| blob | 9b89baa67320d13d5c8adccf1b3b01d04bd2ebb3 |
1 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
3 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
4 (aarch64_ldrstr_offset_compare): New.
5 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
6 load/store orderings.
7 (aarch64_gen_adjusted_ldpstp): Likewise.
9 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
11 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
12 Check for subset of GENERAL_REGS and FP_REGS.
13 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
14 r=w alternative.
16 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
18 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
19 and wi::to_poly_offset. Add the current offset and then check
20 whether the sum fits, rather than using an unchecked addition of
21 a checked term. Check for a shwi rather than a uhwi.
22 * expr.c (get_bit_range): Use tree_to_poly_uint64.
23 (store_constructor): Use poly_int_tree_p.
24 (expand_expr_real_1): Likewise.
25 * function.c (assign_temp): Likewise.
26 * fold-const.c (const_binop): Use poly_int_tree_p and
27 wi::to_poly_offset.
28 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
29 division.
30 * ipa-icf-gimple.c (func_checker::compare_operand): Use
31 to_poly_offset for MEM offsets.
32 * ipa-icf.c (sem_variable::equals): Likewise.
33 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
34 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
35 wi::to_poly_offset for BIT_FIELD_REF offsets.
36 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
37 wi::to_poly_offset.
38 * var-tracking.c (emit_note_insn_var_location): Use
39 tree_to_poly_uint64.
41 2018-05-29 Jim Wilson <jimw@sifive.com>
43 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
45 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
47 PR target/85950
48 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
49 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
50 sequence.
51 (sse4_1_round<mode>2): Use nonimmediate_operand
52 for operand 1 predicate.
54 2018-05-29 Martin Sebor <msebor@redhat.com>
55 Richard Biener <rguenther@suse.de>
57 PR testsuite/85888
58 * calls.c (get_size_range): Call determine_value_range instead
59 of get_value_range..
60 * tree-vrp.h (determine_value_range): Declared new function.
61 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
63 2018-05-29 Richard Biener <rguenther@suse.de>
65 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
66 sure to use non-pattern stmts for get_earlier_stmt arguments.
67 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
68 called on pattern stmts.
69 (get_later_stmt): Likewise.
71 2018-05-29 Martin Liska <mliska@suse.cz>
73 PR gcov-profile/85759
74 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
75 env variables.
77 2018-05-29 Jakub Jelinek <jakub@redhat.com>
79 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
80 VEC_UNPACK_*_EXPR.
81 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
82 VEC_PACK_*_EXPR.
84 PR target/85918
85 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
86 VEC_PACK_FLOAT_EXPR): New tree codes.
87 * tree-pretty-print.c (op_code_prio): Handle
88 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
89 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
90 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
91 * tree-inline.c (estimate_operator_cost): Likewise.
92 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
93 * fold-const.c (const_binop): Likewise.
94 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
95 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
96 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
97 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
98 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
99 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
100 * expr.c (expand_expr_real_2): Likewise.
101 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
102 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
103 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
104 optabs.
105 * optabs.c (expand_widen_pattern_expr): For
106 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
107 sign from result type rather than operand's type.
108 (expand_binop_directly): For vec_packu_float_optab and
109 vec_packs_float_optab allow result type to be different from operand's
110 type.
111 * optabs-tree.c (optab_for_tree_code): Handle
112 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
113 VEC_PACK_FLOAT_EXPR. Formatting fixes.
114 * tree-vect-generic.c (expand_vector_operations_1): Handle
115 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
116 VEC_PACK_FLOAT_EXPR.
117 * tree-vect-stmts.c (supportable_widening_operation): Handle
118 FIX_TRUNC_EXPR.
119 (supportable_narrowing_operation): Handle FLOAT_EXPR.
120 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
121 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
122 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
123 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
124 mode attributes.
125 (vec_pack<floatprefix>_float_<mode>): New expander.
126 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
127 attributes.
128 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
129 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
130 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
131 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
132 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
133 Document.
134 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
135 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
136 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
137 VEC_PACK_FLOAT_EXPR): Document.
139 2018-05-29 Richard Biener <rguenther@suse.de>
141 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
142 member.
143 (stmt_vec_info_vec): Make pointer.
144 (init_stmt_vec_info_vec): Remove.
145 (free_stmt_vec_info_vec): Likewise.
146 (set_stmt_vec_info_vec): New function.
147 (free_stmt_vec_infos): Likewise.
148 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
149 (set_vinfo_for_stmt): Likewise.
150 (get_earlier_stmt): Likewise.
151 (get_later_stmt): Likewise.
152 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
153 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
154 (vec_info::~vec_info): Free stmt_vec_infos.
155 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
156 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
157 (pass_slp_vectorize::execute): Likewise.
158 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
159 (free_stmt_vec_info_vec): Likewise.
160 (set_stmt_vec_info_vec): New function.
161 (free_stmt_vec_infos): Likewise.
162 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
163 the global stmt_vec_info_vec.
164 * tree-parloops.c (gather_scalar_reductions): Use
165 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
166 vector.
168 2018-05-29 Richard Biener <rguenther@suse.de>
170 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
172 2018-05-29 Martin Liska <mliska@suse.cz>
173 David Malcolm <dmalcolm@redhat.com>
175 * vec.c (test_reverse): New.
176 (vec_c_tests): Add new test.
177 * vec.h (vl_ptr>::reverse): New function.
179 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
181 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
183 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
184 and later.
186 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
188 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
190 2018-05-28 Richard Biener <rguenther@suse.de>
192 PR tree-optimization/85933
193 * tree-vect-data-refs.c (vect_record_base_alignments): Only
194 look at stmts marked as vectorizable.
196 2018-05-28 Richard Biener <rguenther@suse.de>
198 PR tree-optimization/85934
199 * tree-vect-generic.c (expand_vector_operations_1): Hoist
200 vector boolean check before scalar optimization.
202 2018-05-28 Jakub Jelinek <jakub@redhat.com>
204 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
205 for armv5te.
207 2018-05-28 Mark Wielaard <mark@klomp.org>
209 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
210 if it is an expression containing a minus sign.
212 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
214 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
216 2018-05-27 Paul Koning <ni1d@arrl.net>
218 * config/pdp11/pdp11.md (truncsihi2): Remove.
220 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
221 Chung-Ju Wu <jasonwucj@gmail.com>
223 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
224 implementation.
225 (unaligned_store_dw): Ditto.
226 * config/nds32/nds32-memory-manipulation.c
227 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
228 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
229 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
230 (emit_setmem_word_loop): Rename to ...
231 (emit_setmem_doubleword_loop): ... this.
232 (nds32_gen_dup_4_byte_to_word_value): New function.
233 (nds32_gen_dup_8_byte_to_double_word_value): New function.
234 (nds32_expand_setmem_loop): Refine implementation.
235 (nds32_expand_setmem_loop_v3m): Ditto.
236 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
237 pattern.
239 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
241 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
243 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
245 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
246 (nds32_init_machine_status): Initialize machine->attr_naked_p and
247 machine->attr_no_prologue_p.
248 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
249 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
250 (nds32_expand_epilogue): Consider attr_naked_p.
251 (nds32_expand_epilogue_v3pop): Likewise.
252 (nds32_can_use_return_insn): Likewise.
253 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
254 attr_no_prologue_p fields.
255 * config/nds32/nds32.opt (mret-in-naked-func): New option.
257 2018-05-27 Jakub Jelinek <jakub@redhat.com>
259 PR target/85918
260 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
261 attributes.
262 * config/i386/sse.md
263 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
264 Rename to ...
265 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
266 ... this.
267 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
268 Rename to ...
269 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
270 ... this.
271 (*<floatsuffix>floatv2div2sf2): Rename to ...
272 (*float<floatunssuffix>v2div2sf2): ... this.
273 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
274 (float<floatunssuffix>v2div2sf2_mask): ... this.
275 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
276 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
277 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
278 to ...
279 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
280 ... this.
281 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
282 Rename to ...
283 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
284 ... this.
285 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
286 Rename to ...
287 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
288 ... this.
289 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
290 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
291 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
292 gen_ufix_truncv8dfv8si2.
293 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
294 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
295 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
296 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
297 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
298 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
299 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
300 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
302 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
304 PR target/85900
305 PR target/85345
306 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
308 2018-05-25 Jim Wilson <jimw@sifive.com>
310 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
311 * config/riscv/riscv.c (struct machine_function): Add
312 interrupt_handler_p and attribute_checked_p fields.
313 (riscv_attribute_table): Add interrupt.
314 (riscv_interrupt_type_p): New.
315 (riscv_save_reg_p): Save extra regs for interrupt handler.
316 (riscv_use_save_libcall): Return false for interrupt handler.
317 (riscv_first_stack_step): Add forward declaration.
318 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
319 for interrupt handler with large frame. Use it for saved reg list.
320 (riscv_expand_prologue): Move flag_stack_usage_info support to
321 eliminate duplication.
322 (riscv_expand_epilogue): Generate mret for interrupt handler.
323 (riscv_epilogue_uses): New.
324 (riscv_can_use_return_insn): Return false for interrupt handler.
325 (riscv_function_ok_for_sibcall): Likewise.
326 (riscv_set_current_function): Add interrupt handler support.
327 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
328 * config/riscv/riscv.md (UNSPECV_MRET): New.
329 (GP_REGNUM): New.
330 (riscv_frflags, riscv_fsflags): Use tab after opcode.
331 (riscv_mret): New.
332 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
334 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
336 PR tree-optimization/85712
337 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
338 this candidate has already been replaced in-situ by a copy.
340 2018-05-25 Jason Merrill <jason@redhat.com>
342 PR c++/80485 - inline function non-zero address.
343 * symtab.c (nonzero_address): Check DECL_COMDAT.
345 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
347 PR target/83628
348 * config/alpha/alpha.md (ashlsi3): New insn pattern.
349 (*ashlsi_se): Rename from *ashldi_se. Define as sign
350 extension of SImode operation. Use const123_operand predicate.
351 (*saddsi_1): Remove.
352 (*saddl_se_1): Ditto.
353 (*ssubsi_1): Ditto.
354 (*ssubl_se_1): Ditto.
355 * config/alpha/predicates.md (const123_operand): New predicate.
356 * config/alpha/constraints.md (P): Use IN_RANGE.
358 2018-05-25 Richard Biener <rguenther@suse.de>
360 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
361 defaulted to true.
362 (ref_maybe_used_by_stmt_p): Likewise.
363 (stmt_may_clobber_ref_p): Likewise.
364 (stmt_may_clobber_ref_p_1): Likewise.
365 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
366 and pass it along.
367 (ref_maybe_used_by_stmt_p): Likewise.
368 (stmt_may_clobber_ref_p): Likewise.
369 (stmt_may_clobber_ref_p_1): Likewise.
370 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
371 the alias oracle to disambiguate DRs with stmts DR analysis
372 couldn't handle.
373 (vect_analyze_data_refs): Do not give up on not analyzable
374 DRs for BB vectorization. Remove code truncating the dataref
375 vector.
377 2018-05-25 Jakub Jelinek <jakub@redhat.com>
379 PR target/85832
380 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
381 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
382 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
384 2018-05-25 Richard Biener <rguenther@suse.de>
386 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
387 function, combining stmt data ref gathering and fatal analysis
388 parts.
389 (vect_analyze_data_refs): Remove now redudnant code and simplify.
390 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
391 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
392 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
393 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
395 2018-05-25 Bin Cheng <bin.cheng@arm.com>
397 PR tree-optimization/85720
398 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
399 SCC if all partitions are builtins.
400 (version_loop_by_alias_check): New parameter. Generate cancelable
401 runtime alias check if all partitions are builtins.
402 (distribute_loop): Update call to above function.
404 2018-05-25 Bin Cheng <bin.cheng@arm.com>
406 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
407 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
408 (parm_default_def_partition_arg): Ditto.
409 (set_parm_default_def_partition): Ditto.
410 (get_parm_default_def_partitions): Ditto and make it static.
411 (get_undefined_value_partitions): Ditto and make it static.
412 (remove_ssa_form): Refactor call to init_var_map here.
413 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
414 computation for loop region.
415 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
416 (register_default_def): Delete.
417 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
418 (parm_default_def_partition_arg): Ditto.
419 (set_parm_default_def_partition): Ditto.
420 (get_parm_default_def_partitions): Ditto and make it static.
421 (get_undefined_value_partitions): Ditto and make it static.
422 (coalesce_with_default, coalesce_with_default): Update comment.
423 (create_coalesce_list_for_region): New func factored out from
424 create_outofssa_var_map.
425 (populate_coalesce_list_for_outofssa): New func factored out from
426 create_outofssa_var_map and coalesce_ssa_name.
427 (create_outofssa_var_map): Delete.
428 (coalesce_ssa_name): Refactor to support live range computation.
429 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
430 (get_parm_default_def_partitions): Delete.
431 (get_undefined_value_partitions): Ditto.
432 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
433 computation for loop region.
434 (new_tree_live_info, loe_visit_block): Ditto.
435 (live_worklist, set_var_live_on_entry): Ditto.
436 (calculate_live_on_exit, verify_live_on_entry): Ditto.
437 * tree-ssa-live.h (struct _var_map): New fields.
438 (init_var_map): Change decl.
439 (region_contains_p): New.
441 2018-05-25 Bin Cheng <bin.cheng@arm.com>
443 * tree-ssa-live.h (live_merge_and_clear): Delete.
445 2018-05-25 Richard Biener <rguenther@suse.de>
447 PR c++/85912
448 * tree-dump.c (dequeue_and_dump): Remove access to removed
449 operand 2 of a SWITCH_EXPR.
451 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
453 * doc/sourcebuild.texi (vect_double_cond_arith): Include
454 multiplication and division.
455 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
456 (cond_udiv@var{m}, cond_umod@var{m}): Document.
457 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
458 (cond_udiv_optab, cond_umod_optab): New optabs.
459 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
460 (IFN_COND_RDIV): New internal functions.
461 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
462 TRUNC_MOD_EXPR and RDIV_EXPR.
463 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
464 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
465 New unspecs.
466 (SVE_INT_BINARY): Include mult.
467 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
468 (optab, sve_int_op): Handle mult.
469 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
470 UNSPEC_COND_DIV.
471 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
472 for SVE_INT_BINARY_SD.
474 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
476 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
477 (optab, sve_int_op): Handle div and udiv.
478 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
479 for SVE_INT_BINARY_SD.
480 (*<optab><mode>3): New insn for the same.
482 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
484 * tree-vect-patterns.c: Include predict.h.
485 (vect_recog_divmod_pattern): Restrict check for division support
486 to when optimizing for size.
488 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
490 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
491 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
492 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
493 (gimple_match_op::set_op): Likewise.
494 (gimple_resimplify4): Declare.
495 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
496 (expr::gen_transform): Likewise.
497 (decision_tree::gen): Generate a simplification routine for 4 operands.
498 * gimple-match-head.c (gimple_simplify): Add an overload for
499 4 operands. In the top-level function, handle up to 4 call
500 arguments and call gimple_resimplify4.
501 (gimple_resimplify4): New function.
502 (build_call_internal): Pass a fourth operand.
503 (maybe_push_to_seq): Likewise.
504 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
505 Fold VEC_COND_EXPRs of an operation and a default value into
506 an IFN_COND_* function if possible.
507 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
508 New unspecs.
509 (SVE_COND_FP_BINARY): Include them.
510 (optab, sve_fp_op): Handle them.
511 (SVE_INT_BINARY_REV): New code iterator.
512 (SVE_COND_FP_BINARY_REV): New int iterator.
513 (commutative): New int attribute.
514 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
515 Declare.
516 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
517 function.
518 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
519 (*cond_<optab><mode>): New patterns for reversed operands.
521 2018-05-25 Richard Biener <rguenther@suse.de>
523 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
524 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
525 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
526 (STMT_VINFO_GROUPED_ACCESS): Adjust.
527 * tree-vect-data-refs.c (everywhere): Adjust users.
528 * tree-vect-loop.c (everywhere): Likewise.
529 * tree-vect-slp.c (everywhere): Likewise.
530 * tree-vect-stmts.c (everywhere): Likewise.
531 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
533 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
535 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
536 Rename to...
537 (gcc_cv_as_section_exclude): ... this.
538 Try Solaris as #exclude syntax.
539 * configure: Regenerate.
540 * config.in: Regenerate.
541 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
542 SECTION_EXCLUDE.
543 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
544 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
546 * varasm.c (default_elf_asm_named_section): Don't check if
547 HAVE_GAS_SECTION_EXCLUDE is defined.
549 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
551 * doc/md.texi: Update the documentation of the cond_* optabs
552 to mention the new final operand. Fix GET_MODE_NUNITS call.
553 Describe the scalar case too.
554 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
555 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
556 instead of 2.
557 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
558 (get_conditional_internal_fn): Update comment.
559 * tree-vect-loop.c (vectorizable_reduction): Pass the original
560 accumulator value as a final argument to conditional functions.
561 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
562 a define_expand and add an "else" operand. Assert for now that
563 the else operand is equal to operand 2. Use SVE_INT_BINARY and
564 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
565 (*cond_<optab><mode>): New patterns.
566 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
567 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
568 (UNSPEC_COND_EOR): Delete.
569 (optab): Remove associated mappings.
570 (SVE_INT_BINARY): New code iterator.
571 (sve_int_op): Remove int attribute and add "minus" to the code
572 attribute.
573 (SVE_COND_INT_OP): Delete.
574 (SVE_COND_FP_OP): Rename to...
575 (SVE_COND_FP_BINARY): ...this.
577 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
579 * optabs.c (can_reuse_operands_p): New function.
580 (maybe_legitimize_operands): Try to reuse the results for
581 earlier operands.
583 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
585 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
586 Add {q} suffix to insn mnemonic.
588 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
590 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
591 (msp430_warn_func_return): New.
593 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
595 * fold-const.c (tree_nonzero_bits): New function.
596 * fold-const.h (tree_nonzero_bits): Likewise.
597 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
598 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
600 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
602 PR target/85900
603 PR target/85345
604 * varasm.c (assemble_alias): Check ifunc_resolver only on
605 FUNCTION_DECL.
607 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
609 PR target/85903
610 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
611 when memory input operand is handled.
613 2018-05-24 Luis Machado <luis.machado@linaro.org>
615 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
616 global.
617 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
619 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
621 * match.pd: Delay FMA folds until after vectorization.
623 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
625 PR target/83009
626 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
627 address check not strict.
629 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
631 * gimple-match.h (gimple_match_op): New class.
632 (mprts_hook): Replace parameters with a gimple_match_op *.
633 (maybe_build_generic_op): Likewise.
634 (gimple_simplified_result_is_gimple_val): Replace parameters with
635 a const gimple_match_op *.
636 (gimple_simplify): Replace code_helper * and tree * parameters with
637 a gimple_match_op * parameter.
638 (gimple_resimplify1): Replace code_helper *, tree and tree *
639 parameters with a gimple_match_op * parameter.
640 (gimple_resimplify2): Likewise.
641 (gimple_resimplify3): Likewise.
642 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
643 parameters with a gimple_match_op * parameter.
644 * gimple-match-head.c (gimple_simplify): Change prototypes of
645 auto-generated functions to take a gimple_match_op * instead of
646 separate code_helper * and tree * parameters. Make the same
647 change in the top-level overload and update calls to the
648 gimple_resimplify routines. Update calls to the auto-generated
649 functions and to maybe_push_res_to_seq in the publicly-facing
650 operation-specific gimple_simplify overloads.
651 (gimple_match_op::MAX_NUM_OPS): Define.
652 (gimple_resimplify1): Replace rcode and ops with a single res_op
653 parameter. Update call to gimple_simplify.
654 (gimple_resimplify2): Likewise.
655 (gimple_resimplify3): Likewise.
656 (mprts_hook): Replace parameters with a gimple_match_op *.
657 (maybe_build_generic_op): Likewise.
658 (build_call_internal): Replace type, nargs and ops with
659 a gimple_match_op *.
660 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
661 with a single gimple_match_op *. Update calls to mprts_hook,
662 build_call_internal and gimple_simplified_result_is_gimple_val.
663 Factor out code that is common to the tree_code and combined_fn cases.
664 * genmatch.c (expr::gen_transform): Replace tem_code and
665 tem_ops with a gimple_match_op called tem_op. Update calls
666 to the gimple_resimplify functions and maybe_push_res_to_seq.
667 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
668 res_ops. Update call to the gimple_resimplify functions.
669 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
670 (decision_tree::gen): Make the functions take a gimple_match_op *
671 called res_op instead of separate res_code and res_ops parameters.
672 Update call accordingly.
673 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
674 and ops with a single res_op parameter. Update calls to
675 maybe_build_generic_op and maybe_push_res_to_seq.
676 (fold_stmt_1): Update calls to gimple_simplify and
677 replace_stmt_with_simplification.
678 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
679 and gimple_simplified_result_is_gimple_val.
680 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
681 gimple_simplify.
682 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
683 with a gimple_match_op *.
684 (vn_nary_build_or_lookup): Likewise. Update call to
685 vn_nary_build_or_lookup_1.
686 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
687 gimple_match_op *. Update calls to the gimple_resimplify routines
688 and to gimple_simplified_result_is_gimple_val.
689 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
690 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
691 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
692 (visit_nary_op): Likewise.
693 (visit_reference_op_load): Likewise.
695 2018-05-23 Luis Machado <luis.machado@linaro.org>
697 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
698 modifier for printing the step amount.
700 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
702 PR target/78849
703 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
704 types.
706 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
708 * doc/sourcebuild.texi (Endianness): New subsubsection.
710 2018-05-23 Luis Machado <luis.machado@linaro.org>
712 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
713 <prefetch_dynamic_strides>: New const bool field.
714 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
715 prefetch_dynamic_strides.
716 (exynosm1_prefetch_tune): Likewise.
717 (thunderxt88_prefetch_tune): Likewise.
718 (thunderx_prefetch_tune): Likewise.
719 (thunderx2t99_prefetch_tune): Likewise.
720 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
721 false.
722 (aarch64_override_options_internal): Update to set
723 PARAM_PREFETCH_DYNAMIC_STRIDES.
724 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
725 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
726 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
727 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
728 prefetch-dynamic-strides setting.
730 2018-05-23 Luis Machado <luis.machado@linaro.org>
732 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
733 <minimum_stride>: New const int field.
734 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
735 minimum_stride field defaulting to -1.
736 (exynosm1_prefetch_tune): Likewise.
737 (thunderxt88_prefetch_tune): Likewise.
738 (thunderx_prefetch_tune): Likewise.
739 (thunderx2t99_prefetch_tune): Likewise.
740 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
741 <default_opt_level>: Set to 3.
742 (aarch64_override_options_internal): Update to set
743 PARAM_PREFETCH_MINIMUM_STRIDE.
744 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
745 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
746 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
747 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
748 stride is constant and is below the minimum stride threshold.
750 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
752 * config/arm/arm-cpus.in (mode26): Delete.
753 (armv4): Delete mode26 reference.
754 * config/arm/arm.c (arm_configure_build_target): Delete use of
755 isa_bit_mode26.
757 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
759 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
760 New insn pattern.
761 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
762 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
763 for non-SSE modes.
764 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
765 (floatunsdidf2): Ditto.
767 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
769 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
770 (fixuns_trunc<mode>si2_avx512f): Ditto.
771 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
772 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
773 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
775 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
777 PR rtl-optimization/79985
778 * df-scan.c (df_insn_refs_collect): Remove special case for
779 global registers and asm statements.
781 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
783 * extend.texi (Global Register Variables): Rewrite the bullet list.
784 Note that the register is available for allocation. Note that access
785 via inline asm must use constraints. Add note about async-signal
786 handlers. Remove paragraph about automagic register selection.
788 2018-05-23 Richard Biener <rguenther@suse.de>
790 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
791 of fixed offset from memset VN.
793 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
795 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
796 first_interp field.
797 (alloc_cand_and_find_basis): Initialize first_interp field.
798 (slsr_process_mul): Modify first_interp field.
799 (slsr_process_add): Likewise.
800 (slsr_process_cast): Modify first_interp field for each new
801 interpretation.
802 (slsr_process_copy): Likewise.
803 (dump_candidate): Dump first_interp field.
804 (replace_mult_candidate): Process all interpretations, not just
805 subsequent ones.
806 (replace_rhs_if_not_dup): Likewise.
807 (replace_one_candidate): Likewise.
809 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
811 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
812 Add new boolean.
813 (aarch64_needs_frame_chain): New function.
814 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
816 2018-05-23 Sudakshina Das <sudi.das@arm.com>
818 PR target/84882
819 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
820 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
821 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
822 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
823 as true for strict-align.
824 (aarch64_can_inline_p): Perform checks even when callee has no
825 attributes to check for strict alignment.
826 * doc/extend.texi (AArch64 Function Attributes): Document
827 no-strict-align.
828 * doc/invoke.texi: (AArch64 Options): Likewise.
830 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
832 PR tree-optimization/85853
833 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
834 the handling of the root of the node to...
835 (vect_slp_analyze_node_operations_1): ...this new function,
836 and run the whole thing with the child nodes' def types
837 set according to their SLP node's def type.
839 2018-05-23 Richard Biener <rguenther@suse.de>
841 PR middle-end/85874
842 * tree-data-ref.c (create_runtime_alias_checks): Defer
843 and ignore overflow warnings.
845 2018-05-23 Yury Gribov <tetra2005@gmail.com>
847 PR tree-optimization/85822
848 * tree-vrp.c (is_masked_range_test): Fix handling of negative
849 constants.
851 2018-05-23 Richard Biener <rguenther@suse.de>
853 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
854 memset constants via native_interpret_expr.
856 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
858 PR target/85345
859 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
860 attribute.
861 (cgraph_node::create_alias): Likewise.
862 (cgraph_node::get_availability): Check ifunc_resolver instead
863 of looking up ifunc attribute.
864 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
865 * varasm.c (do_assemble_alias): Likewise.
866 (assemble_alias): Likewise.
867 (default_binds_local_p_3): Likewise.
868 * cgraph.h (cgraph_node): Add ifunc_resolver.
869 (cgraph_node::only_called_directly_or_aliased_p): Return false
870 for IFUNC resolver.
871 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
872 attribute.
873 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
874 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
875 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
876 instead of looking up ifunc attribute.
878 2018-05-22 Luis Machado <luis.machado@linaro.org>
880 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
882 2018-05-22 Martin Sebor <msebor@redhat.com>
884 PR middle-end/85359
885 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
886 only when expasion succeeds.
887 (expand_builtin_strcmp): Same.
888 (expand_builtin_strncmp): Same.
890 2018-05-22 Martin Sebor <msebor@redhat.com>
892 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
894 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
895 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
897 * config/aarch64/aarch64-ldpstp.md: Replace uses of
898 aarch64_mem_pair_operand with memory_operand and delete operand swapping
899 code.
900 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
901 Add check for legitimate_address.
902 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
903 (aarch64_swap_ldrstr_operands): New.
904 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
905 Define prototype.
907 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
908 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
910 * config/aarch64/aarch64.md: New patterns to generate stp
911 and ldp.
912 (store_pair_sw, store_pair_dw): New patterns to generate stp for
913 single words and double words.
914 (load_pair_sw, load_pair_dw): Likewise.
915 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
916 Delete.
917 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
918 Delete.
919 * config/aarch64/aarch64-ldpstp.md: Modify peephole
920 for different mode ldpstp and add peephole for merged zero stores.
921 Likewise for loads.
922 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
923 Add size check.
924 (aarch64_gen_store_pair): Rename calls to match new patterns.
925 (aarch64_gen_load_pair): Rename calls to match new patterns.
926 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
927 (load_pair<DREG:mode><DREG2:mode>): ... This.
928 (store_pair<mode>): Rename to...
929 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
930 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
931 New mode iterators.
932 (V_INT_EQUIV): Handle SImode.
933 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
934 New predicate.
936 2018-05-22 Martin Sebor <msebor@redhat.com>
938 PR c/85623
939 * calls.c (maybe_warn_nonstring_arg): Use string length to set
940 or ajust the presumed bound on an operation to avoid unnecessary
941 warnings.
943 2018-05-22 Martin Sebor <msebor@redhat.com>
945 PR tree-optimization/85826
946 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
947 assuming that a DECL necesarily has a constant size.
949 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
951 PR middle-end/85862
952 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
954 2018-05-22 Richard Biener <rguenther@suse.de>
956 PR tree-optimization/85834
957 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
958 non-constant and non-zero memset arguments.
960 2018-05-22 Martin Liska <mliska@suse.cz>
962 PR ipa/85607
963 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
965 2018-05-22 Richard Biener <rguenther@suse.de>
967 PR tree-optimization/85863
968 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
969 comparisons when vectype is specified.
970 (vectorizable_condition): Do not specify vectype for
971 vect_is_simple_cond when SLP vectorizing.
973 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
975 PR target/85657
976 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
977 define __ibm128 as long double.
978 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
979 as a distinct type when IEEE 128-bit support is enabled.
980 (init_float128_ieee): Fix up conversions between IFmode and IEEE
981 128-bit types to use the correct functions.
982 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
983 convert between 128-bit floating point types that have different
984 modes but the same representation, instead of using gen_lowpart to
985 makean alias.
986 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
987 KFmode.
988 (IFKF_reg): New attributes to give the register constraints for
989 IFmode and KFmode.
990 (extend<mode>tf2_internal): New insns to mark an explicit
991 conversion between 128-bit floating point types that have a
992 different mode but share the same representation.
994 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
996 PR tree-optimization/85814
997 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
998 a null return from get_strinfo when unsharing the next
999 strinfo in the chain.
1001 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1003 PR gcc/84923
1004 * varasm.c (weak_finish): Clean up weak_decls.
1006 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1008 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
1009 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
1010 UNSPEC_UADALP values.
1011 * config/aarch64/iterators.md (ABAL): New int iterator.
1012 (ABDL2): Likewise.
1013 (ADALP): Likewise.
1014 (sur): Add mappings for the above.
1015 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
1016 New define_insn.
1017 (aarch64_<sur>abal<mode>_4): Likewise.
1018 (aarch64_<sur>adalp<mode>_3): Likewise.
1019 (<sur>sadv16qi): New define_expand.
1021 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
1023 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
1024 (*movdf_internal): Ditto.
1025 (*rcpsf2_sse): Ditto.
1026 (*rsqrtsf2_sse): Ditto.
1027 (*sqrt<mode>2_sse): Ditto.
1029 2018-05-21 Tamar Christina <tamar.christina@arm.com>
1031 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
1032 eor3q<mode>4.
1033 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
1034 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
1035 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
1036 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
1037 vbcaxq_s64): New.
1038 * config/aarch64/arm_neon.h: Likewise.
1039 * config/aarch64/iterators.md (VQ_I): New.
1041 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
1043 * config.gcc: Add arc/t-multilib-linux to tmake_file for
1044 arc*-*-linux*.
1045 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
1046 MULTILIB_DIRNAMES
1048 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1050 * config/nds32/constraints.md (S): New constraint.
1051 * config/nds32/nds32.md (call_internal): Use constraint S.
1052 (call_value_internal): Likewise.
1053 (sibcall_internal): Likewise.
1054 (sibcall_value_internal): Likewise.
1056 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1057 Chung-Ju Wu <jasonwucj@gmail.com>
1059 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
1060 into consideration.
1062 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1063 Chung-Ju Wu <jasonwucj@gmail.com>
1065 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
1066 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
1067 (nds32_rtx_costs_impl): Simplify.
1068 (nds32_address_cost_impl): Simplify.
1069 (nds32_init_rtx_costs): New function.
1070 (nds32_rtx_costs_speed_prefer): Likewise.
1071 (nds32_rtx_costs_size_prefer): Likewise.
1072 (nds32_address_cost_speed_prefer): Likewise.
1073 (nds32_address_cost_speed_fwprop): Likewise.
1074 (nds32_address_cost_size_prefer): Likewise.
1075 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
1076 * config/nds32/nds32.c (nds32_option_override): Use
1077 nds32_init_rtx_costs function.
1079 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1081 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
1082 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
1083 (TARGET_PIPELINE_N8): Likewise.
1084 (TARGET_PIPELINE_N10): Likewise.
1085 (TARGET_PIPELINE_N13): Likewise.
1086 (TARGET_PIPELINE_GRAYWOLF): Likewise.
1088 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
1090 * config/nds32/nds32-fpu.md: Update copyright year.
1092 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1094 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
1096 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1098 * config/nds32/nds32.c
1099 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
1100 * config/nds32/nds32.opt (minline-asm-r15): New option.
1102 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1104 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
1105 MASK_HW_ABS.
1106 * config/nds32/nds32.md (abssi2): New pattern.
1108 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
1110 * config/i386/i386.md (rex64namesuffix): New mode attribute.
1111 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
1112 Merge insn pattern from sse_cvtsi2ss<round_name> and
1113 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
1114 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
1115 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
1116 using SWI48 mode iterator.
1117 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
1118 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
1119 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
1120 pattern from sse_cvttss2si<round_saeonly_name>
1121 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
1122 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
1123 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
1124 using SWI48 mode iterator.
1125 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
1126 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
1127 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
1128 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
1129 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
1130 using SWI48 mode iterator.
1131 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
1132 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
1133 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
1134 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
1135 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
1136 SWI48 mode iterator.
1137 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
1138 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
1139 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
1140 pattern from sse_cvttsd2si<round_saeonly_name>
1141 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
1143 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1145 * config/nds32/nds32-md-auxiliary.c
1146 (nds32_valid_smw_lwm_base_p): Refine.
1147 (nds32_output_smw_single_word): Refine.
1148 (nds32_output_smw_double_word): New.
1149 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
1151 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1153 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
1154 (nds32_output_stack_pop): Refine.
1155 (nds32_expand_unaligned_load): Refine.
1156 (nds32_expand_unaligned_store): Refine.
1158 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1159 Chung-Ju Wu <jasonwucj@gmail.com>
1161 * config/nds32/constants.md: Add TP_REGNUM constant.
1162 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
1163 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
1164 UNSPEC_ADD32.
1165 * config/nds32/nds32-doubleword.md: Consider flag_pic.
1166 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
1167 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
1168 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
1169 and PIC code generation.
1170 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
1171 code generation.
1172 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
1173 optimization.
1174 * config/nds32/nds32.md: Support TLS and PIC.
1175 * config/nds32/nds32.c: Support TLS and PIC.
1176 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
1177 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
1178 predicate.
1180 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1182 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
1183 mode with E_ prefix.
1185 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1186 Chung-Ju Wu <jasonwucj@gmail.com>
1188 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
1189 * config/nds32/nds32-md-auxiliary.c
1190 (symbolic_reference_mentioned_p): New.
1191 (nds32_legitimize_ict_address): New.
1192 (nds32_expand_ict_move): New.
1193 (nds32_indirect_call_referenced_p): New.
1194 (nds32_symbol_binds_local_p): Delete.
1195 (nds32_long_call_p): Modify.
1196 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
1197 * config/nds32/nds32-protos.h
1198 (symbolic_reference_mentioned_p): Declare.
1199 (nds32_legitimize_ict_address): Declare.
1200 (nds32_expand_ict_move): Declare.
1201 (nds32_indirect_call_referenced_p): Declare.
1202 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
1203 (nds32_relax_group): Use nds32_ict_const_p as condition.
1204 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
1205 (nds32_asm_file_start): Output ict_model directive in asm code.
1206 (nds32_legitimate_address_p): Consider indirect call.
1207 (nds32_print_operand): Consider indirect call.
1208 (nds32_print_operand_address): Consider indirect call.
1209 (nds32_insert_attributes): Handle "indirect_call" attribute.
1210 (TARGET_LEGITIMATE_ADDRESS_P): Define.
1211 (TARGET_LEGITIMATE_CONSTANT_P): Define.
1212 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
1213 (TARGET_DELEGITIMIZE_ADDRESS): Define.
1214 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
1215 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
1216 (TARGET_ICT_MODEL_SMALL): Define.
1217 (TARGET_ICT_MODEL_LARGE): Define.
1218 * config/nds32/nds32.md (movsi): Consider ict model.
1219 (call, call_value): Consider ict model.
1220 (sibcall, sibcall_value): Consider ict model.
1221 * config/nds32/nds32.opt (mict-model): New option.
1222 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
1223 model.
1225 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
1226 Monk Chiang <sh.chiang04@gmail.com>
1227 Jim Wilson <jimw@sifive.com>
1229 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
1230 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
1231 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
1232 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
1233 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
1234 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
1235 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
1236 compute save_libcall_adjustment properly.
1237 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
1238 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
1239 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
1240 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
1241 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
1242 (ABI_SPEC): Handle mabi=ilp32e.
1243 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
1244 (RVE): Add RVE mask.
1245 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
1246 <-march>: Add rv32e as an example.
1248 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
1250 PR c++/82899
1251 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
1252 (intra_create_variable_infos): Handle C++ constructors.
1254 2018-05-18 Martin Liska <mliska@suse.cz>
1256 * passes.def: Remove a redundant pass.
1258 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
1260 PR bootstrap/85838
1261 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
1263 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1265 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
1266 (ARMv4): Update.
1267 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
1268 (ARMv6m): Update.
1269 (armv2, armv2a, armv3, armv3m): Delete architectures.
1270 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
1271 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
1272 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
1273 Delete cpus.
1274 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
1275 (*mulsidi3adddi): Likewise.
1276 (mulsidi3): Likewise.
1277 (*mulsidi3_nov6): Likewise.
1278 (umulsidi3): Likewise.
1279 (umulsidi3_nov6): Likewise.
1280 (umaddsidi4): Likewise.
1281 (*umulsidi3adddi): Likewise.
1282 (smulsi3_highpart): Likewise.
1283 (*smulsi3_highpart_nov6): Likewise.
1284 (umulsi3_highpart): Likewise.
1285 (*umulsi3_highpart_nov6): Likewise.
1286 * config/arm/arm.h (arm_arch3m): Delete.
1287 * config/arm/arm.c (arm_arch3m): Delete.
1288 (arm_option_override_internal): Update armv3-related comment.
1289 (arm_configure_build_target): Delete use of isa_bit_mode32.
1290 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
1291 (arm_rtx_costs_internal): Delete check of arm_arch3m.
1292 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
1293 (mulsa3): Likewise.
1294 (mulusa3): Likewise.
1295 * config/arm/arm-protos.h (arm_arch3m): Delete.
1296 * config/arm/arm-tables.opt: Regenerate.
1297 * config/arm/arm-tune.md: Likewise.
1298 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
1299 deleted architectures.
1301 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1303 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
1304 (armv5t, armv5te): New features.
1305 (ARMv5, ARMv5e): Delete fgroups.
1306 (ARMv5t, ARMv5te): Adjust for above changes.
1307 (ARMv6m): Likewise.
1308 (armv5, armv5e): Delete arches.
1309 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
1310 arm_arch5.
1311 (*call_reg_arm): Likewise.
1312 (*call_value_reg_armv5): Likewise.
1313 (*call_value_reg_arm): Likewise.
1314 (*call_symbol): Likewise.
1315 (*call_value_symbol): Likewise.
1316 (*sibcall_insn): Likewise.
1317 (*sibcall_value_insn): Likewise.
1318 (clzsi2): Likewise.
1319 (prefetch): Likewise.
1320 (define_split and define_peephole2 dependent on arm_arch5):
1321 Likewise.
1322 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
1323 arm_arch5e.
1324 (TARGET_ARM_QBIT): Likewise.
1325 (TARGET_DSP_MULTIPLY): Likewise.
1326 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
1327 (arm_arch5, arm_arch5e): Delete.
1328 (arm_arch5t, arm_arch5te): Declare.
1329 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
1330 (arm_arch5t): Declare.
1331 (arm_option_reconfigure_globals): Update for the above.
1332 (arm_options_perform_arch_sanity_checks): Update comment, replace
1333 use of arm_arch5 with arm_arch5t.
1334 (use_return_insn): Likewise.
1335 (arm_emit_call_insn): Likewise.
1336 (output_return_instruction): Likewise.
1337 (arm_final_prescan_insn): Likewise.
1338 (arm_coproc_builtin_available): Likewise.
1339 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
1340 arm_arch5e with arm_arch5t and arm_arch5te.
1341 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
1342 (arm_arch5t, arm_arch5te): Declare.
1343 * config/arm/arm-tables.opt: Regenerate.
1344 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
1345 * config/arm/t-multilib: Likewise.
1346 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
1347 instead of arm_arch5.
1348 (*call_reg_thumb1): Likewise.
1349 (*call_value_reg_thumb1_v5): Likewise.
1350 (*call_value_reg_thumb1): Likewise.
1351 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
1352 unreachable path.
1353 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
1355 2018-05-18 Martin Liska <mliska@suse.cz>
1357 PR gcov-profile/84846
1358 * doc/gcov.texi: Document -t option of gcov tool.
1360 2018-05-18 Martin Liska <mliska@suse.cz>
1362 PR gcov-profile/84846
1363 * gcov.c (print_usage): Add new -t option.
1364 (process_args): Handle the option.
1365 (generate_results): Use stdout as output when requested by
1366 the option.
1368 2018-05-18 Martin Liska <mliska@suse.cz>
1370 PR gcov-profile/84846
1371 * coverage.c (coverage_init): Write PWD to .gcno file.
1372 * doc/gcov.texi: Document how working directory is printed.
1373 * gcov-dump.c (dump_gcov_file): Print PWD.
1374 * gcov.c (output_intermediate_file): Likewise.
1375 (read_graph_file): Read PWD string.
1376 (output_lines): Print PWD.
1378 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1380 PR middle-end/85817
1381 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
1382 for retval and return false if all args to phi are zero.
1384 2018-05-18 Richard Biener <rguenther@suse.de>
1386 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
1387 method.
1388 (evrp_dom_walker::before_dom_children): Call it.
1390 2018-05-18 Richard Biener <rguenther@suse.de>
1392 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
1393 results when processing array refs with variable index.
1395 2018-05-18 Toon Moene <toon@moene.org>
1397 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
1398 directly after that of -floop-interchange. Indicate that both
1399 options are enabled by default when specifying -O3.
