8841 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8844 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
8845 check for operands that need to go through aarch64_sve_reload_be.
8847 2018-02-01 Jakub Jelinek <jakub@redhat.com>
8849 PR tree-optimization/81661
8850 PR tree-optimization/84117
8851 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
8852 * tree-eh.c: Include gimplify.h.
8853 (find_trapping_overflow, replace_trapping_overflow,
8854 rewrite_to_non_trapping_overflow): New functions.
8855 * tree-vect-loop.c: Include tree-eh.h.
8856 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
8857 * tree-data-ref.c: Include tree-eh.h.
8858 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
8860 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
8862 PR rtl-optimization/84123
8863 * combine.c (change_zero_ext): Check if hard register satisfies
8864 can_change_dest_mode before calling gen_lowpart_SUBREG.
8866 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
8869 * ira.c (ira_init_register_move_cost): Remove assert.
8871 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8873 PR rtl-optimization/84071
8874 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
8875 * doc/tm.texi: Regenerate.
8877 2018-01-31 Richard Biener <rguenther@suse.de>
8879 PR tree-optimization/84132
8880 * tree-data-ref.c (analyze_miv_subscript): Properly
8881 check whether evolution_function_is_affine_multivariate_p
8882 before calling gcd_of_steps_may_divide_p.
8884 2018-01-31 Julia Koval <julia.koval@intel.com>
8887 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
8888 * config/i386/i386.md (rdpid_rex64) New.
8889 (rdpid): Make 32bit only.
8891 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
8894 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
8895 an IDENTIFIER_NODE for FUNCTION_TYPE's.
8897 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8900 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8902 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8904 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8906 PR rtl-optimization/84071
8907 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
8908 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
8910 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8912 * config/arc/arc.c (arc_handle_aux_attribute): New function.
8913 (arc_attribute_table): Add 'aux' attribute.
8914 (arc_in_small_data_p): Consider aux like variables.
8915 (arc_is_aux_reg_p): New function.
8916 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
8917 (arc_get_aux_arg): New function.
8918 (prepare_move_operands): Handle aux-register access.
8919 (arc_handle_aux_attribute): New function.
8920 * doc/extend.texi (ARC Variable attributes): Add subsection.
8922 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8924 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
8925 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
8926 (arc_attribute_table): Add 'uncached' attribute.
8927 (arc_print_operand): Print '.di' flag for uncached memory
8929 (arc_in_small_data_p): Do not consider for small data the uncached
8931 (arc_is_uncached_mem_p): New function.
8932 * config/arc/predicates.md (compact_store_memory_operand): Check
8933 for uncached memory accesses.
8934 (nonvol_nonimm_operand): Likewise.
8935 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
8937 2018-01-31 Jakub Jelinek <jakub@redhat.com>
8940 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
8941 falign-loops=): Add Optimization flag.
8943 2018-01-30 Jeff Law <law@redhat.com>
8946 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
8947 INT_REGISTERS_SAVED. Check it prior to calling
8948 get_scratch_register_on_entry.
8949 (ix86_adjust_stack_and_probe): Similarly.
8950 (ix86_emit_probe_stack_range): Similarly.
8951 (ix86_expand_prologue): Corresponding changes.
8953 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8956 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
8957 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
8959 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
8962 * lra-constraints.c (curr_insn_transform): Process AND in the
8965 2018-01-30 Jakub Jelinek <jakub@redhat.com>
8967 PR rtl-optimization/83986
8968 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
8969 dependence against last_pending_memory_flush in addition to
8972 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
8974 PR tree-optimization/81611
8975 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
8978 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8981 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
8984 2018-01-30 Richard Biener <rguenther@suse.de>
8985 Jakub Jelinek <jakub@redhat.com>
8987 PR tree-optimization/84111
8988 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
8989 inner loops added during recursion, as they don't have up-to-date
8992 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8995 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
8996 (can_inline_edge_by_limits_p): ... here.
8997 (can_early_inline_edge_p, check_callers,
8998 update_caller_keys, update_callee_keys, recursive_inlining,
8999 add_new_edges_to_heap, speculation_useful_p,
9000 inline_small_functions,
9001 inline_small_functions, flatten_function,
9002 inline_to_all_callers_1): Update.
9004 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9006 * profile-count.c (profile_count::combine_with_ipa_count): Handle
9009 2018-01-30 Richard Biener <rguenther@suse.de>
9011 PR tree-optimization/83008
9012 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
9013 invariant and constant vector uses in stmts when they need
9016 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9019 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
9020 * configure: Regenerate.
9022 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9024 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
9026 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
9027 Use gen_rtx_REG rather than gen_lowpart.
9029 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9031 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
9032 rather than 0 when creating partial subregs.
9034 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9036 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
9039 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
9042 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
9043 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
9044 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
9045 flags. This restores the settings used before the 2017-07-24.
9046 Turning off pre increment/decrement/modify allows IVOPTS to
9047 optimize DF/SF loops where the index is an int.
9049 2018-01-29 Richard Biener <rguenther@suse.de>
9050 Kelvin Nilsen <kelvin@gcc.gnu.org>
9053 * tree-vect-stmts.c (vectorizable_call): Don't call
9054 targetm.vectorize_builtin_md_vectorized_function if callee is
9057 2018-01-22 Carl Love <cel@us.ibm.com>
9059 * doc/extend.tex: Fix typo in second arg in
9060 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
9062 2018-01-29 Richard Biener <rguenther@suse.de>
9064 PR tree-optimization/84086
9065 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
9066 (flush_ssaname_freelist): When SSA names were released reset
9067 the SCEV hash table.
9069 2018-01-29 Richard Biener <rguenther@suse.de>
9071 PR tree-optimization/84057
9072 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
9073 removed paths when removing edges.
9075 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
9077 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
9078 -mfunction-return=@var{choice}.
9080 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
9083 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
9085 (layout::print_source_line): Likewise.
9086 (test_get_line_width_without_trailing_whitespace): Add test cases.
9088 2018-01-27 Jakub Jelinek <jakub@redhat.com>
9091 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
9094 2018-01-26 Jim Wilson <jimw@sifive.com>
9096 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
9098 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
9101 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9103 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
9104 and CMP + SUB-immediate -> SUBS.
9106 2018-01-26 Martin Sebor <msebor@redhat.com>
9108 PR tree-optimization/83896
9109 * tree-ssa-strlen.c (get_string_len): Rename...
9110 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
9111 Avoid assuming length is constant.
9112 (handle_char_store): Use HOST_WIDE_INT for string length.
9114 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
9117 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
9118 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
9120 2018-01-26 Richard Biener <rguenther@suse.de>
9122 PR rtl-optimization/84003
9123 * dse.c (record_store): Only record redundant stores when
9124 the earlier store aliases at least all accesses the later one does.
9126 2018-01-26 Jakub Jelinek <jakub@redhat.com>
9128 PR rtl-optimization/83985
9129 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
9130 REG_CFA_RESTORE insns.
9131 (delete_unmarked_insns): Don't ignore separate shrink wrapping
9132 REG_CFA_RESTORE insns here.
9135 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
9136 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
9138 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9140 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
9141 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
9142 (arc_init): Likewise.
9143 (arc_override_options): Likewise.
9144 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
9146 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
9148 * config/arc/arc.h (TARGET_DBNZ): Define.
9149 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
9150 properly set the tune attribute.
9151 (dbnz): Use TARGET_DBNZ guard.
9152 * config/arc/arc.opt (mtune): Add core3 option.
9154 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9156 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
9157 recognize new pic like addresses.
9158 (arc_delegitimize_address): Clean up.
9160 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9162 * config/arc/arc-arches.def: Option mrf16 valid for all
9164 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
9165 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
9166 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
9167 * config/arc/arc-tables.opt: Regenerate.
9168 * config/arc/arc.c (arc_conditional_register_usage): Handle
9169 reduced register file case.
9170 (arc_file_start): Set must have build attributes.
9171 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
9173 * config/arc/arc.opt (mrf16): Add new option.
9174 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
9175 * config/arc/genmultilib.awk: Handle new mrf16 option.
9176 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
9177 * config/arc/t-multilib: Regenerate.
9178 * doc/invoke.texi (ARC Options): Document mrf16 option.
9180 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9182 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
9183 * config/arc/arc.c (arc_handle_secure_attribute): New function.
9184 (arc_attribute_table): Add 'secure_call' attribute.
9185 (arc_print_operand): Print secure call operand.
9186 (arc_function_ok_for_sibcall): Don't optimize tail calls when
9188 (arc_is_secure_call_p): New function. * config/arc/arc.md
9189 (call_i): Add support for sjli instruction.
9190 (call_value_i): Likewise.
9191 * config/arc/constraints.md (Csc): New constraint.
9193 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9194 John Eric Martin <John.Martin@emmicro-us.com>
9196 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
9197 * config/arc/arc.c (_arc_jli_section): New struct.
9198 (arc_jli_section): New type.
9199 (rc_jli_sections): New static variable.
9200 (arc_handle_jli_attribute): New function.
9201 (arc_attribute_table): Add jli_always and jli_fixed attribute.
9202 (arc_file_end): New function.
9203 (TARGET_ASM_FILE_END): Define.
9204 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
9205 (arc_add_jli_section): New function.
9206 (jli_call_scan): Likewise.
9207 (arc_reorg): Call jli_call_scan.
9208 (arc_output_addsi): Remove 'S' from printing asm operand.
9209 (arc_is_jli_call_p): New function.
9210 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
9212 (movhi_insn): Likewise.
9213 (movsi_insn): Likewise.
9214 (movsi_set_cc_insn): Likewise.
9215 (loadqi_update): Likewise.
9216 (load_zeroextendqisi_update): Likewise.
9217 (load_signextendqisi_update): Likewise.
9218 (loadhi_update): Likewise.
9219 (load_zeroextendhisi_update): Likewise.
9220 (load_signextendhisi_update): Likewise.
9221 (loadsi_update): Likewise.
9222 (loadsf_update): Likewise.
9223 (movsicc_insn): Likewise.
9224 (bset_insn): Likewise.
9225 (bxor_insn): Likewise.
9226 (bclr_insn): Likewise.
9227 (bmsk_insn): Likewise.
9228 (bicsi3_insn): Likewise.
9229 (cmpsi_cc_c_insn): Likewise.
9230 (movsi_ne): Likewise.
9231 (movsi_cond_exec): Likewise.
9232 (clrsbsi2): Likewise.
9243 (call_i): Remove 'S' asm letter, add jli instruction.
9244 (call_value_i): Likewise.
9245 * config/arc/arc.op (mjli-always): New option.
9246 * config/arc/constraints.md (Cji): New constraint.
9247 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
9249 (subsf3_fpx): Likewise.
9250 (mulsf3_fpx): Likewise.
9251 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
9253 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
9255 * doc/invoke.texi (ARC): Document mjli-always option.
9257 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
9259 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
9260 avoid addition with 0 and use incw and decw where possible.
9262 2018-01-26 Richard Biener <rguenther@suse.de>
9264 PR tree-optimization/81082
9265 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
9266 association if it requires casting to unsigned.
9267 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
9268 from fold_plusminus_mult_expr to catch important cases late when
9269 range info is available.
9271 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9273 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
9274 * configure.ac (hidden_linkonce): New test.
9275 * configure: Regenerate.
9276 * config.in: Regenerate.
9278 2018-01-26 Julia Koval <julia.koval@intel.com>
9280 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
9281 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
9282 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
9283 _mm_mask_bitshuffle_epi64_mask): Fix type.
9284 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
9285 USI_FTYPE_V4DI_V4DI_USI): Remove.
9286 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
9287 __builtin_ia32_vpshufbitqmb256_mask,
9288 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
9289 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
9290 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
9292 2018-01-26 Alan Modra <amodra@gmail.com>
9295 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
9296 UNSPEC_VBPERMQ. Sort other unspecs.
9298 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
9300 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
9302 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
9305 * predict.c (drop_profile): Do not push/pop cfun; update also
9307 (handle_missing_profiles): Fix logic looking for zero profiles.
9309 2018-01-25 Jakub Jelinek <jakub@redhat.com>
9312 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
9313 on functions with #pragma omp declare simd or functions with simd
9315 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
9316 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
9317 Remove trailing \n from warning_at calls.
9319 2018-01-25 Tom de Vries <tom@codesourcery.com>
9322 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9323 for neutered workers.
9325 2018-01-24 Joseph Myers <joseph@codesourcery.com>
9328 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
9329 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
9331 2018-01-24 Jeff Law <law@redhat.com>
9334 * i386.c (get_probe_interval): Move to earlier point.
9335 (ix86_compute_frame_layout): If -fstack-clash-protection and
9336 the frame is larger than the probe interval, then use pushes
9337 to save registers rather than reg->mem moves.
9338 (ix86_expand_prologue): Remove conditional for int_registers_saved
9341 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
9344 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
9345 min/max for never referenced object.
9347 2018-01-24 Jakub Jelinek <jakub@redhat.com>
9350 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
9352 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
9353 attributes from DECL_ATTRIBUTES (decl) without affecting
9354 DECL_ATTRIBUTES (current_function_decl).
9355 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
9356 functions with non-NULL DECL_ABSTRACT_ORIGIN.
9358 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
9360 PR tree-optimization/83979
9361 * fold-const.c (fold_comparison): Use constant_boolean_node
9362 instead of boolean_{true,false}_node.
9364 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
9366 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
9369 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9371 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9372 Simplify the clause that sets the length attribute.
9373 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9374 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
9375 clause that sets the length attribute.
9376 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9378 2018-01-24 Tom de Vries <tom@codesourcery.com>
9381 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
9382 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
9383 Add strict parameter.
9384 (prevent_branch_around_nothing): Insert dummy insn between branch to
9385 label and label with no ptx insn inbetween.
9386 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
9388 2018-01-24 Tom de Vries <tom@codesourcery.com>
9391 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9392 for neutered threads in warp.
9393 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
9395 2018-01-24 Richard Biener <rguenther@suse.de>
9397 PR tree-optimization/83176
9398 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
9401 2018-01-24 Richard Biener <rguenther@suse.de>
9403 PR tree-optimization/82819
9404 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
9405 code generating pluses that are no-ops in the target precision.
9407 2018-01-24 Richard Biener <rguenther@suse.de>
9410 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
9412 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9414 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
9415 to merge probabilities.
9416 * predict.c (probably_never_executed): Also mark as cold functions
9417 with global 0 profile and guessed local profile.
9418 * profile-count.c (profile_probability::combine_with_count): New
9420 * profile-count.h (profile_probability::operator*,
9421 profile_probability::operator*=, profile_probability::operator/,
9422 profile_probability::operator/=): Reduce precision to adjusted
9423 and set value to guessed on contradictory divisions.
9424 (profile_probability::combine_with_freq): Remove.
9425 (profile_probability::combine_wiht_count): Declare.
9426 (profile_count::force_nonzero):: Set to adjusted.
9427 (profile_count::probability_in):: Set quality to adjusted.
9428 * tree-ssa-tail-merge.c (replace_block_by): Use
9431 2018-01-23 Andrew Waterman <andrew@sifive.com>
9432 Jim Wilson <jimw@sifive.com>
9434 * config/riscv/riscv.c (riscv_stack_boundary): New.
9435 (riscv_option_override): Set riscv_stack_boundary. Handle
9436 riscv_preferred_stack_boundary_arg.
9437 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
9438 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
9439 (STACK_BOUNDARY): Set to riscv_stack_boundary.
9440 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
9441 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
9442 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
9444 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
9447 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
9448 of struct ix86_frame.
9449 (ix86_expand_epilogue): Likewise. Add a local variable for
9450 the reg_save_offset field in struct ix86_frame.
9452 2018-01-23 Bin Cheng <bin.cheng@arm.com>
9454 PR tree-optimization/82604
9455 * tree-loop-distribution.c (enum partition_kind): New enum item
9456 PKIND_PARTIAL_MEMSET.
9457 (partition_builtin_p): Support above new enum item.
9458 (generate_code_for_partition): Ditto.
9459 (compute_access_range): Differentiate cases that equality can be
9460 proven at all loops, the innermost loops or no loops.
9461 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
9462 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
9463 (finalize_partitions, distribute_loop): Don't fuse partition of
9464 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
9465 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
9468 2018-01-23 Martin Liska <mliska@suse.cz>
9470 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
9471 order to ignore the predictor.
9472 (PRED_POLYMORPHIC_CALL): Likewise.
9473 (PRED_RECURSIVE_CALL): Likewise.
9475 2018-01-23 Martin Liska <mliska@suse.cz>
9477 * tree-profile.c (tree_profiling): Print function header to
9478 aware reader which function we are working on.
9479 * value-prof.c (gimple_find_values_to_profile): Do not print
9480 not interesting value histograms.
9482 2018-01-23 Martin Liska <mliska@suse.cz>
9484 * profile-count.h (enum profile_quality): Add
9485 profile_uninitialized as the first value. Do not number values
9486 as they are zero based.
9487 (profile_count::verify): Update sanity check.
9488 (profile_probability::verify): Likewise.
9490 2018-01-23 Nathan Sidwell <nathan@acm.org>
9492 * doc/invoke.texi (ffor-scope): Deprecate.
9494 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9496 PR tree-optimization/83510
9497 * domwalk.c (set_all_edges_as_executable): New function.
9498 (dom_walker::dom_walker): Convert bool param
9499 "skip_unreachable_blocks" to enum reachability. Move setup of
9500 edge flags to set_all_edges_as_executable and only do it when
9501 reachability is REACHABLE_BLOCKS.
9502 * domwalk.h (enum dom_walker::reachability): New enum.
9503 (dom_walker::dom_walker): Convert bool param
9504 "skip_unreachable_blocks" to enum reachability.
9505 (set_all_edges_as_executable): New decl.
9506 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
9507 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
9509 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
9510 but converting true to REACHABLE_BLOCKS.
9511 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
9513 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
9514 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
9515 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
9517 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
9518 if check_all_array_refs will be called.
9520 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9522 * tree.c (selftest::test_location_wrappers): Add more test
9525 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9527 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
9528 (selftest::test_bit_in_range): Likewise.
9530 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9533 * doc/sourcebuild.texi (vect_float): Say that the selector
9534 only describes the situation when -funsafe-math-optimizations is on.
9535 (vect_float_strict): Document.
9537 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9539 PR tree-optimization/83965
9540 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
9541 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
9542 instead of checking only for a reduction.
9543 (vect_recog_widen_sum_pattern): Likewise.
9545 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9547 * predict.c (probably_never_executed): Only use precise profile info.
9548 (compute_function_frequency): Skip after inlining hack since we now
9549 have quality checking.
9551 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9553 * profile-count.h (profile_probability::very_unlikely,
9554 profile_probability::unlikely, profile_probability::even): Set
9555 precision to guessed.
9557 2018-01-23 Richard Biener <rguenther@suse.de>
9559 PR tree-optimization/83963
9560 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
9561 Properly terminate dominator walk when crossing the exit edge not
9562 when visiting its source block.
9564 2018-01-23 Jakub Jelinek <jakub@redhat.com>
9567 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
9568 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
9570 2018-01-22 Jakub Jelinek <jakub@redhat.com>
9572 PR tree-optimization/83957
9573 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
9574 semicolon after for body surrounded by braces.
9576 PR tree-optimization/83081
9577 * profile-count.h (profile_probability::split): New method.
9578 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
9579 Use profile_probability::split.
9580 (do_compare_rtx_and_jump): Fix adjustment of probabilities
9581 when splitting a single conditional jump into 2.
9583 2018-01-22 David Malcolm <dmalcolm@redhat.com>
9585 PR tree-optimization/69452
9586 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
9589 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9591 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
9592 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
9593 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
9595 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9597 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
9598 * config/rl78/rl78.md (movdi): New define_expand.
9599 * config/rl78/rl78.c (rl78_split_movdi): New function.
9601 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
9604 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
9606 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
9607 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
9608 128-bit to produce an UNSPEC move to get the double word with the
9609 signbit and then a shift directly to do signbit.
9610 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
9611 implementation with a new version that just does either a direct
9612 move or a regular move. Move memory interface to separate insns.
9613 Move insns so they are next to the expander.
9614 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
9615 with signbit move. Split big and little endian case.
9616 (signbit<mode>2_dm_mem_le): Likewise.
9617 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
9618 (signbit<mode>2_dm2): Likewise.
9620 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9622 * config/rl78/rl78.md (anddi3): New define_expand.
9624 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9626 * config/rl78/rl78.md (umindi3): New define_expand.
9628 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9630 * config/rl78/rl78.md (smindi3): New define_expand.
9632 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9634 * config/rl78/rl78.md (smaxdi3): New define_expand.
9636 2018-01-22 Carl Love <cel@us.ibm.com>
9638 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
9639 LVX_V1TI): Add macro expansion.
9640 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
9641 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
9642 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
9643 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
9644 Change check to determine if the instruction is a byte reversing
9645 entry. Fix typo in comment.
9646 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
9647 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
9648 Add def_builtin calls for new builtins.
9649 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
9650 Add define_insn expansion.
9652 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9654 * config/rl78/rl78.md (umaxdi3): New define_expand.
9656 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9658 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
9659 for non-QImode registers.
9661 2018-01-22 Richard Biener <rguenther@suse.de>
9663 PR tree-optimization/83963
9664 * graphite-scop-detection.c (scop_detection::get_sese): Delay
9665 including the loop exit block.
9666 (scop_detection::merge_sese): Likewise.
9667 (scop_detection::add_scop): Do it here instead.
9669 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9671 * doc/sourcebuild.texi (arm_softfloat): Document.
9673 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
9676 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
9677 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
9678 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
9680 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9681 David Edelsohn <dje.gcc@gmail.com>
9684 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9685 Change "crset eq" to "crset 2".
9686 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9687 (*call_indirect_aix<mode>_nospec): Likewise.
9688 (*call_value_indirect_aix<mode>_nospec): Likewise.
9689 (*call_indirect_elfv2<mode>_nospec): Likewise.
9690 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
9691 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
9692 change assembly output from . to $.
9693 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9694 (indirect_jump<mode>_nospec): Change assembly output from . to $.
9695 (*tablejump<mode>_internal1_nospec): Likewise.
9697 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
9700 * config/sh/sh_optimize_sett_clrt.cc:
9701 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
9703 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
9705 PR tree-optimization/83940
9706 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
9707 offset_dt to vect_constant_def rather than vect_unknown_def_type.
9708 (vect_check_load_store_mask): Add a mask_dt_out parameter and
9709 use it to pass back the definition type.
9710 (vect_check_store_rhs): Likewise rhs_dt_out.
9711 (vect_build_gather_load_calls): Add a mask_dt argument and use
9712 it instead of a call to vect_is_simple_use.
9713 (vectorizable_store): Update calls to vect_check_load_store_mask
9714 and vect_check_store_rhs. Use the dt returned by the latter instead
9715 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
9716 instead of calls to vect_is_simple_use. Pass the scalar rather
9717 than the vector operand to vect_is_simple_use when handling
9718 second and subsequent copies of an rhs value.