1401 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1403 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
1404 iterator. Delete separate integer-mode vec_set<mode> expander.
1405 (aarch64_simd_vec_setv2di): Delete.
1406 (vec_setv2di): Delete.
1407 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
1408 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
1409 the "w, r" alternative.
1411 2018-05-18 Martin Liska <mliska@suse.cz>
1413 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
1414 * tree-pass.h (make_pass_lower_switch_O0): New function.
1415 * tree-switch-conversion.c (node_has_low_bound): Remove.
1416 (node_has_high_bound): Likewise.
1417 (node_is_bounded): Likewise.
1418 (class pass_lower_switch): Make it a template type and create
1419 two instances.
1420 (pass_lower_switch::execute): Add template argument.
1421 (make_pass_lower_switch): New function.
1422 (make_pass_lower_switch_O0): New function.
1423 (do_jump_if_equal): Remove.
1424 (emit_case_nodes): Simplify to just handle all 3 cases and leave
1425 all the hard work to tree optimization passes.
1427 2018-05-18 Martin Liska <mliska@suse.cz>
1429 * dbgcnt.c (limit_low): Renamed from limit.
1430 (limit_high): New variable.
1431 (dbg_cnt_is_enabled): Check for upper limit.
1432 (dbg_cnt): Adjust dumping.
1433 (dbg_cnt_set_limit_by_index): Add new argument for high
1434 value.
1435 (dbg_cnt_set_limit_by_name): Likewise.
1436 (dbg_cnt_process_single_pair): Parse new format.
1437 (dbg_cnt_process_opt): Use strtok.
1438 (dbg_cnt_list_all_counters): Remove 'value' and add
1439 'limit_high'.
1440 * doc/invoke.texi: Document changes.
1442 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
1444 * doc/sourcebuild.texi (scalar_all_fma): Document.
1445 * tree.def (FMA_EXPR): Delete.
1446 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
1447 * internal-fn.c (ternary_direct): New macro.
1448 (expand_ternary_optab_fn): Likewise.
1449 (direct_ternary_optab_supported_p): Likewise.
1450 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
1451 * builtins.c (fold_builtin_fma): Delete.
1452 (fold_builtin_3): Don't call it.
1453 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
1454 * expr.c (expand_expr_real_2): Likewise.
1455 * fold-const.c (operand_equal_p): Likewise.
1456 (fold_ternary_loc): Likewise.
1457 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
1458 * gimple.c (DEFTREECODE): Likewise.
1459 * gimplify.c (gimplify_expr): Likewise.
1460 * optabs-tree.c (optab_for_tree_code): Likewise.
1461 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
1462 * tree-eh.c (operation_could_trap_p): Likewise.
1463 (stmt_could_throw_1_p): Likewise.
1464 * tree-inline.c (estimate_operator_cost): Likewise.
1465 * tree-pretty-print.c (dump_generic_node): Likewise.
1466 (op_code_prio): Likewise.
1467 * tree-ssa-loop-im.c (stmt_cost): Likewise.
1468 * tree-ssa-operands.c (get_expr_operands): Likewise.
1469 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
1470 * fold-const-call.h (fold_fma): Delete.
1471 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
1472 CFN_FNMA and CFN_FNMS.
1473 (fold_fma): Delete.
1474 * genmatch.c (combined_fn): New enum.
1475 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
1476 (commutative_op): New function.
1477 (commutate): Use it. Handle more than 2 operands.
1478 (dt_operand::gen_gimple_expr): Use commutative_op.
1479 (parser::parse_expr): Allow :c to be used with non-binary
1480 operators if the commutative operand is known.
1481 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
1482 CFN_FMS, CFN_FNMA and CFN_FNMS.
1483 (backprop::process_assign_use): Remove FMA_EXPR handling.
1484 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
1485 (gen_hsa_fma): New function.
1486 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
1487 IFN_FNMA and IFN_FNMS.
1488 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
1489 * gimple-fold.h (follow_all_ssa_edges): Declare.
1490 * gimple-fold.c (follow_all_ssa_edges): New function.
1491 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
1492 gimple_build interface and use follow_all_ssa_edges to fold the result.
1493 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
1494 instead of checking for optabs directly.
1495 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
1496 rather than FMA_EXPRs.
1497 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
1498 call to IFN_FMA instead of an FMA_EXPR.
1500 2018-05-17 Jim Wilson <jimw@sifive.com>
1502 * expr.c (do_tablejump): When converting index to Pmode, if we have a
1503 sign extended promoted subreg, and the range does not have the sign bit
1504 set, then do a sign extend.
1506 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
1507 test, check for sign extended subreg and/or constant operands, and
1508 do a sign extend in that case.
1510 2018-05-17 Steve Ellcey <sellcey@cavium.com>
1512 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
1513 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
1514 Add untyped.
1515 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
1516 Change logics_shift_reg to logics_shift_imm.
1517 (thunderx2t99_fp_loadpair_basic): Delete.
1518 (thunderx2t99_fp_storepair_basic): Delete.
1519 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
1520 (thunderx2t99_asimd_polynomial): Delete.
1521 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
1522 and neon_fp_mul_d_scalar_q.
1523 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
1524 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
1525 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
1526 (thunderx2t99_asimd_lut): Add missing tbl types.
1527 (thunderx2t99_asimd_ext): Delete.
1528 (thunderx2t99_asimd_load1_1_mult): Delete.
1529 (thunderx2t99_asimd_load1_2_mult): Delete.
1530 (thunderx2t99_asimd_load1_ldp): New.
1531 (thunderx2t99_asimd_load1): New.
1532 (thunderx2t99_asimd_load2): Add missing *load2* types.
1533 (thunderx2t99_asimd_load3): New.
1534 (thunderx2t99_asimd_load4): New.
1535 (thunderx2t99_asimd_store1_1_mult): Delete.
1536 (thunderx2t99_asimd_store1_2_mult): Delete.
1537 (thunderx2t99_asimd_store2_mult): Delete.
1538 (thunderx2t99_asimd_store2_onelane): Delete.
1539 (thunderx2t99_asimd_store_stp): New.
1540 (thunderx2t99_asimd_store1): New.
1541 (thunderx2t99_asimd_store2): New.
1542 (thunderx2t99_asimd_store3): New.
1543 (thunderx2t99_asimd_store4): New.
1545 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
1547 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
1548 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
1550 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
1551 Segher Boessenkool <segher@kernel.crashing.org>
1553 PR target/85698
1554 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
1555 operand.
1557 2018-05-17 Richard Biener <rguenther@suse.de>
1559 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
1560 for pruning loop and prune defs feeding only already visited PHIs.
1562 2018-05-17 Richard Biener <rguenther@suse.de>
1564 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
1566 2018-05-17 Bin Cheng <bin.cheng@arm.com>
1567 Richard Biener <rguenther@suse.de>
1569 PR tree-optimization/85793
1570 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
1571 for VMAT_ELEMENTWISE.
1573 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1575 * internal-fn.h (lookup_internal_fn): Declare
1576 * internal-fn.c (lookup_internal_fn): New function.
1577 * gimple.c (gimple_build_call_from_tree): Handle calls to
1578 internal functions.
1579 * gimple-pretty-print.c (dump_gimple_call): Print "." before
1580 internal function names.
1581 * tree-pretty-print.c (dump_generic_node): Likewise.
1582 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
1584 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
1586 * gimple-fold.h (gimple_build): Make the function forms take
1587 combined_fn rather than built_in_function.
1588 (gimple_simplify): Likewise.
1589 * gimple-match-head.c (gimple_simplify): Likewise.
1590 * gimple-fold.c (gimple_build): Likewise.
1591 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
1592 rather than gimple_build_call_internal.
1593 (get_initial_defs_for_reduction): Likewise.
1594 (vect_create_epilog_for_reduction): Likewise.
1595 (vectorizable_live_operation): Likewise.
1597 2018-05-17 Martin Liska <mliska@suse.cz>
1599 * gimple-ssa-sprintf.c (format_directive): Do not use
1600 space in between 'G_' and '('.
1602 2018-05-17 Jakub Jelinek <jakub@redhat.com>
1604 PR target/85323
1605 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
1606 even if the mask is not all ones.
1608 PR target/85323
1609 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
1610 vector.
1611 (ix86_gimple_fold_builtin): Likewise.
1613 PR target/85323
1614 * config/i386/i386.c: Include tree-vector-builder.h.
1615 (ix86_vector_shift_count): New function.
1616 (ix86_fold_builtin): Fold shift builtins by scalar count.
1617 (ix86_gimple_fold_builtin): Likewise.
1619 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
1620 _mm512_setzero): New intrinsics.
1622 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
1623 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1625 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
1626 code generation for cases where splatting a value is not useful.
1627 * simplify-rtx.c (simplify_ternary_operation): Simplify
1628 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
1629 mode to a vec_concat.
1631 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
1633 * config.gcc: Support "goldmont-plus".
1634 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1635 "goldmont-plus".
1636 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1637 PROCESSOR_GOLDMONT_PLUS.
1638 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
1639 (processor_target_table): Add "goldmont-plus".
1640 (PTA_GOLDMONT_PLUS): Define.
1641 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
1642 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
1643 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
1644 (fold_builtin_cpu): Add "goldmont-plus".
1645 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
1646 (ix86_option_override_internal): Add "goldmont-plus".
1647 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
1648 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
1649 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
1650 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
1652 2018-05-17 Richard Biener <rguenther@suse.de>
1654 PR tree-optimization/85757
1655 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
1656 remove defs that only feed that PHI from further processing.
1658 2018-05-16 Jim Wilson <jimw@sifive.com>
1660 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
1661 asterisk to name.
1662 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
1664 2018-05-16 Mark Wielaard <mark@klomp.org>
1666 * dwarf2out.c (count_index_strings): New function.
1667 (output_indirect_strings): Call count_index_strings and generate
1668 header for dwarf_version >= 5.
1670 2018-05-16 Mark Wielaard <mark@klomp.org>
1672 * dwarf2out.c (dwarf_FORM): New function.
1673 (set_indirect_string): Use dwarf_FORM.
1674 (reset_indirect_string): Likewise.
1675 (size_of_die): Likewise.
1676 (value_format): Likewise.
1677 (output_die): Likewise.
1678 (add_skeleton_AT_string): Likewise.
1679 (output_macinfo_op): Likewise.
1680 (index_string): Likewise.
1681 (output_index_string_offset): Likewise.
1682 (output_index_string): Likewise.
1683 (count_index_strings): Likewise.
1685 2018-05-16 Carl Love <cel@us.ibm.com>
1687 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
1688 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
1690 2018-05-16 Martin Jambor <mjambor@suse.cz>
1692 * ipa-prop.c (ipa_free_all_edge_args): Remove.
1693 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
1695 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
1697 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
1698 (fnma<mode>4): Likewise.
1699 (fms<mode>4): Likewise.
1700 (fnms<mode>4): Likewise.
1701 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
1702 (aarch64_fnma<mode>4): Likewise.
1703 (aarch64_fms<mode>4): Likewise.
1704 (aarch64_fnms<mode>4): Likewise.
1705 (aarch64_fnmadd<mode>4): Likewise.
1707 2018-05-16 Jason Merrill <jason@redhat.com>
1709 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
1711 2018-05-16 Richard Biener <rguenther@suse.de>
1713 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
1714 (dump_stmt_cost): Declare.
1715 (add_stmt_cost): Dump cost we add.
1716 (add_stmt_costs): New function.
1717 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
1718 No longer exported.
1719 (vect_analyze_stmt): Adjust prototype.
1720 (vectorizable_condition): Likewise.
1721 (vectorizable_live_operation): Likewise.
1722 (vectorizable_reduction): Likewise.
1723 (vectorizable_induction): Likewise.
1724 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
1725 cost vector to pass to vectorizable_ and record afterwards.
1726 (vect_model_reduction_cost): Take cost vector argument and adjust.
1727 (vect_model_induction_cost): Likewise.
1728 (vectorizable_reduction): Likewise.
1729 (vectorizable_induction): Likewise.
1730 (vectorizable_live_operation): Likewise.
1731 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
1732 SLP_TREE_NUMBER_OF_VEC_STMTS.
1733 (vect_analyze_slp_cost_1): Remove.
1734 (vect_analyze_slp_cost): Likewise.
1735 (vect_slp_analyze_node_operations): Take visited args and
1736 a target cost vector. Avoid processing already visited stmt sets.
1737 (vect_slp_analyze_operations): Use a local cost vector to gather
1738 costs and register those of non-discarded instances.
1739 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
1740 (vect_schedule_slp_instance): Remove copying of
1741 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
1742 zero.
1743 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
1744 adding cost. Record cost entry location.
1745 (vect_prologue_cost_for_slp_op): Function to compute cost of
1746 a constant or invariant generated for SLP vect in the prologue,
1747 split out from vect_analyze_slp_cost_1.
1748 (vect_model_simple_cost): Make static. Adjust for SLP costing.
1749 (vect_model_promotion_demotion_cost): Likewise.
1750 (vect_model_store_cost): Likewise, make static.
1751 (vect_model_load_cost): Likewise.
1752 (vectorizable_bswap): Add cost vector arg and adjust.
1753 (vectorizable_call): Likewise.
1754 (vectorizable_simd_clone_call): Likewise.
1755 (vectorizable_conversion): Likewise.
1756 (vectorizable_assignment): Likewise.
1757 (vectorizable_shift): Likewise.
1758 (vectorizable_operation): Likewise.
1759 (vectorizable_store): Likewise.
1760 (vectorizable_load): Likewise.
1761 (vectorizable_condition): Likewise.
1762 (vectorizable_comparison): Likewise.
1763 (can_vectorize_live_stmts): Likewise.
1764 (vect_analyze_stmt): Likewise.
1765 (vect_transform_stmt): Adjust calls to vectorizable_*.
1766 * tree-vectorizer.c: Include gimple-pretty-print.h.
1767 (dump_stmt_cost): New function.
1769 2018-05-16 Richard Biener <rguenther@suse.de>
1771 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
1772 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
1773 * tree-ssa-dse.c: Include tree-ssa-loop.h.
1774 (check_name): New callback.
1775 (dse_classify_store): Track cycles via a visited bitmap of PHI
1776 defs and simplify handling of in-loop and across loop dead stores
1777 and properly fail for loop-variant refs. Handle byte-tracking with
1778 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
1779 limiting the walk.
1781 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
1783 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
1784 (vect_get_mask_type_for_stmt): Likewise.
1785 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
1786 split out from...
1787 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
1788 to determine the statement's vector type and the vector type that
1789 should be used for calculating nunits. Deal with cases in which
1790 the type has to be deferred.
1791 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
1792 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
1793 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
1794 (vect_determine_vf_for_stmt): New functions, split out from...
1795 (vect_determine_vectorization_factor): ...here.
1796 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
1797 (vect_get_mask_type_for_stmt): New functions, split out from
1798 vect_determine_vectorization_factor.
1800 2018-05-16 Richard Biener <rguenther@suse.de>
1802 * tree-cfg.c (verify_gimple_assign_ternary): Properly
1803 verify the [VEC_]COND_EXPR embedded comparison.
1805 2018-05-15 Martin Sebor <msebor@redhat.com>
1807 PR tree-optimization/85753
1808 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
1809 RECORD_TYPE in addition to ARRAY_TYPE.
1811 2018-05-15 Martin Sebor <msebor@redhat.com>
1813 PR middle-end/85643
1814 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
1816 2018-05-15 Richard Biener <rguenther@suse.de>
1818 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
1819 add by_clobber_p one. Change algorithm to collect all defs
1820 representing uses we need to walk and try reducing them to
1821 a single one before failing.
1822 (dse_dom_walker::dse_optimize_stmt): Adjust.
1824 2018-05-13 Mark Wielaard <mark@klomp.org>
1826 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
1827 (size_of_loc_descr): Likewise.
1828 (output_loc_operands): Likewise.
1829 (output_loc_operands_raw): Likewise.
1830 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
1831 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
1832 (hash_loc_operands): Likewise.
1833 (compare_loc_operands): Likewise.
1835 2018-05-14 Mark Wielaard <mark@klomp.org>
1837 * dwarf2out.c (count_index_addrs): New function.
1838 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
1840 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1842 PR tree-optimization/83648
1843 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
1844 return value as malloc candidate.
1846 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1848 PR ipa/85734
1849 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
1850 as true in call to suggest_attribute.
1852 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
1854 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
1855 -mreadonly-in-sdata.
1857 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1859 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
1860 New pattern.
1861 (aarch64_crypto_aesd_fused): Likewise.
1863 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
1865 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
1866 (movsi_aarch64): Likewise.
1867 (load_pairsi): Likewise.
1868 (load_pairdi): Likewise.
1869 (store_pairsi): Likewise.
1870 (store_pairdi): Likewise.
1871 (load_pairsf): Likewise.
1872 (load_pairdf): Likewise.
1873 (store_pairsf): Likewise.
1874 (store_pairdf): Likewise.
1875 (zero_extend): Likewise.
1876 (trunc): Swap alternatives.
1877 (fcvt_target): Add '?' to prefer w over r.
1879 2018-05-14 Jakub Jelinek <jakub@redhat.com>
1881 PR target/85756
1882 * config/i386/i386.md: Disallow non-commutative arithmetics in
1883 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
1884 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
1885 in the peephole2 before it.
1887 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
1889 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
1890 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
1891 (ix86_handle_option): Handle -mcldemote.
1892 * config.gcc: New header.
1893 * config/i386/cldemoteintrin.h: New file.
1894 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
1895 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
1896 -mcldemote.
1897 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1898 OPTION_MASK_ISA_CLDEMOTE.
1899 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
1900 (ix86_valid_target_attribute_inner_p): Ditto.
1901 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
1902 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
1903 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
1904 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
1905 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
1906 (cldemote): New.
1907 * config/i386/i386.opt: Add -mcldemote.
1908 * config/i386/x86intrin.h: New header.
1909 * doc/invoke.texi: Add -mcldemote.
1911 2018-05-14 Richard Biener <rguenther@suse.de>
1913 * doc/match-and-simplify.texi: Adjust :s documentation.
1915 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
1917 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
1918 intended memcpy size.
1919 (REORDER_45): Likewise.
1921 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
1923 * sort.cc: New file.
1924 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
1925 * vec.c (qsort_chk): Use gcc_qsort.
1926 * Makefile.in (OBJS-libcommon): Add sort.o.
1927 (build/sort.o): New target. Use it...
1928 (BUILD_RTL): ... here, and...
1929 (build/gencfn-macros): ... here, and...
1930 (build/genmatch): ... here.
1932 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1933 Chung-Ju Wu <jasonwucj@gmail.com>
1935 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
1936 * config/nds32/nds32-graywolf.md: New file.
1937 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
1938 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
1939 pipeline.
1940 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
1941 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
1942 * config/nds32/nds32.md (pipeline_model): Add graywolf.
1943 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
1944 * config/nds32/pipelines.md: Include n15 settings.
1946 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1947 Chung-Ju Wu <jasonwucj@gmail.com>
1949 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
1950 * config/nds32/nds32-n13.md: New file.
1951 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
1952 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
1953 pipeline.
1954 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
1955 * config/nds32/nds32.md (pipeline_model): Add n13.
1956 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
1957 * config/nds32/pipelines.md: Include n13 settings.
1959 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
1960 Chung-Ju Wu <jasonwucj@gmail.com>
1962 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
1963 * config/nds32/nds32-n10.md: New file.
1964 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
1965 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
1966 pipeline.
1967 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
1968 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
1969 * config/nds32/nds32.md (pipeline_model): Add n10.
1970 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
1971 * config/nds32/pipelines.md: Include n10 settings.
1973 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
1974 Kito Cheng <kito.cheng@gmail.com>
1975 Chung-Ju Wu <jasonwucj@gmail.com>
1977 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
1978 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1979 Add enum values for DSP extension instructions.
1980 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
1981 New constraints.
1982 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
1983 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
1984 New code iterators.
1985 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
1986 * config/nds32/nds32-dspext.md: New file for DSP implementation.
1987 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
1988 * config/nds32/nds32-intrinsic.md: Likewise.
1989 * config/nds32/nds32_intrinsic.h: Likewise.
1990 * config/nds32/nds32-md-auxiliary.c: Likewise.
1991 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
1992 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
1993 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
1994 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
1995 * config/nds32/nds32-protos.h: New declarations for DSP extension.
1996 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
1997 TYPE_DMAC in switch statement.
1998 * config/nds32/nds32.c: New checking and implementation for DSP
1999 extension instructions.
2000 * config/nds32/nds32.h: Likewise.
2001 * config/nds32/nds32.md: Likewise.
2002 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
2003 * config/nds32/predicates.md: Implement new predicates for DSP
2004 extension.
2006 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
2008 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
2009 Reformat alternatives and attributes so it is easier to identify
2010 which constraints/attributes go with which instruction.
2011 (mov<mode>_hardfloat32, FMOVE64): Likewise.
2012 (mov<mode>_softfloat32, FMOVE64): Likewise.
2013 (mov<mode>_hardfloat64, FMOVE64): Likewise.
2014 (mov<mode>_softfloat64, FMOVE64): Likewise.
2016 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
2018 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2019 subsection.
2020 (Basic PowerPC Built-in Functions): The new name of the
2021 subsection previously known as "PowerPC Built-in Functions".
2022 (Basic PowerPC Built-in Functions Available on all Configurations):
2023 New subsubsection.
2024 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
2025 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
2026 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
2027 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
2029 2018-05-11 Martin Jambor <mjambor@suse.cz>
2031 PR ipa/85655
2032 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
2033 single const.
2035 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2037 PR target/85733
2038 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
2040 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
2042 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
2043 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
2044 (ix86_handle_option): Handle -mwaitpkg.
2045 * config.gcc: New header.
2046 * config/i386/cpuid.h (bit_WAITPKG): New bit.
2047 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
2048 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
2049 function type.
2050 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2051 OPTION_MASK_ISA_WAITPKG.
2052 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
2053 (ix86_option_override_internal): Add PTA_WAITPKG.
2054 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
2055 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
2056 IX86_BUILTIN_TPAUSE.
2057 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
2058 __builtin_ia32_umwait and __builtin_ia32_tpause.
2059 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
2060 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
2061 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
2062 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
2063 UNSPECV_TPAUSE): New.
2064 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
2065 * config/i386/i386.opt: Add -mwaitpkg.
2066 * config/i386/waitpkgintrin.h: New file.
2067 * config/i386/x86intrin.h: New header.
2068 * doc/invoke.texi: Add -mwaitpkg.
2070 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2072 PR target/85606
2073 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
2074 equivalent.
2075 (cortex-m0): Use armv6s-m isa.
2076 (cortex-m0plus): Likewise.
2077 (cortex-m1): Likewise.
2078 (cortex-m0.small-multiply): Likewise.
2079 (cortex-m0plus.small-multiply): Likewise.
2080 (cortex-m1.small-multiply): Likewise.
2082 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
2083 Jakub Jelinek <jakub@redhat.com>
2085 PR tree-optimization/85692
2086 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
2087 source permute as well.
2089 2018-05-11 Martin Liska <mliska@suse.cz>
2091 PR sanitizer/85556
2092 * doc/extend.texi: Document LLVM style format for no_sanitize
2093 attribute.
2095 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
2097 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
2098 mode_supports_vsx_dform_quad to mode_supports_dq_form.
2099 (mode_supports_vsx_dform_quad): Likewise.
2100 (mode_supports_vmx_dform): Move these functions to be next to the
2101 other mode_supports functions.
2102 (mode_supports_dq_form): Likewise.
2103 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
2104 mode_supports_dq_form.
2105 (reg_offset_addressing_ok_p): Likewise.
2106 (offsettable_ok_by_alignment): Likewise.
2107 (rs6000_legitimate_offset_address_p): Likewise.
2108 (legitimate_lo_sum_address_p): Likewise.
2109 (rs6000_legitimize_address): Likewise.
2110 (rs6000_legitimize_reload_address): Likewise.
2111 (rs6000_secondary_reload_inner): Likewise.
2112 (rs6000_preferred_reload_class): Likewise.
2113 (rs6000_output_move_128bit): Likewise.
2115 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2117 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
2118 Generate SImode target register for null target.
2119 <case IX86_BUILTIN_XGETBV>: Ditto.
2120 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
2121 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
2123 2018-05-10 Carl Love <cel@us.ibm.com>
2125 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2126 dcbtt and dcbtstt if operands[2] is 0.
2128 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2130 PR target/85693
2131 * config/i386/sse.md (usadv64qi): New expander.
2133 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
2135 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
2136 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
2137 -maltivec=be support.
2138 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
2139 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
2140 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
2141 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
2142 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
2143 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
2144 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
2145 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
2146 altivec_vsumsws): Adjust.
2147 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
2148 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
2149 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
2150 support.
2151 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
2152 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
2153 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
2154 (altivec_lve<VI_char>x): Delete expand.
2155 (*altivec_lve<VI_char>x_internal): Rename to...
2156 (altivec_lve<VI_char>x): ... this.
2157 (altivec_lvxl_<mode>): Delete expand.
2158 (*altivec_lvxl_<mode>_internal): Rename to ...
2159 (altivec_lvxl_<mode>): ... this.
2160 (altivec_stvxl_<mode>): Delete expand.
2161 (*altivec_stvxl_<mode>_internal): Rename to ...
2162 (altivec_stvxl_<mode>): ... this.
2163 (altivec_stve<VI_char>x): Delete expand.
2164 (*altivec_stve<VI_char>x_internal): Rename to ...
2165 (altivec_stve<VI_char>x): ... this.
2166 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
2167 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
2168 reduc_plus_scal_<mode>): Adjust.
2169 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
2170 comment.
2171 (rs6000_cpu_cpp_builtins): Adjust.
2172 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
2173 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
2174 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
2175 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
2176 -maltivec=be support.
2177 (rs6000_split_vec_extract_var): Adjust.
2178 (rs6000_split_v4si_init): Adjust.
2179 (swap_selector_for_mode): Delete.
2180 (altivec_expand_lvx_be, altivec_expand_stvx_be,
2181 altivec_expand_stvex_be): Delete.
2182 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
2183 -maltivec=be support.
2184 (rs6000_gimple_fold_builtin): Ditto.
2185 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
2186 Adjust.
2187 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
2188 (TARGET_DIRECT_MOVE_64BIT): Adjust.
2189 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
2190 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
2191 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
2192 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
2193 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
2194 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
2195 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
2196 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
2197 anonymous split): Adjust.
2198 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
2199 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
2201 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
2203 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
2204 when --with-gxx-include-dir is also specified.
2205 * configure: Regenerate.
2207 2018-05-09 Jim Wilson <jimw@sifive.com>
2209 PR target/84797
2210 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
2211 * config/riscv/t-withmultilib: New.
2212 * config/riscv/withmultilib.h: New.
2213 * doc/install.texi: Document RISC-V --with-multilib-list support.
2215 2018-05-09 Richard Biener <rguenther@suse.de>
2217 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
2218 vector.
2219 (vect_bb_vectorization_profitable_p): Adjust. Compute
2220 actual scalar cost using the cost vector and the add_stmt_cost
2221 machinery.
2223 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2225 PR rtl-optimization/85645
2226 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
2227 in the REG_CFA_REGISTER note for LR, don't leave it empty.
2229 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2231 PR rtl-optimization/85645
2232 * shrink-wrap.c (spread_components): Return a boolean saying if
2233 anything was changed.
2234 (try_shrink_wrapping_separate): Iterate spread_components until
2235 nothing changes anymore.
2237 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2239 PR rtl-optimization/85645
2240 * regrename.c (build_def_use): Also kill the chains that include the
2241 destination of a REG_CFA_REGISTER note.
2243 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2245 PR rtl-optimization/85645
2246 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
2247 insn that has a REG_CFA_REGISTER note.
2249 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
2251 * cfgexpand.c (expand_clobber): New function.
2252 (expand_gimple_stmt_1): Use it.
2253 * tree-vect-stmts.c (vect_clobber_variable): New function,
2254 split out from...
2255 (vectorizable_simd_clone_call): ...here.
2256 (vectorizable_store): Emit a clobber either side of an
2257 IFN_STORE_LANES sequence.
2258 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
2260 2018-05-09 Tom de Vries <tom@codesourcery.com>
2262 PR target/85626
2263 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
2264 (define_insn "trap_if_false"): Add exit after trap.
2266 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
2268 PR rtl-optimization/85638
2269 * bb-reorder.c: Include common/common-target.h.
2270 (create_forwarder_block): New function extracted from...
2271 (fix_up_crossing_landing_pad): ...here. Rename into...
2272 (dw2_fix_up_crossing_landing_pad): ...this.
2273 (sjlj_fix_up_crossing_landing_pad): New function.
2274 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
2275 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
2276 from both partitions and exit the loop after one iteration.
2278 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2280 Revert:
2281 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2282 subsection.
2283 (Basic PowerPC Built-in Functions): The new name of the
2284 subsection previously known as "PowerPC Built-in Functions".
2285 (Basic PowerPC Built-in Functions Available on all Configurations):
2286 New subsubsection.
2287 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2288 subsubsection.
2289 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2290 subsubsection.
2291 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2292 subsubsection.
2293 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2294 subsubsection.
2296 2018-05-08 Jim Wilson <jimw@sifive.com>
2298 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
2299 (LD_EMUL_SUFFIX): New.
2300 (LINK_SPEC): Use it.
2302 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2304 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2305 subsection.
2306 (Basic PowerPC Built-in Functions): The new name of the
2307 subsection previously known as "PowerPC Built-in Functions".
2308 (Basic PowerPC Built-in Functions Available on all Configurations):
2309 New subsubsection.
2310 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2311 subsubsection.
2312 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2313 subsubsection.
2314 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2315 subsubsection.
2316 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2317 subsubsection.
2319 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2321 PR target/85683
2322 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
2323 after cmpelim optimization.
2325 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
2327 * config.gcc: Support "goldmont".
2328 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
2329 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2330 PROCESSOR_GOLDMONT.
2331 * config/i386/i386.c (m_GOLDMONT): Define.
2332 (processor_target_table): Add "goldmont".
2333 (PTA_GOLDMONT): Define.
2334 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
2335 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
2336 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
2337 (fold_builtin_cpu): Add "goldmont".
2338 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
2339 (ix86_option_override_internal): Add "goldmont".
2340 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
2341 (processor_type): Add PROCESSOR_GOLDMONT.
2342 * config/i386/i386.md: Add CPU "glm".
2343 * config/i386/glm.md: New file.
2344 * config/i386/x86-tune.def: Add m_GOLDMONT.
2345 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
2347 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2349 PR target/85572
2350 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
2351 E_V4DImode.
2352 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
2353 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
2354 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
2356 PR target/85317
2357 * config/i386/i386.c (ix86_fold_builtin): Handle
2358 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
2360 PR target/85480
2361 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
2362 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
2364 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
2366 PR target/85658
2367 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
2368 (check_arch): Likewise.
2369 (check_fpu): Return the result rather than printing it.
2370 (end arch): Fix operator precedence.
2371 (end cpu): Likewise.
2372 (END): Print the result from check_fpu.
2374 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2375 Alan Hayward <alan.hayward@arm.com>
2376 David Sherwood <david.sherwood@arm.com>
2378 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
2379 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
2380 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
2381 (*fcmuo<mode>_and): New patterns.
2383 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2385 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
2386 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
2387 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
2388 (cmp_op, sve_imm_con): New code attributes.
2389 (SVE_COND_INT_CMP, imm_con): Delete.
2390 (cmp_op): Remove above unspecs from int attribute.
2391 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
2392 to...
2393 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
2394 comparison-specific unspecs.
2395 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
2396 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
2397 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
2398 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
2399 (*vec_fcm<cmp_op><mode>): Rename to...
2400 (*fcm<cmp_op><mode>): ...this and adjust likewise.
2401 (*vec_fcmuo<mode>): Rename to...
2402 (*fcmuo<mode>): ...this and adjust likewise.
2403 (*pred_fcm<cmp_op><mode>): New pattern.
2404 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
2405 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
2406 functions.
2407 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
2408 and UNORDERED.
2409 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
2410 (aarch64_emit_sve_predicated_cond): New function.
2411 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
2412 (aarch64_emit_unspec_cond_or): Replace with...
2413 (aarch64_emit_sve_or_conds): ...this new function. Use
2414 aarch64_emit_sve_ptrue_op for the individual comparisons and
2415 aarch64_emit_binop to OR them together.
2416 (aarch64_emit_inverted_unspec_cond): Replace with...
2417 (aarch64_emit_sve_inverted_cond): ...this new function. Use
2418 aarch64_emit_sve_ptrue_op for the comparison and
2419 aarch64_emit_unop to invert the result.
2420 (aarch64_expand_sve_vec_cmp_float): Update after the above
2421 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
2423 2018-05-07 Nathan Sidwell <nathan@acm.org>
2425 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
2426 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
2427 (Backwards Compatibility): Likewise.
2429 2018-05-07 Luis Machado <luis.machado@linaro.org>
2431 PR bootstrap/85681
2432 Revert:
2433 2018-05-07 Luis Machado <luis.machado@linaro.org>
2435 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2436 <prefetch_dynamic_strides>: New const bool field.
2437 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2438 prefetch_dynamic_strides.
2439 (exynosm1_prefetch_tune): Likewise.
2440 (thunderxt88_prefetch_tune): Likewise.
2441 (thunderx_prefetch_tune): Likewise.
2442 (thunderx2t99_prefetch_tune): Likewise.
2443 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2444 (aarch64_override_options_internal): Update to set
2445 PARAM_PREFETCH_DYNAMIC_STRIDES.
2446 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2447 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2448 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2449 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2450 prefetch-dynamic-strides setting.
2452 2018-05-07 Luis Machado <luis.machado@linaro.org>
2454 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2455 <minimum_stride>: New const int field.
2456 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2457 minimum_stride field.
2458 (exynosm1_prefetch_tune): Likewise.
2459 (thunderxt88_prefetch_tune): Likewise.
2460 (thunderx_prefetch_tune): Likewise.
2461 (thunderx2t99_prefetch_tune): Likewise.
2462 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2463 (aarch64_override_options_internal): Update to set
2464 PARAM_PREFETCH_MINIMUM_STRIDE.
2465 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2466 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2467 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2468 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2469 stride is constant and is below the minimum stride threshold.
2471 2018-05-07 Luis Machado <luis.machado@linaro.org>
2473 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
2474 to 512.
2476 2018-05-07 Luis Machado <luis.machado@linaro.org>
2478 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2479 <prefetch_dynamic_strides>: New const bool field.
2480 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2481 prefetch_dynamic_strides.
2482 (exynosm1_prefetch_tune): Likewise.
2483 (thunderxt88_prefetch_tune): Likewise.
2484 (thunderx_prefetch_tune): Likewise.
2485 (thunderx2t99_prefetch_tune): Likewise.
2486 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2487 (aarch64_override_options_internal): Update to set
2488 PARAM_PREFETCH_DYNAMIC_STRIDES.
2489 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2490 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2491 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2492 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2493 prefetch-dynamic-strides setting.
2495 2018-05-07 Luis Machado <luis.machado@linaro.org>
2497 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2498 <minimum_stride>: New const int field.
2499 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2500 minimum_stride field.
2501 (exynosm1_prefetch_tune): Likewise.
2502 (thunderxt88_prefetch_tune): Likewise.
2503 (thunderx_prefetch_tune): Likewise.
2504 (thunderx2t99_prefetch_tune): Likewise.
2505 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2506 (aarch64_override_options_internal): Update to set
2507 PARAM_PREFETCH_MINIMUM_STRIDE.
2508 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2509 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2510 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2511 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2512 stride is constant and is below the minimum stride threshold.
2514 2018-05-06 Jakub Jelinek <jakub@redhat.com>
2516 PR c++/85659
2517 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
2518 the type is addressable. Don't force op into register if it has
2519 BLKmode.
2521 2018-05-05 Roland McGrath <mcgrathr@google.com>
2523 PR other/77609
2524 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
2525 any section for which we don't know a specific type it should have,
2526 regardless of name. Previously this was done only for the exact
2527 names ".init_array", ".fini_array", and ".preinit_array".
2528 (default_elf_asm_named_section): Add comment about
2529 relationship with default_section_type_flags and SECTION_NOTYPE.
2530 (get_section): Don't consider it a type conflict if one side has
2531 SECTION_NOTYPE and the other doesn't, as long as neither has the
2532 SECTION_BSS et al used in the default_section_type_flags logic.
2534 2018-05-05 Tom de Vries <tom@codesourcery.com>
2536 PR target/85653
2537 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
2538 (workaround_barsyncs): New function.
2539 (nvptx_reorg): Use workaround_barsyncs.
2540 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2541 (define_expand "nvptx_membar_cta"): New define_expand.
2542 (define_insn "*nvptx_membar_cta"): New insn.
2544 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
2546 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
2547 To improve optimization opportunities.
2548 * builtin-types.def: The new needed builtin types for the above.
2550 2018-05-04 Richard Biener <rguenther@suse.de>
2552 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
2553 * gimple-ssa-store-merging.c
2554 (imm_store_chain_info::output_merged_store): Remove redundant create,
2555 release split_store vector contents on failure.
2556 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
2557 scalar stmt vector on cache hit.
2559 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
2561 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
2562 Xilinx FP support.
2563 * config.gcc (powerpc-xilinx-eabi*): Remove.
2564 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
2565 support.
2566 (fusion_addis_mem_combo_load): Ditto.
2567 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
2568 FP support.
2569 (rs6000_cpu_cpp_builtins): Ditto.
2570 * config/rs6000/rs6000-linux.c
2571 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
2572 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
2573 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
2574 support.
2575 (rs6000_setup_reg_addr_masks): Ditto.
2576 (rs6000_init_hard_regno_mode_ok): Ditto.
2577 (rs6000_option_override_internal): Ditto.
2578 (legitimate_lo_sum_address_p): Ditto.
2579 (rs6000_legitimize_address): Ditto.
2580 (rs6000_legitimize_reload_address): Ditto.
2581 (rs6000_legitimate_address_p): Ditto.
2582 (abi_v4_pass_in_fpr): Ditto.
2583 (setup_incoming_varargs): Ditto.
2584 (rs6000_gimplify_va_arg): Ditto.
2585 (rs6000_split_multireg_move): Ditto.
2586 (rs6000_savres_strategy): Ditto.
2587 (rs6000_emit_prologue_components): Ditto.
2588 (rs6000_emit_epilogue_components): Ditto.
2589 (rs6000_emit_prologue): Ditto.
2590 (rs6000_emit_epilogue): Ditto.
2591 (rs6000_elf_file_end): Ditto.
2592 (rs6000_function_value): Ditto.
2593 (rs6000_libcall_value): Ditto.
2594 * config/rs6000/rs6000.h: Ditto.
2595 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
2596 (TARGET_MINMAX): ... this. New.
2597 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
2598 * config/rs6000/rs6000.md: Remove Xilinx FP support.
2599 (*movsi_internal1_single): Delete.
2600 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
2601 mfpu=, mxilinx-fpu): Delete.
2602 * config/rs6000/singlefp.h: Delete.
2603 * config/rs6000/sysv4.h: Remove Xilinx FP support.
2604 * config/rs6000/t-rs6000: Ditto.
2605 * config/rs6000/t-xilinx: Delete.
2606 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
2607 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
2608 (VStype_simple): Delete.
2609 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
2610 * config/rs6000/xfpu.h: Delete.
2611 * config/rs6000/xfpu.md: Delete.
2612 * config/rs6000/xilinx.h: Delete.
2613 * config/rs6000/xilinx.opt: Delete.
2614 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
2615 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
2617 2018-05-04 Tom de Vries <tom@codesourcery.com>
2619 PR libgomp/85639
2620 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
2621 if ignore == 0.
2623 2018-05-04 Richard Biener <rguenther@suse.de>
2625 PR middle-end/85627
2626 * tree-complex.c (update_complex_assignment): We are always in SSA form.
2627 (expand_complex_div_wide): Likewise.
2628 (expand_complex_operations_1): Likewise.
2629 (expand_complex_libcall): Preserve EH info of the original stmt.
2630 (tree_lower_complex): Handle removed blocks.
2631 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
2632 on complex multiplication and division libcall builtins.
2634 2018-05-04 Richard Biener <rguenther@suse.de>
2636 PR middle-end/85574
2637 * fold-const.c (negate_expr_p): Restrict negation of operand
2638 zero of a division to when we know that can happen without
2639 overflow.
2640 (fold_negate_expr_1): Likewise.
2642 2018-05-04 Jakub Jelinek <jakub@redhat.com>
2644 PR libstdc++/85466
2645 * real.h (real_nextafter): Declare.
2646 * real.c (real_nextafter): New function.
2647 * fold-const-call.c (fold_const_nextafter): New function.
2648 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
2649 CASE_CFN_NEXTTOWARD.
2650 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
2651 even when arg1_mode is different from arg0_mode.
2653 2018-05-03 Nathan Sidwell <nathan@acm.org>
2655 * doc/extend.texi (Deprecated Features): Remove
2656 -ffriend-injection.
2657 (Backwards Compatibility): Likewise.
2658 * doc/invoke.texi (C++ Language Options): Likewise.
2659 (C++ Dialect Options): Likewise.
2661 2018-05-03 Jakub Jelinek <jakub@redhat.com>
2663 PR target/85530
2664 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
2665 _mm512_mask_mullox_epi64): New intrinsics.
2667 2018-05-03 Tom de Vries <tom@codesourcery.com>
2669 PR testsuite/85106
2670 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2671 dump files): Add offload-tree.
2673 2018-05-03 Richard Biener <rguenther@suse.de>
2675 PR tree-optimization/85615
2676 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
2677 to loops not nested in BBs loop father to avoid creating multi-entry
2678 loops.
2680 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2682 PR tree-optimization/70291
2683 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
2684 arguments. Change return type to tree. Emit libcall as a new
2685 statement rather than replacing existing one when inplace_p is true.
2686 (expand_complex_multiplication_components): New function.
2687 (expand_complex_multiplication): Expand floating-point complex
2688 multiplication using the above.
2689 (expand_complex_division): Rename inner_type parameter to type.
2690 Update expand_complex_libcall call-site.
2691 (expand_complex_operations_1): Update expand_complex_multiplication
2692 and expand_complex_division call-sites.
2694 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2696 PR target/85582
2697 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2698 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2699 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
2700 the highest significant bit of the shift count mask is clear. In
2701 check whether and[sq]i3 is needed verify that all significant bits
2702 of the shift count other than the highest are set.
2704 2018-05-02 Tom de Vries <tom@codesourcery.com>
2706 PR libgomp/82428
2707 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
2708 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
2709 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
2710 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
2711 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
2712 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
2713 __builtin_goacc_parlevel_size.
2715 2018-05-02 Richard Biener <rguenther@suse.de>
2717 PR tree-optimization/85597
2718 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
2719 do not use split vect_get_vec_defs call but call vect_get_slp_defs
2720 directly.
2722 2018-05-02 Tom de Vries <tom@codesourcery.com>
2724 PR testsuite/85106
2725 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2726 dump files): Add ltrans-tree.
2728 2018-05-02 Tom de Vries <tom@codesourcery.com>
2730 PR testsuite/85106
2731 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
2732 dump files): Add wpa-ipa.
2734 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
2736 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
2737 powerpc*-*-linux*paired* target.
2738 * config/rs6000/750cl.h: Delete.
2739 * config/rs6000/paired.h: Delete.
2740 * config/rs6000/paired.md: Delete.
2741 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
2742 float support.
2743 * config/rs6000/rs6000-builtin.def: Remove paired float support.
2744 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
2745 comment. Remove paired float support.
2746 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
2747 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
2748 VECTOR_PAIRED.
2749 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
2750 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
2751 declarations.
2752 * config/rs6000/rs6000.c: Remove paired float support.
2753 (paired_expand_vector_init, paired_expand_vector_move,
2754 paired_emit_vector_compare, paired_emit_vector_cond_expr,
2755 (paired_expand_lv_builtin, paired_expand_stv_builtin,
2756 paired_expand_builtin, paired_expand_predicate_builtin,
2757 paired_init_builtins): Delete.
2758 * config/rs6000/rs6000.h: Remove paired float support.
2759 * config/rs6000/rs6000.md: Remove paired float support.
2760 (move_from_CR_ov_bit): Delete.
2761 * config/rs6000/rs6000.opt (mpaired): Delete.
2762 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
2763 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
2765 2018-05-02 Richard Biener <rguenther@suse.de>
2767 PR middle-end/85567
2768 * gimplify.c (gimplify_save_expr): When in SSA form allow
2769 SAVE_EXPRs to compute to SSA vars.
2771 2018-05-02 Jakub Jelinek <jakub@redhat.com>
2773 PR target/85582
2774 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
2775 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
2776 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
2777 clobber operands[2], instead use a new pseudo. Formatting fixes.
2779 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
2781 PR tree-optimization/85586
2782 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
2783 exit early for statements in the same group if the accesses are
2784 not strided.
2786 2018-05-02 Tom de Vries <tom@codesourcery.com>
2788 PR lto/85451
2789 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
2790 error message.
2792 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
2794 PR tree-optimization/85143
2795 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
2797 2018-05-01 Tom de Vries <tom@codesourcery.com>
2799 PR lto/85451
2800 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
2801 not found" error message.
2803 2018-05-01 Tom de Vries <tom@codesourcery.com>
2805 PR other/83786
2806 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
2807 * vec.c (test_ordered_remove_if): New function.
2808 (vec_c_tests): Call test_ordered_remove_if.
2809 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
2810 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
2811 * tree-vect-patterns.c (vect_pattern_recog_1): Use
2812 VEC_ORDERED_REMOVE_IF.
2814 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2816 PR tree-optimization/82665
2817 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
2818 pointer subtraction where arguments come from a memchr call.
2820 2018-05-01 Jakub Jelinek <jakub@redhat.com>
2822 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
2823 --push-state --as-needed and --pop-state instead of --as-needed and
2824 --no-as-needed if ld supports it.
2825 * configure: Regenerated.
2827 PR web/85578
2828 * doc/install.texi2html: Replace _002d with - and _002a with * in
2829 generated html files using sed.
2831 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2833 PR c++/85523
2834 * gcc-rich-location.c (blank_line_before_p): New function.
2835 (use_new_line): New function.
2836 (gcc_rich_location::add_fixit_insert_formatted): New function.
2837 * gcc-rich-location.h
2838 (gcc_rich_location::add_fixit_insert_formatted): New function.
2840 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2842 * selftest.c (assert_streq): Rename "expected" and "actual" to
2843 "val1" and "val2". Extend NULL-handling to cover both inputs
2844 symmetrically, while still requiring both to be non-NULL for a pass.
2845 * selftest.h (assert_streq): Rename "expected" and "actual" to
2846 "val1" and "val2".
2847 (ASSERT_EQ): Likewise.
2848 (ASSERT_EQ_AT): Likewise.
2849 (ASSERT_KNOWN_EQ): Likewise.
2850 (ASSERT_KNOWN_EQ_AT): Likewise.
2851 (ASSERT_NE): Likewise.
2852 (ASSERT_MAYBE_NE): Likewise.
2853 (ASSERT_MAYBE_NE_AT): Likewise.
2854 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
2855 the assertion to pass.
2856 (ASSERT_STREQ_AT): Likewise.
2858 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2860 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
2861 interaction with -pie.
2863 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2865 * selftest.h: Fix alphabetization of per-source-file selftest
2866 declarations.
2868 2018-04-30 Jason Merrill <jason@redhat.com>
2870 PR c++/61982 - dead stores to destroyed objects.
2871 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
2872 of clobber.
2874 2018-04-30 Jason Merrill <jason@redhat.com>
2876 * tree.c (build_clobber): New.
2877 * tree.h: Declare it.
2878 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
2880 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2882 * diagnostic-show-locus.c (layout::layout): Update for
2883 location_get_source_line returning a char_span.
2884 (struct char_span): Move to input.h.
2885 (struct correction): Update for fields in char_span becoming
2886 private.
2887 (struct source_line): Update for location_get_source_line
2888 returning a char_span.
2889 (layout::print_line): Likewise.
2890 * edit-context.c (edited_file::print_content): Likewise.
2891 (edited_file::print_diff_hunk): Likewise.
2892 (edited_file::print_run_of_changed_lines): Likewise.
2893 (edited_file::get_num_lines): Likewise.
2894 (edited_line::edited_line): Likewise.
2895 * final.c (asm_show_source): Likewise.
2896 * input.c (location_get_source_line): Convert return type
2897 from const char * to char_span, losing the final "line_len"
2898 param.
2899 (dump_location_info): Update for the above.
2900 (get_substring_ranges_for_loc): Likewise. Use a char_span
2901 when handling the literal within the line.
2902 (test_reading_source_line): Update for location_get_source_line
2903 returning a char_span.
2904 * input.h (class char_span): Move here from
2905 diagnostic-show-locus.c, converting from a struct to a class.
2906 Make data members private.
2907 (char_span::operator bool): New.
2908 (char_span::length): New.
2909 (char_span::get_buffer): New.
2910 (char_span::operator[]): New.
2911 (char_span::subspan): Make const.
2912 (char_span::xstrdup): New.
2913 (location_get_source_line): Convert return type from const char *
2914 to char_span, losing the final "line_size" param.
2916 2018-04-30 Jan Hubicka <jh@suse.cz>
2918 * lto-wrapper.c (ltrans_priorities): New static var.
2919 (cmp_priority): New.
2920 (run_gcc): Read priorities and if doing parallel build order
2921 the Makefile by them.
2923 2018-04-30 David Malcolm <dmalcolm@redhat.com>
2925 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
2927 2018-04-30 Richard Biener <rguenther@suse.de>
2929 * tree-cfg.c (verify_address): Remove base argument, add
2930 flag whether to check TREE_ADDRESSABLE and do that.
2931 (verify_expr): Remove.
2932 (verify_types_in_gimple_reference): Add pieces from verify_expr.
2933 (verify_gimple_assign_single): Likewise.
2934 (verify_gimple_switch): Likewise.
2935 (verify_expr_location_1): Dereference tp once. Add (disabled)
2936 piece from verify_expr.
2937 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
2939 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2941 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
2943 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2945 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
2946 (small_data_pattern): Likewise.
2947 (arc_rewrite_small_data): Likewise.
2948 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
2949 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
2950 (get_symbol_alignment): New function.
2951 (legitimate_small_data_address_p): Likewise.
2952 (legitimate_scaled_address): Update, call
2953 legitimate_small_data_address_p.
2954 (output_sdata): New static variable.
2955 (arc_print_operand): Update how we handle small data operands.
2956 (arc_print_operand_address): Likewise.
2957 (arc_legitimate_address_p): Update, use
2958 legitimate_small_data_address_p.
2959 (arc_rewrite_small_data_p): Remove.
2960 (arc_rewrite_small_data_1): Likewise.
2961 (arc_rewrite_small_data): Likewise.
2962 (small_data_pattern): Likewise.
2963 (compact_sda_memory_operand): Update to use
2964 legitimate_small_data_address_p and get_symbol_alignment.
2965 (prepare_move_operands): Don't rewite sdata pattern.
2966 (prepare_extend_operands): Remove.
2967 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
2968 pattern.
2969 (zero_extendqisi2): Likewise.
2970 (zero_extendhisi2): Likewise.
2971 (extendqihi2): Likewise.
2972 (extendqisi2): Likewise.
2973 (extendhisi2): Likewise.
2974 (addsi3): Likewise.
2975 (subsi3): Likewise.
2976 (andsi3): Likewise.
2977 * config/arc/constraints.md (Usd): Change it to memory constraint.
2979 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
2981 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
2982 as source of std instructions.
2983 * config/arc/arc.md (movsi_insn): Update pattern predicate to
2984 allow 6-bit constants as source for store instructions.
2985 (movdi_insn): Update instruction pattern to allow 6-bit constants
2986 as source for store instructions.
2988 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
2990 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
2992 2018-04-30 Nathan Sidwell <nathan@acm.org>
2993 Sandra Loosemore <sandra@codesourcery.com>
2995 * dumpfile.c (dump_open): Allow '-' for stdout.
2996 * doc/invoke.texi (Developer Options): Document dump filename
2997 determination early. Document stdin/stdout selection.
2999 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
3001 Microblaze Target: PIC data text relative
3003 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
3004 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
3005 Add declaration.
3006 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
3007 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
3008 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
3009 New addressing mode for data-text relative position indepenedent code.
3010 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
3011 'ADDRESS_SYMBOLIC_TXT_REL'.
3012 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
3013 (microblaze_legitimate_pic_operand): Exclude function calls from
3014 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
3015 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
3016 addresses cases.
3017 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3018 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3019 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
3020 for 'address + offset'.
3021 (microblaze_expand_prologue): Add new function prologue call for
3022 'r20' assignation.
3023 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
3024 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
3025 table in case of TARGET_PIC_DATA_TEXT_REL.
3026 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
3027 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
3028 Add new macros 'UNSPEC_TEXT',
3029 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
3030 + exclude function calls from 'UNSPEC_PLT' in case of data text
3031 relative mode.
3032 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3033 new target hook for generating address diff vector tables in case of
3034 flag_pic.
3035 * doc/tm.texi : Regenerate.
3036 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
3037 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
3038 of addr diff vector generation.
3039 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3040 target hook definition.
3041 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
3042 Add default function for generate_pic_addr_diff_vec -> flag_pic.
3043 * doc/invoke.texi (Add new pic option): Add new microblaze pic
3044 option for data text relative.
3046 2018-04-30 Richard Biener <rguenther@suse.de>
3048 * tree-chrec.h (evolution_function_is_constant_p): Remove
3049 redundant check.
3050 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
3052 2018-04-30 Richard Biener <rguenther@suse.de>
3054 PR bootstrap/85571
3055 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
3057 2018-04-30 Richard Biener <rguenther@suse.de>
3059 PR tree-optimization/28364
3060 PR tree-optimization/85275
3061 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
3062 copying first exit test.
3064 2018-04-28 Mark Wielaard <mark@klomp.org>
3066 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
3067 dwarf_version >= 5.
3068 (dwarf_AT): Handle DW_AT_addr_base.
3069 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
3071 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
3073 PR target/84431
3074 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
3075 (*ashl<dwi>3_doubleword_mask_1): Ditto.
3076 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
3077 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
3079 2018-04-28 Richard Biener <rguenther@suse.de>
3081 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
3082 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
3083 to reflect use. Only add interesting stmts.
3085 2018-04-27 Martin Jambor <mjambor@suse.cz>
3087 PR ipa/85549
3088 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
3089 the jump function allows for passing through aggregate values.
3091 2018-04-27 David Malcolm <dmalcolm@redhat.com>
3093 * input.h (in_system_header_at): Convert from macro to inline
3094 function.
3095 (from_macro_expansion_at): Likewise.
3096 (from_macro_definition_at): Likewise.
3098 2018-04-27 Jeff Law <law@redhat.com>
3100 * config.gcc: Mark tile* targets as deprecated/obsolete.
3102 2018-04-27 Richard Biener <rguenther@suse.de>
3104 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
3105 fix for ILP32.
3107 2018-04-27 Richard Biener <rguenther@suse.de>
3109 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
3111 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
3113 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
3114 with Yd constraint. Set "preferred_for_speed" attribute from
3115 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
3116 with Yd constraint.
3117 (*movdi_internal): Ditto.
3118 (movti_interunit splitters): Remove
3119 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
3120 (movdi_interunit splitters): Ditto.
3121 * config/i386/constraints.md (Ye): Remove.
3122 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
3124 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3126 PR target/85512
3127 * config/aarch64/constraints.md (Usg): Limit to 31.
3128 (Usj): Limit to 63.
3130 2018-04-27 Jakub Jelinek <jakub@redhat.com>
3132 PR tree-optimization/85529
3133 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
3134 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
3135 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
3136 zero extension or masking of the MSB bit.
3137 (optimize_range_tests): Add FIRST_BB argument, pass it through
3138 to optimize_range_tests_var_bound.
3139 (maybe_optimize_range_tests, reassociate_bb): Adjust
3140 optimize_range_tests callers.
3142 2018-04-26 Richard Biener <rguenther@suse.de>
3143 Jakub Jelinek <jakub@redhat.com>
3145 * cgraph.h (symbol_table): Just declare debug method here.
3146 * symtab.c (symbol_table::debug): Define.
3148 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
3150 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
3152 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
3154 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
3155 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
3156 (*movdi_internal): Substitute Yi and Yj constraint with x
3157 and Ym and Yn constraint with y constraint. Update "isa"
3158 attribute and set "preferred_for_speed" attribute from
3159 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
3160 (*movsi_internal): Ditto.
3161 (*movdf_internal): Ditto.
3162 (*movsf_internal): Ditto.
3163 (*zero_extendsidi2): Ditto.
3164 * config/i386/sse.md (vec_set<mode>_0): Ditto.
3165 (sse2_loadld): Ditto.
3166 (*vec_extract<ssevecmodelower>_0): Ditto.
3167 (*vec_extractv4si_0_zext_sse4): Ditto.
3168 (vec_concatv2di): Ditto.
3169 (*vec_dup<mode>): Ditto.
3170 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
3171 * config/i386/constraints.md (Yi): Remove.
3172 (Yj): Remove.
3173 (Ym): Remove.
3174 (Yn): Remove.
3176 2018-04-26 Nathan Sidwell <nathan@acm.org>
3178 * dumpfile.c (dump_open): New.
3179 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
3180 (dump_finish): Detect stdio/stderr by value not name.
3182 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
3184 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
3186 2018-04-26 Tom de Vries <tom@codesourcery.com>
3188 PR target/84952
3189 * config/nvptx/nvptx.c (verify_neutering_jumps)
3190 (verify_neutering_labels): New function
3191 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
3193 2018-04-26 Tom de Vries <tom@codesourcery.com>
3195 PR target/84025
3196 * config/nvptx/nvptx.c (needs_neutering_p): New function.
3197 (nvptx_single): Use needs_neutering_p to skip over insns that do not
3198 need neutering.
3200 2018-04-26 Richard Biener <rguenther@suse.de>
3201 Tom de Vries <tom@codesourcery.com>
3203 PR lto/85422
3204 * lto-streamer-out.c (output_function): Fixup loops if required to match
3205 discovery done in the reader.
3207 2018-04-26 Richard Biener <rguenther@suse.de>
3209 PR tree-optimization/85116
3210 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
3211 have a loop exit from the single latch predecessor. Remove
3212 case of header with just condition.
3213 (ch_base::copy_headers): Exclude infinite loops from any
3214 processing.
3215 (pass_ch::execute): Record exits.
3217 2018-04-26 Richard Biener <rguenther@suse.de>
3219 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
3220 prologue cost vector and pass it to vect_get_load_cost.
3221 (vect_get_peeling_costs_all_drs): Likewise.
3222 (vect_peeling_hash_get_lowest_cost): Likewise.
3223 (vect_enhance_data_refs_alignment): Likewise.
3225 2018-04-26 Richard Biener <rguenther@suse.de>
3227 PR middle-end/85450
3228 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
3229 checking of integer<->pointer conversions.
3230 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
3231 sign-/zero-extending pointer types.
3232 (expand_omp_for_static_chunk): Likewise.
3234 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
3235 Jean Lee <xiaoyur347@gmail.com>
3237 * config/mips/mips.c (mips_asan_shadow_offset): New function.
3238 (TARGET_ASAN_SHADOW_OFFSET): Define.
3239 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
3240 true for -fsanitize=address.
3242 2018-04-25 Mark Wielaard <mark@klomp.org>
3244 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
3245 shorter ones.
3247 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3249 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
3250 than "alu", remove explicit "memory" and "imm_disp" attributes.
3251 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
3253 PR middle-end/85414
3254 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
3255 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
3256 gen_lowpart_no_emit.
3258 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
3260 PR target/85473
3261 * config/i386/i386.c (ix86_expand_builtin): Change memory
3262 operand to XI, extend p0 to Pmode.
3263 * config/i386/i386.md: Change unspec volatile and operand
3264 1 mode to XI, change operand 0 mode to P.
3266 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3268 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
3269 GET_MODE_MASK before any checking.
3270 (nds32_can_use_bset_p): Likewise.
3271 (nds32_can_use_btgl_p): Likewise.
3273 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3275 * config/nds32/nds32-doubleword.md: New define_split pattern for
3276 illegal register number.
3278 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3280 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
3282 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3284 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
3286 2018-04-25 Richard Biener <rguenther@suse.de>
3288 * lto-streamer.h (LTO_major_version): Bump to 8.
3290 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3292 * BASE-VER: Set to 9.0.0.
3294 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
3296 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
3297 in __abskf2 and __powikf2.
3299 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3301 PR target/85512
3302 * config/aarch64/constraints.md (Usg, Usj): New constraints.
3303 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
3304 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
3305 Use the above on operand 2. Reindent.
3306 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
3308 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3310 PR target/85485
3311 * common/config/i386/i386-common.c (ix86_handle_option): Don't
3312 handle OPT_mcet.
3313 * config/i386/i386.opt (mcet): Removed.
3314 * doc/install.texi: Remove -mcet documentation.
3315 * doc/invoke.texi: Likewise.
3317 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3319 PR target/85485
3320 * doc/install.texi: Remove -mcet from bootstrap-cet.
3322 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3324 PR target/85511
3325 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
3326 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
3327 if TARGET_64BIT.
3329 PR target/85503
3330 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
3331 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
3332 containing a CONST_VECTOR.
3334 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
3336 * doc/install.texi: Update newlib dependency for nvptx.
3338 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3340 PR target/85508
3341 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
3342 instead of INTVAL when shifting x left.
3344 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
3346 PR tree-optimization/85478
3347 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
3348 vect_grouped_store_supported for single element vectors.
3350 2018-04-24 Richard Biener <rguenther@suse.de>
3352 PR target/85491
3353 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
3354 load cost increase to the case of non-constant step.
3356 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3358 PR target/84828
3359 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
3360 destination if any_malformed_asm.
3362 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
3364 PR middle-end/85496
3365 * expr.c (store_field): In the bitfield case, if the value comes from
3366 a function call and is returned in registers by means of a PARALLEL,
3367 do not change the mode of the temporary unless BLKmode and VOIDmode.
3369 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
3371 PR rtl-optimization/85423
3372 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
3373 dependencies to debug insns when the previous insn is non-debug.
3375 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3377 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
3378 enums into a single definition.
3379 (fls): Fix predicates and printing.
3380 (seti): Likewise.
3382 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3384 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
3385 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
3386 and short u6 immediate.
3387 (check_if_valid_sleep_operand): Remove.
3388 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
3390 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3392 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
3393 flag_always_save_lp condition.
3394 * config/nds32/nds32.opt (malways-save-lp): New option.
3396 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3398 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
3399 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
3400 * config/nds32/nds32.h
3401 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
3402 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
3404 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3406 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
3407 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
3409 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3410 Chung-Ju Wu <jasonwucj@gmail.com>
3412 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
3413 Declare.
3414 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
3415 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
3417 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3419 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
3421 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3423 * config/nds32/nds32-protos.h (nds32_data_alignment,
3424 nds32_local_alignment): Declare.
3425 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
3426 nds32_local_alignment): New functions.
3427 (TARGET_CONSTANT_ALIGNMENT): Define.
3428 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
3430 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3432 * config/nds32/nds32.c
3433 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
3434 (TARGET_MODES_TIEABLE_P): Likewise.
3436 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3438 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
3439 level Ofast and Og.
3441 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
3442 Chung-Ju Wu <jasonwucj@gmail.com>
3444 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3445 for unaligned access.
3446 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3447 unaligned access.
3448 * config/nds32/nds32-intrinsic.md: Likewise.
3449 * config/nds32/nds32_intrinsic.h: Likewise.
3450 * config/nds32/nds32.h (nds32_builtins): Likewise.
3451 * config/nds32/nds32.opt (munaligned-access): New option.
3452 * config/nds32/nds32.c (nds32_asm_file_start): Display
3453 flag_unaligned_access status.
3455 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3457 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
3458 -mno-relax is present.
3459 * config/riscv/linux.h (LINK_SPEC): Ditto.
3461 2018-04-20 Martin Sebor <msebor@redhat.com>
3463 PR c/85365
3464 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
3465 for null pointers.
3466 (gimple_fold_builtin_stxcpy_chk): Same.
3467 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
3469 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
3471 PR target/85456
3472 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
3473 __powikf2 when long double is IEEE 128-bit.
3475 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3477 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
3478 step to make sure stack always aligned.
3480 2018-04-20 Carl Love <cel@us.ibm.com>
3482 PR target/83402
3483 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
3484 size check for arg0.
3486 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
3487 Tom de Vries <tom@codesourcery.com>
3489 PR target/85445
3490 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
3491 Emit insns for calls too.
3492 (nvptx_find_par): Always look for worker-level predecessor insn.
3493 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
3494 calls.
3495 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
3496 (nvptx_process_pars): Propagate frames for calls.
3498 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
3500 PR target/85469
3501 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
3502 Removed.
3503 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
3504 (ix86_handle_option): Don't handle OPT_mibt.
3505 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
3506 __SHSTK__.
3507 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
3508 has_ibt and ibt.
3509 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
3510 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
3511 (ix86_target_macros): Define __CET__ with flag_cf_protection
3512 for -fcf-protection.
3513 * config/i386/i386.c (isa2_opts): Remove -mibt.
3514 * config/i386/i386.h (TARGET_IBT): Removed.
3515 (TARGET_IBT_P): Likewise.
3516 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
3517 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
3518 * config/i386/i386.opt (mcet): Update help message.
3519 (mshstk): Likewise.
3520 (mibt): Removed.
3521 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
3522 -mcet as an alias for -mshstk.
3524 2018-04-20 Richard Biener <rguenther@suse.de>
3526 PR middle-end/85475
3527 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
3528 complexity by forcing a single use of the multiply operand.
3530 2018-04-20 Martin Jambor <mjambor@suse.cz>
3532 ipa/85449
3533 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
3534 recursion dependency to only apply to non-clones.
3536 2018-04-20 Martin Jambor <mjambor@suse.cz>
3538 ipa/85447
3539 * ipa-cp.c (create_specialized_node): Check that clones of
3540 self-recursive edges exist during IPA-CP.
3542 2018-04-19 Toon Moene <toon@moene.org>
3544 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
3545 by -O3.
3547 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3549 PR tree-optimization/85467
3550 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
3551 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
3552 VECTOR_CST element to type.
3554 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3556 PR target/85397
3557 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
3558 * config/i386/i386.md (builtin_setjmp_setup): Removed.
3559 (builtin_longjmp): Likewise.
3560 (save_stack_nonlocal): New pattern.
3561 (restore_stack_nonlocal): Likewise.
3563 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3565 PR target/85404
3566 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3567 Replace ASM_OUTPUT_LABEL with fprintf.
3569 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
3571 PR target/85417
3572 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
3573 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
3574 * config/i386/i386-c.c (ix86_target_macros_internal): Also
3575 define __IBT__ and __SHSTK__ for -fcf-protection.
3576 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
3577 TARGET_IBT.
3578 (ix86_trampoline_init): Likewise.
3579 (x86_output_mi_thunk): Likewise.
3580 (ix86_notrack_prefixed_insn_p): Likewise.
3581 (ix86_option_override_internal): Don't disallow -fcf-protection.
3582 * config/i386/i386.md (rdssp<mode>): Also enable for
3583 -fcf-protection.
3584 (incssp<mode>): Likewise.
3585 (nop_endbr): Likewise.
3586 * config/i386/i386.opt (mcet): Change help message to built-in
3587 functions only.
3588 (mibt): Likewise.
3589 (mshstk): Likewise.
3590 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
3591 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
3592 enable CET built-in functions.
3594 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
3596 * common/config/i386/i386-common.c
3597 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
3598 OPTION_MASK_ISA_MOVDIRI_UNSET,
3599 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
3600 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
3601 * config.gcc (movdirintrin.h): New header.
3602 * config/i386/cpuid.h (bit_MOVDIRI,
3603 bit_MOVDIR64B): New bits.
3604 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
3605 and -mmvodir64b.
3606 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
3607 (VOID, PVOID, PCVOID)): New function types.
3608 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
3609 __builtin_ia32_directstoreu_u64,
3610 __builtin_ia32_movdir64b): New builtins.
3611 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
3612 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
3613 and -mmovdiri.
3614 (ix86_valid_target_attribute_inner_p): Ditto.
3615 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
3616 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
3617 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
3618 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
3619 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
3620 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
3621 (movdiri<mode>, movdir64b_<mode>): New.
3622 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
3623 * config/i386/immintrin.h: Include movdirintrin.h.
3624 * config/i386/movdirintrin.h: New file.
3625 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
3627 2018-04-19 Richard Biener <rguenther@suse.de>
3629 PR middle-end/85455
3630 * cfg.c (clear_bb_flags): When loop state says we have
3631 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
3633 2018-04-19 Richard Biener <rguenther@suse.de>
3635 PR tree-optimization/84737
3636 * tree-vect-data-refs.c (vect_copy_ref_info): New function
3637 copying restrict info.
3638 (vect_setup_realignment): Use it.
3639 * tree-vectorizer.h (vect_copy_ref_info): Declare.
3640 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
3641 the first DR to all generated stores.
3642 (vectorizable_load): Likewise for loads.
3644 2018-04-19 Jakub Jelinek <jakub@redhat.com>
3646 PR tree-optimization/85446
3647 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
3648 the integral and pointer types to have the same precision.
3650 * doc/install.texi: Document --disable-cet being the default and
3651 --enable-cet=auto.
3653 2018-04-18 Martin Liska <mliska@suse.cz>
3655 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
3656 style.
3658 2018-04-18 Martin Liska <mliska@suse.cz>
3660 Revert
3661 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3663 PR ipa/83983
3664 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3665 arguments if they are comparable.
3667 2018-04-18 Martin Liska <mliska@suse.cz>
3669 Revert
3670 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3672 PR lto/84805
3673 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3674 incomplete types.
3676 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
3678 PR target/85388
3679 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
3680 ENDBR after calling __morestack.
3682 2018-04-18 David Malcolm <dmalcolm@redhat.com>
3684 PR jit/85384
3685 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
3686 by using gcc_base_ver to generate a gcc_driver_version, and use
3687 it when generating GCC_DRIVER_NAME.
3688 * configure: Regenerate.
3690 2018-04-18 Jakub Jelinek <jakub@redhat.com>
3692 PR target/81084
3693 * config.gcc: Obsolete powerpc*-*-*spe*.
3695 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3697 PR debug/84637
3698 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
3699 (stabstr_D): Change type of unum from unsigned int to
3700 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
3701 type.
3703 2018-04-17 Jim Wilson <jimw@sifive.com>
3705 PR 84856
3706 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
3707 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
3708 Set arg_pointer_offset after using pretend_args_size.
3710 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3712 PR rtl-optimization/85431
3713 * dse.c (record_store): Ignore zero width stores.
3715 PR sanitizer/85230
3716 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
3717 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
3718 __builtin_stack_restore rather than after it.
3719 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
3720 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
3721 argument instead of virtual_dynamic_stack_rtx.
3723 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
3725 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
3726 New prototype.
3727 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
3728 Add note to error message to explain internal mapping of overloaded
3729 built-in function name to non-overloaded built-in function name.
3730 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
3731 function.
3733 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3735 PR target/85424
3736 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
3737 where the inputs overlap with the output.
3739 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3741 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
3742 (=v, v) alternative and explicit "memory" attribute.
3743 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
3744 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3745 attributes.
3746 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
3747 "sselog1" type instead of "sselog".
3748 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
3749 "sselog". Remove explicit "memory" attribute.
3750 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
3751 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3752 attributes.
3753 (vec_extract_hi_v32hi): Merge all alternatives into one, use
3754 "sselog1" type instead of "sselog". Remove explicit "memory"
3755 attribute.
3756 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
3757 use "sselog1" type instead of "sselog". Remove explicit "memory"
3758 attribute.
3759 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
3760 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
3761 attributes.
3762 (vec_extract_hi_v64qi): Merge all alternatives into one, use
3763 "sselog1" type instead of "sselog". Remove explicit "memory"
3764 attribute.
3765 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
3766 use "sselog1" type instead of "sselog". Remove explicit "memory"
3767 attribute.
3769 PR target/85430
3770 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
3772 PR middle-end/85414
3773 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
3774 on a SUBREG.
3776 2018-04-17 Martin Jambor <mjambor@suse.cz>
3778 PR ipa/85421
3779 * ipa-cp.c (create_specialized_node): Call
3780 expand_all_artificial_thunks if necessary.
3782 2018-04-17 Martin Liska <mliska@suse.cz>
3784 PR lto/85405
3785 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
3786 in message, remote space in between '_G' and '('.
3788 2018-04-17 Jakub Jelinek <jakub@redhat.com>
3790 PR target/85281
3791 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
3792 avx512f_vmcmp<mode>3<round_saeonly_name>,
3793 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
3794 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
3795 avx512f_rndscale<mode><round_saeonly_name>,
3796 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
3797 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
3798 Use %<iptr>2 instead of %2 for -masm=intel.
3799 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
3800 avx512f_vcvttss2usi<round_saeonly_name>,
3801 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
3802 -masm=intel.
3803 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
3804 avx512f_vcvttsd2usi<round_saeonly_name>,
3805 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
3806 Use %q1 instead of %1 for -masm=intel.
3807 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
3808 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
3809 of %3 for -masm=intel.
3810 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
3811 -masm=intel.
3812 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
3813 -masm=intel.
3814 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
3815 -masm=intel.
3816 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
3817 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
3818 %g1.
3819 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
3820 -masm=intel.
3821 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
3822 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
3823 %g1 and one with %0 and %1.
3824 (avx512er_vmrcp28<mode><round_saeonly_name>,
3825 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
3826 %1 for -masm=intel.
3827 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
3828 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
3829 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
3830 of %0 and %{%4%} for -masm=intel.
3831 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
3832 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
3833 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
3834 order of %0 and %{%5%}%{z%} for -masm=intel.
3836 2018-04-17 Jan Hubicka <jh@suse.cz>
3838 PR lto/85405
3839 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
3841 2018-04-17 Martin Liska <mliska@suse.cz>
3843 PR ipa/85329
3844 * multiple_target.c (create_dispatcher_calls): Set apostrophes
3845 for target_clone error message. Make default implementation
3846 clone to be a local declaration.
3847 (separate_attrs): Add new argument and check for an empty
3848 string.
3849 (expand_target_clones): Handle it.
3850 (ipa_target_clone): Make redirection just for target_clones
3851 functions.
3853 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
3854 Tom de Vries <tom@codesourcery.com>
3856 PR middle-end/84955
3857 * omp-expand.c (expand_oacc_for): Add dummy false branch for
3858 tiled basic blocks without omp continue statements.
3860 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
3862 PR target/83660
3863 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
3864 vec_extract expression as having side effects to make sure it gets
3865 a cleanup point.
3867 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
3869 PR target/85403
3870 * config/i386/i386.c (get_builtin_code_for_version): Check
3871 error_mark_node.
3873 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
3875 PR target/84331
3876 * gcc/config.gcc: Support "skylake".
3877 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
3878 PROCESSOR_SKYLAKE.
3879 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
3880 (processor_target_table): Add "skylake".
3881 (ix86_option_override_internal): Add "skylake".
3882 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
3883 PROCESSOR_CANNONLAKE.
3884 (get_builtin_code_for_version): Fix priority for
3885 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
3886 PROCESSOR_SKYLAKE-AVX512.
3887 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
3888 (processor_type): Add PROCESSOR_SKYLAKE.
3890 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
3891 Jason Merrill <jason@redhat.com>
3893 PR c++/85112
3894 * convert.c (convert_to_integer_1): Use direct recursion for
3895 enumeral types and types with a precision less than the number
3896 of bits in their mode.
3898 2018-04-16 Julia Koval <julia.koval@intel.com>
3900 PR target/84413
3901 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
3902 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
3904 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
3906 PR target/85293
3907 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
3908 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
3909 and -mno-direct-move.
3911 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
3913 PR target/83402
3914 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
3915 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
3916 Ensure negative shifts result in {0}.
3918 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
3920 PR rtl-optimization/79916
3921 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
3922 regs (if any) to define how to gnerate SD moves when LRA is in
3923 progress.
3925 2018-04-13 Jakub Jelinek <jakub@redhat.com>
3927 PR rtl-optimization/85393
3928 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
3929 * except.c (expand_dw2_landing_pad_for_region): Make static.
3930 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
3931 a label and unconditional jump to old_bb, rather than
3932 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
3933 basic block.
3935 PR rtl-optimization/85376
3936 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
3937 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
3938 instead of a specific value.
3940 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3941 Bin Cheng <bin.cheng@arm.com>
3943 PR tree-optimization/82965
3944 PR tree-optimization/83991
3945 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
3946 by_profile_only parameter.
3947 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
3948 information if the loop was predicted to iterate too many times.
3949 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
3951 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
3953 PR lto/71991
3954 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
3955 always inline.
3957 2018-04-13 Martin Liska <mliska@suse.cz>
3958 Jakub Jelinek <jakub@redhat.com>
3960 PR middle-end/81657
3961 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
3962 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
3963 * builtins.c (expand_builtin_memory_copy_args): Use
3964 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
3965 handle dest_addr == pc_rtx.
3967 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
3969 PR target/85291
3970 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
3971 asked to not generate direct moves.
3972 (fix_trunc<mode>si2_stfiwx): Similar.
3973 (fix_trunc<mode>si2_internal): Similar.
3975 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3977 PR debug/83157
3978 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
3979 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
3980 lookup if dest in some wider mode is known to be const0_rtx and
3981 if so, record permanent equivalence for it to be ZERO_EXTEND of
3982 the narrower mode destination.
3984 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
3986 * lto-streamer-out.c (output_function): Revert 259346.
3987 * omp-expand.c (expand_oacc_for): Likewise.
3989 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
3991 PR rtl-optimization/85354
3992 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
3993 * sel-sched.c (sel_global_init): ... here.
3995 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
3997 PR target/85238
3998 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
3999 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
4000 mode for PE-COFF targets.
4001 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
4002 (i386_pe_asm_lto_end): Likewise.
4003 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
4004 (TARGET_ASM_LTO_END): Likewise.
4005 * config/i386/winnt.c (saved_debug_info_level): New static variable.
4006 (i386_pe_asm_lto_start): New function.
4007 (i386_pe_asm_lto_end): Likewise.
4009 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4010 Richard Biener <rguenther@suse.de>
4012 PR middle-end/84955
4013 * lto-streamer-out.c (output_function): Fix CFG loop state before
4014 streaming out.
4015 * omp-expand.c (expand_oacc_for): Handle calls to internal
4016 functions like regular functions.
4018 2018-04-12 Richard Biener <rguenther@suse.de>
4020 PR lto/85371
4021 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
4022 for the early LTO debug to properly generate references to it
4023 during DIE emission. Do not re-use that for the skeleton for
4024 split-dwarf.
4025 (dwarf2out_early_finish): Likewise.
4027 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4029 PR target/85328
4030 * config/i386/sse.md
4031 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
4032 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
4033 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
4034 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
4035 and output is a reg, avoid creating invalid lowpart subreg, but
4036 instead split into a 512-bit move. Don't split if not AVX512VL,
4037 input is xmm16+ reg and output is a mem.
4038 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
4039 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
4040 xmm16+ reg and output is a mem.
4042 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4044 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
4045 also for flag_dwarf2_cfi_asm.
4047 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4049 PR rtl-optimization/85342
4050 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
4051 a bool scalar var inside of the loop instead. Don't try to update
4052 recog_data.operand after failed apply_change_group.
4054 2018-04-12 Tom de Vries <tom@codesourcery.com>
4056 PR target/85296
4057 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
4058 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
4059 array with flexible array member as array without given dimension.
4060 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
4061 argument for undefined param to true.
4063 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
4065 PR target/85321
4066 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4067 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
4068 from PowerPC section.
4069 * config/rs6000/sysv4.opt (mcall-): Improve help text.
4070 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
4071 help text that is too long.
4072 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
4073 help text that is too long.
4074 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
4075 help text that is too long.
4077 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
4079 * config/alpha/alpha.md (stack_probe_internal): Rename
4080 from "probe_stack". Update all callers.
4082 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4084 PR rtl-optimization/84566
4085 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
4086 sched_macro_fuse_insns.
4088 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4090 PR target/84301
4091 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
4092 (compute_block_dependences): ... from here.
4094 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4096 PR tree-optimization/85331
4097 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
4098 from int to HOST_WIDE_INT.
4100 2018-04-11 Martin Jambor <mjambor@suse.cz>
4102 PR ipa/84149
4103 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
4104 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
4105 not the same as the source val.
4106 (cgraph_edge_brings_value_p): New parameter.
4107 (gather_edges_for_value): Pass destination value to
4108 cgraph_edge_brings_value_p.
4109 (perhaps_add_new_callers): Likewise.
4110 (get_info_about_necessary_edges): Likewise and exclude values brought
4111 only by self-recursive edges.
4112 (create_specialized_node): Redirect only clones of self-calling edges.
4113 (+self_recursive_pass_through_p): New function.
4114 (find_more_scalar_values_for_callers_subset): Use it.
4115 (find_aggregate_values_for_callers_subset): Likewise.
4116 (known_aggs_to_agg_replacement_list): Removed.
4117 (decide_whether_version_node): Re-calculate known constants for all
4118 remaining context clones.
4120 2018-04-11 Richard Biener <rguenther@suse.de>
4122 PR lto/85339
4123 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
4124 from early DWARF output.
4125 (dwarf2out_early_finish): Output line info unconditionally into
4126 early DWARF and add reference to it.
4128 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4130 PR target/85281
4131 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
4132 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
4133 other than V2DFmode using iptr mode attribute.
4134 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
4136 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4138 PR rtl-optimization/84659
4139 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
4141 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4143 PR debug/85302
4144 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
4145 SIZEP is NULL.
4146 (output_loc_list): Pass address of a dummy size variable even in the
4147 locview handling loop.
4148 (index_location_lists): Add comment on why skip_loc_list_entry can't
4149 call size_of_locs.
4151 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
4153 PR target/85261
4154 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
4155 into register.
4157 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
4159 PR target/85321
4160 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4161 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
4162 and -mstring-compare-inline-limit.
4164 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4166 PR target/85287
4167 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
4168 for stack clash protection in a register whenever we need it to be in
4169 a register.
4171 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4173 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
4174 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
4176 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4178 PR target/85321
4179 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
4180 the help text.
4181 (mlong-double-): Ditto.
4182 * config/rs6000/sysv4.opt (msdata=): Ditto.
4183 (mtls-size=): Ditto.
4185 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4187 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4188 erroneous entries for
4189 "vector int vec_ldl (int, long int *)", and
4190 "vector unsigned int vec_ldl (int, unsigned long int *)".
4191 Add comments and entries for
4192 "vector bool char vec_ldl (int, bool char *)",
4193 "vector bool short vec_ldl (int, bool short *)",
4194 "vector bool int vec_ldl (int, bool int *)",
4195 "vector bool long long vec_ldl (int, bool long long *)",
4196 "vector pixel vec_ldl (int, pixel *)",
4197 "vector long long vec_ldl (int, long long *)",
4198 "vector unsigned long long vec_ldl (int, unsigned long long *)".
4199 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
4200 type tree bool_long_long_type_node and correct definition of
4201 bool_V2DI_type_node to make reference to this new type tree.
4202 (rs6000_mangle_type): Replace erroneous reference to
4203 bool_long_type_node with bool_long_long_type_node.
4204 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
4205 comments to emphasize sign distinctions for char and int types and
4206 replace RS6000_BTI_bool_long constant with
4207 RS6000_BTI_bool_long_long constant. Also add comment to restrict
4208 use of RS6000_BTI_pixel.
4209 (bool_long_type_node): Remove this macro definition.