9719 (vectorizable_load): Update calls to vect_check_load_store_mask
9720 and vect_build_gather_load_calls. Use the cached mask_dt and
9721 gs_info.offset_dt instead of calls to vect_is_simple_use.
9723 2018-01-20 Jakub Jelinek <jakub@redhat.com>
9726 * tree-emutls.c: Include gimplify.h.
9727 (lower_emutls_2): New function.
9728 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
9729 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
9730 it before further processing.
9733 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
9734 UINTVAL (trueop1) instead of INTVAL (op1).
9736 2018-01-19 Jakub Jelinek <jakub@redhat.com>
9740 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
9741 INCOMING_FRAME_SP_OFFSET if not defined.
9742 (scan_trace): Add ENTRY argument. If true and
9743 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
9744 emit a note to adjust the CFA offset.
9745 (create_cfi_notes): Adjust scan_trace callers.
9746 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
9747 INCOMING_FRAME_SP_OFFSET in the CIE.
9748 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
9749 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
9751 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
9752 * doc/tm.texi: Regenerated.
9754 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9756 PR rtl-optimization/83147
9757 * lra-constraints.c (remove_inheritance_pseudos): Use
9758 lra_substitute_pseudo_within_insn.
9760 2018-01-19 Tom de Vries <tom@codesourcery.com>
9761 Cesar Philippidis <cesar@codesourcery.com>
9764 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
9766 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
9769 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
9770 spaces for function labels.
9772 2018-01-19 Martin Liska <mliska@suse.cz>
9774 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
9775 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
9776 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
9777 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
9778 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
9779 (PRED_CONST_RETURN): Change from 69 to 65.
9780 (PRED_NULL_RETURN): Change from 91 to 71.
9781 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
9782 (PRED_LOOP_GUARD): Change from 66 to 73.
9784 2018-01-19 Martin Liska <mliska@suse.cz>
9786 * predict.c (predict_insn_def): Add new assert.
9787 (struct branch_predictor): Change type to signed integer.
9788 (test_prediction_value_range): Amend test to cover
9790 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
9791 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
9792 (PRED_LOOP_ITERATIONS_MAX): Likewise.
9793 (PRED_LOOP_IV_COMPARE): Likewise.
9794 * predict.h (PROB_UNINITIALIZED): Define new constant.
9796 2018-01-19 Martin Liska <mliska@suse.cz>
9798 * predict.c (dump_prediction): Add new format for
9799 analyze_brprob.py script which is enabled with -details
9801 * profile-count.h (precise_p): New function.
9803 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9805 PR tree-optimization/83922
9806 * tree-vect-loop.c (vect_verify_full_masking): Return false if
9807 there are no statements that need masking.
9808 (vect_active_double_reduction_p): New function.
9809 (vect_analyze_loop_operations): Use it when handling phis that
9810 are not in the loop header.
9812 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9814 PR tree-optimization/83914
9815 * tree-vect-loop.c (vectorizable_induction): Don't convert
9816 init_expr or apply the peeling adjustment for inductions
9817 that are nested within the vectorized loop.
9819 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9821 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
9824 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9828 * function.h (gimplify_parameters): Add gimple_seq * argument.
9829 * function.c: Include gimple.h and options.h.
9830 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
9831 for the added local temporaries if needed.
9832 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
9833 if there are any parameter cleanups, wrap whole body into a
9834 try/finally with the cleanups.
9836 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
9839 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
9840 Use GET_MODE_CLASS for scalar floating point.
9842 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9846 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
9847 Fix call of call_cgraph_insertion_hooks.
9849 2018-01-18 Martin Sebor <msebor@redhat.com>
9851 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
9853 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9856 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
9859 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
9862 * common.opt: (-ffile-prefix-map): New option.
9863 * opts.c (common_handle_option): Defer it.
9864 * opts-global.c (handle_common_deferred_options): Handle it.
9865 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
9866 * file-prefix-map.h: New file.
9867 (remap_debug_filename, add_debug_prefix_map): ...here.
9868 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
9869 * final.c (debug_prefix_map, add_debug_prefix_map
9870 remap_debug_filename): Move to...
9871 * file-prefix-map.c: New file.
9872 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
9873 generalize, get rid of alloca(), use strrchr() instead of strchr().
9874 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
9875 Implement in terms of add_prefix_map().
9876 (remap_macro_filename, remap_debug_filename): Implement in term of
9878 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
9879 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
9880 * dbxout.c: Include file-prefix-map.h.
9881 * varasm.c: Likewise.
9882 * vmsdbgout.c: Likewise.
9883 * xcoffout.c: Likewise.
9884 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
9885 * doc/cppopts.texi (-fmacro-prefix-map): Document.
9886 * doc/invoke.texi (-ffile-prefix-map): Document.
9887 (-fdebug-prefix-map): Update description.
9889 2018-01-18 Martin Liska <mliska@suse.cz>
9891 * config/i386/i386.c (indirect_thunk_name): Document that also
9893 (output_indirect_thunk): Document why both instructions
9894 (pause and lfence) are generated.
9896 2018-01-18 Richard Biener <rguenther@suse.de>
9898 PR tree-optimization/83887
9899 * graphite-scop-detection.c
9900 (scop_detection::get_nearest_dom_with_single_entry): Remove.
9901 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
9902 (scop_detection::merge_sese): Re-implement with a flood-fill
9903 algorithm that properly finds a SESE region if it exists.
9905 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9908 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
9909 pointer_diff optimizations use view_convert instead of convert.
9911 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9913 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9914 Generate different code for -mno-speculate-indirect-jumps.
9915 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9916 (*call_indirect_aix<mode>): Disable for
9917 -mno-speculate-indirect-jumps.
9918 (*call_indirect_aix<mode>_nospec): New define_insn.
9919 (*call_value_indirect_aix<mode>): Disable for
9920 -mno-speculate-indirect-jumps.
9921 (*call_value_indirect_aix<mode>_nospec): New define_insn.
9922 (*sibcall_nonlocal_sysv<mode>): Generate different code for
9923 -mno-speculate-indirect-jumps.
9924 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9926 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
9928 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
9929 long double type, set the flags for noting the default long double
9930 type, even if we don't pass or return a long double type.
9932 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
9935 * ipa-inline.c (flatten_function): Do not overwrite final inlining
9938 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
9940 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
9941 support for merge[hl].
9942 (fold_mergehl_helper): New helper function.
9943 (tree-vector-builder.h): New #include for tree_vector_builder usage.
9944 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
9945 (altivec_vmrglw_direct): Add xxmrglw insn.
9947 2018-01-17 Andrew Waterman <andrew@sifive.com>
9949 * config/riscv/riscv.c (riscv_conditional_register_usage): If
9950 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
9952 2018-01-17 David Malcolm <dmalcolm@redhat.com>
9955 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
9956 call the lto_location_cache before reading the
9957 DECL_SOURCE_LOCATION of the types.
9959 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
9960 Richard Sandiford <richard.sandiford@linaro.org>
9962 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
9963 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
9964 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
9965 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
9966 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
9968 * config/aarch64/constraints.md (aarch64_movti_operand):
9970 * config/aarch64/predicates.md (Uti): Add new constraint.
9972 2018-01-17 Carl Love <cel@us.ibm.com>
9974 * config/rs6000/vsx.md (define_expand xl_len_r,
9975 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
9976 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
9978 (define_expand, define_insn): Move the shift left from the
9979 define_insn to the define_expand for lxvl and stxvl instructions.
9980 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
9981 and XL_LEN_R definitions to PURE.
9983 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
9985 * config/i386/i386.c (indirect_thunk_name): Declare regno
9986 as unsigned int. Compare regno with INVALID_REGNUM.
9987 (output_indirect_thunk): Ditto.
9988 (output_indirect_thunk_function): Ditto.
9989 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
9990 in the call to output_indirect_thunk_function.
9992 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
9995 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
9996 rather than the size of inner_type to determine the stack slot size
9997 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
9999 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
10002 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
10005 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
10007 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
10008 endian Linux systems to optionally enable multilibs for selecting
10009 the long double type if the user configured an explicit type.
10010 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
10011 have no long double multilibs if not defined.
10012 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
10013 warn if the user used -mabi={ieee,ibm}longdouble and we built
10014 multilibs for long double.
10015 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
10016 appropriate multilib option.
10017 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
10019 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
10020 for building long double multilibs.
10021 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
10023 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
10025 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
10028 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
10030 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
10033 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
10036 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
10039 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10041 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
10042 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
10044 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
10046 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
10047 different rtl trees depending on TARGET_64BIT.
10048 (rs6000_gen_lvx): Likewise.
10050 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10052 * config/visium/visium.md (nop): Tweak comment.
10053 (hazard_nop): Likewise.
10055 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10057 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
10058 -mspeculate-indirect-jumps.
10059 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
10060 for -mno-speculate-indirect-jumps.
10061 (*call_indirect_elfv2<mode>_nospec): New define_insn.
10062 (*call_value_indirect_elfv2<mode>): Disable for
10063 -mno-speculate-indirect-jumps.
10064 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
10065 (indirect_jump): Emit different RTL for
10066 -mno-speculate-indirect-jumps.
10067 (*indirect_jump<mode>): Disable for
10068 -mno-speculate-indirect-jumps.
10069 (*indirect_jump<mode>_nospec): New define_insn.
10070 (tablejump): Emit different RTL for
10071 -mno-speculate-indirect-jumps.
10072 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
10073 (tablejumpsi_nospec): New define_expand.
10074 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
10075 (tablejumpdi_nospec): New define_expand.
10076 (*tablejump<mode>_internal1): Disable for
10077 -mno-speculate-indirect-jumps.
10078 (*tablejump<mode>_internal1_nospec): New define_insn.
10079 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
10082 2018-01-16 Artyom Skrobov tyomitch@gmail.com
10084 * caller-save.c (insert_save): Drop unnecessary parameter. All
10087 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10088 Richard Biener <rguenth@suse.de>
10091 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
10092 return early, inline manually is_gimple_sizepos. Make sure if we
10093 call gimplify_expr we don't end up with a gimple constant.
10094 * tree.c (variably_modified_type_p): Don't return true for
10095 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
10096 * gimplify.h (is_gimple_sizepos): Remove.
10098 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10100 PR tree-optimization/83857
10101 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
10102 vectorizable_live_operation for pure SLP statements.
10103 (vectorizable_live_operation): Handle PHIs.
10105 2018-01-16 Richard Biener <rguenther@suse.de>
10107 PR tree-optimization/83867
10108 * tree-vect-stmts.c (vect_transform_stmt): Precompute
10109 nested_in_vect_loop_p since the scalar stmt may get invalidated.
10111 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10114 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
10115 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
10116 If off is not INTEGER_CST, issue a may not be aligned warning
10117 rather than isn't aligned. Use isn%'t rather than isn't.
10118 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
10120 <case MULT_EXPR>: Improve the case when bottom and one of the
10121 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
10122 operand, in that case check if the other operand is multiple of
10123 bottom divided by the INTEGER_CST operand.
10125 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10128 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
10129 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
10130 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
10131 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
10132 * config/pa/pa.c (pa_function_arg_advance): Likewise.
10133 (pa_function_arg, pa_arg_partial_bytes): Likewise.
10134 (pa_function_arg_size): New function.
10136 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10138 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
10139 in a separate statement.
10141 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10143 PR tree-optimization/83847
10144 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
10145 group gathers and scatters.
10147 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10149 PR rtl-optimization/86620
10150 * params.def (max-sched-ready-insns): Bump minimum value to 1.
10152 PR rtl-optimization/83213
10153 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
10154 to last if both are JUMP_INSNs.
10156 PR tree-optimization/83843
10157 * gimple-ssa-store-merging.c
10158 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
10159 store_immediate_info for bswap/nop orig_stores.
10161 2018-01-15 Andrew Waterman <andrew@sifive.com>
10163 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
10165 <UDIV>: Increase cost if !TARGET_DIV.
10167 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
10169 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
10170 (define_attr "cr_logical_3op"): New.
10171 (cceq_ior_compare): Adjust.
10172 (cceq_ior_compare_complement): Adjust.
10173 (*cceq_rev_compare): Adjust.
10174 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
10175 (is_cracked_insn): Adjust.
10176 (insn_must_be_first_in_group): Adjust.
10177 * config/rs6000/40x.md: Adjust.
10178 * config/rs6000/440.md: Adjust.
10179 * config/rs6000/476.md: Adjust.
10180 * config/rs6000/601.md: Adjust.
10181 * config/rs6000/603.md: Adjust.
10182 * config/rs6000/6xx.md: Adjust.
10183 * config/rs6000/7450.md: Adjust.
10184 * config/rs6000/7xx.md: Adjust.
10185 * config/rs6000/8540.md: Adjust.
10186 * config/rs6000/cell.md: Adjust.
10187 * config/rs6000/e300c2c3.md: Adjust.
10188 * config/rs6000/e500mc.md: Adjust.
10189 * config/rs6000/e500mc64.md: Adjust.
10190 * config/rs6000/e5500.md: Adjust.
10191 * config/rs6000/e6500.md: Adjust.
10192 * config/rs6000/mpc.md: Adjust.
10193 * config/rs6000/power4.md: Adjust.
10194 * config/rs6000/power5.md: Adjust.
10195 * config/rs6000/power6.md: Adjust.
10196 * config/rs6000/power7.md: Adjust.
10197 * config/rs6000/power8.md: Adjust.
10198 * config/rs6000/power9.md: Adjust.
10199 * config/rs6000/rs64.md: Adjust.
10200 * config/rs6000/titan.md: Adjust.
10202 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10204 * config/i386/predicates.md (indirect_branch_operand): Rewrite
10205 ix86_indirect_branch_register logic.
10207 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10209 * config/i386/constraints.md (Bs): Update
10210 ix86_indirect_branch_register check. Don't check
10211 ix86_indirect_branch_register with GOT_memory_operand.
10213 * config/i386/predicates.md (GOT_memory_operand): Don't check
10214 ix86_indirect_branch_register here.
10215 (GOT32_symbol_operand): Likewise.
10217 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10219 * config/i386/predicates.md (constant_call_address_operand):
10220 Rewrite ix86_indirect_branch_register logic.
10221 (sibcall_insn_operand): Likewise.
10223 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10225 * config/i386/constraints.md (Bs): Replace
10226 ix86_indirect_branch_thunk_register with
10227 ix86_indirect_branch_register.
10229 * config/i386/i386.md (indirect_jump): Likewise.
10230 (tablejump): Likewise.
10231 (*sibcall_memory): Likewise.
10232 (*sibcall_value_memory): Likewise.
10233 Peepholes of indirect call and jump via memory: Likewise.
10234 * config/i386/i386.opt: Likewise.
10235 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10236 (GOT_memory_operand): Likewise.
10237 (call_insn_operand): Likewise.
10238 (sibcall_insn_operand): Likewise.
10239 (GOT32_symbol_operand): Likewise.
10241 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10243 PR middle-end/83837
10244 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
10245 type rather than type addr's type points to.
10246 (expand_omp_atomic_mutex): Likewise.
10247 (expand_omp_atomic): Likewise.
10249 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10252 * config/i386/i386.c (output_indirect_thunk_function): Use
10253 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
10254 for __x86_return_thunk.
10256 2018-01-15 Richard Biener <rguenther@suse.de>
10258 PR middle-end/83850
10259 * expmed.c (extract_bit_field_1): Fix typo.
10261 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10264 * config/arm/iterators.md (VF): New mode iterator.
10265 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
10266 Remove integer-related logic from pattern.
10267 (neon_vabd<mode>_3): Likewise.
10269 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10271 PR middle-end/82694
10272 * common.opt (fstrict-overflow): No longer an alias.
10273 (fwrapv-pointer): New option.
10274 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
10275 also for pointer types based on flag_wrapv_pointer.
10276 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
10277 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
10278 opts->x_flag_wrapv got set.
10279 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
10280 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
10281 POINTER_TYPE_OVERFLOW_UNDEFINED.
10282 * match.pd: Likewise in address comparison pattern.
10283 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
10285 2018-01-15 Richard Biener <rguenther@suse.de>
10288 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
10289 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
10290 Reset type names to their identifier if their TYPE_DECL doesn't
10291 have linkage (and thus is used for ODR and devirt).
10292 (save_debug_info_for_decl): Remove.
10293 (save_debug_info_for_type): Likewise.
10294 (add_tree_to_fld_list): Adjust.
10295 * tree-pretty-print.c (dump_generic_node): Make dumping of
10296 type names more robust.
10298 2018-01-15 Richard Biener <rguenther@suse.de>
10300 * BASE-VER: Bump to 8.0.1.
10302 2018-01-14 Martin Sebor <msebor@redhat.com>
10305 * builtins.c (check_access): Avoid warning when the no-warning bit
10308 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
10310 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
10311 * ira-color (allocno_hard_regs_compare): Likewise.
10313 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
10316 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
10317 Use .pushsection/.popsection.
10319 2018-01-14 Martin Sebor <msebor@redhat.com>
10322 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
10324 2018-01-14 Jakub Jelinek <jakub@redhat.com>
10326 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
10327 entry from extra_headers.
10328 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
10329 extra_headers, make the list bitwise identical to the i?86-*-* one.
10331 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10333 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
10334 -mcmodel=large with -mindirect-branch=thunk,
10335 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
10336 -mfunction-return=thunk-extern.
10337 * doc/invoke.texi: Document -mcmodel=large is incompatible with
10338 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
10339 -mfunction-return=thunk and -mfunction-return=thunk-extern.
10341 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10343 * config/i386/i386.c (print_reg): Print the name of the full
10344 integer register without '%'.
10345 (ix86_print_operand): Handle 'V'.
10346 * doc/extend.texi: Document 'V' modifier.
10348 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10350 * config/i386/constraints.md (Bs): Disallow memory operand for
10351 -mindirect-branch-register.
10353 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10354 (GOT_memory_operand): Likewise.
10355 (call_insn_operand): Likewise.
10356 (sibcall_insn_operand): Likewise.
10357 (GOT32_symbol_operand): Likewise.
10358 * config/i386/i386.md (indirect_jump): Call convert_memory_address
10359 for -mindirect-branch-register.
10360 (tablejump): Likewise.
10361 (*sibcall_memory): Likewise.
10362 (*sibcall_value_memory): Likewise.
10363 Disallow peepholes of indirect call and jump via memory for
10364 -mindirect-branch-register.
10365 (*call_pop): Replace m with Bw.
10366 (*call_value_pop): Likewise.
10367 (*sibcall_pop_memory): Replace m with Bs.
10368 * config/i386/i386.opt (mindirect-branch-register): New option.
10369 * doc/invoke.texi: Document -mindirect-branch-register option.
10371 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10373 * config/i386/i386-protos.h (ix86_output_function_return): New.
10374 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
10375 set function_return_type.
10376 (indirect_thunk_name): Add ret_p to indicate thunk for function
10378 (output_indirect_thunk_function): Pass false to
10379 indirect_thunk_name.
10380 (ix86_output_indirect_branch_via_reg): Likewise.
10381 (ix86_output_indirect_branch_via_push): Likewise.
10382 (output_indirect_thunk_function): Create alias for function
10383 return thunk if regno < 0.
10384 (ix86_output_function_return): New function.
10385 (ix86_handle_fndecl_attribute): Handle function_return.
10386 (ix86_attribute_table): Add function_return.
10387 * config/i386/i386.h (machine_function): Add
10388 function_return_type.
10389 * config/i386/i386.md (simple_return_internal): Use
10390 ix86_output_function_return.
10391 (simple_return_internal_long): Likewise.
10392 * config/i386/i386.opt (mfunction-return=): New option.
10393 (indirect_branch): Mention -mfunction-return=.
10394 * doc/extend.texi: Document function_return function attribute.
10395 * doc/invoke.texi: Document -mfunction-return= option.
10397 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10399 * config/i386/i386-opts.h (indirect_branch): New.
10400 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
10401 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
10402 with local indirect jump when converting indirect call and jump.
10403 (ix86_set_indirect_branch_type): New.
10404 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
10405 (indirectlabelno): New.
10406 (indirect_thunk_needed): Likewise.
10407 (indirect_thunk_bnd_needed): Likewise.
10408 (indirect_thunks_used): Likewise.
10409 (indirect_thunks_bnd_used): Likewise.
10410 (INDIRECT_LABEL): Likewise.
10411 (indirect_thunk_name): Likewise.
10412 (output_indirect_thunk): Likewise.
10413 (output_indirect_thunk_function): Likewise.
10414 (ix86_output_indirect_branch_via_reg): Likewise.
10415 (ix86_output_indirect_branch_via_push): Likewise.
10416 (ix86_output_indirect_branch): Likewise.
10417 (ix86_output_indirect_jmp): Likewise.
10418 (ix86_code_end): Call output_indirect_thunk_function if needed.
10419 (ix86_output_call_insn): Call ix86_output_indirect_branch if
10421 (ix86_handle_fndecl_attribute): Handle indirect_branch.
10422 (ix86_attribute_table): Add indirect_branch.
10423 * config/i386/i386.h (machine_function): Add indirect_branch_type
10424 and has_local_indirect_jump.
10425 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
10427 (tablejump): Likewise.
10428 (*indirect_jump): Use ix86_output_indirect_jmp.
10429 (*tablejump_1): Likewise.
10430 (simple_return_indirect_internal): Likewise.
10431 * config/i386/i386.opt (mindirect-branch=): New option.
10432 (indirect_branch): New.
10435 (thunk-inline): Likewise.
10436 (thunk-extern): Likewise.
10437 * doc/extend.texi: Document indirect_branch function attribute.
10438 * doc/invoke.texi: Document -mindirect-branch= option.
10440 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
10443 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
10445 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
10447 * ipa-inline.c (want_inline_small_function_p): Return false if
10448 inlining has already failed with CIF_FINAL_ERROR.
10449 (update_caller_keys): Call want_inline_small_function_p before
10451 (update_callee_keys): Likewise.
10453 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
10455 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
10457 (rs6000_quadword_masked_address_p): Likewise.
10458 (quad_aligned_load_p): Likewise.
10459 (quad_aligned_store_p): Likewise.
10460 (const_load_sequence_p): Add comment to describe the outer-most loop.
10461 (mimic_memory_attributes_and_flags): New function.
10462 (rs6000_gen_stvx): Likewise.
10463 (replace_swapped_aligned_store): Likewise.
10464 (rs6000_gen_lvx): Likewise.
10465 (replace_swapped_aligned_load): Likewise.
10466 (replace_swapped_load_constant): Capitalize argument name in
10467 comment describing this function.
10468 (rs6000_analyze_swaps): Add a third pass to search for vector loads
10469 and stores that access quad-word aligned addresses and replace
10470 with stvx or lvx instructions when appropriate.
10471 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
10472 New function prototype.
10473 (rs6000_quadword_masked_address_p): Likewise.
10474 (rs6000_gen_lvx): Likewise.
10475 (rs6000_gen_stvx): Likewise.
10476 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
10477 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
10478 when memory address is aligned.
10479 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
10480 this split to select lvx instruction when memory address is aligned.
10481 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
10482 instruction when memory address is aligned.
10483 (*vsx_le_perm_load_v16qi): Likewise.
10484 (four unnamed splitters): Modify to select the stvx instruction
10485 when memory is aligned.
10487 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
10489 * predict.c (determine_unlikely_bbs): Handle correctly BBs
10490 which appears in the queue multiple times.
10492 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10493 Alan Hayward <alan.hayward@arm.com>
10494 David Sherwood <david.sherwood@arm.com>
10496 * tree-vectorizer.h (vec_lower_bound): New structure.
10497 (_loop_vec_info): Add check_nonzero and lower_bounds.
10498 (LOOP_VINFO_CHECK_NONZERO): New macro.
10499 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
10500 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
10501 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
10502 fields. Make seg_len the distance travelled, not including the
10504 (dr_direction_indicator): Declare.
10505 (dr_zero_step_indicator): Likewise.
10506 (dr_known_forward_stride_p): Likewise.
10507 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
10509 (runtime_alias_check_p): Allow runtime alias checks with
10511 (operator ==): Compare access_size and align.
10512 (prune_runtime_alias_test_list): Rework for new distinction between
10513 the access_size and seg_len.
10514 (create_intersect_range_checks_index): Likewise. Cope with polynomial
10516 (get_segment_min_max): New function.
10517 (create_intersect_range_checks): Use it.
10518 (dr_step_indicator): New function.
10519 (dr_direction_indicator): Likewise.
10520 (dr_zero_step_indicator): Likewise.
10521 (dr_known_forward_stride_p): Likewise.
10522 * tree-loop-distribution.c (data_ref_segment_size): Return
10523 DR_STEP * (niters - 1).
10524 (compute_alias_check_pairs): Update call to the dr_with_seg_len
10526 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
10527 (vect_preserves_scalar_order_p): New function, split out from...
10528 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
10529 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
10530 (vect_vfa_access_size): New function.
10531 (vect_vfa_align): Likewise.
10532 (vect_compile_time_alias): Take access_size_a and access_b arguments.
10533 (dump_lower_bound): New function.
10534 (vect_check_lower_bound): Likewise.
10535 (vect_small_gap_p): Likewise.
10536 (vectorizable_with_step_bound_p): Likewise.
10537 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
10538 depencies if the vectorization factor is 1. Convert the checks
10539 for nonzero steps into checks on the bounds of DR_STEP. Try using
10540 a bunds check for variable steps if the minimum required step is
10541 relatively small. Update calls to the dr_with_seg_len
10542 constructor and to vect_compile_time_alias.
10543 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
10545 (vect_loop_versioning): Call it.
10546 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
10548 (vect_estimate_min_profitable_iters): Account for any bounds checks.
10550 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10551 Alan Hayward <alan.hayward@arm.com>
10552 David Sherwood <david.sherwood@arm.com>
10554 * doc/sourcebuild.texi (vect_scatter_store): Document.
10555 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
10557 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
10559 * genopinit.c (main): Add supports_vec_scatter_store and
10560 supports_vec_scatter_store_cached to target_optabs.
10561 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
10562 IFN_MASK_SCATTER_STORE.
10563 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
10565 * internal-fn.h (internal_store_fn_p): Declare.
10566 (internal_fn_stored_value_index): Likewise.
10567 * internal-fn.c (scatter_store_direct): New macro.
10568 (expand_scatter_store_optab_fn): New function.
10569 (direct_scatter_store_optab_supported_p): New macro.
10570 (internal_store_fn_p): New function.
10571 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
10572 IFN_MASK_SCATTER_STORE.
10573 (internal_fn_mask_index): Likewise.
10574 (internal_fn_stored_value_index): New function.
10575 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
10576 for scatter stores.
10577 * optabs-query.h (supports_vec_scatter_store_p): Declare.
10578 * optabs-query.c (supports_vec_scatter_store_p): New function.
10579 * tree-vectorizer.h (vect_get_store_rhs): Declare.
10580 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
10581 true for scatter stores.
10582 (vect_gather_scatter_fn_p): Handle scatter stores too.
10583 (vect_check_gather_scatter): Consider using scatter stores if
10584 supports_vec_scatter_store_p.
10585 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
10586 scatter stores too.
10587 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10588 internal_fn_stored_value_index.
10589 (check_load_store_masking): Handle scatter stores too.
10590 (vect_get_store_rhs): Make public.
10591 (vectorizable_call): Use internal_store_fn_p.
10592 (vectorizable_store): Handle scatter store internal functions.
10593 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
10594 when deciding whether the end of the group has been reached.
10595 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
10596 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
10597 (mask_scatter_store<mode>): New insns.
10599 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10600 Alan Hayward <alan.hayward@arm.com>
10601 David Sherwood <david.sherwood@arm.com>
10603 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
10604 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
10605 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
10607 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
10608 Use vect_truncate_gather_scatter_offset if we can't treat the
10609 operation as a normal gather load or scatter store.
10610 (get_group_load_store_type): Take the gather_scatter_info
10611 as argument. Try using a gather load or scatter store for
10612 single-element groups.
10613 (get_load_store_type): Update calls to get_group_load_store_type
10614 and vect_use_strided_gather_scatters_p.
10616 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10617 Alan Hayward <alan.hayward@arm.com>
10618 David Sherwood <david.sherwood@arm.com>
10620 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
10621 optional tree argument.
10622 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
10624 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
10625 but continue to use the current value as a fallback.
10626 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
10627 to compare the updates.
10628 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
10629 (get_load_store_type): Use it when handling a strided access.
10630 (vect_get_strided_load_store_ops): New function.
10631 (vect_get_data_ptr_increment): Likewise.
10632 (vectorizable_load): Handle strided gather loads. Always pass
10633 a step to vect_create_data_ref_ptr and bump_vector_ptr.
10635 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10636 Alan Hayward <alan.hayward@arm.com>
10637 David Sherwood <david.sherwood@arm.com>
10639 * doc/md.texi (gather_load@var{m}): Document.
10640 (mask_gather_load@var{m}): Likewise.
10641 * genopinit.c (main): Add supports_vec_gather_load and
10642 supports_vec_gather_load_cached to target_optabs.
10643 * optabs-tree.c (init_tree_optimization_optabs): Use
10644 ggc_cleared_alloc to allocate target_optabs.
10645 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
10646 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
10648 * internal-fn.h (internal_load_fn_p): Declare.
10649 (internal_gather_scatter_fn_p): Likewise.
10650 (internal_fn_mask_index): Likewise.
10651 (internal_gather_scatter_fn_supported_p): Likewise.
10652 * internal-fn.c (gather_load_direct): New macro.
10653 (expand_gather_load_optab_fn): New function.
10654 (direct_gather_load_optab_supported_p): New macro.
10655 (direct_internal_fn_optab): New function.
10656 (internal_load_fn_p): Likewise.
10657 (internal_gather_scatter_fn_p): Likewise.
10658 (internal_fn_mask_index): Likewise.
10659 (internal_gather_scatter_fn_supported_p): Likewise.
10660 * optabs-query.c (supports_at_least_one_mode_p): New function.
10661 (supports_vec_gather_load_p): Likewise.
10662 * optabs-query.h (supports_vec_gather_load_p): Declare.
10663 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
10664 and memory_type field.
10665 (NUM_PATTERNS): Bump to 15.
10666 * tree-vect-data-refs.c: Include internal-fn.h.
10667 (vect_gather_scatter_fn_p): New function.
10668 (vect_describe_gather_scatter_call): Likewise.
10669 (vect_check_gather_scatter): Try using internal functions for
10670 gather loads. Recognize existing calls to a gather load function.
10671 (vect_analyze_data_refs): Consider using gather loads if
10672 supports_vec_gather_load_p.
10673 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
10674 (vect_get_gather_scatter_offset_type): Likewise.
10675 (vect_convert_mask_for_vectype): Likewise.
10676 (vect_add_conversion_to_patterm): Likewise.
10677 (vect_try_gather_scatter_pattern): Likewise.
10678 (vect_recog_gather_scatter_pattern): New pattern recognizer.
10679 (vect_vect_recog_func_ptrs): Add it.
10680 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10681 internal_fn_mask_index and internal_gather_scatter_fn_p.
10682 (check_load_store_masking): Take the gather_scatter_info as an
10683 argument and handle gather loads.
10684 (vect_get_gather_scatter_ops): New function.
10685 (vectorizable_call): Check internal_load_fn_p.
10686 (vectorizable_load): Likewise. Handle gather load internal
10688 (vectorizable_store): Update call to check_load_store_masking.
10689 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
10690 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
10691 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
10692 (aarch64_gather_scale_operand_d): New predicates.
10693 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
10694 (mask_gather_load<mode>): New insns.
10696 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10697 Alan Hayward <alan.hayward@arm.com>
10698 David Sherwood <david.sherwood@arm.com>
10700 * optabs.def (fold_left_plus_optab): New optab.
10701 * doc/md.texi (fold_left_plus_@var{m}): Document.
10702 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
10703 * internal-fn.c (fold_left_direct): Define.
10704 (expand_fold_left_optab_fn): Likewise.
10705 (direct_fold_left_optab_supported_p): Likewise.
10706 * fold-const-call.c (fold_const_fold_left): New function.
10707 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
10708 * tree-parloops.c (valid_reduction_p): New function.
10709 (gather_scalar_reductions): Use it.
10710 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
10711 (vect_finish_replace_stmt): Declare.
10712 * tree-vect-loop.c (fold_left_reduction_fn): New function.
10713 (needs_fold_left_reduction_p): New function, split out from...
10714 (vect_is_simple_reduction): ...here. Accept reductions that
10715 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
10716 (vect_force_simple_reduction): Also store the reduction type in
10717 the assignment's STMT_VINFO_REDUC_TYPE.
10718 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
10719 (merge_with_identity): New function.
10720 (vect_expand_fold_left): Likewise.
10721 (vectorize_fold_left_reduction): Likewise.
10722 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
10723 scalar phi in place for it. Check for target support and reject
10724 cases that would reassociate the operation. Defer the transform
10725 phase to vectorize_fold_left_reduction.
10726 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
10727 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
10728 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
10730 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10732 * tree-if-conv.c (predicate_mem_writes): Remove redundant
10733 call to ifc_temp_var.
10735 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10736 Alan Hayward <alan.hayward@arm.com>
10737 David Sherwood <david.sherwood@arm.com>
10739 * target.def (legitimize_address_displacement): Take the original
10740 offset as a poly_int.
10741 * targhooks.h (default_legitimize_address_displacement): Update
10743 * targhooks.c (default_legitimize_address_displacement): Likewise.
10744 * doc/tm.texi: Regenerate.
10745 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
10746 as an argument, moving assert of ad->disp == ad->disp_term to...
10747 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
10748 Try calling targetm.legitimize_address_displacement before expanding
10749 the address rather than afterwards, and adjust for the new interface.
10750 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
10751 Match the new hook interface. Handle SVE addresses.
10752 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
10753 new hook interface.
10755 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10757 * Makefile.in (OBJS): Add early-remat.o.
10758 * target.def (select_early_remat_modes): New hook.
10759 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
10760 * doc/tm.texi: Regenerate.
10761 * targhooks.h (default_select_early_remat_modes): Declare.
10762 * targhooks.c (default_select_early_remat_modes): New function.
10763 * timevar.def (TV_EARLY_REMAT): New timevar.
10764 * passes.def (pass_early_remat): New pass.
10765 * tree-pass.h (make_pass_early_remat): Declare.
10766 * early-remat.c: New file.
10767 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
10769 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
10771 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10772 Alan Hayward <alan.hayward@arm.com>
10773 David Sherwood <david.sherwood@arm.com>
10775 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
10776 vfm1 with a bound_epilog parameter.
10777 (vect_do_peeling): Update calls accordingly, and move the prologue
10778 call earlier in the function. Treat the base bound_epilog as 0 for
10779 fully-masked loops and retain vf - 1 for other loops. Add 1 to
10780 this base when peeling for gaps.
10781 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
10782 with fully-masked loops.
10783 (vect_estimate_min_profitable_iters): Handle the single peeled
10784 iteration in that case.
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 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
10791 single-element interleaving even if the size is not a power of 2.
10792 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
10793 accesses for single-element interleaving if the group size is
10796 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10797 Alan Hayward <alan.hayward@arm.com>
10798 David Sherwood <david.sherwood@arm.com>
10800 * doc/md.texi (fold_extract_last_@var{m}): Document.
10801 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
10802 * optabs.def (fold_extract_last_optab): New optab.
10803 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
10804 * internal-fn.c (fold_extract_direct): New macro.
10805 (expand_fold_extract_optab_fn): Likewise.
10806 (direct_fold_extract_optab_supported_p): Likewise.
10807 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
10808 * tree-vect-loop.c (vect_model_reduction_cost): Handle
10809 EXTRACT_LAST_REDUCTION.
10810 (get_initial_def_for_reduction): Do not create an initial vector
10811 for EXTRACT_LAST_REDUCTION reductions.
10812 (vectorizable_reduction): Leave the scalar phi in place for
10813 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
10814 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
10815 epilogue code for EXTRACT_LAST_REDUCTION and defer the
10816 transform phase to vectorizable_condition.
10817 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
10819 (vect_finish_stmt_generation): ...here.
10820 (vect_finish_replace_stmt): New function.
10821 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
10822 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
10824 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
10826 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10827 Alan Hayward <alan.hayward@arm.com>
10828 David Sherwood <david.sherwood@arm.com>
10830 * doc/md.texi (extract_last_@var{m}): Document.
10831 * optabs.def (extract_last_optab): New optab.
10832 * internal-fn.def (EXTRACT_LAST): New internal function.
10833 * internal-fn.c (cond_unary_direct): New macro.
10834 (expand_cond_unary_optab_fn): Likewise.
10835 (direct_cond_unary_optab_supported_p): Likewise.
10836 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
10837 loops using EXTRACT_LAST.
10838 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
10839 (extract_last_<mode>): ...this optab.
10840 (vec_extract<mode><Vel>): Update accordingly.
10842 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10843 Alan Hayward <alan.hayward@arm.com>
10844 David Sherwood <david.sherwood@arm.com>
10846 * target.def (empty_mask_is_expensive): New hook.
10847 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
10848 * doc/tm.texi: Regenerate.
10849 * targhooks.h (default_empty_mask_is_expensive): Declare.
10850 * targhooks.c (default_empty_mask_is_expensive): New function.
10851 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
10852 if the target says that empty masks are expensive.
10853 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
10855 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
10857 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10858 Alan Hayward <alan.hayward@arm.com>
10859 David Sherwood <david.sherwood@arm.com>
10861 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
10862 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
10863 (vect_use_loop_mask_for_alignment_p): New function.
10864 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
10865 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
10866 niters_skip argument. Make sure that the first niters_skip elements
10867 of the first iteration are inactive.
10868 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
10869 Update call to vect_set_loop_masks_directly.
10870 (get_misalign_in_elems): New function, split out from...
10871 (vect_gen_prolog_loop_niters): ...here.
10872 (vect_update_init_of_dr): Take a code argument that specifies whether
10873 the adjustment should be added or subtracted.
10874 (vect_update_init_of_drs): Likewise.
10875 (vect_prepare_for_masked_peels): New function.
10876 (vect_do_peeling): Skip prologue peeling if we're using a mask
10877 instead. Update call to vect_update_inits_of_drs.
10878 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10880 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
10881 alignment. Do not include the number of peeled iterations in
10882 the minimum threshold in that case.
10883 (vectorizable_induction): Adjust the start value down by
10884 LOOP_VINFO_MASK_SKIP_NITERS iterations.
10885 (vect_transform_loop): Call vect_prepare_for_masked_peels.
10886 Take the number of skipped iterations into account when calculating
10888 * tree-vect-stmts.c (vect_gen_while_not): New function.
10890 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10891 Alan Hayward <alan.hayward@arm.com>
10892 David Sherwood <david.sherwood@arm.com>
10894 * doc/sourcebuild.texi (vect_fully_masked): Document.
10895 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
10896 default value to 0.
10897 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
10899 (vect_analyze_loop_2): ...here. Don't check the vectorization
10900 factor against the number of loop iterations if the loop is
10903 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10904 Alan Hayward <alan.hayward@arm.com>
10905 David Sherwood <david.sherwood@arm.com>
10907 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
10908 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
10909 (dump_groups): Update accordingly.
10910 (iv_use::mem_type): New member variable.
10911 (address_p): New function.
10912 (record_use): Add a mem_type argument and initialize the new
10914 (record_group_use): Add a mem_type argument. Use address_p.
10915 Remove obsolete null checks of base_object. Update call to record_use.
10916 (find_interesting_uses_op): Update call to record_group_use.
10917 (find_interesting_uses_cond): Likewise.
10918 (find_interesting_uses_address): Likewise.
10919 (get_mem_type_for_internal_fn): New function.
10920 (find_address_like_use): Likewise.
10921 (find_interesting_uses_stmt): Try find_address_like_use before
10922 calling find_interesting_uses_op.
10923 (addr_offset_valid_p): Use the iv mem_type field as the type
10924 of the addressed memory.
10925 (add_autoinc_candidates): Likewise.
10926 (get_address_cost): Likewise.
10927 (split_small_address_groups_p): Use address_p.
10928 (split_address_groups): Likewise.
10929 (add_iv_candidate_for_use): Likewise.
10930 (autoinc_possible_for_pair): Likewise.
10931 (rewrite_groups): Likewise.
10932 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
10933 (determine_group_iv_cost): Update after split of USE_ADDRESS.
10934 (get_alias_ptr_type_for_ptr_address): New function.
10935 (rewrite_use_address): Rewrite address uses in calls that were
10936 identified by find_address_like_use.
10938 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10939 Alan Hayward <alan.hayward@arm.com>
10940 David Sherwood <david.sherwood@arm.com>
10942 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
10944 * gimple-expr.h (is_gimple_addressable: Likewise.
10945 * gimple-expr.c (is_gimple_address): Likewise.
10946 * internal-fn.c (expand_call_mem_ref): New function.
10947 (expand_mask_load_optab_fn): Use it.
10948 (expand_mask_store_optab_fn): Likewise.
10950 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10951 Alan Hayward <alan.hayward@arm.com>
10952 David Sherwood <david.sherwood@arm.com>
10954 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
10955 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
10956 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
10957 (cond_umax@var{mode}): Document.
10958 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
10959 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
10960 (cond_umin_optab, cond_umax_optab): New optabs.
10961 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
10962 (COND_IOR, COND_XOR): New internal functions.
10963 * internal-fn.h (get_conditional_internal_fn): Declare.
10964 * internal-fn.c (cond_binary_direct): New macro.
10965 (expand_cond_binary_optab_fn): Likewise.
10966 (direct_cond_binary_optab_supported_p): Likewise.
10967 (get_conditional_internal_fn): New function.
10968 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
10969 Cope with reduction statements that are vectorized as calls rather
10971 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
10972 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
10973 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
10974 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
10975 (UNSPEC_COND_EOR): New unspecs.
10976 (optab): Add mappings for them.
10977 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
10978 (sve_int_op, sve_fp_op): New int attributes.
10980 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10981 Alan Hayward <alan.hayward@arm.com>
10982 David Sherwood <david.sherwood@arm.com>
10984 * optabs.def (while_ult_optab): New optab.
10985 * doc/md.texi (while_ult@var{m}@var{n}): Document.
10986 * internal-fn.def (WHILE_ULT): New internal function.
10987 * internal-fn.h (direct_internal_fn_supported_p): New override
10988 that takes two types as argument.
10989 * internal-fn.c (while_direct): New macro.
10990 (expand_while_optab_fn): New function.
10991 (convert_optab_supported_p): Likewise.
10992 (direct_while_optab_supported_p): New macro.
10993 * wide-int.h (wi::udiv_ceil): New function.
10994 * tree-vectorizer.h (rgroup_masks): New structure.
10995 (vec_loop_masks): New typedef.
10996 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
10997 and fully_masked_p.
10998 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
10999 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
11000 (vect_max_vf): New function.
11001 (slpeel_make_loop_iterate_ntimes): Delete.
11002 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
11003 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
11004 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
11005 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
11006 internal-fn.h, stor-layout.h and optabs-query.h.
11007 (vect_set_loop_mask): New function.
11008 (add_preheader_seq): Likewise.
11009 (add_header_seq): Likewise.
11010 (interleave_supported_p): Likewise.
11011 (vect_maybe_permute_loop_masks): Likewise.
11012 (vect_set_loop_masks_directly): Likewise.
11013 (vect_set_loop_condition_masked): Likewise.
11014 (vect_set_loop_condition_unmasked): New function, split out from
11015 slpeel_make_loop_iterate_ntimes.
11016 (slpeel_make_loop_iterate_ntimes): Rename to..
11017 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
11018 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
11019 (vect_do_peeling): Update call accordingly.
11020 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
11022 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11023 mask_compare_type, can_fully_mask_p and fully_masked_p.
11024 (release_vec_loop_masks): New function.
11025 (_loop_vec_info): Use it to free the loop masks.
11026 (can_produce_all_loop_masks_p): New function.
11027 (vect_get_max_nscalars_per_iter): Likewise.
11028 (vect_verify_full_masking): Likewise.
11029 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
11030 retries, and free the mask rgroups before retrying. Check loop-wide
11031 reasons for disallowing fully-masked loops. Make the final decision
11032 about whether use a fully-masked loop or not.
11033 (vect_estimate_min_profitable_iters): Do not assume that peeling
11034 for the number of iterations will be needed for fully-masked loops.
11035 (vectorizable_reduction): Disable fully-masked loops.
11036 (vectorizable_live_operation): Likewise.
11037 (vect_halve_mask_nunits): New function.
11038 (vect_double_mask_nunits): Likewise.
11039 (vect_record_loop_mask): Likewise.
11040 (vect_get_loop_mask): Likewise.
11041 (vect_transform_loop): Handle the case in which the final loop
11042 iteration might handle a partial vector. Call vect_set_loop_condition
11043 instead of slpeel_make_loop_iterate_ntimes.
11044 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
11045 (check_load_store_masking): New function.
11046 (prepare_load_store_mask): Likewise.
11047 (vectorizable_store): Handle fully-masked loops.
11048 (vectorizable_load): Likewise.
11049 (supportable_widening_operation): Use vect_halve_mask_nunits for
11051 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
11052 (vect_gen_while): New function.
11053 * config/aarch64/aarch64.md (umax<mode>3): New expander.
11054 (aarch64_uqdec<mode>): New insn.
11056 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11057 Alan Hayward <alan.hayward@arm.com>
11058 David Sherwood <david.sherwood@arm.com>
11060 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
11061 (reduc_xor_scal_optab): New optabs.