4210 (bool_long_long_type_node): New macro definition
4212 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4214 PR rtl-optimization/85300
4215 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
4216 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
4217 simplify_unary_operation fails.
4219 2018-04-10 Martin Liska <mliska@suse.cz>
4221 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
4222 cgraph_edge and ipa_ref.
4224 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4226 PR target/85177
4227 PR target/85255
4228 * config/i386/sse.md
4229 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
4230 computation of the VEC_MERGE selector from mask.
4231 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
4232 Fix decoding of the VEC_MERGE selector into mask.
4234 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4236 PR tree-optimization/85286
4237 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
4239 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4241 * final.c (final_1): Set insn_last_address as well as
4242 insn_current_address.
4244 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4246 PR target/85173
4247 * explow.c (emit_stack_probe): Call validize_mem on memory location
4248 before passing it to gen_probe_stack. Create address operand and
4249 legitimize it for the probe_stack_address case.
4251 2018-04-09 Jan Hubicka <jh@suse.cz>
4253 PR lto/85078
4254 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
4255 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
4256 * tree.c (free_lang_data_in_type): Fix handling of binfos;
4257 walk basetypes.
4258 (free_lang_data): Rebuild type inheritance graph.
4260 2018-04-09 Martin Sebor <msebor@redhat.com>
4262 * invoke.texi (-finline-small-functions): Mention other optimization
4263 options.
4264 (-findirect-inlining, -fpartial-inlining): Same.
4265 (-finline-functions-called-once): Same.
4266 (-freorder-blocks-and-partition): Same.
4268 2018-04-09 Jan Hubicka <jh@suse.cz>
4270 PR rtl/84058
4271 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
4272 jumps; choose last target that matches the criteria (i.e.
4273 no partition changes for non-crossing jumps).
4274 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
4275 support for redirecting crossing jumps to non-crossing.
4277 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
4279 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
4280 also for naked functions.
4282 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
4284 * config/arc/arc.md (add_shift): New pattern.
4285 (add_shift2): Likewise.
4286 (sub_shift): Likewise.
4287 (sub_shift_cmp0_noout): Likewise.
4288 (compare_si_ashiftsi): Likewise.
4289 (xbfu_cmp0_noout): New combine pattern.
4290 (xbfu_cmp0"): Likewise.
4291 (movsi_set_cc_insn): Place the predicable variant first.
4292 (commutative_binary_cmp0_noout): Remove clobber.
4293 (commutative_binary_cmp0): New pattern.
4294 (noncommutative_binary_cmp0): Likewise.
4295 (noncommutative_binary_cmp0_noout): Likewise.
4296 (noncommutative_binary_comparison_result_used): Removed.
4297 (rsub_cmp0): New pattern.
4298 (rsub_cmp0_noout): Likewise.
4299 (extzvsi): Changed, keep only meaningful variants.
4300 (SQH, SEZ): New iterators.
4301 (SQH_postfix): New mode attribute.
4302 (SEZ_prefix): New code attribute.
4303 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
4304 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
4305 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
4306 of numerical value.
4307 (noncommutative_operator): Check the availability of barrel
4308 shifter option.
4310 2018-04-09 Richard Biener <rguenther@suse.de>
4312 PR tree-optimization/85284
4313 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
4314 Only use the niter constraining form of simple_iv when the exit
4315 is always executed.
4317 2018-04-09 Tom de Vries <tom@codesourcery.com>
4319 PR target/84041
4320 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4321 (define_expand "*memory_barrier"): New define_expand.
4322 (define_insn "memory_barrier"): New insn.
4324 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4326 PR rtl-optimization/80463
4327 PR rtl-optimization/83972
4328 PR rtl-optimization/83480
4330 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
4331 correct producer for the insn.
4332 (tidy_control_flow): Fixup seqnos in case of debug insns.
4334 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4336 PR rtl-optimization/83913
4338 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
4339 different sched-times when merging exprs.
4341 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4343 PR rtl-optimization/83962
4345 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
4346 tidy_fallthru_edge and tidy_control_flow.
4348 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4350 PR rtl-optimization/83530
4352 * sel-sched.c (force_next_insn): New global variable.
4353 (remove_insn_for_debug): When force_next_insn is true, also leave only
4354 next insn in the ready list.
4355 (sel_sched_region): When the region wasn't scheduled, make another pass
4356 over it with force_next_insn set to 1.
4358 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4360 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
4361 into tm_file.
4362 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4363 for interrupt control.
4364 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4365 functions for interrupt control.
4366 * config/nds32/nds32-intrinsic.md: Likewise.
4367 * config/nds32/nds32_intrinsic.h: Likewise.
4368 * config/nds32/nds32.h (nds32_builtins): Likewise.
4370 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4372 * config/nds32/nds32.c (nds32_init_machine_status,
4373 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
4374 strict_aligned_p field.
4375 (nds32_expand_to_rtl_hook): New function.
4376 (TARGET_EXPAND_TO_RTL_HOOK): Define.
4377 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
4379 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4380 Chung-Ju Wu <jasonwucj@gmail.com>
4382 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
4383 * config/nds32/nds32-n7.md: New file.
4384 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
4385 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
4386 pipeline.
4387 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
4388 * config/nds32/nds32.md (pipeline_model): Add n7.
4389 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
4390 * config/nds32/pipelines.md: Include n7 settings.
4392 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4393 Chung-Ju Wu <jasonwucj@gmail.com>
4395 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
4396 * config/nds32/nds32-e8.md: New file.
4397 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
4398 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
4399 pipeline.
4400 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
4401 * config/nds32/nds32.md (pipeline_model): Add e8.
4402 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
4403 * config/nds32/pipelines.md: Include e8 settings.
4405 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4406 Chung-Ju Wu <jasonwucj@gmail.com>
4408 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
4409 * config/nds32/nds32-n8.md: New file.
4410 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
4411 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
4412 pipeline.
4413 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
4414 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
4415 * config/nds32/nds32.md (pipeline_model): Add n8.
4416 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
4417 * config/nds32/pipelines.md: Include n8 settings.
4419 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4420 Chung-Ju Wu <jasonwucj@gmail.com>
4422 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
4423 * config/nds32/nds32-n9-2r1w.md: New file.
4424 * config/nds32/nds32-n9-3r2w.md: New file.
4425 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
4426 nds32_register_ports): New or modify for cpu n9.
4427 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
4428 pipeline.
4429 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
4430 * config/nds32/nds32-utils.c: New file.
4431 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
4432 TARGET_MUL_SLOW): Define.
4433 * config/nds32/nds32.md (pipeline_model): New attribute.
4434 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
4435 New options that support cpu n9.
4436 * config/nds32/pipelines.md: Include n9 settings.
4437 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
4439 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4441 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
4442 information if necessary.
4443 (output_cond_branch_compare_zero): Likewise.
4444 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
4445 (nds32_target_alignment): Refine for alignment.
4446 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
4447 (FUNCTION_BOUNDARY): Modify.
4448 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
4449 align case.
4450 * config/nds32/nds32.opt (malways-align, malign-functions): New.
4452 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4454 * config/nds32/constants.md (unspec_volatile_element): Add values for
4455 TLB operation and data prefetch.
4456 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4457 functions for TLB operation and data prefetch.
4458 * config/nds32/nds32-intrinsic.md: Likewise.
4459 * config/nds32/nds32_intrinsic.h: Likewise.
4460 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
4461 (nds32_print_operand): Likewise.
4462 * config/nds32/nds32.h (nds32_builtins): Likewise.
4464 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
4465 Andrew Pinski <pinsika@gcc.gnu.org>
4467 PR middle-end/82976
4468 * match.pd: Use constant_boolean_node of correct type instead of
4469 boolean_true_node or boolean_false_node for simplifying
4470 pointer comparisons to zero.
4472 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4474 PR tree-optimization/80021
4475 * tree.c (verify_type_variant): Make error call in verify_variant_match
4476 translatable and remove final full stop.
4478 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4480 * config/nds32/constants.md (unspec_volatile_element): Add
4481 UNSPEC_VOLATILE_EH_RETURN.
4482 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
4483 nds32_output_stack_pop): Support dwarf exception handling process.
4484 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
4485 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
4486 exception handling process.
4487 (nds32_compute_stack_frame): Likewise.
4488 (nds32_return_addr_rtx): Likewise.
4489 (nds32_initial_elimination_offset): Likewise.
4490 (nds32_expand_prologue): Likewise.
4491 (nds32_expand_epilogue): Likewise.
4492 (nds32_dynamic_chain_address): New function.
4493 * config/nds32/nds32.h (machine_function): Add fields for dwarf
4494 exception handling.
4495 (DYNAMIC_CHAIN_ADDRESS): Define.
4496 (EH_RETURN_DATA_REGNO): Define.
4497 (EH_RETURN_STACKADJ_RTX): Define.
4498 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
4499 patterns for dwarf exception handling.
4501 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4503 * config/nds32/nds32.h: Clean up obsolete macros.
4505 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4507 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4508 Add enum values for particular instructions.
4509 * config/nds32/nds32-intrinsic.c: Implementation of expanding
4510 particular intrinsic functions.
4511 * config/nds32/nds32-intrinsic.md: Likewise.
4512 * config/nds32/nds32_intrinsic.h: Likewise.
4513 * config/nds32/nds32.h (nds32_builtins): Likewise.
4514 * config/nds32/nds32.md (type): Add pbsad and pbsada.
4515 (btst, ave): New patterns for particular instructions.
4517 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4519 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4520 Add enum values for atomic load/store and memory sync.
4521 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
4522 and memory sync.
4523 * config/nds32/nds32-intrinsic.md: Likewise.
4524 * config/nds32/nds32_intrinsic.h: Likewise.
4525 * config/nds32/nds32.h (nds32_builtins): Likewise.
4527 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4529 PR tree-optimization/85257
4530 * fold-const.c (native_encode_vector): If not all elts could fit
4531 and off is -1, return 0 rather than offset.
4532 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
4533 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
4534 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
4535 adjust buffer in native_interpret_expr call.
4537 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4539 * config/nds32/constants.md (unspec_volatile_element): Add cache
4540 control enum values.
4541 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
4542 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
4543 * config/nds32/nds32.c (nds32_cctl_names): New.
4544 (nds32_print_operand): Handle cache control register names.
4545 * config/nds32/nds32.h (nds32_builtins): New enum values.
4546 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
4547 macros.
4548 * config/nds32/nds32.md (type): Add mmu.
4549 * config/nds32/pipelines.md (simple_insn): Add mmu.
4551 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4553 * config/nds32/nds32.md (type): Remove call.
4554 * config/nds32/pipelines.md (simple_insn): Likewise.
4556 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4558 * config/nds32/constants.md (unspec_volatile_element): Add
4559 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
4560 UNSPEC_VOLATILE_FMFCFG.
4561 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
4562 description for fmfcfg and fmfcsr.
4563 (bdesc_1arg): Add fmtcsr.
4564 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
4565 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
4566 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
4567 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
4568 unspec_fmfcfg): New patterns.
4569 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
4570 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
4571 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
4572 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
4573 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
4574 __nds32__fmfcfg): Define.
4576 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4578 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
4579 intrinsic register names.
4580 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
4581 intrinsic register enum values and macros.
4583 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4585 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
4586 for load/store addressing form.
4587 (nds32_print_operand_address): Likewise.
4589 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
4591 PR target/85196
4592 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
4593 based on LABEL_REF. Remove useless assertion.
4594 (pic_address_needs_scratch): Fix formatting.
4595 (sparc_legitimize_pic_address): Minor tweaks.
4596 (sparc_delegitimize_address): Adjust assertion accordingly.
4597 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
4598 into symbolic_operand.
4599 (movsi_high_pic_label_ref): Likewise.
4600 (movsi_lo_sum_pic_label_ref): Likewise.
4601 (movdi_pic_label_ref): Likewise.
4602 (movdi_high_pic_label_ref): Likewise.
4603 (movdi_lo_sum_pic_label_ref): Likewise.
4605 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
4607 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
4608 custom LIB_SPEC setup.
4610 2018-04-06 Ruslan Bukin <br@bsdpad.com>
4611 Kito Cheng <kito.cheng@gmail.com>
4613 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
4614 * config/riscv/freebsd.h: New.
4616 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4618 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
4619 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
4620 file.
4622 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4623 Kito Cheng <kito.cheng@gmail.com>
4625 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
4626 nds32_output_call, nds32_symbol_binds_local_p): New functions.
4627 * config/nds32/nds32-protos.h (nds32_output_call,
4628 nds32_output_return): Declare.
4629 * config/nds32/nds32.md: Refine all the call and return patterns.
4631 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4633 PR debug/85252
4634 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
4635 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
4637 PR rtl-optimization/84872
4638 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
4639 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
4640 EDGE_CROSSING edge.
4642 2018-04-06 Tamar Christina <tamar.christina@arm.com>
4644 * expr.c (copy_blkmode_to_reg): Revert 254862.
4645 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
4647 2018-04-06 Richard Biener <rguenther@suse.de>
4649 PR middle-end/85244
4650 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
4651 after seeing a component reference with an adjacent field. Treat
4652 refs to arrays at struct end of external decls similar to
4653 refs to unconstrained commons.
4655 2018-04-06 Jakub Jelinek <jakub@redhat.com>
4657 PR sanitizer/85213
4658 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
4659 look through SAVE_EXPRs with non-side-effects argument. Adjust
4660 recursive calls.
4661 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
4662 save_p here.
4664 2018-04-06 Richard Biener <rguenther@suse.de>
4666 PR middle-end/85180
4667 * alias.c (find_base_term): New wrapper around find_base_term
4668 unwinding CSELIB_VAL_PTR changes.
4669 (find_base_term): Do not restore CSELIB_VAL_PTR during the
4670 recursion.
4672 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4674 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
4675 instructions.
4676 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
4677 constant definitions.
4678 ("nop"): lr 0,0 -> nopr r0
4679 ("nop_lr0", "nop_lr1"): New insn definitions.
4681 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
4683 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
4684 NDS32_V3PUSH_AVAILABLE_P macro.
4686 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
4687 Chung-Ju Wu <jasonwucj@gmail.com>
4689 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
4690 (nds32*-*-*): Add float and fpu_config into supported_defaults.
4691 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
4692 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
4693 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
4694 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
4695 * config/nds32/constraints.md: New constraints and checking for hard
4696 float configuration.
4697 * config/nds32/iterators.md: New mode iterator and attribute for hard
4698 float configuration.
4699 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
4700 patterns.
4701 * config/nds32/nds32-fpu.md: New file.
4702 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
4703 deal with hard float code generation.
4704 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
4705 ARCH_V3S.
4706 (abi_type, float_reg_number): New enum type.
4707 * config/nds32/nds32-predicates.c: New predicates for hard float.
4708 * config/nds32/nds32-protos.h: Declare functions for hard float.
4709 * config/nds32/nds32.c: Implementation for hard float configuration.
4710 * config/nds32/nds32.h: Definitions for hard float configuration.
4711 * config/nds32/nds32.md: Include hard float machine description and
4712 modify patterns for hard float configuration.
4713 * config/nds32/nds32.opt: New options for hard float configuration.
4714 * config/nds32/predicates.md: New predicates for hard float
4715 configuration.
4717 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4719 * common/config/nds32/nds32-common.c
4720 (nds32_option_optimization_table): Enable -mreleax-hint by default.
4722 2018-04-05 Jakub Jelinek <jakub@redhat.com>
4724 PR middle-end/85195
4725 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
4726 CONSTRUCTOR_ELT (ctor, ...)->value.
4728 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
4730 PR target/85193
4731 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
4733 2018-04-05 Tom de Vries <tom@codesourcery.com>
4735 PR target/85204
4736 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
4737 cond jump.
4739 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4740 Kito Cheng <kito.cheng@gmail.com>
4742 * config/nds32/constraints.md (U33): Fine-tune checking condition.
4743 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
4744 * config/nds32/nds32.h (nds32_16bit_address_type): Add
4745 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
4747 2018-04-05 Shiva Chen <shiva0217@gmail.com>
4748 Kito Cheng <kito.cheng@gmail.com>
4750 * config/nds32/constraints.md (Ufe): New memory constraint.
4751 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
4752 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
4753 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
4754 operands.
4755 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
4756 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
4758 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4760 * config/nds32/nds32.md: Use optimize_size in the condition for
4761 alu-shift instructions.
4763 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4765 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
4767 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4769 * config/nds32/nds32.md (negsi2): Refine pattern.
4771 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4772 Chung-Ju Wu <jasonwucj@gmail.com>
4774 * config/nds32/iterators.md (shift_rotate): New code iterator.
4775 (shift): New code attribute.
4776 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
4777 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
4778 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
4779 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
4780 bit-wise operations.
4781 (andsi3, *andsi3): Ditto.
4782 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
4783 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
4784 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
4785 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
4786 nds32_ior_operand, nds32_xor_operand): New predicates.
4788 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4790 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
4791 (addsi3, subsi3): ... this.
4793 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4795 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
4797 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4799 * config/nds32/nds32.md: Adjust indention.
4801 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
4803 * config/nds32/nds32.md (feature): New attribute.
4805 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
4807 * config/nds32/nds32.md (subtype): New attribute.
4809 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
4811 PR target/85203
4812 * config/arm/arm-builtins.c (arm_expand_builtin): Change
4813 expansion to perform a bitwise AND of the argument followed by a
4814 boolean negation of the result.
4816 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
4818 PR rtl-optimization/84878
4819 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
4820 the basic block. Assert the use reference is not artificial and that
4821 it has an associated insn.
4823 2018-04-04 Michael Matz <matz@suse.de>
4825 * builtins.c (compute_objsize): Pass correct operand
4826 to array_at_struct_end_p.
4828 2018-04-04 Richard Biener <rguenther@suse.de>
4830 PR lto/85176
4831 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
4832 from contexts for DINFO_LEVEL_TERSE and below.
4834 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4836 * config/nds32/nds32-doubleword.md (move_<mode>): Require
4837 resiter_operand condition.
4838 * config/nds32/nds32.md (*move<mode>): Ditto.
4840 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4841 Monk Chiang <sh.chiang04@gmail.com>
4843 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
4845 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4847 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
4849 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4850 Kito Cheng <kito.cheng@gmail.com>
4852 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
4853 nds32_cond_code_str, output_cond_branch,
4854 output_cond_branch_compare_zero, nds32_expand_cbranch,
4855 nds32_expand_cstore, nds32_expand_movcc,
4856 nds32_output_cbranchsi4_equality_zero,
4857 nds32_output_cbranchsi4_equality_reg,
4858 nds32_output_cbranchsi4_equality_reg_or_const_int,
4859 nds32_output_cbranchsi4_greater_less_zero: New functions.
4860 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
4861 nds32_expand_cstore, nds32_expand_movcc,
4862 nds32_output_cbranchsi4_equality_zero,
4863 nds32_output_cbranchsi4_equality_reg,
4864 nds32_output_cbranchsi4_equality_reg_or_const_int,
4865 nds32_output_cbranchsi4_greater_less_zero): Declare.
4866 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
4867 nds32_rimm11s_operand): New predicates.
4868 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
4869 * config/nds32/nds32.md: Rewrite all the branch and conditional move
4870 patterns.
4872 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
4874 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
4875 * config/nds32/nds32.md: Ditto.
4876 * config/nds32/pipelines.md: Ditto.
4878 2018-04-04 Richard Biener <rguenther@suse.de>
4880 PR tree-optimization/85168
4881 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
4882 propagating abnormals.
4884 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4886 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
4888 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
4889 Kito Cheng <kito.cheng@gmail.com>
4891 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
4892 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
4893 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
4894 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
4895 * config/nds32/nds32.md (sibcall_internal): New.
4896 (sibcall_register): Remove.
4897 (sibcall_immediate): Remove.
4898 (sibcall_value_internal): New.
4899 (sibcall_value_register): Remove.
4900 (sibcall_value_immediate): Remove.
4901 * config/nds32/predicates.md (nds32_general_register_operand): New.
4902 (nds32_call_address_operand): New.
4904 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4906 PR rtl-optimization/85167
4907 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
4908 bb_defs if *split_p, instead preinitialize it to NULL.
4910 PR tree-optimization/85156
4911 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
4912 evaluating the argument multiple times.
4914 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
4916 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
4917 than vector.
4918 (_mm_cvtpd_ps): Likewise.
4919 (_mm_cvttpd_epi32): Likewise.
4920 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
4921 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
4922 vector, pixel, and bool following altivec.h include.
4924 2018-04-03 Martin Sebor <msebor@redhat.com>
4926 * doc/extend.texi (Common Function Attributes): Clarify.
4927 (const attribute): Likewise.
4928 (pure attribute): Likewise.
4930 2018-04-03 Jakub Jelinek <jakub@redhat.com>
4932 PR target/85169
4933 * config/i386/i386.c (ix86_expand_vector_set): Use
4934 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
4936 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
4938 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
4939 instructions when changing rounding bits to preserve precision bits
4940 in the x87 control word.
4942 2018-04-03 Martin Liska <mliska@suse.cz>
4944 PR tree-optimization/82491
4945 * rtl.h (strip_offset_and_add): Replace += suboffset with
4946 poly_uint64 () + suboffset.
4948 2018-03-29 Martin Liska <mliska@suse.cz>
4949 Martin Jambor <mjambor@suse.cz>
4951 PR ipa/84947
4952 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
4953 param_type is not an integral or pointer type.
4955 2018-04-03 Richard Biener <rguenther@suse.de>
4957 * sese.h (recompute_all_dominators): Remove.
4959 2018-04-02 Martin Sebor <msebor@redhat.com>
4961 * doc/invoke.texi (-Wrestrict): Fix typos.
4963 2018-04-02 Jim Wilson <jimw@sifive.com>
4965 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
4966 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
4967 (<optab>di3, <optab>si3_extend): Likewise.
4968 (<optab>si3_mask, <optab>si3_mask_1): New.
4969 (<optab>di3_mask, <optab>di3_mask_1): New.
4970 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
4971 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
4972 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
4974 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
4976 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
4977 example.
4979 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
4981 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
4982 (nds32_canonicalize_comparison): New function.
4984 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
4985 Kito Cheng <kito.cheng@gmail.com>
4986 Kuan-Lin Chen <kuanlinchentw@gmail.com>
4988 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
4989 * config/nds32/constants.md (unspec_volatile_element): Add
4990 UNSPEC_VOLATILE_RELAX_GROUP.
4991 * config/nds32/nds32-relax-opt.c: New file.
4992 * config/nds32/nds32-predicates.c
4993 (nds32_symbol_load_store_p): New function.
4994 * config/nds32/nds32-protos.h
4995 (nds32_symbol_load_store_p): Declare function.
4996 (make_pass_nds32_relax_opt): Declare new rtl pass function.
4997 * config/nds32/nds32.c
4998 (nds32_register_pass): New function to register pass.
4999 (nds32_register_passes): New function to register passes.
5000 * config/nds32/nds32.md (relax_group): New pattern.
5001 * config/nds32/nds32.opt (mrelax-hint): New option.
5002 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
5004 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
5006 * config/nds32/t-nds32: Modify files dependency.
5008 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5010 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
5011 (PROFILE_HOOK): Define its implementation.
5013 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5015 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
5016 type and 32-bit size.
5018 2018-04-01 Jakub Jelinek <jakub@redhat.com>
5020 PR middle-end/85090
5021 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
5022 (V_128_256): New mode iterator.
5023 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
5024 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
5025 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
5026 of V.
5027 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
5028 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
5030 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
5032 PR target/83315
5033 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
5034 NaN inputs correctly.
5036 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
5038 PR target/80546
5039 * config/rs6000/vsx.md (??r): New mode attribute.
5040 (*vsx_mov<mode>_64bit): Use it.
5041 (*vsx_mov<mode>_32bit): Likewise.
5043 2018-03-30 Martin Sebor <msebor@redhat.com>
5045 PR tree-optimization/84818
5046 * builtins.c (check_access): Use warning_n.
5048 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5050 PR target/83822
5051 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
5052 condition.
5053 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
5054 condition.
5056 2018-03-30 Julia Koval <julia.koval@intel.com>
5058 PR target/84413
5059 * x86-tune.def (movx, partial_reg_dependency): Enable for
5060 m_SKYLAKE_AVX512.
5062 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
5064 PR inline-asm/84985
5065 * lra-constraints.c (process_alt_operands): Move setting
5066 this_alternative_matches below.
5068 2018-03-29 Martin Liska <mliska@suse.cz>
5070 PR lto/84995.
5071 * doc/invoke.texi: Document how LTO works with debug info.
5072 Describe auto-load support of binutils. Mention 'x86-64'
5073 as valid option value of -march option.
5075 2018-03-29 Jakub Jelinek <jakub@redhat.com>
5077 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
5079 PR c/85094
5080 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
5081 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
5082 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
5083 checking.
5085 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5087 PR target/84912
5088 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
5089 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
5090 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
5091 for RS6000_BTM_POWERPC64.
5092 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
5093 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
5094 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
5095 definition.
5096 (DIVDE): Use it.
5097 (DIVDEU): Likewise.
5099 2018-03-28 Carl Love <cel@us.ibm.com>
5101 Revert
5102 2017-09-27 Carl Love <cel@us.ibm.com>
5104 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
5105 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
5106 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
5107 fctiw instruction.
5109 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5111 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
5112 instead of __vector bool.
5113 (_mm_max_pu8): Likewise.
5114 (_mm_min_pi16): Likewise.
5116 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5118 PR target/84912
5119 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
5120 (DIVWEUO): Likewise.
5121 (DIVDEO): Likewise.
5122 (DIVDEUO): Likewise.
5123 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
5124 DIVWEUO and DIVDEUO.
5125 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
5126 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
5127 (div_extend): Likewise.
5128 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
5129 builtin function.
5130 (__builtin_divweuo): Likewise.
5131 (__builtin_divdeo): Likewise.
5132 (__builtin_divdeuo): Likewise.
5134 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5136 PR target/85095
5137 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
5138 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
5140 PR tree-optimization/82004
5141 * gimple-match-head.c (optimize_pow_to_exp): New function.
5142 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
5143 Don't fold to exp if optimize_pow_to_exp is false.
5145 2018-03-28 Martin Liska <mliska@suse.cz>
5147 PR other/84819
5148 * calls.c (initialize_argument_information): Fix trailing space.
5149 * common.opt: Fix typo and provide better explanation for
5150 -fsanitize-coverage option.
5151 * config/i386/i386.opt: Fix typo.
5153 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5154 Martin Liska <mliska@suse.cz>
5156 PR sanitizer/85081
5157 * gimplify.c (asan_poison_variable): Don't do the check for
5158 gimplify_omp_ctxp here.
5159 (gimplify_decl_expr): Do it here.
5160 (gimplify_target_expr): Likewise.
5162 2018-03-28 Martin Liska <mliska@suse.cz>
5164 PR target/84988
5165 * config/i386/i386.c (ix86_function_arg_advance): Do not call
5166 chkp_type_bounds_count if MPX is not enabled.
5168 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
5170 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
5172 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
5174 PR target/84914
5175 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
5176 function to create the function decl for complex long double
5177 multiply and divide for -mabi=ieeelongdouble.
5178 (init_float128_ieee): Call it.
5180 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
5182 PR target/85044
5183 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
5184 -fcf-protection=branch -mibt.
5185 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
5187 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5189 PR target/81863
5190 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
5192 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
5194 PR target/85056
5195 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
5196 extern array declarations.
5198 2018-03-27 Richard Biener <rguenther@suse.de>
5200 PR middle-end/84067
5201 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
5202 explicit single_use checks.
5204 2018-03-27 Richard Biener <rguenther@suse.de>
5206 PR tree-optimization/85082
5207 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
5208 Valueize the VUSE.
5210 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5212 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
5213 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
5214 Turn on fasynchronous-unwind-tables and funwind-tables.
5216 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
5218 PR target/85073
5219 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
5220 (*bmi_blsr_<mode>_ccz): Ditto.
5222 2018-03-26 Tom de Vries <tom@codesourcery.com>
5224 PR tree-optimization/85063
5225 * omp-general.c (offloading_function_p): New function. Factor out
5226 of ...
5227 * omp-offload.c (pass_omp_target_link::gate): ... here.
5228 * omp-general.h (offloading_function_p): Declare.
5229 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
5230 with attribute omp declare target for offloading functions.
5232 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
5234 PR tree-optimization/84005
5235 * tree-data-ref.h (get_base_for_alignment): Declare.
5236 * tree-data-ref.c (get_base_for_alignment_1): New function.
5237 (get_base_for_alignment): Likewise.
5238 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
5239 get_base_for_alignment to find a suitable base object, instead
5240 of always using drb->base_address.
5242 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5244 PR inline-asm/85022
5245 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
5246 known size by default.
5248 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
5250 PR inline-asm/85030
5251 * lra-constraints.c (process_alt_operands): Don't match BLKmode
5252 and non BLKmode operands.
5254 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5256 PR target/85026
5257 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
5258 Clean up attributes.
5260 2018-03-23 Richard Biener <rguenther@suse.de>
5262 PR debug/85020
5263 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
5264 we are going to emit early debug for LTO.
5266 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5268 PR inline-asm/85034
5269 * function.c (match_asm_constraints_1): Don't optimize if input
5270 doesn't satisfy general_operand predicate for output's mode.
5272 PR inline-asm/85022
5273 * alias.c (write_dependence_p): Don't require for x_canonicalized
5274 non-VOIDmode if x has VOIDmode.
5276 PR sanitizer/85029
5277 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
5278 just don't try to optimize it rather than assert it never happens.
5280 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5282 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
5283 macro expansions for definition of ST_INTERNAL_<mode> and
5284 LD_INTERNAL_<mode> builtins.
5285 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
5286 Remove prototype.
5287 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
5288 function.
5289 (altivec_expand_st_builtin): Likewise.
5290 (altivec_expand_builtin): Remove calls to deleted functions.
5291 (rs6000_address_for_altivec): Delete this function.
5292 * config/rs6000/vector.md: Remove expands for
5293 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
5295 2018-03-22 Sudakshina Das <sudi.das@arm.com>
5297 PR target/84826
5298 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
5299 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
5300 re-computing once computed.
5301 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
5302 (arm_init_machine_status): Initialize
5303 machine->static_chain_stack_bytes.
5305 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5307 PR target/84760
5308 * doc/extend.texi: Add four new prototypes for vec_ld.
5309 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
5310 definitions for more logical presentation.
5311 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
5312 entries for V1TI variants of __builtin_altivec_ld builtin.
5313 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
5314 handling of V1TI variant of LVX icode pattern.
5315 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
5316 (rs6000_gimple_fold_builtin): Likewise.
5317 (altivec_init_builtins): Add code to define
5318 __builtin_altivec_lvx_v1ti function.
5320 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5322 PR inline-asm/84941
5323 * function.c (match_asm_constraints_1): Don't do the optimization
5324 if input isn't a REG, SUBREG, MEM or constant.
5326 2018-03-22 Tom de Vries <tom@codesourcery.com>
5328 PR tree-optimization/84956
5329 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
5330 bb_has_abnormal_pred.
5332 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5334 PR sanitizer/85018
5335 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
5336 DECL_INITIAL (decl) to decl at the end.
5337 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
5338 adjust the comment.
5340 2018-03-21 Joseph Myers <joseph@codesourcery.com>
5342 * doc/extend.texi (__builtin_tgmath): Document when complex
5343 integer types are treated as _Complex _Float64.
5345 2018-03-21 Tom de Vries <tom@codesourcery.com>
5347 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
5349 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5351 PR tree-optimization/84960
5352 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
5353 if it is ENTRY block, move them into single succ of ENTRY in that case.
5355 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
5357 PR tree-optimization/84811
5358 * poly-int.h (poly_span_traits): Remove the T3 parameter and
5359 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
5360 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
5361 (known_subrange_p): Update accordingly. Cast each value involved
5362 in the size comparison, rather than casting the result of the
5363 subtraction.
5365 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5367 PR tree-optimization/84982
5368 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
5369 by flipping the least significant bit rather than all bits from
5370 bitpos to bitpos + bitsize - 1.
5372 2018-03-21 Nathan Sidwell <nathan@acm.org>
5374 * doc/extend.texi (Deprecated Features): Remove mention of
5375 long-deleted deprecations.
5377 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5379 PR jit/84288
5380 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
5381 * configure: Regenerate.
5383 2018-03-21 Tom de Vries <tom@codesourcery.com>
5385 PR tree-optimization/83126
5386 * tree-parloops.c (num_phis): New function.
5387 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
5389 2018-03-21 Nathan Sidwell <nathan@acm.org>
5391 * doc/extend.texi (Deprecated Features): Update deprecated flags,
5392 mention anon-struct/union members and trailing attributes.
5394 2018-03-21 Bin Cheng <bin.cheng@arm.com>
5396 PR tree-optimization/84969
5397 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
5398 builtin memset partitions if they set different rhs values.
5400 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5402 PR rtl-optimization/84989
5403 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
5404 VEC_DUPLICATE with scalar result mode.
5406 2018-03-21 Martin Liska <mliska@suse.cz>
5408 PR ipa/84963
5409 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
5410 not intended return statement.
5412 2018-03-21 Martin Liska <mliska@suse.cz>
5414 PR target/84988
5415 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
5416 (chkp_find_bound_slots_1): Limit number of iterations.
5418 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
5420 PR target/84838
5421 * Minor grammar fixes for x86 options.
5423 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5425 PR debug/84875
5426 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
5427 holding REG_CFA_RESTORE notes, instead turn them into a USE.
5429 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
5431 PR target/83789
5432 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
5433 (altivec_lvx_<mode>_1op): Likewise.
5434 (altivec_stvx_<mode>_2op): Likewise.
5435 (altivec_stvx_<mode>_1op): Likewise.
5436 (altivec_lvx_<VM2:mode>): New define_expand.
5437 (altivec_stvx_<VM2:mode>): Likewise.
5438 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
5439 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5440 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
5441 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5442 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
5443 (rs6000_gen_lvx): Likewise.
5444 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
5445 (altivec_expand_stv_builtin): Likewise.
5446 (altivec_expand_builtin): Likewise.
5447 * config/rs6000/vector.md: Likewise.
5449 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5451 PR target/82518
5452 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
5453 BYTES_BIG_ENDIAN.
5455 2018-03-20 Richard Biener <rguenther@suse.de>
5457 PR target/84986
5458 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
5459 sign-conversions as zero, fall back to standard scalar_stmt
5460 cost for the rest.
5462 2018-03-20 Martin Liska <mliska@suse.cz>
5464 PR ipa/84825
5465 * predict.c (rebuild_frequencies): Handle case when we have
5466 PROFILE_ABSENT, but flag_guess_branch_prob is false.
5468 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5470 PR target/84990
5471 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
5472 flag_section_anchors.
5473 * varasm.c (use_blocks_for_decl_p): Remove hack for
5474 dw2_force_const_mem.
5476 PR target/84845
5477 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
5478 to ...
5479 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
5480 be created, use lowpart_subreg of operands[0] rather than operands[0]
5481 itself.
5482 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
5483 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
5484 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
5485 and n constraint instead of aarch64_shift_imm_di and Usd.
5486 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
5487 (*aarch64_<optab>_reg_minus<mode>3): ... this.
5489 2018-03-20 Sudakshina Das <sudi.das@arm.com>
5491 PR target/82989
5492 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
5493 to favor GPR over NEON registers.
5494 (<shift>di3_neon): Likewise.
5496 2018-03-20 Tom de Vries <tom@codesourcery.com>
5498 PR target/84952
5499 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
5500 (nvptx_process_pars): Emit bar.sync asap and alap.
5502 2018-03-20 Tom de Vries <tom@codesourcery.com>
5504 PR target/84954
5505 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
5506 seen_label if seen_label is already set.
5508 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5510 PR target/84945
5511 * config/i386/i386.c (fold_builtin_cpu): For features above 31
5512 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
5513 Use 1U instead of 1. Formatting fixes.
5515 PR c/84953
5516 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
5517 instead of TREE_TYPE (s1) for the return value.
5519 2018-03-19 Jakub Jelinek <jakub@redhat.com>
5521 PR tree-optimization/84946
5522 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
5523 bitsize + bitsize in poly_uint64 rather than poly_int64.
5525 PR sanitizer/78651
5526 * dwarf2asm.c: Include fold-const.c.
5527 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
5528 of decl rather than decl itself.
5530 PR rtl-optimization/84643
5531 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
5533 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
5535 PR sanitizer/78651
5536 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
5537 calling assemble_variable.
5539 2018-03-19 Sudakshina Das <sudi.das@arm.com>
5541 PR target/81647
5542 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
5543 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
5545 2018-03-19 Jim Wilson <jimw@sifive.com>
5547 PR bootstrap/84856
5548 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
5549 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
5550 (riscv_first_stack_step): Likewise.
5551 (riscv_option_override): Use STACK_BOUNDARY instead of
5552 MIN_STACK_BOUNDARY.
5553 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
5554 MIN_STACK_BOUNDARY.
5555 (BIGGEST_ALIGNMENT): Set to 128.
5556 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
5557 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
5558 STACK_BOUNDARY.
5560 2018-03-19 Richard Biener <rguenther@suse.de>
5562 PR tree-optimization/84933
5563 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
5564 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
5566 2018-03-19 Richard Biener <rguenther@suse.de>
5568 PR tree-optimization/84859
5569 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
5570 (cond_if_else_store_replacement): Perform sinking operation on
5571 single-store BBs regardless of MAX_STORES_TO_SINK setting.
5572 Generalize what a BB with a single eligible store is.
5574 2018-03-19 Richard Biener <rguenther@suse.de>
5576 PR tree-optimization/84929
5577 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
5578 chrec_is_positive against non-chrec arg.
5580 2018-03-19 Tamar Christina <tamar.christina@arm.com>
5582 PR target/84711
5583 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
5585 2018-03-18 Martin Liska <mliska@suse.cz>
5587 PR rtl-optimization/84635
5588 * regrename.c (build_def_use): Use matches_mode only when
5589 matches >= 0.
5591 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
5593 PR tree-optimization/84913
5594 * tree-vect-loop.c (vectorizable_reduction): Don't try to
5595 vectorize chains of COND_EXPRs.
5597 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5599 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
5601 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5603 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
5605 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
5607 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
5609 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
5610 Kito Cheng <kito.cheng@gmail.com>
5612 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
5613 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
5614 (nds32_adjust_reg_alloc_order): New function.
5615 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
5617 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
5619 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
5620 nds32_print_operand, nds32_print_operand_address): Use
5621 HOST_WIDE_INT_PRINT_DEC instead.
5623 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
5625 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
5627 2018-03-17 Jakub Jelinek <jakub@redhat.com>
5629 PR target/84902
5630 * config/i386/i386.c (initial_ix86_tune_features,
5631 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
5632 unsigned long long.
5633 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
5634 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
5635 rather than 1u << ix86_tune. Formatting fix.
5636 (ix86_option_override_internal): Change ix86_arch_mask from
5637 unsigned int to unsigned HOST_WIDE_INT, initialize to
5638 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
5639 (ix86_function_specific_restore): Likewise.
5641 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5643 PR target/84899
5644 * postreload.c (reload_combine_recognize_pattern): Perform
5645 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
5646 truncate_int_for_mode the result for the destination's mode.
5648 PR c/84909
5649 * hsa-gen.c (mem_type_for_type): Fix comment typo.
5650 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
5651 Likewise.
5652 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
5653 Likewise.
5655 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
5657 PR target/84876
5658 * lra-assigns.c (lra_split_hard_reg_for): Don't use
5659 regno_allocno_class_array and sorted_pseudos.
5660 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
5661 insns where regno is used.
5663 2018-03-16 Martin Liska <mliska@suse.cz>
5665 PR ipa/84833
5666 * multiple_target.c (create_dispatcher_calls): Redirect
5667 reference in the symbol table.
5669 2018-03-16 Martin Liska <mliska@suse.cz>
5671 PR ipa/84722
5672 * multiple_target.c (create_dispatcher_calls): Redirect also
5673 an alias.
5675 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5677 PR c++/79937
5678 PR c++/82410
5679 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
5680 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
5681 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
5683 2018-03-16 Julia Koval <julia.koval@intel.com>
5685 * doc/invoke.texi (Skylake Server): Add CLWB.
5686 Cannonlake): Remove CLWB.
5688 2018-03-16 Jakub Jelinek <jakub@redhat.com>
5690 PR tree-optimization/84841
5691 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
5692 1 << 3.
5693 (FLOAT_ONE_CONST_TYPE): Define.
5694 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
5695 (sort_by_operand_rank): Put entries with higher constant_type last
5696 rather than first to match comments.
5698 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
5700 * config/nios2/nios2.md (movsi_internal): Fix thinko in
5701 split predicate.
5703 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5705 PR c++/79085
5706 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
5707 check and use address of target always.
5709 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
5711 PR target/84574
5712 * config/i386/i386.c (indirect_thunk_needed): Update comments.
5713 (indirect_thunk_bnd_needed): Likewise.
5714 (indirect_thunks_used): Likewise.
5715 (indirect_thunks_bnd_used): Likewise.
5716 (indirect_return_needed): New.
5717 (indirect_return_bnd_needed): Likewise.
5718 (output_indirect_thunk_function): Add a bool argument for
5719 function return.
5720 (output_indirect_thunk_function): Don't generate alias for
5721 function return thunk.
5722 (ix86_code_end): Call output_indirect_thunk_function to generate
5723 function return thunks.
5724 (ix86_output_function_return): Set indirect_return_bnd_needed
5725 and indirect_return_needed instead of indirect_thunk_bnd_needed
5726 and indirect_thunk_needed.
5728 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
5730 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
5731 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
5732 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
5734 2018-03-15 David Malcolm <dmalcolm@redhat.com>
5735 Paul Hua <paul.hua.gm@gmail.com>
5737 PR c/84852
5738 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
5740 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
5742 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
5743 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
5744 resp. SFmode cases.
5746 2018-03-15 Tamar Christina <tamar.christina@arm.com>
5748 PR target/84711
5749 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
5750 instead of GET_MODE_SIZE when comparing Units.
5752 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
5754 PR target/68256
5755 * varasm.c (hash_section): Return an unchangeble hash value
5756 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
5757 Return !aarch64_can_use_per_function_literal_pools_p ().
5759 2018-03-15 Jakub Jelinek <jakub@redhat.com>
5761 PR target/84860
5762 * optabs.c (emit_conditional_move): Pass address of cmode's copy
5763 rather than address of cmode as last argument to prepare_cmp_insn.
5765 2018-03-15 Julia Koval <julia.koval@intel.com>
5767 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
5768 F_AVX512VNNI, F_AVX512BITALG): New.
5770 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
5772 PR target/83451
5773 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
5774 insn for floating-point loads and stores.
5776 2018-03-14 Carl Love <cel@us.ibm.com>
5778 * config/rs6000/rs6000-c.c: Add macro definitions for
5779 ALTIVEC_BUILTIN_VEC_PERMXOR.
5780 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
5781 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
5782 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
5783 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
5784 UNSPEC_VPERMXOR.
5785 * config/doc/extend.texi: Add prototypes for vec_permxor.
5787 2018-03-14 David Malcolm <dmalcolm@redhat.com>
5789 PR c/84852
5790 * diagnostic-show-locus.c (class layout_point): Convert m_line
5791 from int to linenum_type.
5792 (line_span::comparator): Use linenum "compare" function when
5793 comparing line numbers.
5794 (test_line_span): New function.
5795 (layout_range::contains_point): Convert param "row" from int to
5796 linenum_type.
5797 (layout_range::intersects_line_p): Likewise.
5798 (layout::will_show_line_p): Likewise.
5799 (layout::print_source_line): Likewise.
5800 (layout::should_print_annotation_line_p): Likewise.
5801 (layout::print_annotation_line): Likewise.
5802 (layout::print_leading_fixits): Likewise.
5803 (layout::annotation_line_showed_range_p): Likewise.
5804 (struct line_corrections): Likewise for field m_row.
5805 (line_corrections::line_corrections): Likewise for param "row".
5806 (layout::print_trailing_fixits): Likewise.
5807 (layout::get_state_at_point): Likewise.
5808 (layout::get_x_bound_for_row): Likewise.
5809 (layout::print_line): Likewise.
5810 (diagnostic_show_locus): Likewise for locals "last_line" and
5811 "row".
5812 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
5813 * input.c (selftest::test_linenum_comparisons): New function.
5814 (selftest::input_c_tests): Call it.
5815 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
5816 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
5817 * selftest.h (ASSERT_GT): New macro.
5818 (ASSERT_GT_AT): New macro.
5819 (ASSERT_LT): New macro.
5820 (ASSERT_LT_AT): New macro.
5822 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
5824 PR rtl-optimization/84780
5825 * combine.c (distribute_links): Don't make a link based on pc_rtx.
5827 2018-03-14 Martin Liska <mliska@suse.cz>
5829 * tree.c (record_node_allocation_statistics): Use
5830 get_stats_node_kind.
5831 (get_stats_node_kind): New function extracted from
5832 record_node_allocation_statistics.
5833 (free_node): Use get_stats_node_kind.
5835 2018-03-14 Richard Biener <rguenther@suse.de>
5837 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
5838 that the value-set of ANTIC_IN doesn't grow.
5840 Revert
5841 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5842 member.
5843 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5844 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5846 2018-03-14 Julia Koval <julia.koval@intel.com>
5848 * config.gcc (icelake-client, icelake-server): New.
5849 (icelake): Remove.
5850 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
5851 (initial_ix86_arch_features): Ditto.
5852 (PTA_SKYLAKE): Add SGX.
5853 (PTA_ICELAKE): Remove.
5854 (PTA_ICELAKE_CLIENT): New.
5855 (PTA_ICELAKE_SERVER): New.
5856 (ix86_option_override_internal): Split up icelake on icelake client and
5857 icelake server.
5858 (get_builtin_code_for_version): Ditto.
5859 (fold_builtin_cpu): Ditto.
5860 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
5861 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
5862 * config/i386/i386.h (processor_type): Ditto.
5863 * doc/invoke.texi: Ditto.
5865 2018-03-14 Jakub Jelinek <jakub@redhat.com>
5867 PR sanitizer/83392
5868 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
5869 INTEGER_CST offset, add it together with bitpos / 8 and
5870 sign extend based on POINTER_SIZE.
5872 PR target/84844
5873 Revert
5874 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
5876 PR target/78090
5877 * config/i386/constraints.md (Yc): New register constraint.
5878 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
5879 Use Yc constraint for alternative 2 of operand 0. Remove
5880 preferred_for_speed attribute.
5882 2018-03-14 Richard Biener <rguenther@suse.de>
5884 PR tree-optimization/84830
5885 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
5886 with the old one to avoid oscillations.
5888 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
5890 PR target/83712
5891 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
5892 pseudos.
5893 (assign_by_spills): Return a flag of reload assignment failure.
5894 Do not process the reload assignment failures. Do not spill other
5895 reload pseudos if they has the same reg class. Update n if
5896 necessary.
5897 (lra_assign): Add a return arg. Set up from the result of
5898 assign_by_spills call.
5899 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
5900 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
5901 usage_insns if it is not NULL.
5902 (spill_hard_reg_in_range): New function.
5903 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
5904 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
5905 function prototypes.
5906 (lra_assign): Change prototype.
5907 * lra.c (lra): Add code to deal with fails by splitting hard reg
5908 live ranges.
5910 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
5912 * config/riscv/riscv.opt (mrelax): New option.
5913 * config/riscv/riscv.c (riscv_file_start): Emit ".option
5914 "norelax" when riscv_mrelax is disabled.
5915 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
5917 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5919 PR target/84743
5920 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
5921 reassociation for int modes.
5923 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5925 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
5926 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
5927 for big-endian.
5928 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
5929 * config/aarch64/aarch64-sve.md
5930 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
5931 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
5932 (*extend<mode><Vwide>2): Rename to...
5933 (aarch64_sve_extend<mode><Vwide>2): ...this.
5934 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
5935 renaming the old pattern to...
5936 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
5937 unsigned packs.
5938 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
5939 define_expand, renaming the old pattern to...
5940 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
5941 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
5942 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
5943 account when deciding which SVE instruction the optab should use.
5944 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
5946 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5948 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
5949 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
5950 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
5951 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
5952 (tlsdesc_small_<mode>): Turn a define_expand and use
5953 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
5954 (tlsdesc_small_advsimd_<mode>): ...this.
5955 (tlsdesc_small_sve_<mode>): New pattern.
5957 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
5959 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
5960 (UNSPEC_UMUL_HIGHPART): New constants.
5961 (MUL_HIGHPART): New int iteraor.
5962 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
5963 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
5964 define_expand.
5965 (*<su>mul<mode>3_highpart): New define_insn.
5967 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
5969 PR lto/84805
5970 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
5971 incomplete types.
5973 2018-03-13 Martin Liska <mliska@suse.cz>
5975 PR ipa/84658.
5976 * (sem_item_optimizer::sem_item_optimizer): Initialize new
5977 vector.
5978 (sem_item_optimizer::~sem_item_optimizer): Release it.
5979 (sem_item_optimizer::merge_classes): Register variable aliases.
5980 (sem_item_optimizer::fixup_pt_set): New function.
5981 (sem_item_optimizer::fixup_points_to_sets): Likewise.
5982 * ipa-icf.h: Declare new variables and functions.
5984 2018-03-13 Jakub Jelinek <jakub@redhat.com>
5986 PR middle-end/84834
5987 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
5988 integer_pow2p@2 and test integer_pow2p in condition.
5989 (A < 0 ? C : 0): Similarly for @1.
5991 PR middle-end/84831
5992 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
5993 characters starting at p contain '\0' character, don't look beyond
5994 that.
5996 PR target/84827
5997 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
5998 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
6000 PR target/84828
6001 * reg-stack.c (change_stack): Change update_end var from int to
6002 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
6003 also call set_block_for_insn on the newly added insns and rescan.
6005 PR target/84786
6006 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
6007 on the last operand.
6009 PR c++/84704
6010 * tree.c (stabilize_reference_1): Return save_expr (e) for
6011 STATEMENT_LIST even if it doesn't have side-effects.
6013 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
6015 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
6017 2018-03-12 Renlin Li <renlin.li@arm.com>
6019 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
6020 aarch64_output_scalar_simd_mov_immediate.
6022 2018-03-12 Martin Sebor <msebor@redhat.com>
6024 PR tree-optimization/83456
6025 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
6026 for perfectly overlapping calls to memcpy.
6027 (gimple_fold_builtin_memory_chk): Same.
6028 (gimple_fold_builtin_strcpy): Handle no-warning.
6029 (gimple_fold_builtin_stxcpy_chk): Same.
6030 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
6032 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6034 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
6035 parameter. Use it for SFmode.
6036 (rs6000_function_arg_advance_1): Adjust.
6037 (rs6000_function_arg): Adjust.
6038 (rs6000_gimplify_va_arg): Pass false for that new parameter.
6040 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6042 PR rtl-optimization/84169
6043 PR rtl-optimization/84780
6044 * combine.c (can_combine_p): Check for a 2-insn combination whether
6045 the destination register is used between the two insns, too.
6047 2018-03-12 Richard Biener <rguenther@suse.de>
6049 PR tree-optimization/84803
6050 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
6051 for refs DR analysis didn't process.
6053 2018-03-12 Richard Biener <rguenther@suse.de>
6055 PR tree-optimization/84777
6056 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
6057 force-vectorize loops ignore whether we are optimizing for size.
6059 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
6061 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
6062 (TARGET_MD_ASM_ADJUST): Define.
6064 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
6065 Kito Cheng <kito.cheng@gmail.com>
6066 Chung-Ju Wu <jasonwucj@gmail.com>
6068 * config/nds32/nds32.c (nds32_compute_stack_frame,
6069 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
6070 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
6071 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
6072 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
6073 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
6074 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
6075 * config/nds32/nds32.md (prologue, epilogue): Use macro
6076 NDS32_V3PUSH_AVAILABLE_P to do checking.
6078 2018-03-11 Jakub Jelinek <jakub@redhat.com>
6080 PR debug/58150
6081 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
6082 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
6083 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
6084 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
6085 addition of most attributes on !orig_type_die or the attribute not
6086 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
6088 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6089 Chung-Ju Wu <jasonwucj@gmail.com>
6091 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
6092 __NDS32_VH__ macro.
6093 * config/nds32/nds32.opt (mvh): New option.
6095 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6096 Chung-Ju Wu <jasonwucj@gmail.com>
6098 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
6099 function.
6100 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
6101 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
6102 definition.
6104 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6105 Chung-Ju Wu <jasonwucj@gmail.com>
6107 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
6108 function.
6109 * config/nds32/nds32-multiple.md (strlensi): New pattern.
6110 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
6112 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
6113 Kito Cheng <kito.cheng@gmail.com>
6114 Chung-Ju Wu <jasonwucj@gmail.com>
6116 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
6117 UNSPEC_FFMISM and UNSPEC_FLMISM.
6118 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
6119 for ffb, ffmism and flmism.
6120 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
6121 (unspec_ffmism): Ditto.
6122 (unspec_flmism): Ditto.
6123 (nds32_expand_builtin_impl): Check if string extension is available.
6124 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
6125 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
6127 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
6129 Reverting patch:
6130 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6132 PR target/83712
6133 * lra-assigns.c (assign_by_spills): Return a flag of reload
6134 assignment failure. Do not process the reload assignment
6135 failures. Do not spill other reload pseudos if they has the same
6136 reg class.
6137 (lra_assign): Add a return arg. Set up from the result of
6138 assign_by_spills call.
6139 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6140 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6141 usage_insns if it is not NULL.
6142 (spill_hard_reg_in_range): New function.
6143 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6144 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6145 function prototypes.
6146 (lra_assign): Change prototype.
6147 * lra.c (lra): Add code to deal with fails by splitting hard reg
6148 live ranges.
6150 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
6152 PR target/84807
6153 * config/i386/i386.opt: Replace Enforcment with Enforcement.
6155 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
6157 PR debug/84620
6158 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
6159 (dw_val_node): Add val_symbolic_view.
6160 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
6161 (symview_upper_bound): New.
6162 (new_line_info_table): Initialize symviews_since_reset.
6163 (dwarf2out_source_line): Count symviews_since_reset and set
6164 symview_upper_bound.
6165 (dw_val_equal_p): Handle symview.
6166 (add_AT_symview): New.
6167 (print_dw_val): Handle symview.
6168 (attr_checksum, attr_checksum_ordered): Likewise.
6169 (same_dw_val_p, size_of_die): Likewise.
6170 (value_format, output_die): Likewise.
6171 (add_high_low_attributes): Use add_AT_symview for entry_view.
6172 (dwarf2out_finish): Reset symview_upper_bound, clear
6173 zero_view_p.
6175 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
6177 PR target/83969
6178 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
6179 Add strict argument and use it.
6180 (rs6000_split_multireg_move): Update for new strict argument.
6181 (mem_operand_gpr): Disallow all non-offsettable addresses.
6182 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
6184 2018-03-09 Jakub Jelinek <jakub@redhat.com>
6186 PR target/84772
6187 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
6188 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
6189 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
6191 PR c++/84767
6192 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
6193 decl, use remap_type if we want to use the type.
6195 2018-03-09 Martin Sebor <msebor@redhat.com>
6197 PR tree-optimization/84526
6198 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6199 Remove dead code.
6200 (builtin_access::generic_overlap): Be prepared to handle non-array
6201 base objects.
6203 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
6205 PR rtl-optimization/84682
6206 * lra-constraints.c (process_address_1): Check is_address flag
6207 for address constraints.
6208 (process_alt_operands): Likewise.
6209 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
6210 preprocess_constraints.
6211 * recog.h (preprocess_constraints): Add oploc parameter.
6212 Adjust callers.
6213 * recog.c (preprocess_constraints): Test address_operand for
6214 CT_ADDRESS constraints.
6216 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6218 PR target/83712
6219 * lra-assigns.c (assign_by_spills): Return a flag of reload
6220 assignment failure. Do not process the reload assignment
6221 failures. Do not spill other reload pseudos if they has the same
6222 reg class.
6223 (lra_assign): Add a return arg. Set up from the result of
6224 assign_by_spills call.
6225 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6226 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6227 usage_insns if it is not NULL.
6228 (spill_hard_reg_in_range): New function.
6229 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6230 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6231 function prototypes.
6232 (lra_assign): Change prototype.
6233 * lra.c (lra): Add code to deal with fails by splitting hard reg
6234 live ranges.
6236 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6238 PR target/83193
6239 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
6240 Accept complain bool parameter. Only emit errors if it is true.
6241 (arm_parse_cpu_option_name): Likewise.
6242 (arm_target_thumb_only): Adjust callers of the above.
6243 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
6244 prototype to take a default true bool parameter.
6245 (arm_parse_arch_option_name): Likewise.
6247 2018-03-09 David Malcolm <dmalcolm@redhat.com>
6248 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
6250 PR jit/64089
6251 PR jit/84288
6252 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
6253 * configure: Regenerate.
6254 * configure.ac ("linker --version-script option"): New.
6255 ("linker soname option"): New.
6257 2018-03-09 Richard Biener <rguenther@suse.de>
6259 PR tree-optimization/84775
6260 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
6261 immediate uses of predicate stmts and mark them modified.
6263 Revert
6264 PR tree-optimization/84178
6265 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6266 to caller.
6267 (version_loop_for_if_conversion): Delay update_ssa call.
6268 (tree_if_conversion): Delay update_ssa until after predicate
6269 insertion.
6271 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
6273 PR target/84763
6274 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
6275 when the function accesses prior frames.
6277 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6279 PR debug/84456
6280 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
6281 gen_llsym, otherwise call maybe_gen_llsym.
6283 PR inline-asm/84742
6284 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
6285 has ',' character inside of it.
6287 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6289 PR target/84748
6290 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
6291 as clobbering CC_REGNUM.
6293 2018-03-08 Richard Biener <rguenther@suse.de>
6295 PR middle-end/84552
6296 * tree-scalar-evolution.c: Include tree-into-ssa.h.
6297 (follow_copies_to_constant): Do not follow SSA names registered
6298 for update.
6300 2018-03-08 Richard Biener <rguenther@suse.de>
6302 PR tree-optimization/84178
6303 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6304 to caller.
6305 (version_loop_for_if_conversion): Delay update_ssa call.
6306 (tree_if_conversion): Delay update_ssa until after predicate
6307 insertion.
6309 2018-03-08 David Malcolm <dmalcolm@redhat.com>
6311 PR tree-optimization/84178
6312 * tree-if-conv.c (release_bb_predicate): Remove the
6313 the assertion that the stmts have NULL use_ops.
6314 Discard the statements, asserting that they haven't
6315 yet been added to a BB.
6317 2018-03-08 Richard Biener <rguenther@suse.de>
6319 PR tree-optimization/84746
6320 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
6321 (phi_translate): Pass in destination ANTIC_OUT set.
6322 (phi_translate_1): Likewise. For a simplified result lookup
6323 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
6324 (phi_translate_set): Adjust.
6325 (do_pre_regular_insertion): Likewise.
6326 (do_pre_partial_partial_insertion): Likewise.
6328 2018-03-08 Martin Liska <mliska@suse.cz>
6330 PR gcov-profile/84735
6331 * doc/gcov.texi: Document usage of profile files.
6332 * gcov-io.h: Document changes in the format.
6334 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
6336 PR debug/84404
6337 PR debug/84408
6338 * dwarf2out.c (struct dw_line_info_table): Update comments for
6339 view == -1.
6340 (FORCE_RESET_NEXT_VIEW): New.
6341 (FORCE_RESETTING_VIEW_P): New.
6342 (RESETTING_VIEW_P): Check for -1 too.
6343 (ZERO_VIEW_P): Likewise.
6344 (new_line_info_table): Force-reset next view.
6345 (dwarf2out_begin_function): Likewise.
6346 (dwarf2out_source_line): Simplify zero_view_p initialization.
6347 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
6348 view directly. Omit view when omitting .loc at line 0.
6350 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6352 PR tree-optimization/84740
6353 * tree-switch-conversion.c (process_switch): Call build_constructors
6354 only if info.phi_count is non-zero.
6356 PR tree-optimization/84739
6357 * tree-tailcall.c (find_tail_calls): Check call arguments against
6358 DECL_ARGUMENTS (current_function_decl) rather than
6359 DECL_ARGUMENTS (func) when checking for tail recursion.
6361 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6363 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
6364 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
6365 Volker Reichelt's entry and add entries for people that perform
6366 GCC fuzzy testing and report numerous bugs.
6368 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
6370 PR target/82411
6371 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
6372 readonly data in sdata, if that is disabled.
6373 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
6374 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
6375 -mreadonly-in-sdata option.
6377 2018-03-07 Martin Sebor <msebor@redhat.com>
6379 PR tree-optimization/84468
6380 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
6381 basic block when looking for nul assignment.
6383 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
6385 PR target/84277
6386 * except.h (output_function_exception_table): Adjust prototype.
6387 * except.c (output_function_exception_table): Remove FNNAME parameter
6388 and add SECTION parameter. Ouput one part of the table at a time.
6389 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
6390 the first part of the exception table and emit unwind directives.
6391 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
6392 (i386_pe_seh_cold_init): Likewise.
6393 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
6394 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
6395 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
6396 (ix86_output_call_insn): Emit a nop in one more case for SEH.
6397 * config/i386/winnt.c: Include except.h.
6398 (struct seh_frame_state): Add reg_offset, after_prologue and
6399 in_cold_section fields.
6400 (i386_pe_seh_end_prologue): Set seh->after_prologue.
6401 (i386_pe_seh_cold_init): New function.
6402 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
6403 to seh->in_cold_section.
6404 (seh_emit_push): Record the offset of the push.
6405 (seh_emit_save): Record the offet of the save.
6406 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
6407 Test seh->after_prologue to disregard the epilogue.
6408 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
6409 (i386_pe_end_cold_function): New function.
6411 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6413 PR fortran/84565
6414 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
6415 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
6417 PR c++/84704
6418 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
6419 on tmp_var.
6420 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
6421 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
6423 PR middle-end/84723
6424 * multiple_target.c: Include tree-inline.h and intl.h.
6425 (expand_target_clones): Diagnose and fail if node->definition and
6426 !tree_versionable_function_p (node->decl).
6428 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
6430 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
6431 sprint_ul.
6432 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
6433 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
6434 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
6436 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6438 PR target/84710
6439 * combine.c (try_combine): Use reg_or_subregno instead of handling
6440 just paradoxical SUBREGs and REGs.
6442 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
6444 * config/arc/arc.c (arc_finalize_pic): Remove function.
6445 (arc_must_save_register): We use single base PIC register, remove
6446 checks to save/restore the PIC register.
6447 (arc_expand_prologue): Likewise.
6448 * config/arc/arc-protos.h (arc_set_default_type_attributes):
6449 Remove.
6450 (arc_verify_short): Likewise.
6451 (arc_attr_type): Likewise.
6452 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
6453 (walk_stores): Likewise.
6454 (arc_address_cost): Make it static.
6455 (arc_verify_short): Likewise.
6456 (branch_dest): Likewise.
6457 (arc_attr_type): Likewise.
6458 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
6459 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
6460 (arc_final_prescan_insn): Remove inserting the nops due to
6461 hardware hazards. It is done in reorg step.
6462 (insn_length_variant_t): Remove.
6463 (insn_length_parameters_t): Likewise.
6464 (arc_insn_length_parameters): Likewise.
6465 (arc_get_insn_variants): Likewise.
6466 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
6468 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6470 PR inline-asm/84683
6471 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
6472 assertion failure.
6474 PR tree-optimization/84687
6475 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
6476 on new_node->decl.
6477 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
6479 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6481 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
6482 Rename to ppc_speculation_barrier.
6483 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
6484 __builtin_ppc_speculation_barrier.
6486 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6488 PR target/84700
6489 * combine.c (combine_simplify_rtx): Don't try to simplify if
6490 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
6491 are equal to x.
6493 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
6495 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
6496 to 32 bytes when compiling for POWER9.
6498 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6500 PR target/84564
6501 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
6502 regparm >= 3 with no arg reg available also for calls with
6503 flag_force_indirect_call. Pass decl to ix86_function_regparm.
6505 PR target/84524
6506 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
6507 orig,vex.
6508 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
6510 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
6512 PR target/84264
6513 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
6515 2018-03-05 Richard Biener <rguenther@suse.de>
6517 PR tree-optimization/84486
6518 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
6519 When inserting a __builtin_assume_aligned call set the LHS
6520 SSA name alignment info accordingly.
6522 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
6524 PR tree-optimization/84114
6525 * config/aarch64/aarch64.c (aarch64_reassociation_width)
6526 Avoid reassociation of FLOAT_MODE addition.
6528 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
6530 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
6531 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
6532 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
6533 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
6534 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
6535 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
6536 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
6537 and -mwbnoinvd.
6538 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
6539 __builtin_ia32_wbinvd): New builtins.
6540 (SPECIAL_ARGS2): New.
6541 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
6542 (SPECIAL_ARGS2): New.
6543 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
6544 (ix86_valid_target_attribute_inner_p): Ditto.
6545 (ix86_init_mmx_sse_builtins): Add special_args2.
6546 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
6547 TARGET_WBNOINVD_P): New.
6548 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
6549 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
6550 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
6551 * config/i386/immintrin.h (_wbinvd): New intrinsic.
6552 * config/i386/pconfigintrin.h: New file.
6553 * config/i386/wbnoinvdintrin.h: Ditto.
6554 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
6555 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
6557 2018-03-05 Richard Biener <rguenther@suse.de>
6559 PR tree-optimization/84670
6560 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6561 member.
6562 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6563 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6564 (compute_antic_aux): Only assert the number of values in ANTIC_IN
6565 doesn't grow if all successors (recursively) were visited at least
6566 once.
6568 2018-03-05 Richard Biener <rguenther@suse.de>
6570 PR tree-optimization/84650
6571 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
6572 if executed in the loop pipeline.
6574 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
6576 * doc/configfiles.texi (Configuration Files): Move info about
6577 conditionalizing $target-protos.h to...
6578 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
6579 differs from $target-protos.h.
6581 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
6582 Chung-Ju Wu <jasonwucj@gmail.com>
6584 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
6585 * config/nds32/nds32-multiple.md (setmemsi): Define.
6586 * config/nds32/nds32-memory-manipulation.c
6587 (nds32_gen_dup_4_byte_to_word_value): New.
6588 (emit_setmem_word_loop): New.
6589 (emit_setmem_byte_loop): New.
6590 (nds32_expand_setmem_loop): New.
6591 (nds32_expand_setmem_loop_v3m): New.
6592 (nds32_expand_setmem_unroll): New.
6593 (nds32_expand_setmem): New.
6595 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6596 Chung-Ju Wu <jasonwucj@gmail.com>
6598 * config/nds32/nds32-memory-manipulation.c
6599 (nds32_emit_load_store): New.
6600 (nds32_emit_post_inc_load_store): New.
6601 (nds32_emit_mem_move): New.
6602 (nds32_emit_mem_move_block): New.
6603 (nds32_expand_movmemsi_loop_unknown_size): New.
6604 (nds32_expand_movmemsi_loop_known_size): New.
6605 (nds32_expand_movmemsi_loop): New.
6606 (nds32_expand_movmemsi_unroll): New.
6607 (nds32_expand_movmemqi): Rename ...
6608 (nds32_expand_movmemsi): ... to this.
6609 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
6610 (movmemsi): ... to this.
6611 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
6612 (nds32_expand_movmemsi): ... to this.
6614 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6615 Monk Chiang <sh.chiang04@gmail.com>
6616 Chung-Ju Wu <jasonwucj@gmail.com>
6618 * config/nds32/nds32-protos.h
6619 (nds32_expand_load_multiple): New arguments.
6620 (nds32_expand_store_multiple): Ditto.
6621 (nds32_valid_multiple_load_store): Rename ...
6622 (nds32_valid_multiple_load_store_p): ... to this.
6623 * config/nds32/nds32-memory-manipulation.c
6624 (nds32_expand_load_multiple): Refine implementation.
6625 (nds32_expand_store_multiple): Ditto.
6626 * config/nds32/nds32-multiple.md
6627 (load_multiple): Update nds32_expand_load_multiple interface.
6628 (store_multiple): Update nds32_expand_store_multiple interface.
6629 * config/nds32/nds32-predicates.c
6630 (nds32_valid_multiple_load_store): Rename ...
6631 (nds32_valid_multiple_load_store_p): ... to this and refine
6632 implementation.
6633 * config/nds32/predicates.md
6634 (nds32_load_multiple_and_update_address_operation): New predicate.
6635 (nds32_store_multiple_and_update_address_operation): New predicate.
6637 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
6638 Chung-Ju Wu <jasonwucj@gmail.com>
6640 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
6641 (combo): New attribute.
6642 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
6644 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
6646 * config/nds32/nds32.opt: Change -mcmodel= default value.
6648 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
6649 Monk Chiang <sh.chiang04@gmail.com>
6650 Chung-Ju Wu <jasonwucj@gmail.com>
6652 * config/nds32/constants.md (unspec_element): New enum.
6653 * config/nds32/constraints.md (Umw): New constraint.
6654 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
6655 * config/nds32/nds32-intrinsic.md: Likewise.
6656 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
6657 (nds32_valid_smw_lwm_base_p): New.
6658 (nds32_output_smw_single_word): New.
6659 (nds32_output_lmw_single_word): New.
6660 (nds32_expand_unaligned_load): New.
6661 (nds32_expand_unaligned_store): New.
6662 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
6663 (nds32_output_smw_single_word): Declare.
6664 (nds32_output_lmw_single_word): Declare.
6665 (nds32_expand_unaligned_load): Declare.
6666 (nds32_expand_unaligned_store): Declare.
6667 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
6668 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
6669 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
6670 NDS32_BUILTIN_UASTORE_DW.
6671 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
6672 predicate.
6674 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
6675 Kito Cheng <kito.cheng@gmail.com>
6676 Chung-Ju Wu <jasonwucj@gmail.com>
6678 * config/nds32/nds32-intrinsic.c
6679 (nds32_expand_builtin_null_ftype_reg): Delete.
6680 (nds32_expand_builtin_reg_ftype_imm): Ditto.
6681 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
6682 (nds32_read_argument): New.
6683 (nds32_legitimize_target): Ditto.
6684 (nds32_legitimize_argument): Ditto.
6685 (nds32_check_constant_argument): Ditto.
6686 (nds32_expand_unop_builtin): Ditto.
6687 (nds32_expand_unopimm_builtin): Ditto.
6688 (nds32_expand_binop_builtin): Ditto.
6689 (nds32_builtin_decl_impl): Ditto.
6690 (builtin_description): Ditto.
6691 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
6692 (nds32_init_builtins_impl): Ditto.
6693 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
6694 (nds32_builtin_decl): New.
6695 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
6696 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
6698 2018-03-02 Jeff Law <law@redhat.com>
6700 * reorg.c (stop_search_p): Handle DEBUG_INSN.
6701 (redundant_insn, fill_simple_delay_slots): Likewise.
6702 (fill_slots_from_thread): Likewise.
6703 * resource.c (mark_referenced_resources): Likewise.
6704 (mark_set_resources, find_dead_or_set_registers): Likewise.
6706 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6708 * substring-locations.h (format_warning_va): Formatting fix for
6709 ATTRIBUTE_GCC_DIAG.
6710 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
6711 argument.
6712 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
6713 * substring-locations.c: Include intl.h.
6714 (format_warning_va): Turned into small wrapper around
6715 format_warning_n_va, renamed to ...
6716 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
6717 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
6718 use ngettext.
6719 (format_warning_at_substring_n): New function.
6720 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
6721 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
6722 format_warning_at_substring with just a shorter name instead of
6723 const function pointer.
6724 (fmtwarn_n): New function.
6725 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
6726 appropriate, get rid of all the fmtstr temporaries, move conditionals
6727 with G_() wrapped string literals directly into fmtwarn arguments,
6728 cast dir.len to (int), formatting fixes.
6730 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
6732 * doc/invoke.texi: Remove "Cilk Plus" references.
6734 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6735 Richard Biener <rguenther@suse.de>
6737 PR ipa/84628
6738 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
6739 for error or warning attributes if CALL_FROM_THUNK_P is set.
6740 Formatting fixes.
6742 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6744 PR target/56540
6745 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
6746 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
6748 PR target/56540
6749 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
6750 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
6752 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
6753 instead of -1U in last predictors element's probability member.
6755 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
6757 PR ipa/83983
6758 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
6759 arguments if they are comparable.
6761 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6763 PR tree-optimization/84634
6764 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
6765 masks and masked_loop_p with a single loop_masks, making sure it's
6766 null for bb vectorization.
6768 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
6770 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
6771 (vect_analyze_data_ref_access): Use loop->safe_len rather than
6772 loop->force_vectorize to check whether there is no alias.
6774 2018-03-02 Jakub Jelinek <jakub@redhat.com>
6776 PR target/84614
6777 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
6778 prototypes.
6779 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
6780 comments.
6781 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
6782 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
6783 instead of a loop around prev_real_insn.
6784 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
6785 prev_real_insn.
6787 PR inline-asm/84625
6788 * config/i386/i386.c (ix86_print_operand): Use conditional
6789 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
6790 zero vector.
6792 2018-03-02 Richard Biener <rguenther@suse.de>
6794 PR tree-optimization/84427
6795 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
6796 (bitmap_set_subtract_values): Rewrite to handle multiple
6797 exprs per value.
6798 (clean): Likewise.
6799 (prune_clobbered_mems): Likewise.
6800 (phi_translate): Take edge instead of pred/phiblock.
6801 (phi_translate_1): Likewise.
6802 (phi_translate_set): Likewise. Insert all translated
6803 exprs for a value into the set, keeping possibly multiple
6804 expressions per value.
6805 (compute_antic_aux): Adjust for phi_translate changes.
6806 When intersecting union the expressions and prune those
6807 not in the final value set, keeping possibly multiple
6808 expressions per value. Do not use value-insertion
6809 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
6810 all expressions. Add verification that the value-sets
6811 only shrink during iteration.
6812 (compute_partial_antic_aux): Adjust for the phi_translate changes.
6813 (do_pre_regular_insertion): Likewise.
6814 (do_pre_partial_partial_insertion): Likewise.
6816 2018-03-02 Richard Biener <rguenther@suse.de>
6818 PR target/82005
6819 * config/darwin.c (saved_debug_info_level): New static global.
6820 (darwin_asm_lto_start): Disable debug info generation for LTO out.
6821 (darwin_asm_lto_end): Restore debug info generation settings.
6823 2018-03-01 Martin Liska <mliska@suse.cz>
6825 PR sanitizer/82484
6826 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
6827 volatile arguments.
6829 2018-03-01 Richard Biener <rguenther@suse.de>
6831 PR debug/84645
6832 * dwarf2out.c (gen_variable_die): Properly handle late VLA
6833 type annotation with LTO when debug was disabled at compile-time.
6835 2018-03-01 Matthew Fortune <mfortune@gmail.com>
6837 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
6838 XINT with INTVAL.
6839 (mips_final_postscan_insn): Likewise.
6841 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
6843 PR rtl-optimization/84528
6844 * alias.c (init_alias_target): Add commentary.
6845 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
6846 a unique base value if the frame pointer is not eliminated
6847 to the stack pointer.
6849 2018-03-01 Tom de Vries <tom@codesourcery.com>
6851 PR rtl-optimization/83327
6852 * lra-int.h (hard_regs_spilled_into): Declare.
6853 * lra.c (hard_regs_spilled_into): Define.
6854 (init_reg_info): Init hard_regs_spilled_into.
6855 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
6856 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
6857 (process_bb_lives): Handle hard_regs_spilled_into.
6858 (lra_create_live_ranges_1): Before doing liveness propagation, clear
6859 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
6861 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
6863 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
6864 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
6865 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
6866 * config/rs6000/aix72.h: New file.
6868 2018-02-28 Jakub Jelinek <jakub@redhat.com>
6870 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
6871 instead of warning_at with conditional singular and plural messages
6872 where possible.
6874 PR target/52991
6875 * stor-layout.c (update_alignment_for_field): For
6876 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
6877 && !DECL_PACKED (field), do the alignment update, just use
6878 only desired_align instead of MAX (type_align, desired_align)
6879 as the alignment.
6880 (place_field): Don't do known_align < desired_align handling
6881 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
6882 is non-NULL, instead do it after rli->prev_field handling and
6883 only if not within a bitfield word. For DECL_PACKED (field)
6884 use type_align of BITS_PER_UNIT.
6886 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
6888 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
6889 superfluous parentheses and trailing spaces.
6891 2018-02-28 Richard Biener <rguenther@suse.de>
6893 PR tree-optimization/84584
6894 * graphite-scop-detection.c (scop_detection::add_scop): Discard
6895 SCoPs with fake exit edge.
6897 2018-02-28 Martin Liska <mliska@suse.cz>
6899 PR testsuite/84597
6900 * timevar.c (timer::print): Fix format to properly print 100%
6901 values.
6903 2018-02-28 Richard Biener <rguenther@suse.de>
6905 PR middle-end/84607
6906 * genmatch.c (capture_info::walk_match): Do not mark
6907 captured expressions without operands as expr_p given
6908 they act more like predicates and should be subject to
6909 "lost tail" side-effect preserving.
6911 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
6913 PR rtl-optimization/81611
6914 * auto-inc-dec.c (attempt_change): Move dead note from
6915 mem_insn if it's the next use of regno
6916 (find_address): Take address use of reg holding
6917 non-incremented value. Add parm to limit search to the named
6918 reg only.
6919 (merge_in_block): Attempt to use a mem insn that is the next
6920 use of the original regno.
6922 2018-02-27 Martin Sebor <msebor@redhat.com>
6924 PR c++/83871
6925 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
6926 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
6928 2018-02-27 Martin Sebor <msebor@redhat.com>
6930 PR translation/84207
6931 * diagnostic-core.h (warning_n, error_n, inform_n): Change
6932 n argument to unsigned HOST_WIDE_INT.
6933 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
6934 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
6935 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
6936 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
6938 2018-02-27 Richard Biener <rguenther@suse.de>
6940 PR tree-optimization/84512
6941 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
6942 Do not use the estimate returned from record_stmt_cost for
6943 the scalar iteration cost but sum properly using add_stmt_cost.
6945 2018-02-27 Richard Biener <rguenther@suse.de>
6947 PR tree-optimization/84466
6948 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6949 Adjust last change to less strictly validate use operands.
6951 2018-02-27 Martin Liska <mliska@suse.cz>
6953 PR gcov-profile/84548
6954 * gcov.c (process_file): Allow partial overlap and consider it
6955 also as group functions.
6956 (output_lines): Properly calculate range of lines for a group.
6958 2018-02-27 Martin Liska <mliska@suse.cz>
6960 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
6961 'ggc' suffixes. Change first column width.
6962 (timer::print): Fix formatting of the column.
6964 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
6966 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
6967 preserve inline entry blocks for the sake of debug inline
6968 entry point markers alone.
6969 (remove_unused_locals): Suggest in comments a better place to
6970 force the preservation of inline entry blocks that are
6971 otherwise unused, but do not preserve them.
6973 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6975 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
6977 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
6979 PR target/84039
6980 * config/i386/constraints.md (Bs): Replace
6981 ix86_indirect_branch_register with
6982 TARGET_INDIRECT_BRANCH_REGISTER.
6983 (Bw): Likewise.
6984 * config/i386/i386.md (indirect_jump): Likewise.
6985 (tablejump): Likewise.
6986 (*sibcall_memory): Likewise.
6987 (*sibcall_value_memory): Likewise.
6988 Peepholes of indirect call and jump via memory: Likewise.
6989 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
6990 (*sibcall_value_GOT_32): Likewise.
6991 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6992 (GOT_memory_operand): Likewise.
6993 (call_insn_operand): Likewise.
6994 (sibcall_insn_operand): Likewise.
6995 (GOT32_symbol_operand): Likewise.
6996 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
6998 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7000 PR rtl-optimization/83496
7001 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
7002 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
7003 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
7004 redundant insn, if any.
7005 (relax_delay_slots): Likewise.
7006 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
7008 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
7010 PR tree-optimization/83965
7011 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
7012 that grouped statements are part of a reduction chain. Return
7013 true if the statement is not marked as a reduction itself but
7014 is part of a group.
7015 (vect_recog_dot_prod_pattern): Don't check whether the statement
7016 is part of a group here.
7017 (vect_recog_sad_pattern): Likewise.
7018 (vect_recog_widen_sum_pattern): Likewise.
7020 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7022 PR debug/84545
7023 * final.c (rest_of_clean_state): Also look for calls inside sequences.
7025 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7027 PR target/84530
7028 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
7029 the bool argument.
7030 (ix86_output_indirect_function_return): New prototype.
7031 (ix86_split_simple_return_pop_internal): Likewise.
7032 * config/i386/i386.c (indirect_return_via_cx): New.
7033 (indirect_return_via_cx_bnd): Likewise.
7034 (indirect_thunk_name): Handle return va CX_REG.
7035 (output_indirect_thunk_function): Create alias for
7036 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
7037 (ix86_output_indirect_jmp): Remove the bool argument.
7038 (ix86_output_indirect_function_return): New function.
7039 (ix86_split_simple_return_pop_internal): Likewise.
7040 * config/i386/i386.md (*indirect_jump): Don't pass false
7041 to ix86_output_indirect_jmp.
7042 (*tablejump_1): Likewise.
7043 (simple_return_pop_internal): Change it to define_insn_and_split.
7044 Call ix86_split_simple_return_pop_internal to split it for
7045 -mfunction-return=.
7046 (simple_return_indirect_internal): Call
7047 ix86_output_indirect_function_return instead of
7048 ix86_output_indirect_jmp.
7050 2018-02-26 Jakub Jelinek <jakub@redhat.com>
7052 PR bootstrap/84405
7053 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
7054 memset and value initialization afterwards.
7056 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
7058 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
7060 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7062 PR target/84521
7063 * common/config/aarch64/aarch64-common.c
7064 (aarch_option_optimization_table[]): Switch
7065 off fomit-frame-pointer
7067 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7068 Chung-Ju Wu <jasonwucj@gmail.com>
7070 * config/nds32/nds32-multiple.md (load_multiple): Disallow
7071 volatile memory.
7072 (store_multiple): Ditto.
7074 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7076 * config.gcc: Add --with-cpu support for nds32 target.
7077 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
7078 * config/nds32/nds32.opt: Add -mcpu= option.
7080 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
7082 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
7083 isel=yes): Warn for these deprecated options.
7085 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
7087 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
7088 ISA_2_5_MASKS_EMBEDDED.
7090 2018-02-23 Jakub Jelinek <jakub@redhat.com>
7092 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
7093 p->max as pointers rather than using iterative_hash_expr.
7095 2018-02-23 Carl Love <cel@us.ibm.com>
7097 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
7098 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
7099 BU_P8V_OVERLOAD_2.
7100 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
7101 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
7102 P8V_BUILTIN_VEC_VUNSIGNED2.
7104 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
7106 PR target/81572
7107 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
7108 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
7109 LRA_UNKNOWN_ALT.
7110 * lra-constraints.c (curr_insn_transform): Set up
7111 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
7112 LRA_UNKNOWN_ALT.
7113 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
7114 * lra-eliminations.c (spill_pseudos): Ditto.
7115 (process_insn_for_elimination): Ditto.
7116 * lra-lives.c (reg_early_clobber_p): Use the new macros.
7117 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
7118 LRA_NON_CLOBBERED_ALT.
7120 2018-02-22 Martin Sebor <msebor@redhat.com>
7122 PR tree-optimization/84480
7123 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
7124 to maybe_diag_stxncpy_trunc. Call it.
7125 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
7126 from gimple_fold_builtin_strcpy. Print inlining stack.
7127 (handle_builtin_stxncpy): Print inlining stack.