11062 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
11063 (reduc_xor_scal_@var{m}): Document.
11064 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
11065 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
11066 internal functions.
11067 * fold-const-call.c (fold_const_call): Handle them.
11068 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
11069 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
11070 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
11071 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
11072 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
11073 (UNSPEC_XORV): New unspecs.
11074 (optab): Add entries for them.
11075 (BITWISEV): New int iterator.
11076 (bit_reduc_op): New int attributes.
11078 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11079 Alan Hayward <alan.hayward@arm.com>
11080 David Sherwood <david.sherwood@arm.com>
11082 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
11083 * internal-fn.def (VEC_SHL_INSERT): New internal function.
11084 * optabs.def (vec_shl_insert_optab): New optab.
11085 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
11086 (duplicate_and_interleave): Likewise.
11087 * tree-vect-loop.c: Include internal-fn.h.
11088 (neutral_op_for_slp_reduction): New function, split out from
11089 get_initial_defs_for_reduction.
11090 (get_initial_def_for_reduction): Handle option 2 for variable-length
11091 vectors by loading the neutral value into a vector and then shifting
11092 the initial value into element 0.
11093 (get_initial_defs_for_reduction): Replace the code argument with
11094 the neutral value calculated by neutral_op_for_slp_reduction.
11095 Use gimple_build_vector for constant-length vectors.
11096 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
11097 but the first group_size elements have a neutral value.
11098 Use duplicate_and_interleave otherwise.
11099 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
11100 Update call to get_initial_defs_for_reduction. Handle SLP
11101 reductions for variable-length vectors by creating one vector
11102 result for each scalar result, with the elements associated
11103 with other scalar results stubbed out with the neutral value.
11104 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
11105 Require IFN_VEC_SHL_INSERT for double reductions on
11106 variable-length vectors, or SLP reductions that have
11107 a neutral value. Require can_duplicate_and_interleave_p
11108 support for variable-length unchained SLP reductions if there
11109 is no neutral value, such as for MIN/MAX reductions. Also require
11110 the number of vector elements to be a multiple of the number of
11111 SLP statements when doing variable-length unchained SLP reductions.
11112 Update call to vect_create_epilog_for_reduction.
11113 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
11114 and remove initial values.
11115 (duplicate_and_interleave): Make public.
11116 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
11117 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
11119 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11120 Alan Hayward <alan.hayward@arm.com>
11121 David Sherwood <david.sherwood@arm.com>
11123 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
11124 (can_duplicate_and_interleave_p): New function.
11125 (vect_get_and_check_slp_defs): Take the vector of statements
11126 rather than just the current one. Remove excess parentheses.
11127 Restriction rejectinon of vect_constant_def and vect_external_def
11128 for variable-length vectors to boolean types, or types for which
11129 can_duplicate_and_interleave_p is false.
11130 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
11131 (duplicate_and_interleave): New function.
11132 (vect_get_constant_vectors): Use gimple_build_vector for
11133 constant-length vectors and suitable variable-length constant
11134 vectors. Use duplicate_and_interleave for other variable-length
11135 vectors. Don't defer the update when inserting new statements.
11137 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11138 Alan Hayward <alan.hayward@arm.com>
11139 David Sherwood <david.sherwood@arm.com>
11141 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
11142 min_profitable_iters doesn't go negative.
11144 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11145 Alan Hayward <alan.hayward@arm.com>
11146 David Sherwood <david.sherwood@arm.com>
11148 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
11149 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
11150 * optabs.def (vec_mask_load_lanes_optab): New optab.
11151 (vec_mask_store_lanes_optab): Likewise.
11152 * internal-fn.def (MASK_LOAD_LANES): New internal function.
11153 (MASK_STORE_LANES): Likewise.
11154 * internal-fn.c (mask_load_lanes_direct): New macro.
11155 (mask_store_lanes_direct): Likewise.
11156 (expand_mask_load_optab_fn): Handle masked operations.
11157 (expand_mask_load_lanes_optab_fn): New macro.
11158 (expand_mask_store_optab_fn): Handle masked operations.
11159 (expand_mask_store_lanes_optab_fn): New macro.
11160 (direct_mask_load_lanes_optab_supported_p): Likewise.
11161 (direct_mask_store_lanes_optab_supported_p): Likewise.
11162 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
11164 (vect_load_lanes_supported): Likewise.
11165 * tree-vect-data-refs.c (strip_conversion): New function.
11166 (can_group_stmts_p): Likewise.
11167 (vect_analyze_data_ref_accesses): Use it instead of checking
11168 for a pair of assignments.
11169 (vect_store_lanes_supported): Take a masked_p parameter.
11170 (vect_load_lanes_supported): Likewise.
11171 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
11172 vect_store_lanes_supported and vect_load_lanes_supported.
11173 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11174 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
11175 parameter. Don't allow gaps for masked accesses.
11176 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
11177 and vect_load_lanes_supported.
11178 (get_load_store_type): Take a masked_p parameter and update
11179 call to get_group_load_store_type.
11180 (vectorizable_store): Update call to get_load_store_type.
11181 Handle IFN_MASK_STORE_LANES.
11182 (vectorizable_load): Update call to get_load_store_type.
11183 Handle IFN_MASK_LOAD_LANES.
11185 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11186 Alan Hayward <alan.hayward@arm.com>
11187 David Sherwood <david.sherwood@arm.com>
11189 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
11191 * config/aarch64/aarch64-protos.h
11192 (aarch64_sve_struct_memory_operand_p): Declare.
11193 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
11194 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
11195 (VPRED, vpred): Handle SVE structure modes.
11196 * config/aarch64/constraints.md (Utx): New constraint.
11197 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
11198 (aarch64_sve_struct_nonimmediate_operand): New predicates.
11199 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
11200 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
11201 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
11202 structure modes. Split into pieces after RA.
11203 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
11204 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
11206 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
11207 SVE structure modes.
11208 (aarch64_classify_address): Likewise.
11209 (sizetochar): Move earlier in file.
11210 (aarch64_print_operand): Handle SVE register lists.
11211 (aarch64_array_mode): New function.
11212 (aarch64_sve_struct_memory_operand_p): Likewise.
11213 (TARGET_ARRAY_MODE): Redefine.
11215 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11216 Alan Hayward <alan.hayward@arm.com>
11217 David Sherwood <david.sherwood@arm.com>
11219 * target.def (array_mode): New target hook.
11220 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
11221 * doc/tm.texi: Regenerate.
11222 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
11223 * hooks.c (hook_optmode_mode_uhwi_none): New function.
11224 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
11225 targetm.array_mode.
11226 * stor-layout.c (mode_for_array): Likewise. Support polynomial
11229 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11230 Alan Hayward <alan.hayward@arm.com>
11231 David Sherwood <david.sherwood@arm.com>
11233 * fold-const.c (fold_binary_loc): Check the argument types
11234 rather than the result type when testing for a vector operation.
11236 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11238 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
11239 * doc/tm.texi: Regenerate.
11241 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11242 Alan Hayward <alan.hayward@arm.com>
11243 David Sherwood <david.sherwood@arm.com>
11245 * doc/invoke.texi (-msve-vector-bits=): Document new option.
11246 (sve): Document new AArch64 extension.
11247 * doc/md.texi (w): Extend the description of the AArch64
11248 constraint to include SVE vectors.
11249 (Upl, Upa): Document new AArch64 predicate constraints.
11250 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
11252 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
11253 (msve-vector-bits=): New option.
11254 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
11255 SVE when these are disabled.
11256 (sve): New extension.
11257 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
11258 modes. Adjust their number of units based on aarch64_sve_vg.
11259 (MAX_BITSIZE_MODE_ANY_MODE): Define.
11260 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
11261 aarch64_addr_query_type.
11262 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
11263 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
11264 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
11265 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
11266 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
11267 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
11268 (aarch64_simd_imm_zero_p): Delete.
11269 (aarch64_check_zero_based_sve_index_immediate): Declare.
11270 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11271 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11272 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11273 (aarch64_sve_float_mul_immediate_p): Likewise.
11274 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11275 rather than an rtx.
11276 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
11277 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
11278 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
11279 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
11280 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
11281 (aarch64_regmode_natural_size): Likewise.
11282 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
11283 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
11285 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
11286 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
11287 for VG and the SVE predicate registers.
11288 (V_ALIASES): Add a "z"-prefixed alias.
11289 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
11290 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
11291 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
11292 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
11293 (REG_CLASS_NAMES): Add entries for them.
11294 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
11295 and the predicate registers.
11296 (aarch64_sve_vg): Declare.
11297 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
11298 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
11299 (REGMODE_NATURAL_SIZE): Define.
11300 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
11302 * config/aarch64/aarch64.c: Include cfgrtl.h.
11303 (simd_immediate_info): Add a constructor for series vectors,
11304 and an associated step field.
11305 (aarch64_sve_vg): New variable.
11306 (aarch64_dbx_register_number): Handle VG and the predicate registers.
11307 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
11308 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
11309 (VEC_ANY_DATA, VEC_STRUCT): New constants.
11310 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
11311 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
11312 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
11313 (aarch64_get_mask_mode): New functions.
11314 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
11315 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
11316 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
11317 predicate modes and predicate registers. Explicitly restrict
11318 GPRs to modes of 16 bytes or smaller. Only allow FP registers
11319 to store a vector mode if it is recognized by
11320 aarch64_classify_vector_mode.
11321 (aarch64_regmode_natural_size): New function.
11322 (aarch64_hard_regno_caller_save_mode): Return the original mode
11324 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
11325 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
11326 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
11327 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
11329 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
11330 does not overlap dest if the function is frame-related. Handle
11332 (aarch64_split_add_offset): New function.
11333 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
11334 them aarch64_add_offset.
11335 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
11336 and update call to aarch64_sub_sp.
11337 (aarch64_add_cfa_expression): New function.
11338 (aarch64_expand_prologue): Pass extra temporary registers to the
11339 functions above. Handle the case in which we need to emit new
11340 DW_CFA_expressions for registers that were originally saved
11341 relative to the stack pointer, but now have to be expressed
11342 relative to the frame pointer.
11343 (aarch64_output_mi_thunk): Pass extra temporary registers to the
11345 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
11346 IP0 and IP1 values for SVE frames.
11347 (aarch64_expand_vec_series): New function.
11348 (aarch64_expand_sve_widened_duplicate): Likewise.
11349 (aarch64_expand_sve_const_vector): Likewise.
11350 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
11351 Handle SVE constants. Use emit_move_insn to move a force_const_mem
11352 into the register, rather than emitting a SET directly.
11353 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
11354 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
11355 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
11356 (offset_9bit_signed_scaled_p): New functions.
11357 (aarch64_replicate_bitmask_imm): New function.
11358 (aarch64_bitmask_imm): Use it.
11359 (aarch64_cannot_force_const_mem): Reject expressions involving
11360 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
11361 (aarch64_classify_index): Handle SVE indices, by requiring
11362 a plain register index with a scale that matches the element size.
11363 (aarch64_classify_address): Handle SVE addresses. Assert that
11364 the mode of the address is VOIDmode or an integer mode.
11365 Update call to aarch64_classify_symbol.
11366 (aarch64_classify_symbolic_expression): Update call to
11367 aarch64_classify_symbol.
11368 (aarch64_const_vec_all_in_range_p): New function.
11369 (aarch64_print_vector_float_operand): Likewise.
11370 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
11371 "vN" for FP registers with SVE modes. Handle (const ...) vectors
11372 and the FP immediates 1.0 and 0.5.
11373 (aarch64_print_address_internal): Handle SVE addresses.
11374 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
11375 (aarch64_regno_regclass): Handle predicate registers.
11376 (aarch64_secondary_reload): Handle big-endian reloads of SVE
11378 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
11379 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
11380 (aarch64_convert_sve_vector_bits): New function.
11381 (aarch64_override_options): Use it to handle -msve-vector-bits=.
11382 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11383 rather than an rtx.
11384 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
11385 Handle SVE vector and predicate modes. Accept VL-based constants
11386 that need only one temporary register, and VL offsets that require
11387 no temporary registers.
11388 (aarch64_conditional_register_usage): Mark the predicate registers
11389 as fixed if SVE isn't available.
11390 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
11391 Return true for SVE vector and predicate modes.
11392 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
11393 rather than an unsigned int. Handle SVE modes.
11394 (aarch64_preferred_simd_mode): Update call accordingly. Handle
11396 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
11398 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11399 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11400 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11401 (aarch64_sve_float_mul_immediate_p): New functions.
11402 (aarch64_sve_valid_immediate): New function.
11403 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
11404 Explicitly reject structure modes. Check for INDEX constants.
11405 Handle PTRUE and PFALSE constants.
11406 (aarch64_check_zero_based_sve_index_immediate): New function.
11407 (aarch64_simd_imm_zero_p): Delete.
11408 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
11409 vector modes. Accept constants in the range of CNT[BHWD].
11410 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
11411 ask for an Advanced SIMD mode.
11412 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
11413 (aarch64_simd_vector_alignment): Handle SVE predicates.
11414 (aarch64_vectorize_preferred_vector_alignment): New function.
11415 (aarch64_simd_vector_alignment_reachable): Use it instead of
11417 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
11418 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
11420 (MAX_VECT_LEN): Delete.
11421 (expand_vec_perm_d): Add a vec_flags field.
11422 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
11423 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
11424 (aarch64_evpc_ext): Don't apply a big-endian lane correction
11426 (aarch64_evpc_rev): Rename to...
11427 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
11428 (aarch64_evpc_rev_global): New function.
11429 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
11430 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
11432 (aarch64_evpc_sve_tbl): New function.
11433 (aarch64_expand_vec_perm_const_1): Update after rename of
11434 aarch64_evpc_rev. Handle SVE permutes too, trying
11435 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
11436 than aarch64_evpc_tbl.
11437 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
11438 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
11439 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
11440 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
11441 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
11442 (aarch64_expand_sve_vcond): New functions.
11443 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
11444 of aarch64_vector_mode_p.
11445 (aarch64_dwarf_poly_indeterminate_value): New function.
11446 (aarch64_compute_pressure_classes): Likewise.
11447 (aarch64_can_change_mode_class): Likewise.
11448 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
11449 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
11450 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
11451 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
11452 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
11453 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
11454 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
11455 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
11457 (Dn, Dl, Dr): Accept const as well as const_vector.
11458 (Dz): Likewise. Compare against CONST0_RTX.
11459 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
11460 of "vector" where appropriate.
11461 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
11462 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
11463 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
11464 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
11465 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
11466 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
11467 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
11468 (v_int_equiv): Extend to SVE modes.
11469 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
11471 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
11472 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
11473 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
11474 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
11475 (SVE_COND_FP_CMP): New int iterators.
11476 (perm_hilo): Handle the new unpack unspecs.
11477 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
11479 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
11480 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
11481 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
11482 (aarch64_equality_operator, aarch64_constant_vector_operand)
11483 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
11484 (aarch64_sve_nonimmediate_operand): Likewise.
11485 (aarch64_sve_general_operand): Likewise.
11486 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
11487 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
11488 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
11489 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
11490 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
11491 (aarch64_sve_float_arith_immediate): Likewise.
11492 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
11493 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
11494 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
11495 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
11496 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
11497 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
11498 (aarch64_sve_float_arith_operand): Likewise.
11499 (aarch64_sve_float_arith_with_sub_operand): Likewise.
11500 (aarch64_sve_float_mul_operand): Likewise.
11501 (aarch64_sve_vec_perm_operand): Likewise.
11502 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
11503 (aarch64_mov_operand): Accept const_poly_int and const_vector.
11504 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
11505 as well as const_vector.
11506 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
11507 in file. Use CONST0_RTX and CONSTM1_RTX.
11508 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
11509 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
11510 Use aarch64_simd_imm_zero.
11511 * config/aarch64/aarch64-sve.md: New file.
11512 * config/aarch64/aarch64.md: Include it.
11513 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
11514 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
11515 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
11516 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
11517 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
11518 (sve): New attribute.
11519 (enabled): Disable instructions with the sve attribute unless
11521 (movqi, movhi): Pass CONST_POLY_INT operaneds through
11522 aarch64_expand_mov_immediate.
11523 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
11524 CNT[BHSD] immediates.
11525 (movti): Split CONST_POLY_INT moves into two halves.
11526 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
11527 Split additions that need a temporary here if the destination
11528 is the stack pointer.
11529 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
11530 (*add<mode>3_poly_1): New instruction.
11531 (set_clobber_cc): New expander.
11533 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11535 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
11536 parameter and use it instead of GET_MODE_SIZE (innermode). Use
11537 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
11538 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
11539 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
11540 Change innermode from fixed_mode_size to machine_mode.
11541 (simplify_subreg): Update call accordingly. Handle a constant-sized
11542 subreg of a variable-length CONST_VECTOR.
11544 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11545 Alan Hayward <alan.hayward@arm.com>
11546 David Sherwood <david.sherwood@arm.com>
11548 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
11549 (add_offset_to_base): New function, split out from...
11550 (create_mem_ref): ...here. When handling a scale other than 1,
11551 check first whether the address is valid without the offset.
11552 Add it into the base if so, leaving the index and scale as-is.
11554 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11557 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
11558 fold_for_warn before checking if arg2 is INTEGER_CST.
11560 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
11562 * config/rs6000/predicates.md (load_multiple_operation): Delete.
11563 (store_multiple_operation): Delete.
11564 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
11565 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
11566 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
11567 guarded by TARGET_STRING.
11568 (rs6000_output_load_multiple): Delete.
11569 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
11570 OPTION_MASK_STRING / TARGET_STRING handling.
11571 (print_operand) <'N', 'O'>: Add comment that these are unused now.
11572 (const rs6000_opt_masks) <"string">: Change mask to 0.
11573 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
11574 (MASK_STRING): Delete.
11575 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
11577 (load_multiple): Delete.
11584 (store_multiple): Delete.
11591 (movmemsi_8reg): Delete.
11592 (corresponding unnamed define_insn): Delete.
11593 (movmemsi_6reg): Delete.
11594 (corresponding unnamed define_insn): Delete.
11595 (movmemsi_4reg): Delete.
11596 (corresponding unnamed define_insn): Delete.
11597 (movmemsi_2reg): Delete.
11598 (corresponding unnamed define_insn): Delete.
11599 (movmemsi_1reg): Delete.
11600 (corresponding unnamed define_insn): Delete.
11601 * config/rs6000/rs6000.opt (mno-string): New.
11602 (mstring): Replace by deprecation warning stub.
11603 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
11605 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11607 * regrename.c (regrename_do_replace): If replacing the same
11608 reg multiple times, try to reuse last created gen_raw_REG.
11611 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
11612 main to workaround a bug in GDB.
11614 2018-01-12 Tom de Vries <tom@codesourcery.com>
11617 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
11619 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
11621 PR rtl-optimization/80481
11622 * ira-color.c (get_cap_member): New function.
11623 (allocnos_conflict_by_live_ranges_p): Use it.
11624 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
11625 (setup_slot_coalesced_allocno_live_ranges): Ditto.
11627 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
11630 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
11631 (*saddl_se_1): Ditto.
11632 (*ssubsi_1): Ditto.
11633 (*ssubl_se_1): Ditto.
11635 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11637 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
11638 rather than wi::to_widest for DR_INITs.
11639 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
11640 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
11641 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
11643 (vect_analyze_group_access_1): Note that here.
11645 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11647 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
11648 polynomial type sizes.
11650 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11652 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
11653 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
11654 (gimple_add_tmp_var): Likewise.
11656 2018-01-12 Martin Liska <mliska@suse.cz>
11658 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
11659 (gimple_alloc_sizes): Likewise.
11660 (dump_gimple_statistics): Use PRIu64 in printf format.
11661 * gimple.h: Change uint64_t to int.
11663 2018-01-12 Martin Liska <mliska@suse.cz>
11665 * tree-core.h: Use uint64_t instead of int.
11666 * tree.c (tree_node_counts): Likewise.
11667 (tree_node_sizes): Likewise.
11668 (dump_tree_statistics): Use PRIu64 in printf format.
11670 2018-01-12 Martin Liska <mliska@suse.cz>
11672 * Makefile.in: As qsort_chk is implemented in vec.c, add
11673 vec.o to linkage of gencfn-macros.
11674 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
11675 passing the info to record_node_allocation_statistics.
11676 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
11678 * ggc-common.c (struct ggc_usage): Add operator== and use
11679 it in operator< and compare function.
11680 * mem-stats.h (struct mem_usage): Likewise.
11681 * vec.c (struct vec_usage): Remove operator< and compare
11682 function. Can be simply inherited.
11684 2018-01-12 Martin Jambor <mjambor@suse.cz>
11687 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
11688 * tree-ssa-math-opts.c: Include domwalk.h.
11689 (convert_mult_to_fma_1): New function.
11690 (fma_transformation_info): New type.
11691 (fma_deferring_state): Likewise.
11692 (cancel_fma_deferring): New function.
11693 (result_of_phi): Likewise.
11694 (last_fma_candidate_feeds_initial_phi): Likewise.
11695 (convert_mult_to_fma): Added deferring logic, split actual
11696 transformation to convert_mult_to_fma_1.
11697 (math_opts_dom_walker): New type.
11698 (math_opts_dom_walker::after_dom_children): New method, body moved
11699 here from pass_optimize_widening_mul::execute, added deferring logic
11701 (pass_optimize_widening_mul::execute): Moved most of code to
11702 math_opts_dom_walker::after_dom_children.
11703 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
11704 * config/i386/i386.c (ix86_option_override_internal): Added
11705 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
11707 2018-01-12 Richard Biener <rguenther@suse.de>
11710 * dwarf2out.c (gen_variable_die): Do not reset old_die for
11711 inline instance vars.
11713 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
11716 * config/rx/rx.c (rx_is_restricted_memory_address):
11717 Handle SUBREG case.
11719 2018-01-12 Richard Biener <rguenther@suse.de>
11721 PR tree-optimization/80846
11722 * target.def (split_reduction): New target hook.
11723 * targhooks.c (default_split_reduction): New function.
11724 * targhooks.h (default_split_reduction): Declare.
11725 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
11726 target requests first reduce vectors by combining low and high
11728 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
11729 (get_vectype_for_scalar_type_and_size): Export.
11730 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
11731 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
11732 * doc/tm.texi: Regenerate.
11733 * config/i386/i386.c (ix86_split_reduction): Implement
11734 TARGET_VECTORIZE_SPLIT_REDUCTION.
11736 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11739 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
11740 in PIC mode except for TARGET_VXWORKS_RTP.
11741 * config/sparc/sparc.c: Include cfgrtl.h.
11742 (TARGET_INIT_PIC_REG): Define.
11743 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
11744 (sparc_pic_register_p): New predicate.
11745 (sparc_legitimate_address_p): Use it.
11746 (sparc_legitimize_pic_address): Likewise.
11747 (sparc_delegitimize_address): Likewise.
11748 (sparc_mode_dependent_address_p): Likewise.
11749 (gen_load_pcrel_sym): Remove 4th parameter.
11750 (load_got_register): Adjust call to above. Remove obsolete stuff.
11751 (sparc_expand_prologue): Do not call load_got_register here.
11752 (sparc_flat_expand_prologue): Likewise.
11753 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
11754 (sparc_use_pseudo_pic_reg): New function.
11755 (sparc_init_pic_reg): Likewise.
11756 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11757 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
11759 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
11761 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
11762 Add item for branch_cost.
11764 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11766 PR rtl-optimization/83565
11767 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
11768 not extend the result to a larger mode for rotate operations.
11769 (num_sign_bit_copies1): Likewise.
11771 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11774 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
11776 Use values-Xc.o for -pedantic.
11777 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
11779 2018-01-12 Martin Liska <mliska@suse.cz>
11782 * ipa-devirt.c (final_warning_record::grow_type_warnings):
11784 (possible_polymorphic_call_targets): Use it.
11785 (ipa_devirt): Likewise.
11787 2018-01-12 Martin Liska <mliska@suse.cz>
11789 * profile-count.h (enum profile_quality): Use 0 as invalid
11790 enum value of profile_quality.
11792 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
11794 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
11795 -mext-string options.
11797 2018-01-12 Richard Biener <rguenther@suse.de>
11799 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
11800 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
11801 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
11803 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
11805 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
11807 * configure.ac (--with-long-double-format): Add support for the
11808 configuration option to change the default long double format on
11810 * config.gcc (powerpc*-linux*-*): Likewise.
11811 * configure: Regenerate.
11812 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
11813 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
11814 used without modification.
11816 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11818 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
11819 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
11820 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
11821 MISC_BUILTIN_SPEC_BARRIER.
11822 (rs6000_init_builtins): Likewise.
11823 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
11825 (speculation_barrier): New define_insn.
11826 * doc/extend.texi: Document __builtin_speculation_barrier.
11828 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11831 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
11832 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
11833 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
11835 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
11836 integral modes instead of "ss" and "sd".
11837 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
11838 vectors with 32-bit and 64-bit elements.
11839 (vecdupssescalarmodesuffix): New mode attribute.
11840 (vec_dup<mode>): Use it.
11842 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
11845 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
11846 frame if argument is passed on stack.
11848 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11851 * ree.c (combine_reaching_defs): Optimize also
11852 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
11853 reg2=any_extend(exp); reg1=reg2;, formatting fix.
11855 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11857 PR middle-end/83189
11858 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
11860 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11862 PR middle-end/83718
11863 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
11864 after they are computed.
11866 2018-01-11 Bin Cheng <bin.cheng@arm.com>
11868 PR tree-optimization/83695
11869 * gimple-loop-linterchange.cc
11870 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
11871 reset cached scev information after interchange.
11872 (pass_linterchange::execute): Remove call to scev_reset_htab.
11874 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11876 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
11877 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
11878 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
11879 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
11880 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
11881 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
11882 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
11883 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
11884 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
11885 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
11886 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
11887 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
11888 (V_lane_reg): Likewise.
11889 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
11891 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
11892 (vfmal_lane_low<mode>_intrinsic,
11893 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
11894 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
11895 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
11896 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
11897 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
11898 vfmsl_lane_high<mode>_intrinsic): New define_insns.
11900 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11902 * config/arm/arm-cpus.in (fp16fml): New feature.
11903 (ALL_SIMD): Add fp16fml.
11904 (armv8.2-a): Add fp16fml as an option.
11905 (armv8.3-a): Likewise.
11906 (armv8.4-a): Add fp16fml as part of fp16.
11907 * config/arm/arm.h (TARGET_FP16FML): Define.
11908 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
11910 * config/arm/arm-modes.def (V2HF): Define.
11911 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
11912 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
11913 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
11914 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
11915 vfmsl_low, vfmsl_high): New set of builtins.
11916 * config/arm/iterators.md (PLUSMINUS): New code iterator.
11917 (vfml_op): New code attribute.
11918 (VFMLHALVES): New int iterator.
11919 (VFML, VFMLSEL): New mode attributes.
11920 (V_reg): Define mapping for V2HF.
11921 (V_hi, V_lo): New mode attributes.
11922 (VF_constraint): Likewise.
11923 (vfml_half, vfml_half_selector): New int attributes.
11924 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
11926 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
11927 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
11929 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
11930 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
11931 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
11932 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
11934 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
11935 Document new effective target and option set.
11937 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11939 * config/arm/arm-cpus.in (armv8_4): New feature.
11940 (ARMv8_4a): New fgroup.
11941 (armv8.4-a): New arch.
11942 * config/arm/arm-tables.opt: Regenerate.
11943 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
11944 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
11945 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
11946 Add matching rules for -march=armv8.4-a and extensions.
11947 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
11949 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
11952 * config/rx/rx.md (BW): New mode attribute.
11953 (sync_lock_test_and_setsi): Add mode suffix to insn output.
11955 2018-01-11 Richard Biener <rguenther@suse.de>
11957 PR tree-optimization/83435
11958 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
11959 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
11960 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
11962 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11963 Alan Hayward <alan.hayward@arm.com>
11964 David Sherwood <david.sherwood@arm.com>
11966 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
11968 (aarch64_classify_address): Initialize it. Track polynomial offsets.
11969 (aarch64_print_address_internal): Use it to check for a zero offset.
11971 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11972 Alan Hayward <alan.hayward@arm.com>
11973 David Sherwood <david.sherwood@arm.com>
11975 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
11976 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
11977 Return a poly_int64 rather than a HOST_WIDE_INT.
11978 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
11979 rather than a HOST_WIDE_INT.
11980 * config/aarch64/aarch64.h (aarch64_frame): Protect with
11981 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
11982 hard_fp_offset, frame_size, initial_adjust, callee_offset and
11983 final_offset from HOST_WIDE_INT to poly_int64.
11984 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
11985 to_constant when getting the number of units in an Advanced SIMD
11987 (aarch64_builtin_vectorized_function): Check for a constant number
11989 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
11991 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
11992 attribute instead of GET_MODE_NUNITS.
11993 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
11994 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
11995 GET_MODE_SIZE for fixed-size registers.
11996 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
11997 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
11998 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
11999 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
12000 (aarch64_print_operand, aarch64_print_address_internal)
12001 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
12002 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
12003 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
12004 Handle polynomial GET_MODE_SIZE.
12005 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
12006 wider than SImode without modification.
12007 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
12008 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
12009 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
12010 passing and returning SVE modes.
12011 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
12012 rather than GEN_INT.
12013 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
12014 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
12015 (aarch64_allocate_and_probe_stack_space): Likewise.
12016 (aarch64_layout_frame): Cope with polynomial offsets.
12017 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
12018 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
12019 polynomial offsets.
12020 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
12021 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
12022 poly_int64 rather than a HOST_WIDE_INT.
12023 (aarch64_get_separate_components, aarch64_process_components)
12024 (aarch64_expand_prologue, aarch64_expand_epilogue)
12025 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
12026 (aarch64_anchor_offset): New function, split out from...
12027 (aarch64_legitimize_address): ...here.
12028 (aarch64_builtin_vectorization_cost): Handle polynomial
12029 TYPE_VECTOR_SUBPARTS.
12030 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
12032 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
12033 number of elements from the PARALLEL rather than the mode.
12034 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
12035 rather than GET_MODE_BITSIZE.
12036 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
12037 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
12038 (aarch64_expand_vec_perm_const_1): Handle polynomial
12039 d->perm.length () and d->perm elements.
12040 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
12041 Apply to_constant to d->perm elements.
12042 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
12043 polynomial CONST_VECTOR_NUNITS.
12044 (aarch64_move_pointer): Take amount as a poly_int64 rather
12046 (aarch64_progress_pointer): Avoid temporary variable.
12047 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
12048 the mode attribute instead of GET_MODE.
12050 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12051 Alan Hayward <alan.hayward@arm.com>
12052 David Sherwood <david.sherwood@arm.com>
12054 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
12055 x exists before using it.
12056 (aarch64_add_constant_internal): Rename to...
12057 (aarch64_add_offset_1): ...this. Replace regnum with separate
12058 src and dest rtxes. Handle the case in which they're different,
12059 including when the offset is zero. Replace scratchreg with an rtx.
12060 Use 2 additions if there is no spare register into which we can
12061 move a 16-bit constant.
12062 (aarch64_add_constant): Delete.
12063 (aarch64_add_offset): Replace reg with separate src and dest
12064 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
12065 Use aarch64_add_offset_1.
12066 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
12067 an rtx rather than an int. Take the delta as a poly_int64
12068 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
12069 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
12070 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
12071 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
12072 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
12073 and aarch64_add_sp.
12074 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
12075 aarch64_add_constant.
12077 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12079 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
12080 Use scalar_float_mode.
12082 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12084 * config/aarch64/aarch64-simd.md
12085 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
12086 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
12087 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
12088 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
12089 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
12090 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
12091 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
12092 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
12093 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
12094 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
12096 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12099 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
12100 targ_options->x_arm_arch_string is non NULL.
12102 2018-01-11 Tamar Christina <tamar.christina@arm.com>
12104 * config/aarch64/aarch64.h
12105 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
12107 2018-01-11 Sudakshina Das <sudi.das@arm.com>
12110 * expmed.c (emit_store_flag_force): Swap if const op0
12111 and change VOIDmode to mode of op0.
12113 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12115 PR rtl-optimization/83761
12116 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
12117 than bytes to mode_for_size.
12119 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12121 PR middle-end/83189
12122 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
12123 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
12126 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12128 PR middle-end/83575
12129 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
12130 when in layout mode.
12131 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
12132 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
12135 2018-01-10 Michael Collison <michael.collison@arm.com>
12137 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
12138 * config/aarch64/aarch64-option-extension.def: Add
12139 AARCH64_OPT_EXTENSION of 'fp16fml'.
12140 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12141 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
12142 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
12143 * config/aarch64/constraints.md (Ui7): New constraint.
12144 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
12145 (VFMLA_SEL_W): Ditto.
12148 (VFMLA16_LOW): New int iterator.
12149 (VFMLA16_HIGH): Ditto.
12150 (UNSPEC_FMLAL): New unspec.
12151 (UNSPEC_FMLSL): Ditto.
12152 (UNSPEC_FMLAL2): Ditto.
12153 (UNSPEC_FMLSL2): Ditto.
12154 (f16mac): New code attribute.
12155 * config/aarch64/aarch64-simd-builtins.def
12156 (aarch64_fmlal_lowv2sf): Ditto.
12157 (aarch64_fmlsl_lowv2sf): Ditto.
12158 (aarch64_fmlalq_lowv4sf): Ditto.
12159 (aarch64_fmlslq_lowv4sf): Ditto.
12160 (aarch64_fmlal_highv2sf): Ditto.
12161 (aarch64_fmlsl_highv2sf): Ditto.
12162 (aarch64_fmlalq_highv4sf): Ditto.
12163 (aarch64_fmlslq_highv4sf): Ditto.
12164 (aarch64_fmlal_lane_lowv2sf): Ditto.
12165 (aarch64_fmlsl_lane_lowv2sf): Ditto.
12166 (aarch64_fmlal_laneq_lowv2sf): Ditto.
12167 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
12168 (aarch64_fmlalq_lane_lowv4sf): Ditto.
12169 (aarch64_fmlsl_lane_lowv4sf): Ditto.
12170 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
12171 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
12172 (aarch64_fmlal_lane_highv2sf): Ditto.
12173 (aarch64_fmlsl_lane_highv2sf): Ditto.
12174 (aarch64_fmlal_laneq_highv2sf): Ditto.
12175 (aarch64_fmlsl_laneq_highv2sf): Ditto.
12176 (aarch64_fmlalq_lane_highv4sf): Ditto.
12177 (aarch64_fmlsl_lane_highv4sf): Ditto.
12178 (aarch64_fmlalq_laneq_highv4sf): Ditto.
12179 (aarch64_fmlsl_laneq_highv4sf): Ditto.
12180 * config/aarch64/aarch64-simd.md:
12181 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
12182 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12183 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
12184 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12185 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
12186 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
12187 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
12188 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
12189 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
12190 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
12191 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
12192 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
12193 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
12194 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
12195 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
12196 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
12197 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
12198 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
12199 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
12200 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
12201 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
12202 (vfmlsl_low_u32): Ditto.
12203 (vfmlalq_low_u32): Ditto.
12204 (vfmlslq_low_u32): Ditto.
12205 (vfmlal_high_u32): Ditto.
12206 (vfmlsl_high_u32): Ditto.
12207 (vfmlalq_high_u32): Ditto.
12208 (vfmlslq_high_u32): Ditto.
12209 (vfmlal_lane_low_u32): Ditto.
12210 (vfmlsl_lane_low_u32): Ditto.
12211 (vfmlal_laneq_low_u32): Ditto.
12212 (vfmlsl_laneq_low_u32): Ditto.
12213 (vfmlalq_lane_low_u32): Ditto.
12214 (vfmlslq_lane_low_u32): Ditto.
12215 (vfmlalq_laneq_low_u32): Ditto.
12216 (vfmlslq_laneq_low_u32): Ditto.
12217 (vfmlal_lane_high_u32): Ditto.
12218 (vfmlsl_lane_high_u32): Ditto.
12219 (vfmlal_laneq_high_u32): Ditto.
12220 (vfmlsl_laneq_high_u32): Ditto.
12221 (vfmlalq_lane_high_u32): Ditto.
12222 (vfmlslq_lane_high_u32): Ditto.
12223 (vfmlalq_laneq_high_u32): Ditto.
12224 (vfmlslq_laneq_high_u32): Ditto.
12225 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
12226 (AARCH64_FL_FOR_ARCH8_4): New.
12227 (AARCH64_ISA_F16FML): New ISA flag.
12228 (TARGET_F16FML): New feature flag for fp16fml.
12229 (doc/invoke.texi): Document new fp16fml option.
12231 2018-01-10 Michael Collison <michael.collison@arm.com>
12233 * config/aarch64/aarch64-builtins.c:
12234 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
12235 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12236 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
12237 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
12238 (AARCH64_ISA_SHA3): New ISA flag.
12239 (TARGET_SHA3): New feature flag for sha3.
12240 * config/aarch64/iterators.md (sha512_op): New int attribute.
12241 (CRYPTO_SHA512): New int iterator.
12242 (UNSPEC_SHA512H): New unspec.
12243 (UNSPEC_SHA512H2): Ditto.
12244 (UNSPEC_SHA512SU0): Ditto.
12245 (UNSPEC_SHA512SU1): Ditto.
12246 * config/aarch64/aarch64-simd-builtins.def
12247 (aarch64_crypto_sha512hqv2di): New builtin.
12248 (aarch64_crypto_sha512h2qv2di): Ditto.
12249 (aarch64_crypto_sha512su0qv2di): Ditto.
12250 (aarch64_crypto_sha512su1qv2di): Ditto.
12251 (aarch64_eor3qv8hi): Ditto.
12252 (aarch64_rax1qv2di): Ditto.
12253 (aarch64_xarqv2di): Ditto.
12254 (aarch64_bcaxqv8hi): Ditto.
12255 * config/aarch64/aarch64-simd.md:
12256 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
12257 (aarch64_crypto_sha512su0qv2di): Ditto.
12258 (aarch64_crypto_sha512su1qv2di): Ditto.
12259 (aarch64_eor3qv8hi): Ditto.
12260 (aarch64_rax1qv2di): Ditto.
12261 (aarch64_xarqv2di): Ditto.
12262 (aarch64_bcaxqv8hi): Ditto.
12263 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
12264 (vsha512h2q_u64): Ditto.
12265 (vsha512su0q_u64): Ditto.
12266 (vsha512su1q_u64): Ditto.
12267 (veor3q_u16): Ditto.
12268 (vrax1q_u64): Ditto.
12269 (vxarq_u64): Ditto.
12270 (vbcaxq_u16): Ditto.
12271 * config/arm/types.md (crypto_sha512): New type attribute.
12272 (crypto_sha3): Ditto.
12273 (doc/invoke.texi): Document new sha3 option.
12275 2018-01-10 Michael Collison <michael.collison@arm.com>
12277 * config/aarch64/aarch64-builtins.c:
12278 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
12279 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12280 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
12281 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
12282 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
12283 (AARCH64_ISA_SM4): New ISA flag.
12284 (TARGET_SM4): New feature flag for sm4.
12285 * config/aarch64/aarch64-simd-builtins.def
12286 (aarch64_sm3ss1qv4si): Ditto.
12287 (aarch64_sm3tt1aq4si): Ditto.
12288 (aarch64_sm3tt1bq4si): Ditto.
12289 (aarch64_sm3tt2aq4si): Ditto.
12290 (aarch64_sm3tt2bq4si): Ditto.
12291 (aarch64_sm3partw1qv4si): Ditto.
12292 (aarch64_sm3partw2qv4si): Ditto.
12293 (aarch64_sm4eqv4si): Ditto.
12294 (aarch64_sm4ekeyqv4si): Ditto.
12295 * config/aarch64/aarch64-simd.md:
12296 (aarch64_sm3ss1qv4si): Ditto.
12297 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
12298 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
12299 (aarch64_sm4eqv4si): Ditto.
12300 (aarch64_sm4ekeyqv4si): Ditto.
12301 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
12302 (sm3part_op): Ditto.
12303 (CRYPTO_SM3TT): Ditto.
12304 (CRYPTO_SM3PART): Ditto.
12305 (UNSPEC_SM3SS1): New unspec.
12306 (UNSPEC_SM3TT1A): Ditto.
12307 (UNSPEC_SM3TT1B): Ditto.
12308 (UNSPEC_SM3TT2A): Ditto.
12309 (UNSPEC_SM3TT2B): Ditto.
12310 (UNSPEC_SM3PARTW1): Ditto.
12311 (UNSPEC_SM3PARTW2): Ditto.
12312 (UNSPEC_SM4E): Ditto.
12313 (UNSPEC_SM4EKEY): Ditto.
12314 * config/aarch64/constraints.md (Ui2): New constraint.
12315 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
12316 * config/arm/types.md (crypto_sm3): New type attribute.
12317 (crypto_sm4): Ditto.
12318 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
12319 (vsm3tt1aq_u32): Ditto.
12320 (vsm3tt1bq_u32): Ditto.
12321 (vsm3tt2aq_u32): Ditto.
12322 (vsm3tt2bq_u32): Ditto.
12323 (vsm3partw1q_u32): Ditto.
12324 (vsm3partw2q_u32): Ditto.
12325 (vsm4eq_u32): Ditto.
12326 (vsm4ekeyq_u32): Ditto.
12327 (doc/invoke.texi): Document new sm4 option.
12329 2018-01-10 Michael Collison <michael.collison@arm.com>
12331 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
12332 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
12333 (AARCH64_FL_FOR_ARCH8_4): New.
12334 (AARCH64_FL_V8_4): New flag.
12335 (doc/invoke.texi): Document new armv8.4-a option.
12337 2018-01-10 Michael Collison <michael.collison@arm.com>
12339 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12340 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
12341 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
12342 * config/aarch64/aarch64-option-extension.def: Add
12343 AARCH64_OPT_EXTENSION of 'sha2'.
12344 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
12345 (crypto): Disable sha2 and aes if crypto disabled.
12346 (crypto): Enable aes and sha2 if enabled.
12347 (simd): Disable sha2 and aes if simd disabled.
12348 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
12350 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
12351 (TARGET_SHA2): New feature flag for sha2.
12352 (TARGET_AES): New feature flag for aes.
12353 * config/aarch64/aarch64-simd.md:
12354 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
12355 conditional on TARGET_AES.
12356 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
12357 (aarch64_crypto_sha1hsi): Make pattern conditional
12359 (aarch64_crypto_sha1hv4si): Ditto.
12360 (aarch64_be_crypto_sha1hv4si): Ditto.
12361 (aarch64_crypto_sha1su1v4si): Ditto.
12362 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
12363 (aarch64_crypto_sha1su0v4si): Ditto.
12364 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
12365 (aarch64_crypto_sha256su0v4si): Ditto.
12366 (aarch64_crypto_sha256su1v4si): Ditto.
12367 (doc/invoke.texi): Document new aes and sha2 options.
12369 2018-01-10 Martin Sebor <msebor@redhat.com>
12371 PR tree-optimization/83781
12372 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
12375 2018-01-11 Martin Sebor <msebor@gmail.com>
12376 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12378 PR tree-optimization/83501
12379 PR tree-optimization/81703
12381 * tree-ssa-strlen.c (get_string_cst): Rename...
12382 (get_string_len): ...to this. Handle global constants.
12383 (handle_char_store): Adjust.
12385 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
12386 Jim Wilson <jimw@sifive.com>
12388 * config/riscv/riscv-protos.h (riscv_output_return): New.
12389 * config/riscv/riscv.c (struct machine_function): New naked_p field.
12390 (riscv_attribute_table, riscv_output_return),
12391 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
12392 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
12393 (riscv_compute_frame_info): Only compute frame->mask if not a naked
12395 (riscv_expand_prologue): Add early return for naked function.
12396 (riscv_expand_epilogue): Likewise.
12397 (riscv_function_ok_for_sibcall): Return false for naked function.
12398 (riscv_set_current_function): New.
12399 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
12400 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
12401 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
12402 * doc/extend.texi (RISC-V Function Attributes): New.
12404 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
12406 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
12407 check for 128-bit long double before checking TCmode.
12408 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
12409 128-bit long doubles before checking TFmode or TCmode.
12410 (FLOAT128_IBM_P): Likewise.
12412 2018-01-10 Martin Sebor <msebor@redhat.com>
12414 PR tree-optimization/83671
12415 * builtins.c (c_strlen): Unconditionally return zero for the empty
12417 Use -Warray-bounds for warnings.
12418 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
12419 for non-constant array indices with COMPONENT_REF, arrays of
12420 arrays, and pointers to arrays.
12421 (gimple_fold_builtin_strlen): Determine and set length range for
12422 non-constant character arrays.
12424 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
12426 PR middle-end/81897
12427 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
12430 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
12432 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
12434 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12437 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
12438 VECTOR_MEM_ALTIVEC_OR_VSX_P.
12439 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
12440 indexed_or_indirect_operand predicate.
12441 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
12442 (*vsx_le_perm_load_v8hi): Likewise.
12443 (*vsx_le_perm_load_v16qi): Likewise.
12444 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
12445 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
12446 (*vsx_le_perm_store_v8hi): Likewise.
12447 (*vsx_le_perm_store_v16qi): Likewise.
12448 (eight unnamed splitters): Likewise.
12450 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12452 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
12453 * config/rs6000/emmintrin.h: Likewise.
12454 * config/rs6000/mmintrin.h: Likewise.
12455 * config/rs6000/xmmintrin.h: Likewise.
12457 2018-01-10 David Malcolm <dmalcolm@redhat.com>
12460 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
12462 * tree.c (tree_nop_conversion): Return true for location wrapper
12464 (maybe_wrap_with_location): New function.