7128 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
7130 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
7132 PR target/84176
7133 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
7134 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
7135 and -fcheck-pointer-bounds are used together.
7136 (indirect_thunk_prefix): New enum.
7137 (indirect_thunk_need_prefix): New function.
7138 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
7139 "_nt" instead of "_bnd" for NOTRACK prefix.
7140 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
7141 (output_indirect_thunk_function): Likewise.
7142 (): Likewise.
7143 (ix86_code_end): Update output_indirect_thunk_function calls.
7144 (ix86_output_indirect_branch_via_reg): Replace
7145 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
7146 (ix86_output_indirect_branch_via_push): Likewise.
7147 (ix86_output_function_return): Likewise.
7148 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
7149 incompatible with -fcf-protection=branch and
7150 -fcheck-pointer-bounds.
7152 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7154 PR target/83335
7155 * config/aarch64/aarch64.c (aarch64_print_address_internal):
7156 Change gcc_assert call to output_operand_lossage.
7158 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7160 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
7162 2018-02-22 DJ Delorie <dj@redhat.com>
7163 Sebastian Perta <sebastian.perta@renesas.com>
7164 Oleg Endo <olegendo@gcc.gnu.org>
7166 * config/rx/rx.c (rx_rtx_costs): New function.
7167 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
7169 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
7171 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
7173 2018-02-22 Martin Liska <mliska@suse.cz>
7175 PR driver/83193
7176 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
7177 Add "native" as a possible value.
7179 2018-02-22 Martin Liska <mliska@suse.cz>
7181 PR driver/83193
7182 * config/i386/i386.c (ix86_option_override_internal):
7183 Add "native" as a possible value for -march and -mtune.
7185 2018-02-22 Jakub Jelinek <jakub@redhat.com>
7187 PR target/84502
7188 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
7189 to all type variants.
7191 PR tree-optimization/84503
7192 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
7193 width as info->bitpos + info->bitsize - start.
7194 (merged_store_group::merge_overlapping): Simplify width computation.
7195 (check_no_overlap): New function.
7196 (imm_store_chain_info::try_coalesce_bswap): Compute expected
7197 start + width and last_order of the group, fail if check_no_overlap
7198 fails.
7199 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
7200 to group if check_no_overlap fails.
7202 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7204 * config/rs6000/altivec.md: Delete contraint arguments to
7205 define_expand, define_split, and define_peephole2, and in
7206 define_insn_and_split if always unused.
7207 * config/rs6000/darwin.md: Ditto.
7208 * config/rs6000/dfp.md: Ditto.
7209 * config/rs6000/rs6000.md: Ditto.
7210 * config/rs6000/sync.md: Ditto.
7211 * config/rs6000/vector.md: Ditto.
7212 * config/rs6000/vsx.md: Ditto.
7214 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7216 * config/rs6000/altivec.md: Write output control strings as braced
7217 blocks instead of double-quoted strings.
7218 * config/rs6000/darwin.md: Ditto.
7219 * config/rs6000/rs6000.md: Ditto.
7220 * config/rs6000/vector.md: Ditto.
7221 * config/rs6000/vsx.md: Ditto.
7223 2018-02-21 Jason Merrill <jason@redhat.com>
7225 PR c++/84314 - ICE with templates and fastcall attribute.
7226 * attribs.c (build_type_attribute_qual_variant): Remove assert.
7228 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7230 * ipa-cp.c (determine_versionability): Fix comment typos.
7232 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7234 PR c/84229
7235 * ipa-cp.c (determine_versionability): Do not version functions caling
7236 va_arg_pack.
7238 2018-02-21 Martin Liska <mliska@suse.cz>
7240 PR driver/83193
7241 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
7242 Add "native" as a possible value.
7243 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
7244 the macro when native cpu detection is available.
7246 2018-02-21 Martin Liska <mliska@suse.cz>
7248 PR driver/83193
7249 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
7250 Add "native" as a possible value.
7251 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
7252 when native cpu detection is available.
7254 2018-02-21 Jakub Jelinek <jakub@redhat.com>
7255 Martin Sebor <msebor@redhat.com>
7257 PR tree-optimization/84478
7258 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
7259 false.
7260 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
7261 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
7262 support which is conservatively correct, for 2 only stay conservative
7263 for maxlen. Formatting and comment capitalization fixes. Add STRICT
7264 argument to the 2 argument get_range_strlen, adjust 6 arg
7265 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
7266 false.
7267 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
7268 (gimple_fold_builtin_strlen): Pass true as last argument to
7269 get_range_strlen.
7271 2018-02-20 Martin Sebor <msebor@redhat.com>
7273 PR middle-end/84095
7274 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
7275 (builtin_memref::set_base_and_offset): Same. Handle inner references.
7276 (builtin_memref::builtin_memref): Factor out parts into
7277 set_base_and_offset and call it.
7279 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7281 PR middle-end/84406
7282 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
7283 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
7284 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
7285 search at the associated MODE_INT.
7287 2018-02-20 Jeff Law <law@redhat.com>
7289 PR middle-end/82123
7290 PR tree-optimization/81592
7291 PR middle-end/79257
7292 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
7293 for range data rather than using global data.
7294 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
7295 range data rather than using global data.
7296 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
7297 pass it to children as needed.
7298 (struct directive::fmtresult): Similarly.
7299 (struct directive::set_width): Similarly.
7300 (struct directive::set_precision): Similarly.
7301 (format_integer, format_directive, parse_directive): Similarly.
7302 (format_none): Accept unnamed vr_values parameter.
7303 (format_percent, format_floating, format_character): Similarly.
7304 (format_string, format_plain): Similarly.
7305 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
7306 the EVRP range analyzer for range data rather than using global data.
7307 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
7308 gimple-ssa-evrp-analyze.h
7309 (class sprintf_dom_walker): Add after_dom_children member function.
7310 Add evrp_range_analyzer member.
7311 (sprintf_dom_walker::before_dom_children): Call into the EVRP
7312 range analyzer as needed.
7313 (sprintf_dom_walker::after_dom_children): New member function.
7314 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
7315 if not optimizing.
7316 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
7317 (evrp_range_analyzer::pop_to_marker): Likewise.
7319 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7321 PR tree-optimization/84419
7322 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
7323 with the required type if its current type is compatible but
7324 different.
7326 2018-02-20 Jakub Jelinek <jakub@redhat.com>
7328 PR middle-end/82004
7329 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
7330 after vectorization.
7332 2018-02-20 Martin Liska <mliska@suse.cz>
7334 PR driver/83193
7335 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
7336 possible values if we don't have a hint.
7338 2018-02-20 Martin Liska <mliska@suse.cz>
7340 PR c/84310
7341 PR target/79747
7342 * final.c (shorten_branches): Build align_tab array with one
7343 more element.
7344 * opts.c (finish_options): Add alignment option limit check.
7345 (MAX_CODE_ALIGN): Likewise.
7346 (MAX_CODE_ALIGN_VALUE): Likewise.
7347 * doc/invoke.texi: Document maximum allowed option value for
7348 all -falign-* options.
7350 2018-02-19 Jakub Jelinek <jakub@redhat.com>
7352 PR target/84146
7353 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
7354 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
7355 * var-tracking.c (emit_note_insn_var_location): Remove all references
7356 to NOTE_INSN_CALL_ARG_LOCATION.
7357 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
7358 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
7359 Use copy_rtx_if_shared.
7360 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
7361 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
7362 (dwarf2out_var_location): Remove handling of
7363 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
7364 on call_insn.
7365 * final.c (final_scan_insn): Remove all references to
7366 NOTE_INSN_CALL_ARG_LOCATION.
7367 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
7368 before dumping final insns.
7369 * except.c (emit_note_eh_region_end): Remove all references to
7370 NOTE_INSN_CALL_ARG_LOCATION.
7371 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
7372 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
7373 * config/arc/arc.c (hwloop_optimize): Likewise.
7374 * config/arm/arm.c (create_fix_barrier): Likewise.
7375 * config/s390/s390.c (s390_chunkify_start): Likewise.
7376 * config/sh/sh.c (find_barrier): Likewise.
7377 * config/i386/i386.c (rest_of_insert_endbranch,
7378 ix86_seh_fixup_eh_fallthru): Likewise.
7379 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
7380 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
7381 * config/frv/frv.c (frv_function_prologue): Likewise.
7382 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
7383 reg note.
7384 (note_outside_basic_block_p): Remove all references to
7385 NOTE_INSN_CALL_ARG_LOCATION.
7386 * gengtype.c (adjust_field_rtx_def): Likewise.
7387 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
7388 Likewise.
7389 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
7390 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
7392 PR c++/84444
7393 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
7394 is ADDR_EXPR.
7396 PR tree-optimization/84452
7397 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
7398 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
7399 is NULL.
7401 2018-02-19 Martin Liska <mliska@suse.cz>
7403 PR sanitizer/82183
7404 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
7406 2018-02-19 Martin Liska <mliska@suse.cz>
7407 Richard Sandiford <richard.sandiford@linaro.org>
7409 PR tree-optimization/82491
7410 * gimple-fold.c (get_base_constructor): Make earlier bail out
7411 to prevent ubsan.
7413 2018-02-19 Carl Love <cel@us.ibm.com>
7415 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
7416 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
7417 BU_P8V_OVERLOAD_1.
7418 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
7419 P8V_BUILTIN_VEC_NEG.
7421 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
7423 * config/rl78/rl78.md (movdf): New define expand.
7425 2018-02-19 Martin Liska <mliska@suse.cz>
7427 PR other/80589
7428 * doc/invoke.texi: Fix typo.
7429 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
7431 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
7433 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
7434 handle rs6000_single_float and rs6000_double_float specially for
7435 e500 family CPUs.
7437 2018-02-16 Jeff Law <law@redhat.com>
7439 * config/rx/rx.c (add_pop_cfi_notes): New function.;
7440 (pop_regs): Use it.
7442 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7444 PR ipa/84425
7445 * ipa-inline.c (inline_small_functions): Fix a typo.
7447 2018-02-16 Nathan Sidwell <nathan@acm.org>
7449 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
7451 2018-02-16 Carl Love <cel@us.ibm.com>
7453 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
7454 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
7455 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
7456 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
7457 expansion to P8V_BUILTIN_VEC_FLOAT2.
7459 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
7461 PR rtl-optimization/70023
7462 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
7463 src_regno into account.
7465 2018-02-16 Carl Love <cel@us.ibm.com>
7467 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
7468 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
7469 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
7470 * config/rs6000/rs6000.c: Remove case statements for
7471 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
7472 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
7473 and P9V_BUILTIN_VEC_VINSERT4B.
7474 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
7475 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
7476 * config/rs6000/vsx.md:
7477 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
7478 vec_insert4b.
7480 2018-02-16 Carl Love <cel@us.ibm.com>
7482 * config/rs6000/altivec.h: Add builtin names vec_extract4b
7483 vec_insert4b.
7484 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
7485 definitions.
7486 * config/rs6000/rs6000-c.c: Add the definitions for
7487 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
7488 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
7489 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
7490 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
7491 definition for insert4b and define insn *insert3b_internal.
7492 * doc/extend.texi: Add documentation for vec_extract4b.
7494 2018-02-16 Nathan Sidwell <nathan@acm.org>
7496 * doc/extend.texi (Backwards Compatibility): Mention friend
7497 injection. Note for-scope is deprecated.
7498 * doc/invoke.texi (-ffriend-injection): Deprecate.
7500 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
7502 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
7503 that moved to I2, also allow destinations that are a paradoxical
7504 subreg (instead of a normal reg).
7506 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
7508 PR target/83831
7509 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
7510 to QImode.
7512 2018-02-16 Richard Biener <rguenther@suse.de>
7514 PR tree-optimization/84037
7515 PR tree-optimization/84016
7516 PR target/82862
7517 * config/i386/i386.c (ix86_builtin_vectorization_cost):
7518 Adjust vec_construct for the fact we need additional higher latency
7519 128bit inserts for AVX256 and AVX512 vector builds.
7520 (ix86_add_stmt_cost): Scale vector construction cost for
7521 elementwise loads.
7523 2018-02-16 Richard Biener <rguenther@suse.de>
7525 PR tree-optimization/84417
7526 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
7527 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
7528 (non_rewritable_lvalue_p): Likewise, use poly-ints.
7530 2018-02-16 Martin Liska <mliska@suse.cz>
7532 PR sanitizer/84307
7533 * internal-fn.def (ASAN_CHECK): Set proper flags.
7534 (ASAN_MARK): Likewise.
7536 2018-02-16 Julia Koval <julia.koval@intel.com>
7538 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
7539 from PTA_CANNONLAKE.
7541 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7543 PR target/84272
7544 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
7545 Use ++iter rather than iter++ for std::list iterators.
7546 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
7547 defer deleting them until all nodes in the forest are processed. Do
7548 free even leaf nodes. Change to_process into auto_vec.
7550 PR bootstrap/84405
7551 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
7552 * vec.h (vec_default_construct): Use memset instead of placement new
7553 if BROKEN_VALUE_INITIALIZATION is defined.
7554 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
7555 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
7556 is defined.
7558 PR rtl-optimization/83723
7559 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
7560 * lra.c (lra_substitute_pseudo): Likewise. If true, use
7561 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
7562 recursive calls.
7563 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
7564 callers.
7565 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
7567 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
7569 PR rtl-optimization/81443
7570 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
7571 from inner REGs to paradoxical SUBREGs.
7573 2018-02-16 Richard Biener <rguenther@suse.de>
7575 PR tree-optimization/84399
7576 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7577 For operands we can analyze at their definition make sure we can
7578 analyze them at each use as well.
7580 2018-02-16 Richard Biener <rguenther@suse.de>
7582 PR tree-optimization/84190
7583 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
7584 volatile accesses if the decl isn't volatile.
7586 2018-02-15 Jason Merrill <jason@redhat.com>
7588 PR c++/84314 - ICE with templates and fastcall attribute.
7589 * attribs.c (build_type_attribute_qual_variant): Don't clobber
7590 TYPE_CANONICAL on an existing type.
7592 2018-02-15 Jakub Jelinek <jakub@redhat.com>
7594 PR tree-optimization/84383
7595 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
7596 dstoff nor call operand_equal_p if dstbase is NULL.
7598 PR tree-optimization/84334
7599 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
7600 also a CONSTANT_CLASS_P, punt.
7602 2018-02-14 Jim Wilson <jimw@sifive.com>
7604 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
7605 first SMALL_OPERAND check. New local min_second_step. Move assert
7606 to where locals are set. Add TARGET_RVC support.
7607 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
7609 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
7611 * doc/invoke.texi: Correct -Wformat-overflow code sample.
7613 2018-02-14 Martin Sebor <msebor@redhat.com>
7615 PR tree-optimization/83698
7616 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
7617 arrays constrain the offset range to their bounds.
7618 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
7619 (builtin_access::overlap): Avoid setting the size of overlap if it's
7620 already been set.
7621 (maybe_diag_overlap): Also consider arrays when deciding what values
7622 of offsets to include in diagnostics.
7624 2018-02-14 Martin Sebor <msebor@redhat.com>
7626 PR c/84108
7627 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
7628 that correspond to the kind of a declaration.
7630 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
7632 PR target/83984
7633 * config/pa/pa.md: Load address of PIC label using the linkage table
7634 if the label is nonlocal.
7636 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
7638 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
7639 warning message if user requests -maltivec=be.
7640 * doc/invoke.texi: Document deprecation of -maltivec=be.
7642 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
7644 PR target/84220
7645 * config/rs6000/rs6000-c.c: Update definitions for
7646 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
7647 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
7649 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7651 PR target/84239
7652 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
7653 add _get_ssp intrinsics. Remove argument from
7654 __builtin_ia32_rdssp[d|q].
7655 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
7656 * config/i386/i386-builtin.def: Remove argument from
7657 __builtin_ia32_rdssp[d|q].
7658 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
7659 ix86_expand_special_args_builtin for _rdssp[d|q].
7660 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
7661 Clear register before usage.
7662 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
7663 Add documentation for new _get_ssp and _inc_ssp intrinsics.
7665 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
7667 PR tree-optimization/84357
7668 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
7669 operand 1 of an ARRAY_REF too.
7671 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
7673 PR target/83831
7674 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
7675 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7676 declarations.
7677 (set_of_reg): New struct.
7678 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
7679 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
7680 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
7681 functions.
7682 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
7683 Split into bitclr, bitset, bitinvert patterns if appropriate.
7684 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
7685 use rx_fuse_in_memory_bitop.
7686 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
7687 to named insn, correct maximum insn length.
7689 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
7691 PR target/79242
7692 * machmode.def: Define a complex mode for PARTIAL_INT.
7693 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
7694 MODE_PARTIAL_INT.
7695 * doc/rtl.texi: Document CSPImode.
7696 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
7697 handling.
7698 (msp430_hard_regno_nregs_with_padding): Likewise.
7700 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
7702 PR target/84279
7703 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
7705 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
7707 PR rtl-optimization/84169
7708 * combine.c (try_combine): New variable split_i2i3. Set it to true if
7709 we generated a parallel as new i3 and we split that to new i2 and i3
7710 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
7711 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
7712 those to i2, not i1. Partially rewrite this scan code.
7714 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7716 PR c/82210
7717 * stor-layout.c (place_field): For variable length fields, adjust
7718 offset_align afterwards not just based on the field's alignment,
7719 but also on the size.
7721 PR middle-end/84309
7722 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
7723 of exps and logs in the use_exp2 case.
7725 2018-02-13 Jeff Law <law@redhat.com>
7727 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
7728 entry for "vector".
7730 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
7731 ARGS as unused.
7733 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
7735 PR debug/84342
7736 PR debug/84319
7737 * common.opt (gas-loc-support, gas-locview-support): New.
7738 (ginline-points, ginternal-reset-location-views): New.
7739 * doc/invoke.texi: Document them. Use @itemx where intended.
7740 (gvariable-location-views): Adjust.
7741 * target.def (reset_location_view): New.
7742 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
7743 (TARGET_RESET_LOCATION_VIEW): New.
7744 * doc/tm.texi: Rebuilt.
7745 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
7746 (dwarf2out_default_as_locview_support): New.
7747 (output_asm_line_debug_info): Use option variables.
7748 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
7749 (output_loc_list): Likewise.
7750 (add_high_low_attributes): Check option variables.
7751 Don't output entry view attribute in strict mode.
7752 (gen_inlined_subroutine_die): Check option variables.
7753 (dwarf2out_inline_entry): Likewise.
7754 (init_sections_and_labels): Likewise.
7755 (dwarf2out_early_finish): Likewise.
7756 (maybe_reset_location_view): New, from...
7757 (dwarf2out_var_location): ... here. Call it.
7758 * debug.h (dwarf2out_default_as_loc_support): Declare.
7759 (dwarf2out_default_as_locview_support): Declare.
7760 * hooks.c (hook_int_rtx_insn_0): New.
7761 * hooks.h (hook_int_rtx_insn_0): Declare.
7762 * toplev.c (process_options): Take -gas-loc-support and
7763 -gas-locview-support from dwarf2out. Enable
7764 -gvariable-location-views by default only with locview
7765 assembler support. Enable -ginternal-reset-location-views by
7766 default only if the target defines the corresponding hook.
7767 Enable -ginline-points by default if location views are
7768 enabled; force it disabled if statement frontiers are
7769 disabled.
7770 * tree-inline.c (expand_call_inline): Check option variables.
7771 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
7773 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
7775 PR tree-optimization/84321
7776 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
7777 handling. Also check whether the anti-range contains any values
7778 that satisfy the mask; switch to a VR_RANGE if not.
7780 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
7782 PR sanitizer/84340
7783 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
7785 2018-02-13 Martin Jambor <mjambor@suse.cz>
7787 PR c++/83990
7788 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
7789 of call statements, also set location of a load to a temporary.
7791 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
7793 * config/rl78/rl78.c (add_vector_labels): New function.
7794 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
7795 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
7796 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
7797 which checks that no arguments are passed.
7798 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
7799 * doc/extend.texi: Documentation for the new attribute.
7801 2018-02-13 Andreas Schwab <schwab@suse.de>
7803 * config/riscv/linux.h (CPP_SPEC): Define.
7805 2018-02-13 Jakub Jelinek <jakub@redhat.com>
7807 PR target/84335
7808 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
7809 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
7810 OPTION_MASK_ISA_AES as first argument to def_builtin_const
7811 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
7812 instead of OPTION_MASK_ISA_PCLMUL as first argument to
7813 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
7814 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
7815 temporarily for AES and PCLMUL builtins.
7817 PR tree-optimization/84339
7818 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
7819 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
7820 Formatting fixes.
7822 PR middle-end/84309
7823 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
7824 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
7825 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
7826 inline function.
7827 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
7828 inline function.
7829 * omp-simd-clone.h: New file.
7830 * omp-simd-clone.c: Include omp-simd-clone.h.
7831 (expand_simd_clones): No longer static.
7832 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
7833 cgraph.h and omp-simd-clone.h.
7834 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
7835 (vect_recog_widen_shift_pattern): Formatting fix.
7836 (vect_pattern_recog_1): Don't check optab for calls.
7838 PR target/84336
7839 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
7840 operands[2] into a REG before using gen_lowpart on it.
7842 2018-02-12 Jeff Law <law@redhat.com>
7844 PR target/83760
7845 * config/sh/sh.c (find_barrier): Consider a sibling call
7846 a barrier as well.
7848 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
7849 successfully back substituting a reg.
7851 2018-02-12 Richard Biener <rguenther@suse.de>
7853 PR tree-optimization/84037
7854 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
7855 parameter, move visited init to caller.
7856 (vect_slp_analyze_operations): Separate cost from validity
7857 check, initialize visited once for all instances.
7858 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
7859 for all instances.
7860 * tree-vect-stmts.c (vect_model_simple_cost): Make early
7861 out an assert.
7862 (vect_model_promotion_demotion_cost): Likewise.
7863 (vectorizable_bswap): Guard cost modeling with !slp_node
7864 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
7865 SLP stmts.
7866 (vectorizable_call): Likewise.
7867 (vectorizable_conversion): Likewise.
7868 (vectorizable_assignment): Likewise.
7869 (vectorizable_shift): Likewise.
7870 (vectorizable_operation): Likewise.
7871 (vectorizable_store): Likewise.
7872 (vectorizable_load): Likewise.
7873 (vectorizable_condition): Likewise.
7874 (vectorizable_comparison): Likewise.
7876 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
7878 PR sanitizer/84307
7879 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
7880 (ASAN_MARK): Fix fnspec to account for return value, change pointer
7881 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
7883 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
7885 PR middle-end/83665
7886 * params.def (inline-min-speedup): Increase from 8 to 15.
7887 (max-inline-insns-auto): Decrease from 40 to 30.
7888 * ipa-split.c (consider_split): Add some buffer for function to
7889 be considered inlining candidate.
7890 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
7891 default values.
7893 2018-02-12 Richard Biener <rguenther@suse.de>
7895 PR tree-optimization/84037
7896 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
7897 matched stmts if we cannot swap the non-matched ones.
7899 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7901 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
7902 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
7903 _mm_maskz_scalef_round_ss): New intrinsics.
7904 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
7905 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
7906 __builtin_ia32_scalefss_round): Remove.
7907 (__builtin_ia32_scalefsd_mask_round,
7908 __builtin_ia32_scalefss_mask_round): New intrinsics.
7909 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
7910 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
7911 ((match_operand:VF_128 2 "<round_nimm_predicate>"
7912 "<round_constraint>")): Changed to ...
7913 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
7914 "<round_scalar_constraint>")): ... this.
7915 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
7916 %0, %1, %2<round_op3>}"): Changed to ...
7917 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
7918 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
7919 %2<round_scalar_mask_op3>}"): ... this.
7920 * config/i386/subst.md (round_scalar_nimm_predicate): New.
7922 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
7924 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
7925 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
7926 (_mm_maskz_sqrt_round_ss): New intrinsics.
7927 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
7928 (__builtin_ia32_sqrtsd_mask_round)
7929 (__builtin_ia32_sqrtss_mask_round): New builtins.
7930 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
7931 (__builtin_ia32_sqrtss_round): Remove.
7932 (__builtin_ia32_sqrtsd_mask_round)
7933 (__builtin_ia32_sqrtss_mask_round): New builtins.
7934 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
7935 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
7936 ((match_operand:VF_128 1 "vector_operand"
7937 "xBm,<round_constraint>")): Changed to ...
7938 ((match_operand:VF_128 1 "vector_operand"
7939 "xBm,<round_scalar_constraint>")): ... this.
7940 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
7941 %0, %2, %<iptr>1<round_op3>}): Changed to ...
7942 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
7943 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
7944 %<iptr>1<round_scalar_mask_op3>}): ... this.
7945 ((set_attr "prefix" "<round_prefix>")): Changed to ...
7946 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
7948 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
7950 PR target/84266
7951 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
7952 Cast vec_cmpeq result to correct type.
7953 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
7954 Cast vec_cmpgt result to correct type.
7956 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
7958 * final.c (final_scan_insn_1): Renamed from...
7959 (final_scan_insn): ... this. New wrapper, to recover
7960 seen from the outermost call in recursive ones.
7961 * config/sparc/sparc.c (output_return): Drop seen from call.
7962 (output_sibcall): Likewise.
7963 * config/visium/visium.c (output_branch): Likewise.
7965 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
7967 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
7968 function label.
7970 2018-02-10 Alan Modra <amodra@gmail.com>
7972 PR target/84300
7973 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
7974 Specify LR as an input.
7976 2018-02-10 Jakub Jelinek <jakub@redhat.com>
7978 PR sanitizer/83987
7979 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
7980 remove_member_access_dummy_vars): New functions.
7981 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
7982 lower_omp_1, execute_lower_omp): Use them.
7984 PR rtl-optimization/84308
7985 * shrink-wrap.c (spread_components): Release todo vector.
7987 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
7989 PR rtl-optimization/57193
7990 * ira-color.c (struct allocno_color_data): Add member
7991 conflict_allocno_hard_prefs.
7992 (update_conflict_allocno_hard_prefs): New.
7993 (bucket_allocno_compare_func): Add a preference based on
7994 conflict_allocno_hard_prefs.
7995 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
7996 (color_allocnos): Remove a dead code. Initiate
7997 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
7999 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8001 PR target/84226
8002 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
8003 constraint from =wa to wa. Avoid a subreg on the output operand,
8004 instead use a pseudo and subreg it in a move.
8005 (p9_xxbrd_<mode>): Changed to ...
8006 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
8007 (p9_xxbrd_v2df): New expander.
8008 (p9_xxbrw_<mode>): Changed to ...
8009 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
8010 (p9_xxbrw_v4sf): New expander.
8012 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8014 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
8016 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
8018 PR target/83926
8019 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
8020 multiply in 32-bit mode.
8021 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
8022 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
8023 mode.
8025 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8027 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
8028 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
8029 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
8030 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
8032 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
8034 PR lto/84213
8035 * dwarf2out.c (is_trivial_indirect_ref): New function.
8036 (dwarf2out_late_global_decl): Do not generate a location
8037 attribute for variables that have a non-trivial DECL_VALUE_EXPR
8038 and that are not defined in the current unit.
8040 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8042 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
8043 instead of a libcall for UNORDERED.
8045 2018-02-09 Tamar Christina <tamar.christina@arm.com>
8047 PR target/82641
8048 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
8049 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
8051 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8053 PR target/PR84295
8054 * config/s390/s390.c (s390_set_current_function): Invoke
8055 s390_indirect_branch_settings also if fndecl didn't change.
8057 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8059 * config/rs6000/rs6000.md (blockage): Set length to zero.
8061 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8063 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
8065 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8067 PR sanitizer/84285
8068 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
8069 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
8070 -static-lib*san.
8072 PR debug/84252
8073 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
8074 PARALLEL incoming that failed vt_get_decl_and_offset check.
8076 PR middle-end/84237
8077 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
8078 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
8079 TREE_READONLY bit.
8080 (get_variable_section): For decls in named .bss* sections pass true as
8081 second argument to bss_initializer_p.
8083 2018-02-09 Marek Polacek <polacek@redhat.com>
8084 Jakub Jelinek <jakub@redhat.com>
8086 PR c++/83659
8087 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
8088 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
8089 Sync some changes from cxx_fold_indirect_ref.
8091 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8093 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
8094 markers.
8095 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
8096 (BLOCK_INLINE_ENTRY_LABEL): New.
8097 (dwarf2out_var_location): Disregard inline entry markers.
8098 (inline_entry_data): New struct.
8099 (inline_entry_data_hasher): New hashtable type.
8100 (inline_entry_data_hasher::hash): New.
8101 (inline_entry_data_hasher::equal): New.
8102 (inline_entry_data_table): New variable.
8103 (add_high_low_attributes): Add DW_AT_entry_pc and
8104 DW_AT_GNU_entry_view attributes if a pending entry is found
8105 in inline_entry_data_table. Add old entry_pc attribute only
8106 if debug nonbinding markers are disabled.
8107 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
8108 markers are enabled.
8109 (block_within_block_p, dwarf2out_inline_entry): New.
8110 (dwarf2out_finish): Check that no entries remained in
8111 inline_entry_data_table.
8112 * final.c (reemit_insn_block_notes): Handle inline entry notes.
8113 (final_scan_insn, notice_source_line): Likewise.
8114 (rest_of_clean_state): Skip inline entry markers.
8115 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
8116 markers.
8117 * gimple.c (gimple_build_debug_inline_entry): New.
8118 * gimple.h (enum gimple_debug_subcode): Add
8119 GIMPLE_DEBUG_INLINE_ENTRY.
8120 (gimple_build_debug_inline_entry): Declare.
8121 (gimple_debug_inline_entry_p): New.
8122 (gimple_debug_nonbind_marker_p): Adjust.
8123 * insn-notes.def (INLINE_ENTRY): New.
8124 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
8125 inline entry marker notes.
8126 (print_insn): Likewise.
8127 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
8128 (INSN_DEBUG_MARKER_KIND): Likewise.
8129 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
8130 * tree-inline.c (expand_call_inline): Build and insert
8131 debug_inline_entry stmt.
8132 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
8133 inline entry blocks early, if nonbind markers are enabled.
8134 (dump_scope_block): Dump fragment info.
8135 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
8136 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
8137 (gimple_build_debug_inline_entry): New.
8138 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
8139 Enable/disable inline entry points too.
8140 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
8141 (DEBUG_INSN): Describe inline entry markers.
8143 * common.opt (gvariable-location-views): New.
8144 (gvariable-location-views=incompat5): New.
8145 * config.in: Rebuilt.
8146 * configure: Rebuilt.
8147 * configure.ac: Test assembler for view support.
8148 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
8149 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
8150 * dwarf2out.c (var_loc_view): New typedef.
8151 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
8152 (dwarf2out_locviews_in_attribute): New.
8153 (dwarf2out_locviews_in_loclist): New.
8154 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
8155 (enum dw_line_info_opcode): Add LI_adv_address.
8156 (struct dw_line_info_table): Add view.
8157 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
8158 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
8159 (zero_view_p): New variable.
8160 (ZERO_VIEW_P): New macro.
8161 (output_asm_line_debug_info): New.
8162 (struct var_loc_node): Add view.
8163 (add_AT_view_list, AT_loc_list): New.
8164 (add_var_loc_to_decl): Add view param. Test it against last.
8165 (new_loc_list): Add view params. Record them.
8166 (AT_loc_list_ptr): Handle loc and view lists.
8167 (view_list_to_loc_list_val_node): New.
8168 (print_dw_val): Handle dw_val_class_view_list.
8169 (size_of_die): Likewise.
8170 (value_format): Likewise.
8171 (loc_list_has_views): New.
8172 (gen_llsym): Set vl_symbol too.
8173 (maybe_gen_llsym, skip_loc_list_entry): New.
8174 (dwarf2out_maybe_output_loclist_view_pair): New.
8175 (output_loc_list): Output view list or entries too.
8176 (output_view_list_offset): New.
8177 (output_die): Handle dw_val_class_view_list.
8178 (output_dwarf_version): New.
8179 (output_compilation_unit_header): Use it.
8180 (output_skeleton_debug_sections): Likewise.
8181 (output_rnglists, output_line_info): Likewise.
8182 (output_pubnames, output_aranges): Update version comments.
8183 (output_one_line_info_table): Output view numbers in asm comments.
8184 (dw_loc_list): Determine current endview, pass it to new_loc_list.
8185 Call maybe_gen_llsym.
8186 (loc_list_from_tree_1): Adjust.
8187 (add_AT_location_description): Create view list attribute if
8188 needed, check it's absent otherwise.
8189 (convert_cfa_to_fb_loc_list): Adjust.
8190 (maybe_emit_file): Call output_asm_line_debug_info for test.
8191 (dwarf2out_var_location): Reset views as needed. Precompute
8192 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
8193 attribute. Set view.
8194 (new_line_info_table): Reset next view.
8195 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
8196 (dwarf2out_source_line): Likewise. Output view resets and labels to
8197 the assembler, or select appropriate line info opcodes.
8198 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
8199 (optimize_string_length): Catch it. Adjust.
8200 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
8201 dw_val_class_view_list, and remove it if no longer needed.
8202 (hash_loc_list): Hash view numbers.
8203 (loc_list_hasher::equal): Compare them.
8204 (optimize_location_lists): Check whether a view list symbol is
8205 needed, and whether the locview attribute is present, and
8206 whether they match. Remove the locview attribute if no longer
8207 needed.
8208 (index_location_lists): Call skip_loc_list_entry for test.
8209 (dwarf2out_finish): Call output_asm_line_debug_info for test.
8210 Use output_dwarf_version.
8211 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
8212 (struct dw_val_node): Add val_view_list.
8213 * final.c (SEEN_NEXT_VIEW): New.
8214 (set_next_view_needed): New.
8215 (clear_next_view_needed): New.
8216 (maybe_output_next_view): New.
8217 (final_start_function): Rename to...
8218 (final_start_function_1): ... this. Take pointer to FIRST,
8219 add SEEN parameter. Emit param bindings in the initial view.
8220 (final_start_function): Reintroduce SEEN-less interface.
8221 (final): Rename to...
8222 (final_1): ... this. Take SEEN parameter. Output final pending
8223 next view at the end.
8224 (final): Reintroduce seen-less interface.
8225 (final_scan_insn): Output pending next view before switching
8226 sections or ending a block. Mark the next view as needed when
8227 outputting variable locations. Notify debug backend of section
8228 changes, and of location view changes.
8229 (rest_of_handle_final): Adjust.
8230 * toplev.c (process_options): Autodetect value for debug variable
8231 location views option. Warn on incompat5 without -gdwarf-5.
8232 * doc/invoke.texi (gvariable-location-views): New.
8233 (gvariable-location-views=incompat5): New.
8234 (gno-variable-location-views): New.
8236 2018-02-08 David Malcolm <dmalcolm@redhat.com>
8238 PR tree-optimization/84136
8239 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
8240 that the result of find_edge is non-NULL.
8242 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8244 PR target/83008
8245 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
8246 storing integer register in SImode. Fix cost of 256 and 512
8247 byte aligned SSE register store.
8249 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8251 * config/i386/i386.c (ix86_multiplication_cost): Fix
8252 multiplication cost for TARGET_AVX512DQ.
8254 2018-02-08 Marek Polacek <polacek@redhat.com>
8256 PR tree-optimization/84238
8257 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
8258 get_range_strlen.
8260 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8262 PR tree-optimization/84265
8263 * tree-vect-stmts.c (vectorizable_store): Don't treat
8264 VMAT_CONTIGUOUS accesses as grouped.
8265 (vectorizable_load): Likewise.
8267 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8269 PR tree-optimization/81635
8270 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
8271 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
8272 (test_round_for_mask): New functions.
8273 (wide_int_cc_tests): Call test_round_for_mask.
8274 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
8275 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
8276 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
8277 range returned by get_range_info.
8279 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8281 PR ipa/81360
8282 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
8283 * symtab.c: Include builtins.h
8284 (symtab_node::output_to_lto_symbol_table_p): Move here
8285 from lto-streamer-out.c:output_symbol_p.
8286 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
8287 (output_symbol_p): Move all logic to symtab.c
8288 (produce_symtab): Update.
8290 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8292 * config/s390/s390-opts.h (enum indirect_branch): Define.
8293 * config/s390/s390-protos.h (s390_return_addr_from_memory)
8294 (s390_indirect_branch_via_thunk)
8295 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
8296 (enum s390_indirect_branch_type): Define.
8297 * config/s390/s390.c (struct s390_frame_layout, struct
8298 machine_function): Remove.
8299 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
8300 (indirect_branch_table_label_no, indirect_branch_table_name):
8301 Define variables.
8302 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
8303 (enum s390_indirect_branch_option): Define.
8304 (s390_return_addr_from_memory): New function.
8305 (s390_handle_string_attribute): New function.
8306 (s390_attribute_table): Add new attribute handler.
8307 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
8308 (s390_indirect_branch_via_thunk): New function.
8309 (s390_indirect_branch_via_inline_thunk): New function.
8310 (s390_function_ok_for_sibcall): When jumping via thunk disallow
8311 sibling call optimization for non z10 compiles.
8312 (s390_emit_call): Force indirect branch target to be a single
8313 register. Add r1 clobber for non-z10 compiles.
8314 (s390_emit_epilogue): Emit return jump via return_use expander.
8315 (s390_reorg): Handle JUMP_INSNs as execute targets.
8316 (s390_option_override_internal): Perform validity checks for the
8317 new command line options.
8318 (s390_indirect_branch_attrvalue): New function.
8319 (s390_indirect_branch_settings): New function.
8320 (s390_set_current_function): Invoke s390_indirect_branch_settings.
8321 (s390_output_indirect_thunk_function): New function.
8322 (s390_code_end): Implement target hook.
8323 (s390_case_values_threshold): Implement target hook.
8324 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
8325 macros.
8326 * config/s390/s390.h (struct s390_frame_layout)
8327 (struct machine_function): Move here from s390.c.
8328 (TARGET_INDIRECT_BRANCH_NOBP_RET)
8329 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
8330 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
8331 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
8332 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
8333 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
8334 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
8335 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
8336 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
8337 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
8338 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
8339 (mnemonic attribute): Add values which aren't recognized
8340 automatically.
8341 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
8342 pattern for branch conversion. Fix mnemonic attribute.
8343 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
8344 indirect branch via thunk if requested.
8345 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
8346 ("*indirect_jump"): Disable for branch conversion using out of
8347 line thunks.
8348 ("indirect_jump_via_thunk<mode>_z10")
8349 ("indirect_jump_via_thunk<mode>")
8350 ("indirect_jump_via_inlinethunk<mode>_z10")
8351 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
8352 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
8353 ("casesi_jump_via_inlinethunk<mode>_z10")
8354 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
8355 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
8356 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
8357 ("*indirect2_jump"): Disable for branch conversion.
8358 ("casesi_jump"): Turn into expander and expand patterns for branch
8359 conversion.
8360 ("return_use"): New expander.
8361 ("*return"): Emit return via thunk and rename it to ...
8362 ("*return<mode>"): ... this one.
8363 * config/s390/s390.opt: Add new options and and enum for the
8364 option values.
8366 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8368 * lra-constraints.c (match_reload): Unconditionally use
8369 gen_lowpart_SUBREG, rather than selecting between that
8370 and equivalent gen_rtx_SUBREG code.
8372 2018-02-08 Richard Biener <rguenther@suse.de>
8374 PR tree-optimization/84233
8375 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
8376 changed flag instead of boguously re-using phi_inserted.
8378 2018-02-08 Martin Jambor <mjambor@suse.cz>
8380 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
8381 static local variables.
8383 2018-02-08 Richard Biener <rguenther@suse.de>
8385 PR tree-optimization/84278
8386 * tree-vect-stmts.c (vectorizable_store): When looking for
8387 smaller vector types to perform grouped strided loads/stores
8388 make sure the mode is supported by the target.
8389 (vectorizable_load): Likewise.
8391 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8393 * config/aarch64/aarch64.c (aarch64_components_for_bb):
8394 Increase LDP/STP opportunities by adding adjacent callee-saves.
8396 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8398 PR rtl-optimization/84068
8399 PR rtl-optimization/83459
8400 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
8402 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
8404 PR tree-optimization/84224
8405 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
8406 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
8407 non-zero arguments.
8409 2018-02-07 Iain Sandoe <iain@codesourcery.com>
8411 PR target/84113
8412 * config/rs6000/altivec.md (*restore_world): Remove LR use.
8413 * config/rs6000/predicates.md (restore_world_operation): Adjust op
8414 count, remove one USE.
8416 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
8418 * doc/install.texi (Configuration): Document the
8419 --with-long-double-format={ibm,ieee} PowerPC configuration
8420 options.
8422 PR target/84154
8423 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
8424 Convert from define_expand to be define_insn_and_split. Rework
8425 float/double/_Float128 conversions to QI/HI/SImode to work with
8426 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
8427 conversions to QI/HImode types did a store and then a load to
8428 truncate the value. For conversions to VSX registers, don't split
8429 the insn, instead emit the code directly. Use the code iterator
8430 any_fix to combine signed and unsigned conversions.
8431 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
8432 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
8433 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8434 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
8435 (fix_<mode>di2_hw): Likewise.
8436 (fixuns_<mode>di2_hw): Likewise.
8437 (fix_<mode>si2_hw): Likewise.
8438 (fixuns_<mode>si2_hw): Likewise.
8439 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
8440 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8441 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
8442 fix<uns>_trunc<SFDF:mode>si2_p8.
8443 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
8444 used.
8445 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
8446 (fix<uns>_<mode>_mem): Likewise.
8447 (fctiw<u>z_<mode>_mem): Likewise.
8448 (fix<uns>_<mode>_mem): Likewise.
8449 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
8450 the register allocator from doing a direct move to the GPRs to do
8451 a store, and instead use the ISA 3.0 store byte/half-word from
8452 vector register instruction. For IEEE 128-bit floating point,
8453 also optimize stores of 32-bit ints.