12465 (selftest::check_strip_nops): New function.
12466 (selftest::test_location_wrappers): New function.
12467 (selftest::tree_c_tests): Call it.
12468 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
12469 (maybe_wrap_with_location): New decl.
12470 (EXPR_LOCATION_WRAPPER_P): New macro.
12471 (location_wrapper_p): New inline function.
12472 (tree_strip_any_location_wrapper): New inline function.
12474 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
12477 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
12478 stack_realign_offset for the largest alignment of stack slot
12480 (ix86_find_max_used_stack_alignment): New function.
12481 (ix86_finalize_stack_frame_flags): Use it. Set
12482 max_used_stack_alignment if we don't realign stack.
12483 * config/i386/i386.h (machine_function): Add
12484 max_used_stack_alignment.
12486 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
12488 * config/arm/arm.opt (-mbranch-cost): New option.
12489 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
12492 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
12495 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
12496 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
12498 2018-01-10 Richard Biener <rguenther@suse.de>
12501 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
12502 early out so it also covers the case where we have a non-NULL
12505 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12507 PR tree-optimization/83753
12508 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
12509 for non-strided grouped accesses if the number of elements is 1.
12511 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12514 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
12515 * i386.h (TARGET_USE_GATHER): Define.
12516 * x86-tune.def (X86_TUNE_USE_GATHER): New.
12518 2018-01-10 Martin Liska <mliska@suse.cz>
12521 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
12522 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
12524 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
12525 CLEANUP_NO_PARTITIONING is not set.
12527 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12529 * doc/rtl.texi: Remove documentation of (const ...) wrappers
12530 for vectors, as a partial revert of r254296.
12531 * rtl.h (const_vec_p): Delete.
12532 (const_vec_duplicate_p): Don't test for vector CONSTs.
12533 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
12534 * expmed.c (make_tree): Likewise.
12537 * common.md (E, F): Use CONSTANT_P instead of checking for
12539 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
12540 checking for CONST_VECTOR.
12542 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12544 PR middle-end/83575
12545 * predict.c (force_edge_cold): Handle in more sane way edges
12546 with no prediction.
12548 2018-01-09 Carl Love <cel@us.ibm.com>
12550 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
12552 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
12553 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
12554 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
12555 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
12556 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
12557 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
12558 * config/rs6000/rs6000-protos.h: Add extern defition for
12559 rs6000_generate_float2_double_code.
12560 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
12562 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
12563 (float2_v2df): Add define_expand.
12565 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
12568 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
12569 op_mode in the force_to_mode call.
12571 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12573 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
12574 instead of checking each element individually.
12575 (aarch64_evpc_uzp): Likewise.
12576 (aarch64_evpc_zip): Likewise.
12577 (aarch64_evpc_ext): Likewise.
12578 (aarch64_evpc_rev): Likewise.
12579 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
12580 instead of checking each element individually. Return true without
12582 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
12583 whether all selected elements come from the same input, instead of
12584 checking each element individually. Remove calls to gen_rtx_REG,
12585 start_sequence and end_sequence and instead assert that no rtl is
12588 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12590 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
12591 order of HIGH and CONST checks.
12593 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12595 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
12596 if the destination isn't an SSA_NAME.
12598 2018-01-09 Richard Biener <rguenther@suse.de>
12600 PR tree-optimization/83668
12601 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
12603 (canonicalize_loop_form): ... here, renamed from ...
12604 (canonicalize_loop_closed_ssa_form): ... this and amended to
12605 swap successor edges for loop exit blocks to make us use
12606 the RPO order we need for initial schedule generation.
12608 2018-01-09 Joseph Myers <joseph@codesourcery.com>
12610 PR tree-optimization/64811
12611 * match.pd: When optimizing comparisons with Inf, avoid
12612 introducing or losing exceptions from comparisons with NaN.
12614 2018-01-09 Martin Liska <mliska@suse.cz>
12617 * asan.c (shadow_mem_size): Add gcc_assert.
12619 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
12621 Don't save registers in main().
12624 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
12625 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
12626 * config/avr/avr.c (avr_set_current_function): Don't error if
12627 naked, OS_task or OS_main are specified at the same time.
12628 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
12630 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
12632 * common/config/avr/avr-common.c (avr_option_optimization_table):
12633 Switch on -mmain-is-OS_task for optimizing compilations.
12635 2018-01-09 Richard Biener <rguenther@suse.de>
12637 PR tree-optimization/83572
12638 * graphite.c: Include cfganal.h.
12639 (graphite_transform_loops): Connect infinite loops to exit
12640 and remove fake edges at the end.
12642 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12644 * ipa-inline.c (edge_badness): Revert accidental checkin.
12646 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12649 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
12650 symbols; not inline clones.
12652 2018-01-09 Jakub Jelinek <jakub@redhat.com>
12655 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
12656 hard registers. Formatting fixes.
12658 PR preprocessor/83722
12659 * gcc.c (try_generate_repro): Pass
12660 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
12661 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
12664 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
12665 Kito Cheng <kito.cheng@gmail.com>
12667 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
12668 (riscv_leaf_function_p): Delete.
12669 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
12671 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12673 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
12675 (do_ifelse): New function.
12676 (do_isel): New function.
12677 (do_sub3): New function.
12678 (do_add3): New function.
12679 (do_load_mask_compare): New function.
12680 (do_overlap_load_compare): New function.
12681 (expand_compare_loop): New function.
12682 (expand_block_compare): Call expand_compare_loop() when appropriate.
12683 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
12684 option description.
12685 (-mblock-compare-inline-loop-limit): New option.
12687 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12690 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
12691 Reverse order of second and third operands in first alternative.
12692 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
12693 of first and second elements in UNSPEC_VPERMR vector.
12694 (altivec_expand_vec_perm_le): Likewise.
12696 2018-01-08 Jeff Law <law@redhat.com>
12698 PR rtl-optimizatin/81308
12699 * tree-switch-conversion.c (cfg_altered): New file scoped static.
12700 (process_switch): If group_case_labels makes a change, then set
12702 (pass_convert_switch::execute): If a switch is converted, then
12703 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
12705 PR rtl-optimization/81308
12706 * recog.c (split_all_insns): Conditionally cleanup the CFG after
12709 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
12711 PR target/83663 - Revert r255946
12712 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
12713 generation for cases where splatting a value is not useful.
12714 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
12715 across a vec_duplicate and a paradoxical subreg forming a vector
12716 mode to a vec_concat.
12718 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12720 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
12721 -march=armv8.3-a variants.
12722 * config/arm/t-multilib: Likewise.
12723 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
12725 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12727 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
12729 (cceq_ior_compare_complement): Give it a name so I can use it, and
12730 change boolean_or_operator predicate to boolean_operator so it can
12731 be used to generate a crand.
12732 (eqne): New code iterator.
12733 (bd/bd_neg): New code_attrs.
12734 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
12735 a single define_insn.
12736 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
12737 decrement (bdnzt/bdnzf/bdzt/bdzf).
12738 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
12739 with the new names of the branch decrement patterns, and added the
12740 names of the branch decrement conditional patterns.
12742 2018-01-08 Richard Biener <rguenther@suse.de>
12744 PR tree-optimization/83563
12745 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
12748 2018-01-08 Richard Biener <rguenther@suse.de>
12750 PR middle-end/83713
12751 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
12753 2018-01-08 Richard Biener <rguenther@suse.de>
12755 PR tree-optimization/83685
12756 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
12757 references to abnormals.
12759 2018-01-08 Richard Biener <rguenther@suse.de>
12762 * dwarf2out.c (output_indirect_strings): Handle empty
12763 skeleton_debug_str_hash.
12764 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
12766 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12768 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
12769 (emit_store_direct): Likewise.
12770 (arc_trampoline_adjust_address): Likewise.
12771 (arc_asm_trampoline_template): New function.
12772 (arc_initialize_trampoline): Use asm_trampoline_template.
12773 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
12774 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
12775 * config/arc/arc.md (flush_icache): Delete pattern.
12777 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12779 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
12780 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
12783 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12786 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
12787 by not USED_FOR_TARGET.
12788 (make_pass_resolve_sw_modes): Likewise.
12790 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12792 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
12795 2018-01-08 Richard Biener <rguenther@suse.de>
12797 PR middle-end/83580
12798 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
12800 2018-01-08 Richard Biener <rguenther@suse.de>
12802 PR middle-end/83517
12803 * match.pd ((t * 2) / 2) -> t): Add missing :c.
12805 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
12807 PR middle-end/81897
12808 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
12809 basic blocks with a small number of successors.
12810 (convert_control_dep_chain_into_preds): Improve handling of
12812 (dump_predicates): Split apart into...
12813 (dump_pred_chain): ...here...
12814 (dump_pred_info): ...and here.
12815 (can_one_predicate_be_invalidated_p): Add debugging printfs.
12816 (can_chain_union_be_invalidated_p): Improve check for invalidation
12818 (uninit_uses_cannot_happen): Avoid unnecessary if
12819 convert_control_dep_chain_into_preds yielded nothing.
12821 2018-01-06 Martin Sebor <msebor@redhat.com>
12823 PR tree-optimization/83640
12824 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
12825 subtracting negative offset from size.
12826 (builtin_access::overlap): Adjust offset bounds of the access to fall
12827 within the size of the object if possible.
12829 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
12831 PR rtl-optimization/83699
12832 * expmed.c (extract_bit_field_1): Restrict the vector usage of
12833 extract_bit_field_as_subreg to cases in which the extracted
12834 value is also a vector.
12836 * lra-constraints.c (process_alt_operands): Test for the equivalence
12837 substitutions when detecting a possible reload cycle.
12839 2018-01-06 Jakub Jelinek <jakub@redhat.com>
12842 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
12843 by default if flag_selective_schedling{,2}. Formatting fixes.
12845 PR rtl-optimization/83682
12846 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
12847 if it has non-VECTOR_MODE element mode.
12848 (vec_duplicate_p): Likewise.
12850 PR middle-end/83694
12851 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
12852 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
12854 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12857 * config/i386/i386-builtin.def
12858 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
12859 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
12860 Require also OPTION_MASK_ISA_AVX512F in addition to
12861 OPTION_MASK_ISA_GFNI.
12862 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
12863 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
12864 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
12865 to OPTION_MASK_ISA_GFNI.
12866 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
12867 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
12868 OPTION_MASK_ISA_AVX512BW.
12869 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
12870 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
12871 addition to OPTION_MASK_ISA_GFNI.
12872 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
12873 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
12874 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
12875 to OPTION_MASK_ISA_GFNI.
12876 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
12877 a requirement for all ISAs rather than any of them with a few
12879 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
12881 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
12882 bitmasks to be enabled with 3 exceptions, instead of requiring any
12883 enabled ISA with lots of exceptions.
12884 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
12885 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
12886 Change avx512bw in isa attribute to avx512f.
12887 * config/i386/sgxintrin.h: Add license boilerplate.
12888 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
12889 to __AVX512F__ and __AVX512VL to __AVX512VL__.
12890 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
12891 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
12893 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
12894 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
12895 temporarily sse2 rather than sse if not enabled already.
12898 * config/i386/sse.md (VI248_VLBW): Rename to ...
12899 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
12900 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
12901 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
12902 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
12903 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
12904 mode iterator instead of VI248_VLBW.
12906 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
12908 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
12909 (record_modified): Skip clobbers; add debug output.
12910 (param_change_prob): Use sreal frequencies.
12912 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12914 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
12915 punt for user-aligned variables.
12917 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12919 * tree-chrec.c (chrec_contains_symbols): Return true for
12922 2018-01-05 Sudakshina Das <sudi.das@arm.com>
12925 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
12926 of (x|y) == x for BICS pattern.
12928 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12930 PR tree-optimization/83605
12931 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
12932 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
12935 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
12937 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
12938 * config/epiphany/rtems.h: New file.
12940 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12941 Uros Bizjak <ubizjak@gmail.com>
12944 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
12945 QIreg_operand instead of register_operand predicate.
12946 * config/i386/i386.c (ix86_rop_should_change_byte_p,
12947 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
12948 comments instead of -fmitigate[-_]rop.
12950 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12953 * cgraphunit.c (symbol_table::compile): Switch to text_section
12954 before calling assembly_start debug hook.
12955 * run-rtl-passes.c (run_rtl_passes): Likewise.
12958 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12960 * tree-vrp.c (extract_range_from_binary_expr_1): Check
12961 range_int_cst_p rather than !symbolic_range_p before calling
12962 extract_range_from_multiplicative_op_1.
12964 2018-01-04 Jeff Law <law@redhat.com>
12966 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
12967 redundant test in assertion.
12969 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12971 * doc/rtl.texi: Document machine_mode wrapper classes.
12973 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12975 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
12976 using tree_to_uhwi.
12978 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12980 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
12981 the VEC_PERM_EXPR fold to fail.
12983 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12986 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
12987 to switched_sections.
12989 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12992 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
12993 test for d.testing.
12995 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
12998 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
12999 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
13001 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13004 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
13005 is BLKmode and bitpos not zero or mode change is needed.
13007 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13010 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
13013 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
13016 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
13017 instead of MULT rtx. Update all corresponding splitters.
13018 (*saddl_se): Ditto.
13019 (*ssub<modesuffix>): Ditto.
13020 (*ssubl_se): Ditto.
13021 (*cmp_sadd_di): Update split patterns.
13022 (*cmp_sadd_si): Ditto.
13023 (*cmp_sadd_sidi): Ditto.
13024 (*cmp_ssub_di): Ditto.
13025 (*cmp_ssub_si): Ditto.
13026 (*cmp_ssub_sidi): Ditto.
13027 * config/alpha/predicates.md (const23_operand): New predicate.
13028 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
13029 Look for ASHIFT, not MULT inner operand.
13030 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
13032 2018-01-04 Martin Liska <mliska@suse.cz>
13034 PR gcov-profile/83669
13035 * gcov.c (output_intermediate_file): Add version to intermediate
13037 * doc/gcov.texi: Document new field 'version' in intermediate
13038 file format. Fix location of '-k' option of gcov command.
13040 2018-01-04 Martin Liska <mliska@suse.cz>
13043 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
13045 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13047 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
13049 2018-01-03 Martin Sebor <msebor@redhat.com>
13051 PR tree-optimization/83655
13052 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
13053 checking calls with invalid arguments.
13055 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13057 * tree-vect-stmts.c (vect_get_store_rhs): New function.
13058 (vectorizable_mask_load_store): Delete.
13059 (vectorizable_call): Return false for masked loads and stores.
13060 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
13061 instead of gimple_assign_rhs1.
13062 (vectorizable_load): Handle IFN_MASK_LOAD.
13063 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
13065 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13067 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
13069 (vectorizable_mask_load_store): ...here.
13070 (vectorizable_load): ...and here.
13072 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13074 * tree-vect-stmts.c (vect_build_all_ones_mask)
13075 (vect_build_zero_merge_argument): New functions, split out from...
13076 (vectorizable_load): ...here.
13078 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13080 * tree-vect-stmts.c (vect_check_store_rhs): New function,
13082 (vectorizable_mask_load_store): ...here.
13083 (vectorizable_store): ...and here.
13085 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13087 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
13089 (vectorizable_mask_load_store): ...here.
13091 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13093 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
13094 (vect_model_store_cost): Take a vec_load_store_type instead of a
13096 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
13097 (vect_model_store_cost): Take a vec_load_store_type instead of a
13099 (vectorizable_mask_load_store): Update accordingly.
13100 (vectorizable_store): Likewise.
13101 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
13103 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13105 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
13106 IFN_MASK_LOAD calls here rather than...
13107 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
13109 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13110 Alan Hayward <alan.hayward@arm.com>
13111 David Sherwood <david.sherwood@arm.com>
13113 * expmed.c (extract_bit_field_1): For vector extracts,
13114 fall back to extract_bit_field_as_subreg if vec_extract
13117 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13118 Alan Hayward <alan.hayward@arm.com>
13119 David Sherwood <david.sherwood@arm.com>
13121 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
13122 they are variable or constant sized.
13123 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
13124 slots for constant-sized data.
13126 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13127 Alan Hayward <alan.hayward@arm.com>
13128 David Sherwood <david.sherwood@arm.com>
13130 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
13131 handling COND_EXPRs with boolean comparisons, try to find a better
13132 basis for the mask type than the boolean itself.
13134 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13136 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
13137 is calculated and how it can be overridden.
13138 * genmodes.c (max_bitsize_mode_any_mode): New variable.
13139 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
13141 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
13144 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13145 Alan Hayward <alan.hayward@arm.com>
13146 David Sherwood <david.sherwood@arm.com>
13148 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
13149 Remove the mode argument.
13150 (aarch64_simd_valid_immediate): Remove the mode and inverse
13152 * config/aarch64/iterators.md (bitsize): New iterator.
13153 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
13154 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
13155 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
13156 aarch64_simd_valid_immediate.
13157 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
13158 (aarch64_reg_or_bic_imm): Likewise.
13159 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
13160 with an insn_type enum and msl with a modifier_type enum.
13161 Replace element_width with a scalar_mode. Change the shift
13162 to unsigned int. Add constructors for scalar_float_mode and
13163 scalar_int_mode elements.
13164 (aarch64_vect_float_const_representable_p): Delete.
13165 (aarch64_can_const_movi_rtx_p)
13166 (aarch64_simd_scalar_immediate_valid_for_move)
13167 (aarch64_simd_make_constant): Update call to
13168 aarch64_simd_valid_immediate.
13169 (aarch64_advsimd_valid_immediate_hs): New function.
13170 (aarch64_advsimd_valid_immediate): Likewise.
13171 (aarch64_simd_valid_immediate): Remove mode and inverse
13172 arguments. Rewrite to use the above. Use const_vec_duplicate_p
13173 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
13174 and aarch64_float_const_representable_p on the result.
13175 (aarch64_output_simd_mov_immediate): Remove mode argument.
13176 Update call to aarch64_simd_valid_immediate and use of
13177 simd_immediate_info.
13178 (aarch64_output_scalar_simd_mov_immediate): Update call
13181 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13182 Alan Hayward <alan.hayward@arm.com>
13183 David Sherwood <david.sherwood@arm.com>
13185 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
13186 (mode_nunits): Likewise CONST_MODE_NUNITS.
13187 * machmode.def (ADJUST_NUNITS): Document.
13188 * genmodes.c (mode_data::need_nunits_adj): New field.
13189 (blank_mode): Update accordingly.
13190 (adj_nunits): New variable.
13191 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
13193 (emit_mode_size_inline): Set need_bytesize_adj for all modes
13194 listed in adj_nunits.
13195 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
13196 listed in adj_nunits. Don't emit case statements for such modes.
13197 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
13198 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
13199 nothing if adj_nunits is nonnull.
13200 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
13201 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
13202 (emit_mode_fbit): Update use of print_maybe_const_decl.
13203 (emit_move_size): Likewise. Treat the array as non-const
13205 (emit_mode_adjustments): Handle adj_nunits.
13207 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13209 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
13210 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
13211 (VECTOR_MODES): Use it.
13212 (make_vector_modes): Take the prefix as an argument.
13214 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13215 Alan Hayward <alan.hayward@arm.com>
13216 David Sherwood <david.sherwood@arm.com>
13218 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
13219 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
13220 for MODE_VECTOR_BOOL.
13221 * machmode.def (VECTOR_BOOL_MODE): Document.
13222 * genmodes.c (VECTOR_BOOL_MODE): New macro.
13223 (make_vector_bool_mode): New function.
13224 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
13226 * lto-streamer-in.c (lto_input_mode_table): Likewise.
13227 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
13229 * stor-layout.c (int_mode_for_mode): Likewise.
13230 * tree.c (build_vector_type_for_mode): Likewise.
13231 * varasm.c (output_constant_pool_2): Likewise.
13232 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
13233 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
13234 for MODE_VECTOR_BOOL.
13235 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
13236 of mode class checks.
13237 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
13238 instead of a list of mode class checks.
13239 (expand_vector_scalar_condition): Likewise.
13240 (type_for_widest_vector_mode): Handle BImode as an inner mode.
13242 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13243 Alan Hayward <alan.hayward@arm.com>
13244 David Sherwood <david.sherwood@arm.com>
13246 * machmode.h (mode_size): Change from unsigned short to
13248 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
13249 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13250 or if measurement_type is not polynomial.
13251 (fixed_size_mode::includes_p): Check for constant-sized modes.
13252 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
13253 return a poly_uint16 rather than an unsigned short.
13254 (emit_mode_size): Change the type of mode_size from unsigned short
13255 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
13256 (emit_mode_adjustments): Cope with polynomial vector sizes.
13257 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13259 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13261 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
13262 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
13263 * caller-save.c (setup_save_areas): Likewise.
13264 (replace_reg_with_saved_mem): Likewise.
13265 * calls.c (emit_library_call_value_1): Likewise.
13266 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
13267 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
13268 (gen_lowpart_for_combine): Likewise.
13269 * convert.c (convert_to_integer_1): Likewise.
13270 * cse.c (equiv_constant, cse_insn): Likewise.
13271 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
13272 (cselib_subst_to_values): Likewise.
13273 * dce.c (word_dce_process_block): Likewise.
13274 * df-problems.c (df_word_lr_mark_ref): Likewise.
13275 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
13276 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
13277 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
13278 (rtl_for_decl_location): Likewise.
13279 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
13280 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
13281 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
13282 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
13283 (expand_expr_real_1): Likewise.
13284 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
13285 (pad_below): Likewise.
13286 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13287 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
13288 * ira.c (get_subreg_tracking_sizes): Likewise.
13289 * ira-build.c (ira_create_allocno_objects): Likewise.
13290 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
13291 (ira_sort_regnos_for_alter_reg): Likewise.
13292 * ira-costs.c (record_operand_costs): Likewise.
13293 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
13294 (resolve_simple_move): Likewise.
13295 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
13296 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
13297 (lra_constraints): Likewise.
13298 (CONST_POOL_OK_P): Reject variable-sized modes.
13299 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
13300 (add_pseudo_to_slot, lra_spill): Likewise.
13301 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13302 * optabs-query.c (get_best_extraction_insn): Likewise.
13303 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13304 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
13305 (expand_mult_highpart, valid_multiword_target_p): Likewise.
13306 * recog.c (offsettable_address_addr_space_p): Likewise.
13307 * regcprop.c (maybe_mode_change): Likewise.
13308 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
13309 * regrename.c (build_def_use): Likewise.
13310 * regstat.c (dump_reg_info): Likewise.
13311 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
13312 (find_reloads, find_reloads_subreg_address): Likewise.
13313 * reload1.c (eliminate_regs_1): Likewise.
13314 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
13315 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
13316 (simplify_binary_operation_1, simplify_subreg): Likewise.
13317 * targhooks.c (default_function_arg_padding): Likewise.
13318 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
13319 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
13320 (verify_gimple_assign_ternary): Likewise.
13321 * tree-inline.c (estimate_move_cost): Likewise.
13322 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13323 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
13324 (get_address_cost_ainc): Likewise.
13325 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
13326 (vect_supportable_dr_alignment): Likewise.
13327 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13328 (vectorizable_reduction): Likewise.
13329 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
13330 (vectorizable_operation, vectorizable_load): Likewise.
13331 * tree.c (build_same_sized_truth_vector_type): Likewise.
13332 * valtrack.c (cleanup_auto_inc_dec): Likewise.
13333 * var-tracking.c (emit_note_insn_var_location): Likewise.
13334 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
13335 (ADDR_VEC_ALIGN): Likewise.
13337 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13338 Alan Hayward <alan.hayward@arm.com>
13339 David Sherwood <david.sherwood@arm.com>
13341 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
13343 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13344 or if measurement_type is polynomial.
13345 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
13346 * combine.c (make_extraction): Likewise.
13347 * dse.c (find_shift_sequence): Likewise.
13348 * dwarf2out.c (mem_loc_descriptor): Likewise.
13349 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
13350 (extract_bit_field, extract_low_bits): Likewise.
13351 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
13352 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
13353 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
13354 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
13355 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13356 * reload.c (find_reloads): Likewise.
13357 * reload1.c (alter_reg): Likewise.
13358 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
13359 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
13360 * tree-if-conv.c (predicate_mem_writes): Likewise.
13361 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
13362 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
13363 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
13364 * valtrack.c (dead_debug_insert_temp): Likewise.
13365 * varasm.c (mergeable_constant_section): Likewise.
13366 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
13368 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13369 Alan Hayward <alan.hayward@arm.com>
13370 David Sherwood <david.sherwood@arm.com>
13372 * expr.c (expand_assignment): Cope with polynomial mode sizes
13373 when assigning to a CONCAT.
13375 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13376 Alan Hayward <alan.hayward@arm.com>
13377 David Sherwood <david.sherwood@arm.com>
13379 * machmode.h (mode_precision): Change from unsigned short to
13381 (mode_to_precision): Return a poly_uint16 rather than an unsigned
13383 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
13384 or if measurement_type is not polynomial.
13385 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
13386 in which the mode is already known to be a scalar_int_mode.
13387 * genmodes.c (emit_mode_precision): Change the type of mode_precision
13388 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
13390 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13391 for GET_MODE_PRECISION.
13392 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13393 for GET_MODE_PRECISION.
13394 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
13396 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
13397 (expand_field_assignment, make_extraction): Likewise.
13398 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
13399 (get_last_value): Likewise.
13400 * convert.c (convert_to_integer_1): Likewise.
13401 * cse.c (cse_insn): Likewise.
13402 * expr.c (expand_expr_real_1): Likewise.
13403 * lra-constraints.c (simplify_operand_subreg): Likewise.
13404 * optabs-query.c (can_atomic_load_p): Likewise.
13405 * optabs.c (expand_atomic_load): Likewise.
13406 (expand_atomic_store): Likewise.
13407 * ree.c (combine_reaching_defs): Likewise.
13408 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
13409 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
13410 * tree.h (type_has_mode_precision_p): Likewise.
13411 * ubsan.c (instrument_si_overflow): Likewise.
13413 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13414 Alan Hayward <alan.hayward@arm.com>
13415 David Sherwood <david.sherwood@arm.com>
13417 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
13418 polynomial numbers of units.
13419 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
13420 (valid_vector_subparts_p): New function.
13421 (build_vector_type): Remove temporary shim and take the number
13422 of units as a poly_uint64 rather than an int.
13423 (build_opaque_vector_type): Take the number of units as a
13424 poly_uint64 rather than an int.
13425 * tree.c (build_vector_from_ctor): Handle polynomial
13426 TYPE_VECTOR_SUBPARTS.
13427 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
13428 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
13429 (build_vector_from_val): If the number of units is variable,
13430 use build_vec_duplicate_cst for constant operands and
13431 VEC_DUPLICATE_EXPR otherwise.
13432 (make_vector_type): Remove temporary is_constant ().
13433 (build_vector_type, build_opaque_vector_type): Take the number of
13434 units as a poly_uint64 rather than an int.
13435 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
13437 * cfgexpand.c (expand_debug_expr): Likewise.
13438 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
13439 (store_constructor, expand_expr_real_1): Likewise.
13440 (const_scalar_mask_from_tree): Likewise.
13441 * fold-const-call.c (fold_const_reduction): Likewise.
13442 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
13443 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
13444 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
13445 (fold_relational_const): Likewise.
13446 (native_interpret_vector): Likewise. Change the size from an
13447 int to an unsigned int.
13448 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
13449 TYPE_VECTOR_SUBPARTS.
13450 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
13451 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
13452 duplicating a non-constant operand into a variable-length vector.
13453 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
13454 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
13455 * ipa-icf.c (sem_variable::equals): Likewise.
13456 * match.pd: Likewise.
13457 * omp-simd-clone.c (simd_clone_subparts): Likewise.
13458 * print-tree.c (print_node): Likewise.
13459 * stor-layout.c (layout_type): Likewise.
13460 * targhooks.c (default_builtin_vectorization_cost): Likewise.
13461 * tree-cfg.c (verify_gimple_comparison): Likewise.
13462 (verify_gimple_assign_binary): Likewise.
13463 (verify_gimple_assign_ternary): Likewise.
13464 (verify_gimple_assign_single): Likewise.
13465 * tree-pretty-print.c (dump_generic_node): Likewise.
13466 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13467 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
13468 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
13469 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
13470 (vect_shift_permute_load_chain): Likewise.
13471 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
13472 (expand_vector_condition, optimize_vector_constructor): Likewise.
13473 (lower_vec_perm, get_compute_type): Likewise.
13474 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13475 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
13476 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
13477 (vect_recog_mask_conversion_pattern): Likewise.
13478 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
13479 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
13480 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13481 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
13482 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
13483 (vectorizable_shift, vectorizable_operation, vectorizable_store)
13484 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
13485 (supportable_widening_operation): Likewise.
13486 (supportable_narrowing_operation): Likewise.
13487 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
13489 * varasm.c (output_constant): Likewise.
13491 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13492 Alan Hayward <alan.hayward@arm.com>
13493 David Sherwood <david.sherwood@arm.com>
13495 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
13496 so that both the length == 3 and length != 3 cases set up their
13497 own permute vectors. Add comments explaining why we know the
13498 number of elements is constant.
13499 (vect_permute_load_chain): Likewise.
13501 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13502 Alan Hayward <alan.hayward@arm.com>
13503 David Sherwood <david.sherwood@arm.com>
13505 * machmode.h (mode_nunits): Change from unsigned char to
13507 (ONLY_FIXED_SIZE_MODES): New macro.
13508 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
13509 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
13510 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
13512 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
13513 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
13514 or if measurement_type is not polynomial.
13515 * genmodes.c (ZERO_COEFFS): New macro.
13516 (emit_mode_nunits_inline): Make mode_nunits_inline return a
13518 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
13519 Use ZERO_COEFFS when emitting initializers.
13520 * data-streamer.h (bp_pack_poly_value): New function.
13521 (bp_unpack_poly_value): Likewise.
13522 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13523 for GET_MODE_NUNITS.
13524 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13525 for GET_MODE_NUNITS.
13526 * tree.c (make_vector_type): Remove temporary shim and make
13527 the real function take the number of units as a poly_uint64
13528 rather than an int.
13529 (build_vector_type_for_mode): Handle polynomial nunits.
13530 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
13531 * emit-rtl.c (const_vec_series_p_1): Likewise.
13532 (gen_rtx_CONST_VECTOR): Likewise.
13533 * fold-const.c (test_vec_duplicate_folding): Likewise.
13534 * genrecog.c (validate_pattern): Likewise.
13535 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
13536 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13537 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
13538 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
13539 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
13540 * rtlanal.c (subreg_get_info): Likewise.
13541 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13542 (vect_grouped_load_supported): Likewise.
13543 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
13544 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
13545 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
13546 (simplify_const_unary_operation, simplify_binary_operation_1)
13547 (simplify_const_binary_operation, simplify_ternary_operation)
13548 (test_vector_ops_duplicate, test_vector_ops): Likewise.
13549 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
13550 instead of CONST_VECTOR_NUNITS.
13551 * varasm.c (output_constant_pool_2): Likewise.
13552 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
13553 explicit-encoded elements in the XVEC for variable-length vectors.
13555 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13557 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
13559 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13560 Alan Hayward <alan.hayward@arm.com>
13561 David Sherwood <david.sherwood@arm.com>
13563 * coretypes.h (fixed_size_mode): Declare.
13564 (fixed_size_mode_pod): New typedef.
13565 * builtins.h (target_builtins::x_apply_args_mode)
13566 (target_builtins::x_apply_result_mode): Change type to
13567 fixed_size_mode_pod.
13568 * builtins.c (apply_args_size, apply_result_size, result_vector)
13569 (expand_builtin_apply_args_1, expand_builtin_apply)
13570 (expand_builtin_return): Update accordingly.
13572 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13574 * cse.c (hash_rtx_cb): Hash only the encoded elements.
13575 * cselib.c (cselib_hash_rtx): Likewise.
13576 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
13577 CONST_VECTOR encoding.
13579 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13580 Jeff Law <law@redhat.com>
13583 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
13584 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
13585 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
13586 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
13589 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
13590 explicitly probe *sp in a noreturn function if there were any callee
13591 register saves or frame pointer is needed.
13593 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13596 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
13597 BLKmode for ternary, binary or unary expressions.
13600 * var-tracking.c (delete_vta_debug_insn): New inline function.
13601 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
13602 insns from get_insns () to NULL instead of each bb separately.
13603 Use delete_vta_debug_insn. No longer static.
13604 (vt_debug_insns_local, variable_tracking_main_1): Adjust
13605 delete_vta_debug_insns callers.
13606 * rtl.h (delete_vta_debug_insns): Declare.
13607 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
13608 instead of variable_tracking_main.
13610 2018-01-03 Martin Sebor <msebor@redhat.com>
13612 PR tree-optimization/83603
13613 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
13614 arguments past the endof the argument list in functions declared
13615 without a prototype.
13616 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
13617 Avoid checking when arguments are null.
13619 2018-01-03 Martin Sebor <msebor@redhat.com>
13622 * doc/extend.texi (attribute const): Fix a typo.
13623 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
13624 issuing -Wsuggest-attribute for void functions.
13626 2018-01-03 Martin Sebor <msebor@redhat.com>
13628 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
13629 offset_int::from instead of wide_int::to_shwi.
13630 (maybe_diag_overlap): Remove assertion.
13631 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
13632 * gimple-ssa-sprintf.c (format_directive): Same.
13633 (parse_directive): Same.
13634 (sprintf_dom_walker::compute_format_length): Same.
13635 (try_substitute_return_value): Same.
13637 2018-01-03 Jeff Law <law@redhat.com>
13639 PR middle-end/83654
13640 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
13641 non-constant residual for zero at runtime and avoid probing in
13642 that case. Reorganize code for trailing problem to mirror handling
13645 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13647 PR tree-optimization/83501
13648 * tree-ssa-strlen.c (get_string_cst): New.
13649 (handle_char_store): Call get_string_cst.
13651 2018-01-03 Martin Liska <mliska@suse.cz>
13653 PR tree-optimization/83593
13654 * tree-ssa-strlen.c: Include tree-cfg.h.
13655 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
13656 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
13657 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
13659 (strlen_dom_walker::before_dom_children): Call
13660 gimple_purge_dead_eh_edges. Dump tranformation with details
13662 (strlen_dom_walker::before_dom_children): Update call by adding
13663 new argument cleanup_eh.
13664 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
13666 2018-01-03 Martin Liska <mliska@suse.cz>
13669 * cif-code.def (VARIADIC_THUNK): New enum value.
13670 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
13673 2018-01-03 Jan Beulich <jbeulich@suse.com>
13675 * sse.md (mov<mode>_internal): Tighten condition for when to use
13676 vmovdqu<ssescalarsize> for TI and OI modes.
13678 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13680 Update copyright years.
13682 2018-01-03 Martin Liska <mliska@suse.cz>
13685 * ipa-visibility.c (function_and_variable_visibility): Skip
13686 functions with noipa attribure.
13688 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13690 * gcc.c (process_command): Update copyright notice dates.
13691 * gcov-dump.c (print_version): Ditto.
13692 * gcov.c (print_version): Ditto.
13693 * gcov-tool.c (print_version): Ditto.
13694 * gengtype.c (create_file): Ditto.
13695 * doc/cpp.texi: Bump @copying's copyright year.
13696 * doc/cppinternals.texi: Ditto.
13697 * doc/gcc.texi: Ditto.
13698 * doc/gccint.texi: Ditto.
13699 * doc/gcov.texi: Ditto.
13700 * doc/install.texi: Ditto.
13701 * doc/invoke.texi: Ditto.
13703 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13705 * vector-builder.h (vector_builder::m_full_nelts): Change from
13706 unsigned int to poly_uint64.
13707 (vector_builder::full_nelts): Update prototype accordingly.
13708 (vector_builder::new_vector): Likewise.
13709 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
13710 (vector_builder::operator ==): Likewise.
13711 (vector_builder::finalize): Likewise.
13712 * int-vector-builder.h (int_vector_builder::int_vector_builder):
13713 Take the number of elements as a poly_uint64 rather than an
13715 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
13716 from unsigned int to poly_uint64.
13717 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
13718 (vec_perm_indices::new_vector): Likewise.
13719 (vec_perm_indices::length): Likewise.
13720 (vec_perm_indices::nelts_per_input): Likewise.
13721 (vec_perm_indices::input_nelts): Likewise.
13722 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
13723 number of elements per input as a poly_uint64 rather than an
13724 unsigned int. Use the original encoding for variable-length
13725 vectors, rather than clamping each individual element.
13726 For the second and subsequent elements in each pattern,
13727 clamp the step and base before clamping their sum.
13728 (vec_perm_indices::series_p): Handle polynomial element counts.
13729 (vec_perm_indices::all_in_range_p): Likewise.
13730 (vec_perm_indices_to_tree): Likewise.
13731 (vec_perm_indices_to_rtx): Likewise.
13732 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
13733 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
13734 (tree_vector_builder::new_binary_operation): Handle polynomial
13735 element counts. Return false if we need to know the number
13736 of elements at compile time.
13737 * fold-const.c (fold_vec_perm): Punt if the number of elements
13738 isn't known at compile time.
13740 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13742 * vec-perm-indices.h (vec_perm_builder): Change element type
13743 from HOST_WIDE_INT to poly_int64.
13744 (vec_perm_indices::element_type): Update accordingly.
13745 (vec_perm_indices::clamp): Handle polynomial element_types.
13746 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13747 (vec_perm_indices::all_in_range_p): Likewise.
13748 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
13750 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
13751 polynomial vec_perm_indices element types.
13752 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
13753 * fold-const.c (fold_vec_perm): Likewise.
13754 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
13755 * tree-vect-generic.c (lower_vec_perm): Likewise.
13756 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13757 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
13758 element type to HOST_WIDE_INT.
13760 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13761 Alan Hayward <alan.hayward@arm.com>
13762 David Sherwood <david.sherwood@arm.com>
13764 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
13765 rather than an int. Use plus_constant.
13766 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
13767 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
13769 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13770 Alan Hayward <alan.hayward@arm.com>
13771 David Sherwood <david.sherwood@arm.com>
13773 * calls.c (emit_call_1, expand_call): Change struct_value_size from
13774 a HOST_WIDE_INT to a poly_int64.
13776 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13777 Alan Hayward <alan.hayward@arm.com>
13778 David Sherwood <david.sherwood@arm.com>
13780 * calls.c (load_register_parameters): Cope with polynomial
13781 mode sizes. Require a constant size for BLKmode parameters
13782 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
13783 forces a parameter to be padded at the lsb end in order to
13784 fill a complete number of words, require the parameter size
13785 to be ordered wrt UNITS_PER_WORD.
13787 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13788 Alan Hayward <alan.hayward@arm.com>
13789 David Sherwood <david.sherwood@arm.com>
13791 * reload1.c (spill_stack_slot_width): Change element type
13792 from unsigned int to poly_uint64_pod.
13793 (alter_reg): Treat mode sizes as polynomial.
13795 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13796 Alan Hayward <alan.hayward@arm.com>
13797 David Sherwood <david.sherwood@arm.com>
13799 * reload.c (complex_word_subreg_p): New function.
13800 (reload_inner_reg_of_subreg, push_reload): Use it.
13802 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13803 Alan Hayward <alan.hayward@arm.com>
13804 David Sherwood <david.sherwood@arm.com>
13806 * lra-constraints.c (process_alt_operands): Reject matched
13807 operands whose sizes aren't ordered.
13808 (match_reload): Refer to this check here.
13810 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13811 Alan Hayward <alan.hayward@arm.com>
13812 David Sherwood <david.sherwood@arm.com>
13814 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
13815 that the mode size is in the set {1, 2, 4, 8, 16}.
13817 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13818 Alan Hayward <alan.hayward@arm.com>
13819 David Sherwood <david.sherwood@arm.com>
13821 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
13822 Use plus_constant instead of gen_rtx_PLUS.
13824 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13825 Alan Hayward <alan.hayward@arm.com>
13826 David Sherwood <david.sherwood@arm.com>
13828 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
13829 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
13830 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
13831 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
13832 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
13833 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
13834 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
13835 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
13836 * config/i386/i386.c (ix86_push_rounding): ...this new function.
13837 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
13839 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
13840 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
13841 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
13842 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
13843 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
13844 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
13845 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
13846 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
13847 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
13848 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
13850 * expr.c (emit_move_resolve_push): Treat the input and result
13851 of PUSH_ROUNDING as a poly_int64.
13852 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
13853 (emit_push_insn): Likewise.
13854 * lra-eliminations.c (mark_not_eliminable): Likewise.
13855 * recog.c (push_operand): Likewise.
13856 * reload1.c (elimination_effects): Likewise.
13857 * rtlanal.c (nonzero_bits1): Likewise.
13858 * calls.c (store_one_arg): Likewise. Require the padding to be
13859 known at compile time.
13861 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13862 Alan Hayward <alan.hayward@arm.com>
13863 David Sherwood <david.sherwood@arm.com>
13865 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
13866 Use plus_constant instead of gen_rtx_PLUS.
13868 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13869 Alan Hayward <alan.hayward@arm.com>
13870 David Sherwood <david.sherwood@arm.com>
13872 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
13873 rather than an int.
13875 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13876 Alan Hayward <alan.hayward@arm.com>
13877 David Sherwood <david.sherwood@arm.com>
13879 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
13880 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
13881 via stack temporaries. Treat the mode size as polynomial too.
13883 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13884 Alan Hayward <alan.hayward@arm.com>
13885 David Sherwood <david.sherwood@arm.com>
13887 * expr.c (expand_expr_real_2): When handling conversions involving
13888 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
13889 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
13890 as a poly_uint64 too.
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 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
13898 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13899 Alan Hayward <alan.hayward@arm.com>
13900 David Sherwood <david.sherwood@arm.com>
13902 * combine.c (can_change_dest_mode): Handle polynomial
13903 REGMODE_NATURAL_SIZE.
13904 * expmed.c (store_bit_field_1): Likewise.
13905 * expr.c (store_constructor): Likewise.
13906 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
13907 and polynomial REGMODE_NATURAL_SIZE.
13908 (gen_lowpart_common): Likewise.
13909 * reginfo.c (record_subregs_of_mode): Likewise.
13910 * rtlanal.c (read_modify_subreg_p): Likewise.
13912 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13913 Alan Hayward <alan.hayward@arm.com>
13914 David Sherwood <david.sherwood@arm.com>
13916 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
13917 numbers of elements.
13919 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13920 Alan Hayward <alan.hayward@arm.com>
13921 David Sherwood <david.sherwood@arm.com>
13923 * match.pd: Cope with polynomial numbers of vector elements.
13925 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13926 Alan Hayward <alan.hayward@arm.com>
13927 David Sherwood <david.sherwood@arm.com>
13929 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
13930 in a POINTER_PLUS_EXPR.
13932 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13933 Alan Hayward <alan.hayward@arm.com>
13934 David Sherwood <david.sherwood@arm.com>
13936 * omp-simd-clone.c (simd_clone_subparts): New function.
13937 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
13938 (ipa_simd_modify_function_body): Likewise.
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-generic.c (nunits_for_known_piecewise_op): New function.
13945 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
13946 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
13947 (expand_vector_condition, vector_element): Likewise.
13948 (subparts_gt): New function.
13949 (get_compute_type): Use subparts_gt.
13950 (count_type_subparts): Delete.
13951 (expand_vector_operations_1): Use subparts_gt instead of
13952 count_type_subparts.
13954 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13955 Alan Hayward <alan.hayward@arm.com>
13956 David Sherwood <david.sherwood@arm.com>
13958 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
13959 (vect_compile_time_alias): ...this new function. Do the calculation
13960 on poly_ints rather than trees.
13961 (vect_prune_runtime_alias_test_list): Update call accordingly.
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-slp.c (vect_build_slp_tree_1): Handle polynomial
13969 (vect_schedule_slp_instance): Likewise.
13971 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13972 Alan Hayward <alan.hayward@arm.com>
13973 David Sherwood <david.sherwood@arm.com>
13975 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
13976 constant and extern definitions for variable-length vectors.
13977 (vect_get_constant_vectors): Note that the number of units
13978 is known to be constant.
13980 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13981 Alan Hayward <alan.hayward@arm.com>
13982 David Sherwood <david.sherwood@arm.com>
13984 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
13985 of units as polynomial. Choose between WIDE and NARROW based
13988 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13989 Alan Hayward <alan.hayward@arm.com>
13990 David Sherwood <david.sherwood@arm.com>
13992 * tree-vect-stmts.c (simd_clone_subparts): New function.
13993 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
13995 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13996 Alan Hayward <alan.hayward@arm.com>
13997 David Sherwood <david.sherwood@arm.com>
13999 * tree-vect-stmts.c (vectorizable_call): Treat the number of
14000 vectors as polynomial. Use build_index_vector for
14001 IFN_GOMP_SIMD_LANE.
14003 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14004 Alan Hayward <alan.hayward@arm.com>
14005 David Sherwood <david.sherwood@arm.com>
14007 * tree-vect-stmts.c (get_load_store_type): Treat the number of
14008 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
14009 for variable-length vectors.
14010 (vectorizable_mask_load_store): Treat the number of units as
14011 polynomial, asserting that it is constant if the condition has
14012 already been enforced.
14013 (vectorizable_store, vectorizable_load): Likewise.
14015 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14016 Alan Hayward <alan.hayward@arm.com>
14017 David Sherwood <david.sherwood@arm.com>
14019 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
14020 of units as polynomial. Punt if we can't tell at compile time
14021 which vector contains the final result.