8454 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
8456 2018-02-07 Alan Hayward <alan.hayward@arm.com>
8458 * genextract.c (push_pathstr_operand): New function to support
8459 [a-zA-Z].
8460 (walk_rtx): Call push_pathstr_operand.
8461 (print_path): Support [a-zA-Z].
8463 2018-02-07 Richard Biener <rguenther@suse.de>
8465 PR tree-optimization/84037
8466 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
8467 (cse_and_gimplify_to_preheader): Declare.
8468 (vect_get_place_in_interleaving_chain): Likewise.
8469 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8470 ivexpr_map.
8471 (_loop_vec_info::~_loop_vec_info): Delete it.
8472 (cse_and_gimplify_to_preheader): New function.
8473 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
8474 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
8475 (vectorizable_load): Likewise. For grouped stores always base
8476 the IV on the first element.
8477 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
8478 condition before gimplifying.
8480 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8482 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
8483 *DIV_EXPR and *MOD_EXPR.
8485 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
8487 PR target/84248
8488 * config/i386/i386.c (ix86_option_override_internal): Mask out
8489 the CF_SET bit when checking -fcf-protection.
8491 2018-02-07 Tom de Vries <tom@codesourcery.com>
8493 PR libgomp/84217
8494 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
8495 enough.
8497 2018-02-07 Richard Biener <rguenther@suse.de>
8499 PR tree-optimization/84204
8500 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
8501 this place.
8503 PR tree-optimization/84205
8504 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
8505 special-case isl_ast_op_zdiv_r.
8507 PR tree-optimization/84223
8508 * graphite-scop-detection.c (gather_bbs::before_dom_children):
8509 Only add conditions from within the region.
8510 (gather_bbs::after_dom_children): Adjust.
8512 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
8514 PR target/84209
8515 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
8516 * config/avr/avr.md: Only post-reload split REG-REG moves if
8517 either register is GENERAL_REG_P.
8519 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8521 PR tree-optimization/84235
8522 * tree-ssa-scopedtables.c
8523 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
8524 if the subtraction is performed in floating point type where NaNs are
8525 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
8526 build 1. Formatting fix.
8528 2018-02-06 Jakub Jelinek <jakub@redhat.com>
8530 PR target/84146
8531 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
8532 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
8533 and skip it regardless of bb boundaries. Use CALL_P macro,
8534 don't test INSN_P (insn) together with CALL_P or JUMP_P check
8535 unnecessarily, formatting fix.
8537 2018-02-06 Michael Collison <michael.collison@arm.com>
8539 * config/arm/thumb2.md:
8540 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
8541 (*thumb_mov_notscc): Ditto.
8543 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
8545 PR target/84154
8546 * config/rs6000/rs6000.md (su code attribute): Use "u" for
8547 unsigned_fix, not "s".
8549 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8551 * configure.ac (gcc_fn_eh_frame_ro): New function.
8552 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
8553 correct .eh_frame permissions.
8554 * configure: Regenerate.
8556 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
8558 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
8559 irrelevant options.
8561 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8563 * config/rs6000/rs6000.c (rs6000_option_override_internal):
8564 Display warning message for -mno-speculate-indirect-jumps.
8566 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
8568 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
8569 Undocumented.
8570 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
8572 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
8574 PR tree-optimization/84225
8575 * tree-eh.c (find_trapping_overflow): Only call
8576 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
8578 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8580 PR target/84145
8581 * config/i386/i386.c: Reimplement the check of possible options
8582 -mibt/-mshstk conbination. Change error messages.
8583 * doc/invoke.texi: Fix a typo: remove extra '='.
8585 2018-02-06 Marek Polacek <polacek@redhat.com>
8587 PR tree-optimization/84228
8588 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
8590 2018-02-06 Tamar Christina <tamar.christina@arm.com>
8592 PR target/82641
8593 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
8594 emitted arch directives.
8595 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
8596 __ARM_FEATURE_COPROC before changing architectures.
8598 2018-02-06 Richard Biener <rguenther@suse.de>
8600 * config/i386/i386.c (print_reg): Fix typo.
8601 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
8603 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
8605 * configure: Regenerate.
8607 2018-02-05 Martin Sebor <msebor@redhat.com>
8609 PR tree-optimization/83369
8610 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
8611 inlining context.
8613 2018-02-05 Martin Liska <mliska@suse.cz>
8615 * doc/invoke.texi: Cherry-pick upstream r323995.
8617 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
8619 * ira.c (ira_init_register_move_cost): Adjust comment.
8621 2018-02-05 Martin Liska <mliska@suse.cz>
8623 PR gcov-profile/84137
8624 * doc/gcov.texi: Fix typo in documentation.
8626 2018-02-05 Martin Liska <mliska@suse.cz>
8628 PR gcov-profile/83879
8629 * doc/gcov.texi: Document necessity of --dynamic-list-data when
8630 using dlopen functionality.
8632 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
8634 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
8635 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
8636 _mm_maskz_range_ss, _mm_mask_range_round_ss,
8637 _mm_maskz_range_round_ss): New intrinsics.
8638 (__builtin_ia32_rangesd128_round)
8639 (__builtin_ia32_rangess128_round): Remove.
8640 (__builtin_ia32_rangesd128_mask_round,
8641 __builtin_ia32_rangess128_mask_round): New builtins.
8642 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
8643 __builtin_ia32_rangess128_round): Remove.
8644 (__builtin_ia32_rangesd128_mask_round,
8645 __builtin_ia32_rangess128_mask_round): New builtins.
8646 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
8647 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
8648 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
8649 "<round_saeonly_constraint>")): Changed to ...
8650 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
8651 "<round_saeonly_scalar_constraint>")): ... this.
8652 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
8653 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
8654 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
8655 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
8656 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
8658 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
8660 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
8661 options.
8662 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
8663 Remove all values except native, 8540 and 8548.
8665 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
8667 * config/i386/i386.c (ix86_output_function_return): Pass
8668 INVALID_REGNUM, instead of -1, as invalid register number to
8669 indirect_thunk_name and output_indirect_thunk.
8671 2018-02-02 Julia Koval <julia.koval@intel.com>
8673 * config.gcc: Add -march=icelake.
8674 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
8675 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
8676 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
8677 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
8678 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
8679 (processor_target_table): Add icelake.
8680 (ix86_option_override_internal): Handle new PTAs.
8681 (get_builtin_code_for_version): Handle icelake.
8682 (M_INTEL_COREI7_ICELAKE): New.
8683 (fold_builtin_cpu): Handle icelake.
8684 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
8685 * doc/invoke.texi: Add -march=icelake.
8687 2018-02-02 Julia Koval <julia.koval@intel.com>
8689 * config/i386/i386.c (ix86_option_override_internal): Change flags type
8690 to wide_int_bitmask.
8691 * wide-int-bitmask.h: New.
8693 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8695 PR target/84066
8696 * config/i386/i386.md: Replace Pmode with word_mode in
8697 builtin_setjmp_setup and builtin_longjmp to support x32.
8699 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
8701 PR target/56010
8702 PR target/83743
8703 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
8704 #include "opts.h".
8705 (rs6000_supported_cpu_names): New static variable.
8706 (linux_cpu_translation_table): Likewise.
8707 (elf_platform) <cpu>: Define new static variable and use it.
8708 Translate kernel AT_PLATFORM name to canonical name if needed.
8709 Error if platform name is unknown.
8711 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
8713 PR target/84089
8714 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
8716 2018-02-01 Jeff Law <law@redhat.com>
8718 PR target/84128
8719 * config/i386/i386.c (release_scratch_register_on_entry): Add new
8720 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
8721 the scratch if RELEASE_VIA_POP is false.
8722 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
8723 If we have to save a temporary register, decrement SIZE appropriately.
8724 Pass new arguments to release_scratch_register_on_entry.
8725 (ix86_adjust_stack_and_probe): Likewise.
8726 (ix86_emit_probe_stack_range): Pass new arguments to
8727 release_scratch_register_on_entry.
8729 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
8731 PR rtl-optimization/84157
8732 * combine.c (change_zero_ext): Use REG_P predicate in
8733 front of HARD_REGISTER_P predicate.
8735 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
8737 * config/avr/avr.c (avr_option_override): Move disabling of
8738 -fdelete-null-pointer-checks to...
8739 * common/config/avr/avr-common.c (avr_option_optimization_table):
8740 ...here.
8742 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8744 PR tree-optimization/81635
8745 * tree-data-ref.c (split_constant_offset_1): For types that
8746 wrap on overflow, try to use range info to prove that wrapping
8747 cannot occur.
8749 2018-02-01 Renlin Li <renlin.li@arm.com>
8751 PR target/83370
8752 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
8753 TAILCALL_ADDR_REGS.
8754 (aarch64_register_move_cost): Likewise.
8755 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
8756 TAILCALL_ADDR_REGS.
8757 (REG_CLASS_NAMES): Likewise.
8758 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
8759 TAILCALL_ADDR_REGS. Remove IP registers.
8760 * config/aarch64/aarch64.md (Ucs): Update register constraint.
8762 2018-02-01 Richard Biener <rguenther@suse.de>
8764 * domwalk.h (dom_walker::dom_walker): Add additional constructor
8765 for specifying RPO order and allow NULL for that.
8766 * domwalk.c (dom_walker::dom_walker): Likewise.
8767 (dom_walker::walk): Handle NULL RPO order.
8768 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
8769 in RPO order.
8770 (rewrite_update_dom_walker): Likewise.
8771 (mark_def_dom_walker): Likewise.
8773 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8775 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
8776 (aarch64_maybe_expand_sve_subreg_move): Declare.
8777 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
8778 * config/aarch64/predicates.md (aarch64_any_register_operand): New
8779 predicate.
8780 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
8781 that are semantically a reverse operation.
8782 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
8783 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
8784 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
8785 functions.
8786 (aarch64_can_change_mode_class): For big-endian, forbid changes
8787 between two SVE modes if they have different element sizes.
8789 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8791 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
8792 the TImode handling for big-endian targets.
8794 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8796 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
8797 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
8798 not just bytes.
8799 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
8800 Remove BSWAP handing for big-endian targets and use the form of
8801 LD1RQ appropariate for the mode.
8803 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8805 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
8806 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
8807 duplicated element.
8809 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
8811 PR tearget/83845
8812 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
8813 check for operands that need to go through aarch64_sve_reload_be.
8815 2018-02-01 Jakub Jelinek <jakub@redhat.com>
8817 PR tree-optimization/81661
8818 PR tree-optimization/84117
8819 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
8820 * tree-eh.c: Include gimplify.h.
8821 (find_trapping_overflow, replace_trapping_overflow,
8822 rewrite_to_non_trapping_overflow): New functions.
8823 * tree-vect-loop.c: Include tree-eh.h.
8824 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
8825 * tree-data-ref.c: Include tree-eh.h.
8826 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
8828 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
8830 PR rtl-optimization/84123
8831 * combine.c (change_zero_ext): Check if hard register satisfies
8832 can_change_dest_mode before calling gen_lowpart_SUBREG.
8834 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
8836 PR target/82444
8837 * ira.c (ira_init_register_move_cost): Remove assert.
8839 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8841 PR rtl-optimization/84071
8842 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
8843 * doc/tm.texi: Regenerate.
8845 2018-01-31 Richard Biener <rguenther@suse.de>
8847 PR tree-optimization/84132
8848 * tree-data-ref.c (analyze_miv_subscript): Properly
8849 check whether evolution_function_is_affine_multivariate_p
8850 before calling gcd_of_steps_may_divide_p.
8852 2018-01-31 Julia Koval <julia.koval@intel.com>
8854 PR target/83618
8855 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
8856 * config/i386/i386.md (rdpid_rex64) New.
8857 (rdpid): Make 32bit only.
8859 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
8861 PR lto/84105
8862 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
8863 an IDENTIFIER_NODE for FUNCTION_TYPE's.
8865 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8867 Revert
8868 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8870 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8872 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
8874 PR rtl-optimization/84071
8875 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
8876 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
8878 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8880 * config/arc/arc.c (arc_handle_aux_attribute): New function.
8881 (arc_attribute_table): Add 'aux' attribute.
8882 (arc_in_small_data_p): Consider aux like variables.
8883 (arc_is_aux_reg_p): New function.
8884 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
8885 (arc_get_aux_arg): New function.
8886 (prepare_move_operands): Handle aux-register access.
8887 (arc_handle_aux_attribute): New function.
8888 * doc/extend.texi (ARC Variable attributes): Add subsection.
8890 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
8892 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
8893 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
8894 (arc_attribute_table): Add 'uncached' attribute.
8895 (arc_print_operand): Print '.di' flag for uncached memory
8896 accesses.
8897 (arc_in_small_data_p): Do not consider for small data the uncached
8898 types.
8899 (arc_is_uncached_mem_p): New function.
8900 * config/arc/predicates.md (compact_store_memory_operand): Check
8901 for uncached memory accesses.
8902 (nonvol_nonimm_operand): Likewise.
8903 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
8905 2018-01-31 Jakub Jelinek <jakub@redhat.com>
8907 PR c/84100
8908 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
8909 falign-loops=): Add Optimization flag.
8911 2018-01-30 Jeff Law <law@redhat.com>
8913 PR target/84064
8914 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
8915 INT_REGISTERS_SAVED. Check it prior to calling
8916 get_scratch_register_on_entry.
8917 (ix86_adjust_stack_and_probe): Similarly.
8918 (ix86_emit_probe_stack_range): Similarly.
8919 (ix86_expand_prologue): Corresponding changes.
8921 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8923 PR target/40411
8924 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
8925 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
8927 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
8929 PR target/84112
8930 * lra-constraints.c (curr_insn_transform): Process AND in the
8931 address.
8933 2018-01-30 Jakub Jelinek <jakub@redhat.com>
8935 PR rtl-optimization/83986
8936 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
8937 dependence against last_pending_memory_flush in addition to
8938 pending_jump_insns.
8940 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
8942 PR tree-optimization/81611
8943 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
8944 copies.
8946 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8948 PR target/83758
8949 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
8950 a reg rtx.
8952 2018-01-30 Richard Biener <rguenther@suse.de>
8953 Jakub Jelinek <jakub@redhat.com>
8955 PR tree-optimization/84111
8956 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
8957 inner loops added during recursion, as they don't have up-to-date
8958 SSA form.
8960 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8962 PR ipa/81360
8963 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
8964 (can_inline_edge_by_limits_p): ... here.
8965 (can_early_inline_edge_p, check_callers,
8966 update_caller_keys, update_callee_keys, recursive_inlining,
8967 add_new_edges_to_heap, speculation_useful_p,
8968 inline_small_functions,
8969 inline_small_functions, flatten_function,
8970 inline_to_all_callers_1): Update.
8972 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
8974 * profile-count.c (profile_count::combine_with_ipa_count): Handle
8975 zeros correctly.
8977 2018-01-30 Richard Biener <rguenther@suse.de>
8979 PR tree-optimization/83008
8980 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
8981 invariant and constant vector uses in stmts when they need
8982 more than one stmt.
8984 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8986 PR bootstrap/84017
8987 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
8988 * configure: Regenerate.
8990 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8992 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
8993 pattern.
8994 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
8995 Use gen_rtx_REG rather than gen_lowpart.
8997 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
8999 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
9000 rather than 0 when creating partial subregs.
9002 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9004 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
9005 of usage.
9007 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
9009 PR target/81550
9010 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
9011 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
9012 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
9013 flags. This restores the settings used before the 2017-07-24.
9014 Turning off pre increment/decrement/modify allows IVOPTS to
9015 optimize DF/SF loops where the index is an int.
9017 2018-01-29 Richard Biener <rguenther@suse.de>
9018 Kelvin Nilsen <kelvin@gcc.gnu.org>
9020 PR bootstrap/80867
9021 * tree-vect-stmts.c (vectorizable_call): Don't call
9022 targetm.vectorize_builtin_md_vectorized_function if callee is
9023 NULL.
9025 2018-01-22 Carl Love <cel@us.ibm.com>
9027 * doc/extend.tex: Fix typo in second arg in
9028 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
9030 2018-01-29 Richard Biener <rguenther@suse.de>
9032 PR tree-optimization/84086
9033 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
9034 (flush_ssaname_freelist): When SSA names were released reset
9035 the SCEV hash table.
9037 2018-01-29 Richard Biener <rguenther@suse.de>
9039 PR tree-optimization/84057
9040 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
9041 removed paths when removing edges.
9043 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
9045 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
9046 -mfunction-return=@var{choice}.
9048 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
9050 PR diagnostic/84034
9051 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
9052 Handle CR like TAB.
9053 (layout::print_source_line): Likewise.
9054 (test_get_line_width_without_trailing_whitespace): Add test cases.
9056 2018-01-27 Jakub Jelinek <jakub@redhat.com>
9058 PR middle-end/84040
9059 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
9060 debug insns.
9062 2018-01-26 Jim Wilson <jimw@sifive.com>
9064 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
9066 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
9067 specified.
9069 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9071 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
9072 and CMP + SUB-immediate -> SUBS.
9074 2018-01-26 Martin Sebor <msebor@redhat.com>
9076 PR tree-optimization/83896
9077 * tree-ssa-strlen.c (get_string_len): Rename...
9078 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
9079 Avoid assuming length is constant.
9080 (handle_char_store): Use HOST_WIDE_INT for string length.
9082 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
9084 PR target/81763
9085 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
9086 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
9088 2018-01-26 Richard Biener <rguenther@suse.de>
9090 PR rtl-optimization/84003
9091 * dse.c (record_store): Only record redundant stores when
9092 the earlier store aliases at least all accesses the later one does.
9094 2018-01-26 Jakub Jelinek <jakub@redhat.com>
9096 PR rtl-optimization/83985
9097 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
9098 REG_CFA_RESTORE insns.
9099 (delete_unmarked_insns): Don't ignore separate shrink wrapping
9100 REG_CFA_RESTORE insns here.
9102 PR c/83989
9103 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
9104 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
9106 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9108 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
9109 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
9110 (arc_init): Likewise.
9111 (arc_override_options): Likewise.
9112 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
9113 value.
9114 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
9115 support.
9116 * config/arc/arc.h (TARGET_DBNZ): Define.
9117 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
9118 properly set the tune attribute.
9119 (dbnz): Use TARGET_DBNZ guard.
9120 * config/arc/arc.opt (mtune): Add core3 option.
9122 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9124 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
9125 recognize new pic like addresses.
9126 (arc_delegitimize_address): Clean up.
9128 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9130 * config/arc/arc-arches.def: Option mrf16 valid for all
9131 architectures.
9132 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
9133 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
9134 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
9135 * config/arc/arc-tables.opt: Regenerate.
9136 * config/arc/arc.c (arc_conditional_register_usage): Handle
9137 reduced register file case.
9138 (arc_file_start): Set must have build attributes.
9139 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
9140 mrf16 option value.
9141 * config/arc/arc.opt (mrf16): Add new option.
9142 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
9143 * config/arc/genmultilib.awk: Handle new mrf16 option.
9144 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
9145 * config/arc/t-multilib: Regenerate.
9146 * doc/invoke.texi (ARC Options): Document mrf16 option.
9148 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9150 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
9151 * config/arc/arc.c (arc_handle_secure_attribute): New function.
9152 (arc_attribute_table): Add 'secure_call' attribute.
9153 (arc_print_operand): Print secure call operand.
9154 (arc_function_ok_for_sibcall): Don't optimize tail calls when
9155 secure.
9156 (arc_is_secure_call_p): New function. * config/arc/arc.md
9157 (call_i): Add support for sjli instruction.
9158 (call_value_i): Likewise.
9159 * config/arc/constraints.md (Csc): New constraint.
9161 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9162 John Eric Martin <John.Martin@emmicro-us.com>
9164 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
9165 * config/arc/arc.c (_arc_jli_section): New struct.
9166 (arc_jli_section): New type.
9167 (rc_jli_sections): New static variable.
9168 (arc_handle_jli_attribute): New function.
9169 (arc_attribute_table): Add jli_always and jli_fixed attribute.
9170 (arc_file_end): New function.
9171 (TARGET_ASM_FILE_END): Define.
9172 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
9173 (arc_add_jli_section): New function.
9174 (jli_call_scan): Likewise.
9175 (arc_reorg): Call jli_call_scan.
9176 (arc_output_addsi): Remove 'S' from printing asm operand.
9177 (arc_is_jli_call_p): New function.
9178 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
9179 operand.
9180 (movhi_insn): Likewise.
9181 (movsi_insn): Likewise.
9182 (movsi_set_cc_insn): Likewise.
9183 (loadqi_update): Likewise.
9184 (load_zeroextendqisi_update): Likewise.
9185 (load_signextendqisi_update): Likewise.
9186 (loadhi_update): Likewise.
9187 (load_zeroextendhisi_update): Likewise.
9188 (load_signextendhisi_update): Likewise.
9189 (loadsi_update): Likewise.
9190 (loadsf_update): Likewise.
9191 (movsicc_insn): Likewise.
9192 (bset_insn): Likewise.
9193 (bxor_insn): Likewise.
9194 (bclr_insn): Likewise.
9195 (bmsk_insn): Likewise.
9196 (bicsi3_insn): Likewise.
9197 (cmpsi_cc_c_insn): Likewise.
9198 (movsi_ne): Likewise.
9199 (movsi_cond_exec): Likewise.
9200 (clrsbsi2): Likewise.
9201 (norm_f): Likewise.
9202 (normw): Likewise.
9203 (swap): Likewise.
9204 (divaw): Likewise.
9205 (flag): Likewise.
9206 (sr): Likewise.
9207 (kflag): Likewise.
9208 (ffs): Likewise.
9209 (ffs_f): Likewise.
9210 (fls): Likewise.
9211 (call_i): Remove 'S' asm letter, add jli instruction.
9212 (call_value_i): Likewise.
9213 * config/arc/arc.op (mjli-always): New option.
9214 * config/arc/constraints.md (Cji): New constraint.
9215 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
9216 operand.
9217 (subsf3_fpx): Likewise.
9218 (mulsf3_fpx): Likewise.
9219 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
9220 asm operand.
9221 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
9222 function attrbutes.
9223 * doc/invoke.texi (ARC): Document mjli-always option.
9225 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
9227 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
9228 avoid addition with 0 and use incw and decw where possible.
9230 2018-01-26 Richard Biener <rguenther@suse.de>
9232 PR tree-optimization/81082
9233 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
9234 association if it requires casting to unsigned.
9235 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
9236 from fold_plusminus_mult_expr to catch important cases late when
9237 range info is available.
9239 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9241 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
9242 * configure.ac (hidden_linkonce): New test.
9243 * configure: Regenerate.
9244 * config.in: Regenerate.
9246 2018-01-26 Julia Koval <julia.koval@intel.com>
9248 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
9249 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
9250 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
9251 _mm_mask_bitshuffle_epi64_mask): Fix type.
9252 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
9253 USI_FTYPE_V4DI_V4DI_USI): Remove.
9254 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
9255 __builtin_ia32_vpshufbitqmb256_mask,
9256 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
9257 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
9258 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
9260 2018-01-26 Alan Modra <amodra@gmail.com>
9262 PR target/84033
9263 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
9264 UNSPEC_VBPERMQ. Sort other unspecs.
9266 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
9268 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
9270 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
9272 PR middle-end/83055
9273 * predict.c (drop_profile): Do not push/pop cfun; update also
9274 node->count.
9275 (handle_missing_profiles): Fix logic looking for zero profiles.
9277 2018-01-25 Jakub Jelinek <jakub@redhat.com>
9279 PR middle-end/83977
9280 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
9281 on functions with #pragma omp declare simd or functions with simd
9282 attribute.
9283 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
9284 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
9285 Remove trailing \n from warning_at calls.
9287 2018-01-25 Tom de Vries <tom@codesourcery.com>
9289 PR target/84028
9290 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9291 for neutered workers.
9293 2018-01-24 Joseph Myers <joseph@codesourcery.com>
9295 PR target/68467
9296 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
9297 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
9299 2018-01-24 Jeff Law <law@redhat.com>
9301 PR target/83994
9302 * i386.c (get_probe_interval): Move to earlier point.
9303 (ix86_compute_frame_layout): If -fstack-clash-protection and
9304 the frame is larger than the probe interval, then use pushes
9305 to save registers rather than reg->mem moves.
9306 (ix86_expand_prologue): Remove conditional for int_registers_saved
9307 assertion.
9309 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
9311 PR target/84014
9312 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
9313 min/max for never referenced object.
9315 2018-01-24 Jakub Jelinek <jakub@redhat.com>
9317 PR middle-end/83977
9318 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
9319 here.
9320 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
9321 attributes from DECL_ATTRIBUTES (decl) without affecting
9322 DECL_ATTRIBUTES (current_function_decl).
9323 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
9324 functions with non-NULL DECL_ABSTRACT_ORIGIN.
9326 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
9328 PR tree-optimization/83979
9329 * fold-const.c (fold_comparison): Use constant_boolean_node
9330 instead of boolean_{true,false}_node.
9332 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
9334 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
9335 with zero counts.
9337 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9339 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9340 Simplify the clause that sets the length attribute.
9341 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9342 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
9343 clause that sets the length attribute.
9344 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9346 2018-01-24 Tom de Vries <tom@codesourcery.com>
9348 PR target/83589
9349 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
9350 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
9351 Add strict parameter.
9352 (prevent_branch_around_nothing): Insert dummy insn between branch to
9353 label and label with no ptx insn inbetween.
9354 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
9356 2018-01-24 Tom de Vries <tom@codesourcery.com>
9358 PR target/81352
9359 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9360 for neutered threads in warp.
9361 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
9363 2018-01-24 Richard Biener <rguenther@suse.de>
9365 PR tree-optimization/83176
9366 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
9367 operands.
9369 2018-01-24 Richard Biener <rguenther@suse.de>
9371 PR tree-optimization/82819
9372 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
9373 code generating pluses that are no-ops in the target precision.
9375 2018-01-24 Richard Biener <rguenther@suse.de>
9377 PR middle-end/84000
9378 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
9380 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9382 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
9383 to merge probabilities.
9384 * predict.c (probably_never_executed): Also mark as cold functions
9385 with global 0 profile and guessed local profile.
9386 * profile-count.c (profile_probability::combine_with_count): New
9387 member function.
9388 * profile-count.h (profile_probability::operator*,
9389 profile_probability::operator*=, profile_probability::operator/,
9390 profile_probability::operator/=): Reduce precision to adjusted
9391 and set value to guessed on contradictory divisions.
9392 (profile_probability::combine_with_freq): Remove.
9393 (profile_probability::combine_wiht_count): Declare.
9394 (profile_count::force_nonzero):: Set to adjusted.
9395 (profile_count::probability_in):: Set quality to adjusted.
9396 * tree-ssa-tail-merge.c (replace_block_by): Use
9397 combine_with_count.
9399 2018-01-23 Andrew Waterman <andrew@sifive.com>
9400 Jim Wilson <jimw@sifive.com>
9402 * config/riscv/riscv.c (riscv_stack_boundary): New.
9403 (riscv_option_override): Set riscv_stack_boundary. Handle
9404 riscv_preferred_stack_boundary_arg.
9405 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
9406 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
9407 (STACK_BOUNDARY): Set to riscv_stack_boundary.
9408 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
9409 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
9410 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
9412 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
9414 PR target/83905
9415 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
9416 of struct ix86_frame.
9417 (ix86_expand_epilogue): Likewise. Add a local variable for
9418 the reg_save_offset field in struct ix86_frame.
9420 2018-01-23 Bin Cheng <bin.cheng@arm.com>
9422 PR tree-optimization/82604
9423 * tree-loop-distribution.c (enum partition_kind): New enum item
9424 PKIND_PARTIAL_MEMSET.
9425 (partition_builtin_p): Support above new enum item.
9426 (generate_code_for_partition): Ditto.
9427 (compute_access_range): Differentiate cases that equality can be
9428 proven at all loops, the innermost loops or no loops.
9429 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
9430 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
9431 (finalize_partitions, distribute_loop): Don't fuse partition of
9432 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
9433 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
9434 parloop is enabled.
9436 2018-01-23 Martin Liska <mliska@suse.cz>
9438 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
9439 order to ignore the predictor.
9440 (PRED_POLYMORPHIC_CALL): Likewise.
9441 (PRED_RECURSIVE_CALL): Likewise.
9443 2018-01-23 Martin Liska <mliska@suse.cz>
9445 * tree-profile.c (tree_profiling): Print function header to
9446 aware reader which function we are working on.
9447 * value-prof.c (gimple_find_values_to_profile): Do not print
9448 not interesting value histograms.
9450 2018-01-23 Martin Liska <mliska@suse.cz>
9452 * profile-count.h (enum profile_quality): Add
9453 profile_uninitialized as the first value. Do not number values
9454 as they are zero based.
9455 (profile_count::verify): Update sanity check.
9456 (profile_probability::verify): Likewise.
9458 2018-01-23 Nathan Sidwell <nathan@acm.org>
9460 * doc/invoke.texi (ffor-scope): Deprecate.
9462 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9464 PR tree-optimization/83510
9465 * domwalk.c (set_all_edges_as_executable): New function.
9466 (dom_walker::dom_walker): Convert bool param
9467 "skip_unreachable_blocks" to enum reachability. Move setup of
9468 edge flags to set_all_edges_as_executable and only do it when
9469 reachability is REACHABLE_BLOCKS.
9470 * domwalk.h (enum dom_walker::reachability): New enum.
9471 (dom_walker::dom_walker): Convert bool param
9472 "skip_unreachable_blocks" to enum reachability.
9473 (set_all_edges_as_executable): New decl.
9474 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
9475 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
9476 "reachability".
9477 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
9478 but converting true to REACHABLE_BLOCKS.
9479 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
9480 * tree-vrp.c
9481 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
9482 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
9483 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
9484 REACHABLE_BLOCKS.
9485 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
9486 if check_all_array_refs will be called.
9488 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9490 * tree.c (selftest::test_location_wrappers): Add more test
9491 coverage.
9493 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9495 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
9496 (selftest::test_bit_in_range): Likewise.
9498 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9500 PR testsuite/83888
9501 * doc/sourcebuild.texi (vect_float): Say that the selector
9502 only describes the situation when -funsafe-math-optimizations is on.
9503 (vect_float_strict): Document.
9505 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9507 PR tree-optimization/83965
9508 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
9509 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
9510 instead of checking only for a reduction.
9511 (vect_recog_widen_sum_pattern): Likewise.
9513 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9515 * predict.c (probably_never_executed): Only use precise profile info.
9516 (compute_function_frequency): Skip after inlining hack since we now
9517 have quality checking.
9519 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9521 * profile-count.h (profile_probability::very_unlikely,
9522 profile_probability::unlikely, profile_probability::even): Set
9523 precision to guessed.
9525 2018-01-23 Richard Biener <rguenther@suse.de>
9527 PR tree-optimization/83963
9528 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
9529 Properly terminate dominator walk when crossing the exit edge not
9530 when visiting its source block.
9532 2018-01-23 Jakub Jelinek <jakub@redhat.com>
9534 PR c++/83918
9535 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
9536 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
9538 2018-01-22 Jakub Jelinek <jakub@redhat.com>
9540 PR tree-optimization/83957
9541 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
9542 semicolon after for body surrounded by braces.
9544 PR tree-optimization/83081
9545 * profile-count.h (profile_probability::split): New method.
9546 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
9547 Use profile_probability::split.
9548 (do_compare_rtx_and_jump): Fix adjustment of probabilities
9549 when splitting a single conditional jump into 2.
9551 2018-01-22 David Malcolm <dmalcolm@redhat.com>
9553 PR tree-optimization/69452
9554 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
9555 decl.
9557 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9559 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
9560 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
9561 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
9563 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9565 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
9566 * config/rl78/rl78.md (movdi): New define_expand.
9567 * config/rl78/rl78.c (rl78_split_movdi): New function.
9569 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
9571 PR target/83862
9572 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
9573 no longer used.
9574 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
9575 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
9576 128-bit to produce an UNSPEC move to get the double word with the
9577 signbit and then a shift directly to do signbit.
9578 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
9579 implementation with a new version that just does either a direct
9580 move or a regular move. Move memory interface to separate insns.
9581 Move insns so they are next to the expander.
9582 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
9583 with signbit move. Split big and little endian case.
9584 (signbit<mode>2_dm_mem_le): Likewise.
9585 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
9586 (signbit<mode>2_dm2): Likewise.
9588 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9590 * config/rl78/rl78.md (anddi3): New define_expand.
9592 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9594 * config/rl78/rl78.md (umindi3): New define_expand.
9596 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9598 * config/rl78/rl78.md (smindi3): New define_expand.
9600 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9602 * config/rl78/rl78.md (smaxdi3): New define_expand.
9604 2018-01-22 Carl Love <cel@us.ibm.com>
9606 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
9607 LVX_V1TI): Add macro expansion.
9608 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
9609 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
9610 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
9611 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
9612 Change check to determine if the instruction is a byte reversing
9613 entry. Fix typo in comment.
9614 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
9615 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
9616 Add def_builtin calls for new builtins.
9617 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
9618 Add define_insn expansion.
9620 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9622 * config/rl78/rl78.md (umaxdi3): New define_expand.
9624 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
9626 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
9627 for non-QImode registers.
9629 2018-01-22 Richard Biener <rguenther@suse.de>
9631 PR tree-optimization/83963
9632 * graphite-scop-detection.c (scop_detection::get_sese): Delay
9633 including the loop exit block.
9634 (scop_detection::merge_sese): Likewise.
9635 (scop_detection::add_scop): Do it here instead.
9637 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9639 * doc/sourcebuild.texi (arm_softfloat): Document.
9641 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
9643 PR gcc/77734
9644 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
9645 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
9646 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
9648 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9649 David Edelsohn <dje.gcc@gmail.com>
9651 PR target/83946
9652 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9653 Change "crset eq" to "crset 2".
9654 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9655 (*call_indirect_aix<mode>_nospec): Likewise.
9656 (*call_value_indirect_aix<mode>_nospec): Likewise.
9657 (*call_indirect_elfv2<mode>_nospec): Likewise.
9658 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
9659 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
9660 change assembly output from . to $.
9661 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9662 (indirect_jump<mode>_nospec): Change assembly output from . to $.
9663 (*tablejump<mode>_internal1_nospec): Likewise.
9665 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
9667 PR target/80870
9668 * config/sh/sh_optimize_sett_clrt.cc:
9669 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
9671 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
9673 PR tree-optimization/83940
9674 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
9675 offset_dt to vect_constant_def rather than vect_unknown_def_type.
9676 (vect_check_load_store_mask): Add a mask_dt_out parameter and
9677 use it to pass back the definition type.
9678 (vect_check_store_rhs): Likewise rhs_dt_out.
9679 (vect_build_gather_load_calls): Add a mask_dt argument and use
9680 it instead of a call to vect_is_simple_use.
9681 (vectorizable_store): Update calls to vect_check_load_store_mask
9682 and vect_check_store_rhs. Use the dt returned by the latter instead
9683 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
9684 instead of calls to vect_is_simple_use. Pass the scalar rather
9685 than the vector operand to vect_is_simple_use when handling
9686 second and subsequent copies of an rhs value.
9687 (vectorizable_load): Update calls to vect_check_load_store_mask
9688 and vect_build_gather_load_calls. Use the cached mask_dt and
9689 gs_info.offset_dt instead of calls to vect_is_simple_use.
9691 2018-01-20 Jakub Jelinek <jakub@redhat.com>
9693 PR middle-end/83945
9694 * tree-emutls.c: Include gimplify.h.
9695 (lower_emutls_2): New function.
9696 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
9697 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
9698 it before further processing.
9700 PR target/83930
9701 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
9702 UINTVAL (trueop1) instead of INTVAL (op1).
9704 2018-01-19 Jakub Jelinek <jakub@redhat.com>
9706 PR debug/81570
9707 PR debug/83728
9708 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
9709 INCOMING_FRAME_SP_OFFSET if not defined.
9710 (scan_trace): Add ENTRY argument. If true and
9711 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
9712 emit a note to adjust the CFA offset.
9713 (create_cfi_notes): Adjust scan_trace callers.
9714 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
9715 INCOMING_FRAME_SP_OFFSET in the CIE.
9716 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
9717 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
9718 Likewise.
9719 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
9720 * doc/tm.texi: Regenerated.
9722 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9724 PR rtl-optimization/83147
9725 * lra-constraints.c (remove_inheritance_pseudos): Use
9726 lra_substitute_pseudo_within_insn.
9728 2018-01-19 Tom de Vries <tom@codesourcery.com>
9729 Cesar Philippidis <cesar@codesourcery.com>
9731 PR target/83920
9732 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
9734 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
9736 PR target/83790
9737 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
9738 spaces for function labels.
9740 2018-01-19 Martin Liska <mliska@suse.cz>
9742 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
9743 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
9744 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
9745 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
9746 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
9747 (PRED_CONST_RETURN): Change from 69 to 65.
9748 (PRED_NULL_RETURN): Change from 91 to 71.
9749 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
9750 (PRED_LOOP_GUARD): Change from 66 to 73.
9752 2018-01-19 Martin Liska <mliska@suse.cz>
9754 * predict.c (predict_insn_def): Add new assert.
9755 (struct branch_predictor): Change type to signed integer.
9756 (test_prediction_value_range): Amend test to cover
9757 PROB_UNINITIALIZED.
9758 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
9759 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
9760 (PRED_LOOP_ITERATIONS_MAX): Likewise.
9761 (PRED_LOOP_IV_COMPARE): Likewise.
9762 * predict.h (PROB_UNINITIALIZED): Define new constant.
9764 2018-01-19 Martin Liska <mliska@suse.cz>
9766 * predict.c (dump_prediction): Add new format for
9767 analyze_brprob.py script which is enabled with -details
9768 suboption.
9769 * profile-count.h (precise_p): New function.
9771 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9773 PR tree-optimization/83922
9774 * tree-vect-loop.c (vect_verify_full_masking): Return false if
9775 there are no statements that need masking.
9776 (vect_active_double_reduction_p): New function.
9777 (vect_analyze_loop_operations): Use it when handling phis that
9778 are not in the loop header.
9780 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
9782 PR tree-optimization/83914
9783 * tree-vect-loop.c (vectorizable_induction): Don't convert
9784 init_expr or apply the peeling adjustment for inductions
9785 that are nested within the vectorized loop.
9787 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9789 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
9790 instead of NEG.
9792 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9794 PR sanitizer/81715
9795 PR testsuite/83882
9796 * function.h (gimplify_parameters): Add gimple_seq * argument.
9797 * function.c: Include gimple.h and options.h.
9798 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
9799 for the added local temporaries if needed.
9800 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
9801 if there are any parameter cleanups, wrap whole body into a
9802 try/finally with the cleanups.
9804 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
9806 PR target/82964
9807 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
9808 Use GET_MODE_CLASS for scalar floating point.
9810 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9812 PR ipa/82256
9813 patch by PaX Team
9814 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
9815 Fix call of call_cgraph_insertion_hooks.
9817 2018-01-18 Martin Sebor <msebor@redhat.com>
9819 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
9821 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
9823 PR ipa/83619
9824 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
9825 frequencies.
9827 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
9829 PR other/70268
9830 * common.opt: (-ffile-prefix-map): New option.
9831 * opts.c (common_handle_option): Defer it.
9832 * opts-global.c (handle_common_deferred_options): Handle it.
9833 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
9834 * file-prefix-map.h: New file.
9835 (remap_debug_filename, add_debug_prefix_map): ...here.
9836 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
9837 * final.c (debug_prefix_map, add_debug_prefix_map
9838 remap_debug_filename): Move to...
9839 * file-prefix-map.c: New file.
9840 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
9841 generalize, get rid of alloca(), use strrchr() instead of strchr().
9842 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
9843 Implement in terms of add_prefix_map().
9844 (remap_macro_filename, remap_debug_filename): Implement in term of
9845 remap_filename().
9846 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
9847 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
9848 * dbxout.c: Include file-prefix-map.h.
9849 * varasm.c: Likewise.
9850 * vmsdbgout.c: Likewise.
9851 * xcoffout.c: Likewise.
9852 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
9853 * doc/cppopts.texi (-fmacro-prefix-map): Document.
9854 * doc/invoke.texi (-ffile-prefix-map): Document.
9855 (-fdebug-prefix-map): Update description.
9857 2018-01-18 Martin Liska <mliska@suse.cz>
9859 * config/i386/i386.c (indirect_thunk_name): Document that also
9860 lfence is emitted.
9861 (output_indirect_thunk): Document why both instructions
9862 (pause and lfence) are generated.
9864 2018-01-18 Richard Biener <rguenther@suse.de>
9866 PR tree-optimization/83887
9867 * graphite-scop-detection.c
9868 (scop_detection::get_nearest_dom_with_single_entry): Remove.
9869 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
9870 (scop_detection::merge_sese): Re-implement with a flood-fill
9871 algorithm that properly finds a SESE region if it exists.
9873 2018-01-18 Jakub Jelinek <jakub@redhat.com>
9875 PR c/61240
9876 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
9877 pointer_diff optimizations use view_convert instead of convert.
9879 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9881 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9882 Generate different code for -mno-speculate-indirect-jumps.
9883 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9884 (*call_indirect_aix<mode>): Disable for
9885 -mno-speculate-indirect-jumps.
9886 (*call_indirect_aix<mode>_nospec): New define_insn.
9887 (*call_value_indirect_aix<mode>): Disable for
9888 -mno-speculate-indirect-jumps.
9889 (*call_value_indirect_aix<mode>_nospec): New define_insn.
9890 (*sibcall_nonlocal_sysv<mode>): Generate different code for
9891 -mno-speculate-indirect-jumps.
9892 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9894 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
9896 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
9897 long double type, set the flags for noting the default long double
9898 type, even if we don't pass or return a long double type.
9900 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
9902 PR ipa/83051
9903 * ipa-inline.c (flatten_function): Do not overwrite final inlining
9904 failure.
9906 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
9908 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
9909 support for merge[hl].
9910 (fold_mergehl_helper): New helper function.
9911 (tree-vector-builder.h): New #include for tree_vector_builder usage.