14023 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14024 Alan Hayward <alan.hayward@arm.com>
14025 David Sherwood <david.sherwood@arm.com>
14027 * tree-vect-loop.c (vectorizable_induction): Treat the number
14028 of units as polynomial. Punt on SLP inductions. Use an integer
14029 VEC_SERIES_EXPR for variable-length integer reductions. Use a
14030 cast of such a series for variable-length floating-point
14033 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14034 Alan Hayward <alan.hayward@arm.com>
14035 David Sherwood <david.sherwood@arm.com>
14037 * tree.h (build_index_vector): Declare.
14038 * tree.c (build_index_vector): New function.
14039 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
14040 of units as polynomial, forcibly converting it to a constant if
14041 vectorizable_reduction has already enforced the condition.
14042 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
14043 to create a {1,2,3,...} vector.
14044 (vectorizable_reduction): Treat the number of units as polynomial.
14045 Choose vectype_in based on the largest scalar element size rather
14046 than the smallest number of units. Enforce the restrictions
14049 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14050 Alan Hayward <alan.hayward@arm.com>
14051 David Sherwood <david.sherwood@arm.com>
14053 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
14054 number of units as polynomial.
14056 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14057 Alan Hayward <alan.hayward@arm.com>
14058 David Sherwood <david.sherwood@arm.com>
14060 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
14061 * target.def (autovectorize_vector_sizes): Return the vector sizes
14062 by pointer, using vector_sizes rather than a bitmask.
14063 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
14064 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
14065 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
14067 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
14068 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
14069 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
14070 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
14071 * omp-general.c (omp_max_vf): Likewise.
14072 * omp-low.c (omp_clause_aligned_alignment): Likewise.
14073 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
14074 * tree-vect-loop.c (vect_analyze_loop): Likewise.
14075 * tree-vect-slp.c (vect_slp_bb): Likewise.
14076 * doc/tm.texi: Regenerate.
14077 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
14079 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
14080 the vector size as a poly_uint64 rather than an unsigned int.
14081 (current_vector_size): Change from an unsigned int to a poly_uint64.
14082 (get_vectype_for_scalar_type): Update accordingly.
14083 * tree.h (build_truth_vector_type): Take the size and number of
14084 units as a poly_uint64 rather than an unsigned int.
14085 (build_vector_type): Add a temporary overload that takes
14086 the number of units as a poly_uint64 rather than an unsigned int.
14087 * tree.c (make_vector_type): Likewise.
14088 (build_truth_vector_type): Take the number of units as a poly_uint64
14089 rather than an unsigned int.
14091 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14092 Alan Hayward <alan.hayward@arm.com>
14093 David Sherwood <david.sherwood@arm.com>
14095 * target.def (get_mask_mode): Take the number of units and length
14096 as poly_uint64s rather than unsigned ints.
14097 * targhooks.h (default_get_mask_mode): Update accordingly.
14098 * targhooks.c (default_get_mask_mode): Likewise.
14099 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
14100 * doc/tm.texi: Regenerate.
14102 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14103 Alan Hayward <alan.hayward@arm.com>
14104 David Sherwood <david.sherwood@arm.com>
14106 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
14107 * omp-general.c (omp_max_vf): Likewise.
14108 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
14109 (expand_omp_simd): Handle polynomial safelen.
14110 * omp-low.c (omplow_simd_context): Add a default constructor.
14111 (omplow_simd_context::max_vf): Change from int to poly_uint64.
14112 (lower_rec_simd_input_clauses): Update accordingly.
14113 (lower_rec_input_clauses): Likewise.
14115 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14116 Alan Hayward <alan.hayward@arm.com>
14117 David Sherwood <david.sherwood@arm.com>
14119 * tree-vectorizer.h (vect_nunits_for_cost): New function.
14120 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
14121 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
14122 (vect_analyze_slp_cost): Likewise.
14123 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
14124 (vect_model_load_cost): Likewise.
14126 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14127 Alan Hayward <alan.hayward@arm.com>
14128 David Sherwood <david.sherwood@arm.com>
14130 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
14131 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
14132 from an unsigned int * to a poly_uint64_pod *.
14133 (calculate_unrolling_factor): New function.
14134 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
14136 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14137 Alan Hayward <alan.hayward@arm.com>
14138 David Sherwood <david.sherwood@arm.com>
14140 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
14141 from an unsigned int to a poly_uint64.
14142 (_loop_vec_info::slp_unrolling_factor): Likewise.
14143 (_loop_vec_info::vectorization_factor): Change from an int
14145 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
14146 (vect_get_num_vectors): New function.
14147 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
14148 (vect_get_num_copies): Use vect_get_num_vectors.
14149 (vect_analyze_data_ref_dependences): Change max_vf from an int *
14150 to an unsigned int *.
14151 (vect_analyze_data_refs): Change min_vf from an int * to a
14153 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14154 than an unsigned HOST_WIDE_INT.
14155 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
14156 (vect_analyze_data_ref_dependence): Change max_vf from an int *
14157 to an unsigned int *.
14158 (vect_analyze_data_ref_dependences): Likewise.
14159 (vect_compute_data_ref_alignment): Handle polynomial vf.
14160 (vect_enhance_data_refs_alignment): Likewise.
14161 (vect_prune_runtime_alias_test_list): Likewise.
14162 (vect_shift_permute_load_chain): Likewise.
14163 (vect_supportable_dr_alignment): Likewise.
14164 (dependence_distance_ge_vf): Take the vectorization factor as a
14165 poly_uint64 rather than an unsigned HOST_WIDE_INT.
14166 (vect_analyze_data_refs): Change min_vf from an int * to a
14168 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
14169 vfm1 as a poly_uint64 rather than an int. Make the same change
14170 for the returned bound_scalar.
14171 (vect_gen_vector_loop_niters): Handle polynomial vf.
14172 (vect_do_peeling): Likewise. Update call to
14173 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
14174 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
14176 * tree-vect-loop.c (vect_determine_vectorization_factor)
14177 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
14178 (vect_get_known_peeling_cost): Likewise.
14179 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
14180 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
14181 (vect_transform_loop): Likewise. Use the lowest possible VF when
14182 updating the upper bounds of the loop.
14183 (vect_min_worthwhile_factor): Make static. Return an unsigned int
14184 rather than an int.
14185 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
14186 polynomial unroll factors.
14187 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
14188 (vect_make_slp_decision): Likewise.
14189 (vect_supported_load_permutation_p): Likewise, and polynomial
14191 (vect_analyze_slp_cost): Handle polynomial vf.
14192 (vect_slp_analyze_node_operations): Likewise.
14193 (vect_slp_analyze_bb_1): Likewise.
14194 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14195 than an unsigned HOST_WIDE_INT.
14196 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
14197 (vectorizable_load): Handle polynomial vf.
14198 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
14200 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
14202 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14203 Alan Hayward <alan.hayward@arm.com>
14204 David Sherwood <david.sherwood@arm.com>
14206 * match.pd: Handle bit operations involving three constants
14207 and try to fold one pair.
14209 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14211 * tree-vect-loop-manip.c: Include gimple-fold.h.
14212 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
14213 niters_maybe_zero parameters. Handle other cases besides a step of 1.
14214 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
14215 Add a path that uses a step of VF instead of 1, but disable it
14217 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
14218 and niters_no_overflow parameters. Update calls to
14219 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
14220 Create a new SSA name if the latter choses to use a ste other
14221 than zero, and return it via niters_vector_mult_vf_var.
14222 * tree-vect-loop.c (vect_transform_loop): Update calls to
14223 vect_do_peeling, vect_gen_vector_loop_niters and
14224 slpeel_make_loop_iterate_ntimes.
14225 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
14226 (vect_gen_vector_loop_niters): Update declarations after above changes.
14228 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
14230 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
14231 128-bit round to integer instructions.
14232 (ceil<mode>2): Likewise.
14233 (btrunc<mode>2): Likewise.
14234 (round<mode>2): Likewise.
14236 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14238 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
14239 unaligned VSX load/store on P8/P9.
14240 (expand_block_clear): Allow the use of unaligned VSX
14241 load/store on P8/P9.
14243 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14245 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
14247 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
14248 swap associated with both a load and a store.
14250 2018-01-02 Andrew Waterman <andrew@sifive.com>
14252 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
14253 * config/riscv/riscv.md (clear_cache): Use it.
14255 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
14257 * web.c: Remove out-of-date comment.
14259 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14261 * expr.c (fixup_args_size_notes): Check that any existing
14262 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
14263 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
14264 (emit_single_push_insn): ...here.
14266 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14268 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
14269 (const_vector_encoded_nelts): New function.
14270 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
14271 (const_vector_int_elt, const_vector_elt): Declare.
14272 * emit-rtl.c (const_vector_int_elt_1): New function.
14273 (const_vector_elt): Likewise.
14274 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
14275 of CONST_VECTOR_ELT.
14277 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14279 * expr.c: Include rtx-vector-builder.h.
14280 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
14281 directly on the tree encoding.
14282 (const_vector_from_tree): Likewise.
14283 * optabs.c: Include rtx-vector-builder.h.
14284 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
14285 sequence of "u" values.
14286 * vec-perm-indices.c: Include rtx-vector-builder.h.
14287 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
14288 directly on the vec_perm_indices encoding.
14290 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14292 * doc/rtl.texi (const_vector): Describe new encoding scheme.
14293 * Makefile.in (OBJS): Add rtx-vector-builder.o.
14294 * rtx-vector-builder.h: New file.
14295 * rtx-vector-builder.c: Likewise.
14296 * rtl.h (rtx_def::u2): Add a const_vector field.
14297 (CONST_VECTOR_NPATTERNS): New macro.
14298 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
14299 (CONST_VECTOR_DUPLICATE_P): Likewise.
14300 (CONST_VECTOR_STEPPED_P): Likewise.
14301 (CONST_VECTOR_ENCODED_ELT): Likewise.
14302 (const_vec_duplicate_p): Check for a duplicated vector encoding.
14303 (unwrap_const_vec_duplicate): Likewise.
14304 (const_vec_series_p): Check for a non-duplicated vector encoding.
14305 Say that the function only returns true for integer vectors.
14306 * emit-rtl.c: Include rtx-vector-builder.h.
14307 (gen_const_vec_duplicate_1): Delete.
14308 (gen_const_vector): Call gen_const_vec_duplicate instead of
14309 gen_const_vec_duplicate_1.
14310 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
14311 (gen_const_vec_duplicate): Use rtx_vector_builder.
14312 (gen_const_vec_series): Likewise.
14313 (gen_rtx_CONST_VECTOR): Likewise.
14314 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
14315 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14316 Build a new vector rather than modifying a CONST_VECTOR in-place.
14317 (handle_special_swappables): Update call accordingly.
14318 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
14319 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14320 Build a new vector rather than modifying a CONST_VECTOR in-place.
14321 (handle_special_swappables): Update call accordingly.
14323 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14325 * simplify-rtx.c (simplify_const_binary_operation): Use
14326 CONST_VECTOR_ELT instead of XVECEXP.
14328 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14330 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
14331 the selector elements to be different from the data elements
14332 if the selector is a VECTOR_CST.
14333 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
14334 ssizetype for the selector.
14336 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14338 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
14339 before testing each element individually.
14340 * tree-vect-generic.c (lower_vec_perm): Likewise.
14342 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14344 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
14345 * selftest-run-tests.c (selftest::run_tests): Call it.
14346 * vector-builder.h (vector_builder::operator ==): New function.
14347 (vector_builder::operator !=): Likewise.
14348 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
14349 (vec_perm_indices::all_from_input_p): New function.
14350 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14351 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
14352 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
14353 instead of reading the VECTOR_CST directly. Detect whether both
14354 vector inputs are the same before constructing the vec_perm_indices,
14355 and update the number of inputs argument accordingly. Use the
14356 utility functions added above. Only construct sel2 if we need to.
14358 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14360 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
14361 the broadcast of the low byte.
14362 (expand_mult_highpart): Use an explicit encoding for the permutes.
14363 * optabs-query.c (can_mult_highpart_p): Likewise.
14364 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
14365 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14366 (vectorizable_bswap): Likewise.
14367 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
14368 explicit encoding for the power-of-2 permutes.
14369 (vect_permute_store_chain): Likewise.
14370 (vect_grouped_load_supported): Likewise.
14371 (vect_permute_load_chain): Likewise.
14373 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14375 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
14376 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
14377 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
14378 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14379 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
14380 (vect_gen_perm_mask_any): Likewise.
14382 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14384 * int-vector-builder.h: New file.
14385 * vec-perm-indices.h: Include int-vector-builder.h.
14386 (vec_perm_indices): Redefine as an int_vector_builder.
14387 (auto_vec_perm_indices): Delete.
14388 (vec_perm_builder): Redefine as a stand-alone class.
14389 (vec_perm_indices::vec_perm_indices): New function.
14390 (vec_perm_indices::clamp): Likewise.
14391 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
14392 (vec_perm_indices::new_vector): New function.
14393 (vec_perm_indices::new_expanded_vector): Update for new
14394 vec_perm_indices class.
14395 (vec_perm_indices::rotate_inputs): New function.
14396 (vec_perm_indices::all_in_range_p): Operate directly on the
14397 encoded form, without computing elided elements.
14398 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
14399 encoding. Update for new vec_perm_indices class.
14400 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
14401 the given vec_perm_builder.
14402 (expand_vec_perm_var): Update vec_perm_builder constructor.
14403 (expand_mult_highpart): Use vec_perm_builder instead of
14404 auto_vec_perm_indices.
14405 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
14406 vec_perm_indices instead of auto_vec_perm_indices. Use a single
14407 or double series encoding as appropriate.
14408 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
14409 vec_perm_indices instead of auto_vec_perm_indices.
14410 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14411 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14412 (vect_permute_store_chain): Likewise.
14413 (vect_grouped_load_supported): Likewise.
14414 (vect_permute_load_chain): Likewise.
14415 (vect_shift_permute_load_chain): Likewise.
14416 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14417 (vect_transform_slp_perm_load): Likewise.
14418 (vect_schedule_slp_instance): Likewise.
14419 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14420 (vectorizable_mask_load_store): Likewise.
14421 (vectorizable_bswap): Likewise.
14422 (vectorizable_store): Likewise.
14423 (vectorizable_load): Likewise.
14424 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
14425 vec_perm_indices instead of auto_vec_perm_indices. Use
14426 tree_to_vec_perm_builder to read the vector from a tree.
14427 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
14428 vec_perm_builder instead of a vec_perm_indices.
14429 (have_whole_vector_shift): Use vec_perm_builder and
14430 vec_perm_indices instead of auto_vec_perm_indices. Leave the
14431 truncation to calc_vec_perm_mask_for_shift.
14432 (vect_create_epilog_for_reduction): Likewise.
14433 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
14434 from auto_vec_perm_indices to vec_perm_indices.
14435 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14436 instead of changing individual elements.
14437 (aarch64_vectorize_vec_perm_const): Use new_vector to install
14438 the vector in d.perm.
14439 * config/arm/arm.c (expand_vec_perm_d::perm): Change
14440 from auto_vec_perm_indices to vec_perm_indices.
14441 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14442 instead of changing individual elements.
14443 (arm_vectorize_vec_perm_const): Use new_vector to install
14444 the vector in d.perm.
14445 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
14446 Update vec_perm_builder constructor.
14447 (rs6000_expand_interleave): Likewise.
14448 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
14449 (rs6000_expand_interleave): Likewise.
14451 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14453 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
14454 to qimode could truncate the indices.
14455 * optabs.c (expand_vec_perm_var): Likewise.
14457 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14459 * Makefile.in (OBJS): Add vec-perm-indices.o.
14460 * vec-perm-indices.h: New file.
14461 * vec-perm-indices.c: Likewise.
14462 * target.h (vec_perm_indices): Replace with a forward class
14464 (auto_vec_perm_indices): Move to vec-perm-indices.h.
14465 * optabs.h: Include vec-perm-indices.h.
14466 (expand_vec_perm): Delete.
14467 (selector_fits_mode_p, expand_vec_perm_var): Declare.
14468 (expand_vec_perm_const): Declare.
14469 * target.def (vec_perm_const_ok): Replace with...
14470 (vec_perm_const): ...this new hook.
14471 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
14472 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
14473 * doc/tm.texi: Regenerate.
14474 * optabs.def (vec_perm_const): Delete.
14475 * doc/md.texi (vec_perm_const): Likewise.
14476 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
14477 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
14478 expand_vec_perm for constant permutation vectors. Assert that
14479 the mode of variable permutation vectors is the integer equivalent
14480 of the mode that is being permuted.
14481 * optabs-query.h (selector_fits_mode_p): Declare.
14482 * optabs-query.c: Include vec-perm-indices.h.
14483 (selector_fits_mode_p): New function.
14484 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
14485 is defined, instead of checking whether the vec_perm_const_optab
14486 exists. Use targetm.vectorize.vec_perm_const instead of
14487 targetm.vectorize.vec_perm_const_ok. Check whether the indices
14488 fit in the vector mode before using a variable permute.
14489 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
14490 vec_perm_indices instead of an rtx.
14491 (expand_vec_perm): Replace with...
14492 (expand_vec_perm_const): ...this new function. Take the selector
14493 as a vec_perm_indices rather than an rtx. Also take the mode of
14494 the selector. Update call to shift_amt_for_vec_perm_mask.
14495 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
14496 Use vec_perm_indices::new_expanded_vector to expand the original
14497 selector into bytes. Check whether the indices fit in the vector
14498 mode before using a variable permute.
14499 (expand_vec_perm_var): Make global.
14500 (expand_mult_highpart): Use expand_vec_perm_const.
14501 * fold-const.c: Includes vec-perm-indices.h.
14502 * tree-ssa-forwprop.c: Likewise.
14503 * tree-vect-data-refs.c: Likewise.
14504 * tree-vect-generic.c: Likewise.
14505 * tree-vect-loop.c: Likewise.
14506 * tree-vect-slp.c: Likewise.
14507 * tree-vect-stmts.c: Likewise.
14508 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
14510 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
14511 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
14512 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
14513 (aarch64_vectorize_vec_perm_const): ...this new function.
14514 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14515 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14516 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
14517 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
14518 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14519 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14520 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
14522 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
14523 check for NEON modes.
14524 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
14525 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
14526 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
14527 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
14529 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
14530 the old VEC_PERM_CONST conditions.
14531 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
14532 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
14533 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
14534 (ia64_vectorize_vec_perm_const_ok): Merge into...
14535 (ia64_vectorize_vec_perm_const): ...this new function.
14536 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
14537 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
14538 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
14539 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
14540 * config/mips/mips.c (mips_expand_vec_perm_const)
14541 (mips_vectorize_vec_perm_const_ok): Merge into...
14542 (mips_vectorize_vec_perm_const): ...this new function.
14543 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
14544 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
14545 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
14546 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
14547 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
14548 (rs6000_expand_vec_perm_const): Delete.
14549 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
14551 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14552 (altivec_expand_vec_perm_const_le): Take each operand individually.
14553 Operate on constant selectors rather than rtxes.
14554 (altivec_expand_vec_perm_const): Likewise. Update call to
14555 altivec_expand_vec_perm_const_le.
14556 (rs6000_expand_vec_perm_const): Delete.
14557 (rs6000_vectorize_vec_perm_const_ok): Delete.
14558 (rs6000_vectorize_vec_perm_const): New function.
14559 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14560 an element count and rtx array.
14561 (rs6000_expand_extract_even): Update call accordingly.
14562 (rs6000_expand_interleave): Likewise.
14563 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
14564 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
14565 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
14566 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
14567 (rs6000_expand_vec_perm_const): Delete.
14568 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14569 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14570 (altivec_expand_vec_perm_const_le): Take each operand individually.
14571 Operate on constant selectors rather than rtxes.
14572 (altivec_expand_vec_perm_const): Likewise. Update call to
14573 altivec_expand_vec_perm_const_le.
14574 (rs6000_expand_vec_perm_const): Delete.
14575 (rs6000_vectorize_vec_perm_const_ok): Delete.
14576 (rs6000_vectorize_vec_perm_const): New function. Remove stray
14577 reference to the SPE evmerge intructions.
14578 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14579 an element count and rtx array.
14580 (rs6000_expand_extract_even): Update call accordingly.
14581 (rs6000_expand_interleave): Likewise.
14582 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
14583 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
14585 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14587 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14589 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
14590 vector mode and that that mode matches the mode of the data
14592 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
14593 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
14594 directly using expand_vec_perm_1 when forcing selectors into
14596 (expand_vec_perm_var): New function, split out from expand_vec_perm.
14598 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14600 * optabs-query.h (can_vec_perm_p): Delete.
14601 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
14602 * optabs-query.c (can_vec_perm_p): Split into...
14603 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
14604 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
14605 particular selector is valid.
14606 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14607 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14608 (vect_grouped_load_supported): Likewise.
14609 (vect_shift_permute_load_chain): Likewise.
14610 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14611 (vect_transform_slp_perm_load): Likewise.
14612 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14613 (vectorizable_bswap): Likewise.
14614 (vect_gen_perm_mask_checked): Likewise.
14615 * fold-const.c (fold_ternary_loc): Likewise. Don't take
14616 implementations of variable permutation vectors into account
14617 when deciding which selector to use.
14618 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
14619 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
14620 with a false third argument.
14621 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
14622 to test whether the constant selector is valid and can_vec_perm_var_p
14623 to test whether a variable selector is valid.
14625 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14627 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
14628 * optabs-query.c (can_vec_perm_p): Likewise.
14629 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
14630 instead of vec_perm_indices.
14631 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
14632 (vect_gen_perm_mask_checked): Likewise,
14633 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
14634 (vect_gen_perm_mask_checked): Likewise,
14636 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14638 * optabs-query.h (qimode_for_vec_perm): Declare.
14639 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
14640 (qimode_for_vec_perm): ...this new function.
14641 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
14643 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14645 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
14646 does not have a conditional at the top.
14648 2018-01-02 Richard Biener <rguenther@suse.de>
14650 * ipa-inline.c (big_speedup_p): Fix expression.
14652 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14655 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
14658 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14662 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
14663 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
14664 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
14665 cond_taken_branch_cost 3->4.
14667 2018-01-01 Jakub Jelinek <jakub@redhat.com>
14669 PR tree-optimization/83581
14670 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
14671 TODO_cleanup_cfg if any changes have been made.
14673 PR middle-end/83608
14674 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
14675 convert_modes if target mode has the right side, but different mode
14678 PR middle-end/83609
14679 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
14680 last argument when extracting from CONCAT. If either from_real or
14681 from_imag is NULL, use expansion through memory. If result is not
14682 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
14683 the parts directly to inner mode, if even that fails, use expansion
14686 PR middle-end/83623
14687 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
14688 check for bswap in mode rather than HImode and use that in expand_unop
14691 Copyright (C) 2018 Free Software Foundation, Inc.
14693 Copying and distribution of this file, with or without modification,
14694 are permitted in any medium without royalty provided the copyright
14695 notice and this notice are preserved.