9912 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
9913 (altivec_vmrglw_direct): Add xxmrglw insn.
9915 2018-01-17 Andrew Waterman <andrew@sifive.com>
9917 * config/riscv/riscv.c (riscv_conditional_register_usage): If
9918 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
9920 2018-01-17 David Malcolm <dmalcolm@redhat.com>
9922 PR lto/83121
9923 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
9924 call the lto_location_cache before reading the
9925 DECL_SOURCE_LOCATION of the types.
9927 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
9928 Richard Sandiford <richard.sandiford@linaro.org>
9930 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
9931 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
9932 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
9933 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
9934 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
9935 Add declaration.
9936 * config/aarch64/constraints.md (aarch64_movti_operand):
9937 Limit immediates.
9938 * config/aarch64/predicates.md (Uti): Add new constraint.
9940 2018-01-17 Carl Love <cel@us.ibm.com>
9942 * config/rs6000/vsx.md (define_expand xl_len_r,
9943 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
9944 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
9945 lxvll.
9946 (define_expand, define_insn): Move the shift left from the
9947 define_insn to the define_expand for lxvl and stxvl instructions.
9948 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
9949 and XL_LEN_R definitions to PURE.
9951 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
9953 * config/i386/i386.c (indirect_thunk_name): Declare regno
9954 as unsigned int. Compare regno with INVALID_REGNUM.
9955 (output_indirect_thunk): Ditto.
9956 (output_indirect_thunk_function): Ditto.
9957 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
9958 in the call to output_indirect_thunk_function.
9960 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
9962 PR middle-end/83884
9963 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
9964 rather than the size of inner_type to determine the stack slot size
9965 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
9967 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
9969 PR target/83546
9970 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
9971 to PTA_SILVERMONT.
9973 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
9975 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
9976 endian Linux systems to optionally enable multilibs for selecting
9977 the long double type if the user configured an explicit type.
9978 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
9979 have no long double multilibs if not defined.
9980 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
9981 warn if the user used -mabi={ieee,ibm}longdouble and we built
9982 multilibs for long double.
9983 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
9984 appropriate multilib option.
9985 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
9986 multilib options.
9987 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
9988 for building long double multilibs.
9989 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
9991 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
9993 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
9994 copies.
9996 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
9997 64 bits.
9998 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
9999 128 bits.
10001 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
10002 variables.
10004 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
10005 return value.
10007 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10009 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
10010 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
10012 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
10014 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
10015 different rtl trees depending on TARGET_64BIT.
10016 (rs6000_gen_lvx): Likewise.
10018 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10020 * config/visium/visium.md (nop): Tweak comment.
10021 (hazard_nop): Likewise.
10023 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10025 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
10026 -mspeculate-indirect-jumps.
10027 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
10028 for -mno-speculate-indirect-jumps.
10029 (*call_indirect_elfv2<mode>_nospec): New define_insn.
10030 (*call_value_indirect_elfv2<mode>): Disable for
10031 -mno-speculate-indirect-jumps.
10032 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
10033 (indirect_jump): Emit different RTL for
10034 -mno-speculate-indirect-jumps.
10035 (*indirect_jump<mode>): Disable for
10036 -mno-speculate-indirect-jumps.
10037 (*indirect_jump<mode>_nospec): New define_insn.
10038 (tablejump): Emit different RTL for
10039 -mno-speculate-indirect-jumps.
10040 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
10041 (tablejumpsi_nospec): New define_expand.
10042 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
10043 (tablejumpdi_nospec): New define_expand.
10044 (*tablejump<mode>_internal1): Disable for
10045 -mno-speculate-indirect-jumps.
10046 (*tablejump<mode>_internal1_nospec): New define_insn.
10047 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
10048 option.
10050 2018-01-16 Artyom Skrobov tyomitch@gmail.com
10052 * caller-save.c (insert_save): Drop unnecessary parameter. All
10053 callers updated.
10055 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10056 Richard Biener <rguenth@suse.de>
10058 PR libgomp/83590
10059 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
10060 return early, inline manually is_gimple_sizepos. Make sure if we
10061 call gimplify_expr we don't end up with a gimple constant.
10062 * tree.c (variably_modified_type_p): Don't return true for
10063 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
10064 * gimplify.h (is_gimple_sizepos): Remove.
10066 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10068 PR tree-optimization/83857
10069 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
10070 vectorizable_live_operation for pure SLP statements.
10071 (vectorizable_live_operation): Handle PHIs.
10073 2018-01-16 Richard Biener <rguenther@suse.de>
10075 PR tree-optimization/83867
10076 * tree-vect-stmts.c (vect_transform_stmt): Precompute
10077 nested_in_vect_loop_p since the scalar stmt may get invalidated.
10079 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10081 PR c/83844
10082 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
10083 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
10084 If off is not INTEGER_CST, issue a may not be aligned warning
10085 rather than isn't aligned. Use isn%'t rather than isn't.
10086 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
10087 into MULT_EXPR.
10088 <case MULT_EXPR>: Improve the case when bottom and one of the
10089 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
10090 operand, in that case check if the other operand is multiple of
10091 bottom divided by the INTEGER_CST operand.
10093 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10095 PR target/83858
10096 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
10097 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
10098 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
10099 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
10100 * config/pa/pa.c (pa_function_arg_advance): Likewise.
10101 (pa_function_arg, pa_arg_partial_bytes): Likewise.
10102 (pa_function_arg_size): New function.
10104 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10106 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
10107 in a separate statement.
10109 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10111 PR tree-optimization/83847
10112 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
10113 group gathers and scatters.
10115 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10117 PR rtl-optimization/86620
10118 * params.def (max-sched-ready-insns): Bump minimum value to 1.
10120 PR rtl-optimization/83213
10121 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
10122 to last if both are JUMP_INSNs.
10124 PR tree-optimization/83843
10125 * gimple-ssa-store-merging.c
10126 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
10127 store_immediate_info for bswap/nop orig_stores.
10129 2018-01-15 Andrew Waterman <andrew@sifive.com>
10131 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
10132 !TARGET_MUL.
10133 <UDIV>: Increase cost if !TARGET_DIV.
10135 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
10137 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
10138 (define_attr "cr_logical_3op"): New.
10139 (cceq_ior_compare): Adjust.
10140 (cceq_ior_compare_complement): Adjust.
10141 (*cceq_rev_compare): Adjust.
10142 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
10143 (is_cracked_insn): Adjust.
10144 (insn_must_be_first_in_group): Adjust.
10145 * config/rs6000/40x.md: Adjust.
10146 * config/rs6000/440.md: Adjust.
10147 * config/rs6000/476.md: Adjust.
10148 * config/rs6000/601.md: Adjust.
10149 * config/rs6000/603.md: Adjust.
10150 * config/rs6000/6xx.md: Adjust.
10151 * config/rs6000/7450.md: Adjust.
10152 * config/rs6000/7xx.md: Adjust.
10153 * config/rs6000/8540.md: Adjust.
10154 * config/rs6000/cell.md: Adjust.
10155 * config/rs6000/e300c2c3.md: Adjust.
10156 * config/rs6000/e500mc.md: Adjust.
10157 * config/rs6000/e500mc64.md: Adjust.
10158 * config/rs6000/e5500.md: Adjust.
10159 * config/rs6000/e6500.md: Adjust.
10160 * config/rs6000/mpc.md: Adjust.
10161 * config/rs6000/power4.md: Adjust.
10162 * config/rs6000/power5.md: Adjust.
10163 * config/rs6000/power6.md: Adjust.
10164 * config/rs6000/power7.md: Adjust.
10165 * config/rs6000/power8.md: Adjust.
10166 * config/rs6000/power9.md: Adjust.
10167 * config/rs6000/rs64.md: Adjust.
10168 * config/rs6000/titan.md: Adjust.
10170 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10172 * config/i386/predicates.md (indirect_branch_operand): Rewrite
10173 ix86_indirect_branch_register logic.
10175 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10177 * config/i386/constraints.md (Bs): Update
10178 ix86_indirect_branch_register check. Don't check
10179 ix86_indirect_branch_register with GOT_memory_operand.
10180 (Bw): Likewise.
10181 * config/i386/predicates.md (GOT_memory_operand): Don't check
10182 ix86_indirect_branch_register here.
10183 (GOT32_symbol_operand): Likewise.
10185 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10187 * config/i386/predicates.md (constant_call_address_operand):
10188 Rewrite ix86_indirect_branch_register logic.
10189 (sibcall_insn_operand): Likewise.
10191 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10193 * config/i386/constraints.md (Bs): Replace
10194 ix86_indirect_branch_thunk_register with
10195 ix86_indirect_branch_register.
10196 (Bw): Likewise.
10197 * config/i386/i386.md (indirect_jump): Likewise.
10198 (tablejump): Likewise.
10199 (*sibcall_memory): Likewise.
10200 (*sibcall_value_memory): Likewise.
10201 Peepholes of indirect call and jump via memory: Likewise.
10202 * config/i386/i386.opt: Likewise.
10203 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10204 (GOT_memory_operand): Likewise.
10205 (call_insn_operand): Likewise.
10206 (sibcall_insn_operand): Likewise.
10207 (GOT32_symbol_operand): Likewise.
10209 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10211 PR middle-end/83837
10212 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
10213 type rather than type addr's type points to.
10214 (expand_omp_atomic_mutex): Likewise.
10215 (expand_omp_atomic): Likewise.
10217 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10219 PR target/83839
10220 * config/i386/i386.c (output_indirect_thunk_function): Use
10221 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
10222 for __x86_return_thunk.
10224 2018-01-15 Richard Biener <rguenther@suse.de>
10226 PR middle-end/83850
10227 * expmed.c (extract_bit_field_1): Fix typo.
10229 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10231 PR target/83687
10232 * config/arm/iterators.md (VF): New mode iterator.
10233 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
10234 Remove integer-related logic from pattern.
10235 (neon_vabd<mode>_3): Likewise.
10237 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10239 PR middle-end/82694
10240 * common.opt (fstrict-overflow): No longer an alias.
10241 (fwrapv-pointer): New option.
10242 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
10243 also for pointer types based on flag_wrapv_pointer.
10244 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
10245 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
10246 opts->x_flag_wrapv got set.
10247 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
10248 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
10249 POINTER_TYPE_OVERFLOW_UNDEFINED.
10250 * match.pd: Likewise in address comparison pattern.
10251 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
10253 2018-01-15 Richard Biener <rguenther@suse.de>
10255 PR lto/83804
10256 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
10257 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
10258 Reset type names to their identifier if their TYPE_DECL doesn't
10259 have linkage (and thus is used for ODR and devirt).
10260 (save_debug_info_for_decl): Remove.
10261 (save_debug_info_for_type): Likewise.
10262 (add_tree_to_fld_list): Adjust.
10263 * tree-pretty-print.c (dump_generic_node): Make dumping of
10264 type names more robust.
10266 2018-01-15 Richard Biener <rguenther@suse.de>
10268 * BASE-VER: Bump to 8.0.1.
10270 2018-01-14 Martin Sebor <msebor@redhat.com>
10272 PR other/83508
10273 * builtins.c (check_access): Avoid warning when the no-warning bit
10274 is set.
10276 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
10278 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
10279 * ira-color (allocno_hard_regs_compare): Likewise.
10281 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
10283 PR target/83013
10284 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
10285 Use .pushsection/.popsection.
10287 2018-01-14 Martin Sebor <msebor@redhat.com>
10289 PR c++/81327
10290 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
10292 2018-01-14 Jakub Jelinek <jakub@redhat.com>
10294 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
10295 entry from extra_headers.
10296 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
10297 extra_headers, make the list bitwise identical to the i?86-*-* one.
10299 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10301 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
10302 -mcmodel=large with -mindirect-branch=thunk,
10303 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
10304 -mfunction-return=thunk-extern.
10305 * doc/invoke.texi: Document -mcmodel=large is incompatible with
10306 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
10307 -mfunction-return=thunk and -mfunction-return=thunk-extern.
10309 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10311 * config/i386/i386.c (print_reg): Print the name of the full
10312 integer register without '%'.
10313 (ix86_print_operand): Handle 'V'.
10314 * doc/extend.texi: Document 'V' modifier.
10316 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10318 * config/i386/constraints.md (Bs): Disallow memory operand for
10319 -mindirect-branch-register.
10320 (Bw): Likewise.
10321 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10322 (GOT_memory_operand): Likewise.
10323 (call_insn_operand): Likewise.
10324 (sibcall_insn_operand): Likewise.
10325 (GOT32_symbol_operand): Likewise.
10326 * config/i386/i386.md (indirect_jump): Call convert_memory_address
10327 for -mindirect-branch-register.
10328 (tablejump): Likewise.
10329 (*sibcall_memory): Likewise.
10330 (*sibcall_value_memory): Likewise.
10331 Disallow peepholes of indirect call and jump via memory for
10332 -mindirect-branch-register.
10333 (*call_pop): Replace m with Bw.
10334 (*call_value_pop): Likewise.
10335 (*sibcall_pop_memory): Replace m with Bs.
10336 * config/i386/i386.opt (mindirect-branch-register): New option.
10337 * doc/invoke.texi: Document -mindirect-branch-register option.
10339 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10341 * config/i386/i386-protos.h (ix86_output_function_return): New.
10342 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
10343 set function_return_type.
10344 (indirect_thunk_name): Add ret_p to indicate thunk for function
10345 return.
10346 (output_indirect_thunk_function): Pass false to
10347 indirect_thunk_name.
10348 (ix86_output_indirect_branch_via_reg): Likewise.
10349 (ix86_output_indirect_branch_via_push): Likewise.
10350 (output_indirect_thunk_function): Create alias for function
10351 return thunk if regno < 0.
10352 (ix86_output_function_return): New function.
10353 (ix86_handle_fndecl_attribute): Handle function_return.
10354 (ix86_attribute_table): Add function_return.
10355 * config/i386/i386.h (machine_function): Add
10356 function_return_type.
10357 * config/i386/i386.md (simple_return_internal): Use
10358 ix86_output_function_return.
10359 (simple_return_internal_long): Likewise.
10360 * config/i386/i386.opt (mfunction-return=): New option.
10361 (indirect_branch): Mention -mfunction-return=.
10362 * doc/extend.texi: Document function_return function attribute.
10363 * doc/invoke.texi: Document -mfunction-return= option.
10365 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10367 * config/i386/i386-opts.h (indirect_branch): New.
10368 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
10369 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
10370 with local indirect jump when converting indirect call and jump.
10371 (ix86_set_indirect_branch_type): New.
10372 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
10373 (indirectlabelno): New.
10374 (indirect_thunk_needed): Likewise.
10375 (indirect_thunk_bnd_needed): Likewise.
10376 (indirect_thunks_used): Likewise.
10377 (indirect_thunks_bnd_used): Likewise.
10378 (INDIRECT_LABEL): Likewise.
10379 (indirect_thunk_name): Likewise.
10380 (output_indirect_thunk): Likewise.
10381 (output_indirect_thunk_function): Likewise.
10382 (ix86_output_indirect_branch_via_reg): Likewise.
10383 (ix86_output_indirect_branch_via_push): Likewise.
10384 (ix86_output_indirect_branch): Likewise.
10385 (ix86_output_indirect_jmp): Likewise.
10386 (ix86_code_end): Call output_indirect_thunk_function if needed.
10387 (ix86_output_call_insn): Call ix86_output_indirect_branch if
10388 needed.
10389 (ix86_handle_fndecl_attribute): Handle indirect_branch.
10390 (ix86_attribute_table): Add indirect_branch.
10391 * config/i386/i386.h (machine_function): Add indirect_branch_type
10392 and has_local_indirect_jump.
10393 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
10394 to true.
10395 (tablejump): Likewise.
10396 (*indirect_jump): Use ix86_output_indirect_jmp.
10397 (*tablejump_1): Likewise.
10398 (simple_return_indirect_internal): Likewise.
10399 * config/i386/i386.opt (mindirect-branch=): New option.
10400 (indirect_branch): New.
10401 (keep): Likewise.
10402 (thunk): Likewise.
10403 (thunk-inline): Likewise.
10404 (thunk-extern): Likewise.
10405 * doc/extend.texi: Document indirect_branch function attribute.
10406 * doc/invoke.texi: Document -mindirect-branch= option.
10408 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
10410 PR ipa/83051
10411 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
10413 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
10415 * ipa-inline.c (want_inline_small_function_p): Return false if
10416 inlining has already failed with CIF_FINAL_ERROR.
10417 (update_caller_keys): Call want_inline_small_function_p before
10418 can_inline_edge_p.
10419 (update_callee_keys): Likewise.
10421 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
10423 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
10424 New function.
10425 (rs6000_quadword_masked_address_p): Likewise.
10426 (quad_aligned_load_p): Likewise.
10427 (quad_aligned_store_p): Likewise.
10428 (const_load_sequence_p): Add comment to describe the outer-most loop.
10429 (mimic_memory_attributes_and_flags): New function.
10430 (rs6000_gen_stvx): Likewise.
10431 (replace_swapped_aligned_store): Likewise.
10432 (rs6000_gen_lvx): Likewise.
10433 (replace_swapped_aligned_load): Likewise.
10434 (replace_swapped_load_constant): Capitalize argument name in
10435 comment describing this function.
10436 (rs6000_analyze_swaps): Add a third pass to search for vector loads
10437 and stores that access quad-word aligned addresses and replace
10438 with stvx or lvx instructions when appropriate.
10439 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
10440 New function prototype.
10441 (rs6000_quadword_masked_address_p): Likewise.
10442 (rs6000_gen_lvx): Likewise.
10443 (rs6000_gen_stvx): Likewise.
10444 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
10445 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
10446 when memory address is aligned.
10447 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
10448 this split to select lvx instruction when memory address is aligned.
10449 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
10450 instruction when memory address is aligned.
10451 (*vsx_le_perm_load_v16qi): Likewise.
10452 (four unnamed splitters): Modify to select the stvx instruction
10453 when memory is aligned.
10455 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
10457 * predict.c (determine_unlikely_bbs): Handle correctly BBs
10458 which appears in the queue multiple times.
10460 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10461 Alan Hayward <alan.hayward@arm.com>
10462 David Sherwood <david.sherwood@arm.com>
10464 * tree-vectorizer.h (vec_lower_bound): New structure.
10465 (_loop_vec_info): Add check_nonzero and lower_bounds.
10466 (LOOP_VINFO_CHECK_NONZERO): New macro.
10467 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
10468 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
10469 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
10470 fields. Make seg_len the distance travelled, not including the
10471 access size.
10472 (dr_direction_indicator): Declare.
10473 (dr_zero_step_indicator): Likewise.
10474 (dr_known_forward_stride_p): Likewise.
10475 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
10476 tree-ssanames.h.
10477 (runtime_alias_check_p): Allow runtime alias checks with
10478 variable strides.
10479 (operator ==): Compare access_size and align.
10480 (prune_runtime_alias_test_list): Rework for new distinction between
10481 the access_size and seg_len.
10482 (create_intersect_range_checks_index): Likewise. Cope with polynomial
10483 segment lengths.
10484 (get_segment_min_max): New function.
10485 (create_intersect_range_checks): Use it.
10486 (dr_step_indicator): New function.
10487 (dr_direction_indicator): Likewise.
10488 (dr_zero_step_indicator): Likewise.
10489 (dr_known_forward_stride_p): Likewise.
10490 * tree-loop-distribution.c (data_ref_segment_size): Return
10491 DR_STEP * (niters - 1).
10492 (compute_alias_check_pairs): Update call to the dr_with_seg_len
10493 constructor.
10494 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
10495 (vect_preserves_scalar_order_p): New function, split out from...
10496 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
10497 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
10498 (vect_vfa_access_size): New function.
10499 (vect_vfa_align): Likewise.
10500 (vect_compile_time_alias): Take access_size_a and access_b arguments.
10501 (dump_lower_bound): New function.
10502 (vect_check_lower_bound): Likewise.
10503 (vect_small_gap_p): Likewise.
10504 (vectorizable_with_step_bound_p): Likewise.
10505 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
10506 depencies if the vectorization factor is 1. Convert the checks
10507 for nonzero steps into checks on the bounds of DR_STEP. Try using
10508 a bunds check for variable steps if the minimum required step is
10509 relatively small. Update calls to the dr_with_seg_len
10510 constructor and to vect_compile_time_alias.
10511 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
10512 function.
10513 (vect_loop_versioning): Call it.
10514 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
10515 when retrying.
10516 (vect_estimate_min_profitable_iters): Account for any bounds checks.
10518 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10519 Alan Hayward <alan.hayward@arm.com>
10520 David Sherwood <david.sherwood@arm.com>
10522 * doc/sourcebuild.texi (vect_scatter_store): Document.
10523 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
10524 optabs.
10525 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
10526 Document.
10527 * genopinit.c (main): Add supports_vec_scatter_store and
10528 supports_vec_scatter_store_cached to target_optabs.
10529 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
10530 IFN_MASK_SCATTER_STORE.
10531 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
10532 functions.
10533 * internal-fn.h (internal_store_fn_p): Declare.
10534 (internal_fn_stored_value_index): Likewise.
10535 * internal-fn.c (scatter_store_direct): New macro.
10536 (expand_scatter_store_optab_fn): New function.
10537 (direct_scatter_store_optab_supported_p): New macro.
10538 (internal_store_fn_p): New function.
10539 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
10540 IFN_MASK_SCATTER_STORE.
10541 (internal_fn_mask_index): Likewise.
10542 (internal_fn_stored_value_index): New function.
10543 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
10544 for scatter stores.
10545 * optabs-query.h (supports_vec_scatter_store_p): Declare.
10546 * optabs-query.c (supports_vec_scatter_store_p): New function.
10547 * tree-vectorizer.h (vect_get_store_rhs): Declare.
10548 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
10549 true for scatter stores.
10550 (vect_gather_scatter_fn_p): Handle scatter stores too.
10551 (vect_check_gather_scatter): Consider using scatter stores if
10552 supports_vec_scatter_store_p.
10553 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
10554 scatter stores too.
10555 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10556 internal_fn_stored_value_index.
10557 (check_load_store_masking): Handle scatter stores too.
10558 (vect_get_store_rhs): Make public.
10559 (vectorizable_call): Use internal_store_fn_p.
10560 (vectorizable_store): Handle scatter store internal functions.
10561 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
10562 when deciding whether the end of the group has been reached.
10563 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
10564 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
10565 (mask_scatter_store<mode>): New insns.
10567 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10568 Alan Hayward <alan.hayward@arm.com>
10569 David Sherwood <david.sherwood@arm.com>
10571 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
10572 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
10573 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
10574 function.
10575 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
10576 Use vect_truncate_gather_scatter_offset if we can't treat the
10577 operation as a normal gather load or scatter store.
10578 (get_group_load_store_type): Take the gather_scatter_info
10579 as argument. Try using a gather load or scatter store for
10580 single-element groups.
10581 (get_load_store_type): Update calls to get_group_load_store_type
10582 and vect_use_strided_gather_scatters_p.
10584 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10585 Alan Hayward <alan.hayward@arm.com>
10586 David Sherwood <david.sherwood@arm.com>
10588 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
10589 optional tree argument.
10590 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
10591 null target hooks.
10592 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
10593 but continue to use the current value as a fallback.
10594 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
10595 to compare the updates.
10596 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
10597 (get_load_store_type): Use it when handling a strided access.
10598 (vect_get_strided_load_store_ops): New function.
10599 (vect_get_data_ptr_increment): Likewise.
10600 (vectorizable_load): Handle strided gather loads. Always pass
10601 a step to vect_create_data_ref_ptr and bump_vector_ptr.
10603 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10604 Alan Hayward <alan.hayward@arm.com>
10605 David Sherwood <david.sherwood@arm.com>
10607 * doc/md.texi (gather_load@var{m}): Document.
10608 (mask_gather_load@var{m}): Likewise.
10609 * genopinit.c (main): Add supports_vec_gather_load and
10610 supports_vec_gather_load_cached to target_optabs.
10611 * optabs-tree.c (init_tree_optimization_optabs): Use
10612 ggc_cleared_alloc to allocate target_optabs.
10613 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
10614 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
10615 functions.
10616 * internal-fn.h (internal_load_fn_p): Declare.
10617 (internal_gather_scatter_fn_p): Likewise.
10618 (internal_fn_mask_index): Likewise.
10619 (internal_gather_scatter_fn_supported_p): Likewise.
10620 * internal-fn.c (gather_load_direct): New macro.
10621 (expand_gather_load_optab_fn): New function.
10622 (direct_gather_load_optab_supported_p): New macro.
10623 (direct_internal_fn_optab): New function.
10624 (internal_load_fn_p): Likewise.
10625 (internal_gather_scatter_fn_p): Likewise.
10626 (internal_fn_mask_index): Likewise.
10627 (internal_gather_scatter_fn_supported_p): Likewise.
10628 * optabs-query.c (supports_at_least_one_mode_p): New function.
10629 (supports_vec_gather_load_p): Likewise.
10630 * optabs-query.h (supports_vec_gather_load_p): Declare.
10631 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
10632 and memory_type field.
10633 (NUM_PATTERNS): Bump to 15.
10634 * tree-vect-data-refs.c: Include internal-fn.h.
10635 (vect_gather_scatter_fn_p): New function.
10636 (vect_describe_gather_scatter_call): Likewise.
10637 (vect_check_gather_scatter): Try using internal functions for
10638 gather loads. Recognize existing calls to a gather load function.
10639 (vect_analyze_data_refs): Consider using gather loads if
10640 supports_vec_gather_load_p.
10641 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
10642 (vect_get_gather_scatter_offset_type): Likewise.
10643 (vect_convert_mask_for_vectype): Likewise.
10644 (vect_add_conversion_to_patterm): Likewise.
10645 (vect_try_gather_scatter_pattern): Likewise.
10646 (vect_recog_gather_scatter_pattern): New pattern recognizer.
10647 (vect_vect_recog_func_ptrs): Add it.
10648 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
10649 internal_fn_mask_index and internal_gather_scatter_fn_p.
10650 (check_load_store_masking): Take the gather_scatter_info as an
10651 argument and handle gather loads.
10652 (vect_get_gather_scatter_ops): New function.
10653 (vectorizable_call): Check internal_load_fn_p.
10654 (vectorizable_load): Likewise. Handle gather load internal
10655 functions.
10656 (vectorizable_store): Update call to check_load_store_masking.
10657 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
10658 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
10659 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
10660 (aarch64_gather_scale_operand_d): New predicates.
10661 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
10662 (mask_gather_load<mode>): New insns.
10664 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10665 Alan Hayward <alan.hayward@arm.com>
10666 David Sherwood <david.sherwood@arm.com>
10668 * optabs.def (fold_left_plus_optab): New optab.
10669 * doc/md.texi (fold_left_plus_@var{m}): Document.
10670 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
10671 * internal-fn.c (fold_left_direct): Define.
10672 (expand_fold_left_optab_fn): Likewise.
10673 (direct_fold_left_optab_supported_p): Likewise.
10674 * fold-const-call.c (fold_const_fold_left): New function.
10675 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
10676 * tree-parloops.c (valid_reduction_p): New function.
10677 (gather_scalar_reductions): Use it.
10678 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
10679 (vect_finish_replace_stmt): Declare.
10680 * tree-vect-loop.c (fold_left_reduction_fn): New function.
10681 (needs_fold_left_reduction_p): New function, split out from...
10682 (vect_is_simple_reduction): ...here. Accept reductions that
10683 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
10684 (vect_force_simple_reduction): Also store the reduction type in
10685 the assignment's STMT_VINFO_REDUC_TYPE.
10686 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
10687 (merge_with_identity): New function.
10688 (vect_expand_fold_left): Likewise.
10689 (vectorize_fold_left_reduction): Likewise.
10690 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
10691 scalar phi in place for it. Check for target support and reject
10692 cases that would reassociate the operation. Defer the transform
10693 phase to vectorize_fold_left_reduction.
10694 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
10695 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
10696 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
10698 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10700 * tree-if-conv.c (predicate_mem_writes): Remove redundant
10701 call to ifc_temp_var.
10703 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10704 Alan Hayward <alan.hayward@arm.com>
10705 David Sherwood <david.sherwood@arm.com>
10707 * target.def (legitimize_address_displacement): Take the original
10708 offset as a poly_int.
10709 * targhooks.h (default_legitimize_address_displacement): Update
10710 accordingly.
10711 * targhooks.c (default_legitimize_address_displacement): Likewise.
10712 * doc/tm.texi: Regenerate.
10713 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
10714 as an argument, moving assert of ad->disp == ad->disp_term to...
10715 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
10716 Try calling targetm.legitimize_address_displacement before expanding
10717 the address rather than afterwards, and adjust for the new interface.
10718 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
10719 Match the new hook interface. Handle SVE addresses.
10720 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
10721 new hook interface.
10723 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10725 * Makefile.in (OBJS): Add early-remat.o.
10726 * target.def (select_early_remat_modes): New hook.
10727 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
10728 * doc/tm.texi: Regenerate.
10729 * targhooks.h (default_select_early_remat_modes): Declare.
10730 * targhooks.c (default_select_early_remat_modes): New function.
10731 * timevar.def (TV_EARLY_REMAT): New timevar.
10732 * passes.def (pass_early_remat): New pass.
10733 * tree-pass.h (make_pass_early_remat): Declare.
10734 * early-remat.c: New file.
10735 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
10736 function.
10737 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
10739 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10740 Alan Hayward <alan.hayward@arm.com>
10741 David Sherwood <david.sherwood@arm.com>
10743 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
10744 vfm1 with a bound_epilog parameter.
10745 (vect_do_peeling): Update calls accordingly, and move the prologue
10746 call earlier in the function. Treat the base bound_epilog as 0 for
10747 fully-masked loops and retain vf - 1 for other loops. Add 1 to
10748 this base when peeling for gaps.
10749 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
10750 with fully-masked loops.
10751 (vect_estimate_min_profitable_iters): Handle the single peeled
10752 iteration in that case.
10754 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10755 Alan Hayward <alan.hayward@arm.com>
10756 David Sherwood <david.sherwood@arm.com>
10758 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
10759 single-element interleaving even if the size is not a power of 2.
10760 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
10761 accesses for single-element interleaving if the group size is
10762 not a power of 2.
10764 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10765 Alan Hayward <alan.hayward@arm.com>
10766 David Sherwood <david.sherwood@arm.com>
10768 * doc/md.texi (fold_extract_last_@var{m}): Document.
10769 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
10770 * optabs.def (fold_extract_last_optab): New optab.
10771 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
10772 * internal-fn.c (fold_extract_direct): New macro.
10773 (expand_fold_extract_optab_fn): Likewise.
10774 (direct_fold_extract_optab_supported_p): Likewise.
10775 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
10776 * tree-vect-loop.c (vect_model_reduction_cost): Handle
10777 EXTRACT_LAST_REDUCTION.
10778 (get_initial_def_for_reduction): Do not create an initial vector
10779 for EXTRACT_LAST_REDUCTION reductions.
10780 (vectorizable_reduction): Leave the scalar phi in place for
10781 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
10782 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
10783 epilogue code for EXTRACT_LAST_REDUCTION and defer the
10784 transform phase to vectorizable_condition.
10785 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
10786 split out from...
10787 (vect_finish_stmt_generation): ...here.
10788 (vect_finish_replace_stmt): New function.
10789 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
10790 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
10791 pattern.
10792 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
10794 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10795 Alan Hayward <alan.hayward@arm.com>
10796 David Sherwood <david.sherwood@arm.com>
10798 * doc/md.texi (extract_last_@var{m}): Document.
10799 * optabs.def (extract_last_optab): New optab.
10800 * internal-fn.def (EXTRACT_LAST): New internal function.
10801 * internal-fn.c (cond_unary_direct): New macro.
10802 (expand_cond_unary_optab_fn): Likewise.
10803 (direct_cond_unary_optab_supported_p): Likewise.
10804 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
10805 loops using EXTRACT_LAST.
10806 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
10807 (extract_last_<mode>): ...this optab.
10808 (vec_extract<mode><Vel>): Update accordingly.
10810 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10811 Alan Hayward <alan.hayward@arm.com>
10812 David Sherwood <david.sherwood@arm.com>
10814 * target.def (empty_mask_is_expensive): New hook.
10815 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
10816 * doc/tm.texi: Regenerate.
10817 * targhooks.h (default_empty_mask_is_expensive): Declare.
10818 * targhooks.c (default_empty_mask_is_expensive): New function.
10819 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
10820 if the target says that empty masks are expensive.
10821 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
10822 New function.
10823 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
10825 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10826 Alan Hayward <alan.hayward@arm.com>
10827 David Sherwood <david.sherwood@arm.com>
10829 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
10830 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
10831 (vect_use_loop_mask_for_alignment_p): New function.
10832 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
10833 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
10834 niters_skip argument. Make sure that the first niters_skip elements
10835 of the first iteration are inactive.
10836 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
10837 Update call to vect_set_loop_masks_directly.
10838 (get_misalign_in_elems): New function, split out from...
10839 (vect_gen_prolog_loop_niters): ...here.
10840 (vect_update_init_of_dr): Take a code argument that specifies whether
10841 the adjustment should be added or subtracted.
10842 (vect_update_init_of_drs): Likewise.
10843 (vect_prepare_for_masked_peels): New function.
10844 (vect_do_peeling): Skip prologue peeling if we're using a mask
10845 instead. Update call to vect_update_inits_of_drs.
10846 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10847 mask_skip_niters.
10848 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
10849 alignment. Do not include the number of peeled iterations in
10850 the minimum threshold in that case.
10851 (vectorizable_induction): Adjust the start value down by
10852 LOOP_VINFO_MASK_SKIP_NITERS iterations.
10853 (vect_transform_loop): Call vect_prepare_for_masked_peels.
10854 Take the number of skipped iterations into account when calculating
10855 the loop bounds.
10856 * tree-vect-stmts.c (vect_gen_while_not): New function.
10858 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10859 Alan Hayward <alan.hayward@arm.com>
10860 David Sherwood <david.sherwood@arm.com>
10862 * doc/sourcebuild.texi (vect_fully_masked): Document.
10863 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
10864 default value to 0.
10865 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
10866 split out from...
10867 (vect_analyze_loop_2): ...here. Don't check the vectorization
10868 factor against the number of loop iterations if the loop is
10869 fully-masked.
10871 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10872 Alan Hayward <alan.hayward@arm.com>
10873 David Sherwood <david.sherwood@arm.com>
10875 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
10876 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
10877 (dump_groups): Update accordingly.
10878 (iv_use::mem_type): New member variable.
10879 (address_p): New function.
10880 (record_use): Add a mem_type argument and initialize the new
10881 mem_type field.
10882 (record_group_use): Add a mem_type argument. Use address_p.
10883 Remove obsolete null checks of base_object. Update call to record_use.
10884 (find_interesting_uses_op): Update call to record_group_use.
10885 (find_interesting_uses_cond): Likewise.
10886 (find_interesting_uses_address): Likewise.
10887 (get_mem_type_for_internal_fn): New function.
10888 (find_address_like_use): Likewise.
10889 (find_interesting_uses_stmt): Try find_address_like_use before
10890 calling find_interesting_uses_op.
10891 (addr_offset_valid_p): Use the iv mem_type field as the type
10892 of the addressed memory.
10893 (add_autoinc_candidates): Likewise.
10894 (get_address_cost): Likewise.
10895 (split_small_address_groups_p): Use address_p.
10896 (split_address_groups): Likewise.
10897 (add_iv_candidate_for_use): Likewise.
10898 (autoinc_possible_for_pair): Likewise.
10899 (rewrite_groups): Likewise.
10900 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
10901 (determine_group_iv_cost): Update after split of USE_ADDRESS.
10902 (get_alias_ptr_type_for_ptr_address): New function.
10903 (rewrite_use_address): Rewrite address uses in calls that were
10904 identified by find_address_like_use.
10906 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10907 Alan Hayward <alan.hayward@arm.com>
10908 David Sherwood <david.sherwood@arm.com>
10910 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
10911 TARGET_MEM_REFs.
10912 * gimple-expr.h (is_gimple_addressable: Likewise.
10913 * gimple-expr.c (is_gimple_address): Likewise.
10914 * internal-fn.c (expand_call_mem_ref): New function.
10915 (expand_mask_load_optab_fn): Use it.
10916 (expand_mask_store_optab_fn): Likewise.
10918 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10919 Alan Hayward <alan.hayward@arm.com>
10920 David Sherwood <david.sherwood@arm.com>
10922 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
10923 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
10924 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
10925 (cond_umax@var{mode}): Document.
10926 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
10927 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
10928 (cond_umin_optab, cond_umax_optab): New optabs.
10929 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
10930 (COND_IOR, COND_XOR): New internal functions.
10931 * internal-fn.h (get_conditional_internal_fn): Declare.
10932 * internal-fn.c (cond_binary_direct): New macro.
10933 (expand_cond_binary_optab_fn): Likewise.
10934 (direct_cond_binary_optab_supported_p): Likewise.
10935 (get_conditional_internal_fn): New function.
10936 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
10937 Cope with reduction statements that are vectorized as calls rather
10938 than assignments.
10939 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
10940 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
10941 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
10942 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
10943 (UNSPEC_COND_EOR): New unspecs.
10944 (optab): Add mappings for them.
10945 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
10946 (sve_int_op, sve_fp_op): New int attributes.
10948 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10949 Alan Hayward <alan.hayward@arm.com>
10950 David Sherwood <david.sherwood@arm.com>
10952 * optabs.def (while_ult_optab): New optab.
10953 * doc/md.texi (while_ult@var{m}@var{n}): Document.
10954 * internal-fn.def (WHILE_ULT): New internal function.
10955 * internal-fn.h (direct_internal_fn_supported_p): New override
10956 that takes two types as argument.
10957 * internal-fn.c (while_direct): New macro.
10958 (expand_while_optab_fn): New function.
10959 (convert_optab_supported_p): Likewise.
10960 (direct_while_optab_supported_p): New macro.
10961 * wide-int.h (wi::udiv_ceil): New function.
10962 * tree-vectorizer.h (rgroup_masks): New structure.
10963 (vec_loop_masks): New typedef.
10964 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
10965 and fully_masked_p.
10966 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
10967 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
10968 (vect_max_vf): New function.
10969 (slpeel_make_loop_iterate_ntimes): Delete.
10970 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
10971 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
10972 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
10973 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
10974 internal-fn.h, stor-layout.h and optabs-query.h.
10975 (vect_set_loop_mask): New function.
10976 (add_preheader_seq): Likewise.
10977 (add_header_seq): Likewise.
10978 (interleave_supported_p): Likewise.
10979 (vect_maybe_permute_loop_masks): Likewise.
10980 (vect_set_loop_masks_directly): Likewise.
10981 (vect_set_loop_condition_masked): Likewise.
10982 (vect_set_loop_condition_unmasked): New function, split out from
10983 slpeel_make_loop_iterate_ntimes.
10984 (slpeel_make_loop_iterate_ntimes): Rename to..
10985 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
10986 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
10987 (vect_do_peeling): Update call accordingly.
10988 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
10989 loops.
10990 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10991 mask_compare_type, can_fully_mask_p and fully_masked_p.
10992 (release_vec_loop_masks): New function.
10993 (_loop_vec_info): Use it to free the loop masks.
10994 (can_produce_all_loop_masks_p): New function.
10995 (vect_get_max_nscalars_per_iter): Likewise.
10996 (vect_verify_full_masking): Likewise.
10997 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
10998 retries, and free the mask rgroups before retrying. Check loop-wide
10999 reasons for disallowing fully-masked loops. Make the final decision
11000 about whether use a fully-masked loop or not.
11001 (vect_estimate_min_profitable_iters): Do not assume that peeling
11002 for the number of iterations will be needed for fully-masked loops.
11003 (vectorizable_reduction): Disable fully-masked loops.
11004 (vectorizable_live_operation): Likewise.
11005 (vect_halve_mask_nunits): New function.
11006 (vect_double_mask_nunits): Likewise.
11007 (vect_record_loop_mask): Likewise.
11008 (vect_get_loop_mask): Likewise.
11009 (vect_transform_loop): Handle the case in which the final loop
11010 iteration might handle a partial vector. Call vect_set_loop_condition
11011 instead of slpeel_make_loop_iterate_ntimes.
11012 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
11013 (check_load_store_masking): New function.
11014 (prepare_load_store_mask): Likewise.
11015 (vectorizable_store): Handle fully-masked loops.
11016 (vectorizable_load): Likewise.
11017 (supportable_widening_operation): Use vect_halve_mask_nunits for
11018 booleans.
11019 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
11020 (vect_gen_while): New function.
11021 * config/aarch64/aarch64.md (umax<mode>3): New expander.
11022 (aarch64_uqdec<mode>): New insn.
11024 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11025 Alan Hayward <alan.hayward@arm.com>
11026 David Sherwood <david.sherwood@arm.com>
11028 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
11029 (reduc_xor_scal_optab): New optabs.
11030 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
11031 (reduc_xor_scal_@var{m}): Document.
11032 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
11033 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
11034 internal functions.
11035 * fold-const-call.c (fold_const_call): Handle them.
11036 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
11037 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
11038 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
11039 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
11040 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
11041 (UNSPEC_XORV): New unspecs.
11042 (optab): Add entries for them.
11043 (BITWISEV): New int iterator.
11044 (bit_reduc_op): New int attributes.
11046 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11047 Alan Hayward <alan.hayward@arm.com>
11048 David Sherwood <david.sherwood@arm.com>
11050 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
11051 * internal-fn.def (VEC_SHL_INSERT): New internal function.
11052 * optabs.def (vec_shl_insert_optab): New optab.
11053 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
11054 (duplicate_and_interleave): Likewise.
11055 * tree-vect-loop.c: Include internal-fn.h.
11056 (neutral_op_for_slp_reduction): New function, split out from
11057 get_initial_defs_for_reduction.
11058 (get_initial_def_for_reduction): Handle option 2 for variable-length
11059 vectors by loading the neutral value into a vector and then shifting
11060 the initial value into element 0.
11061 (get_initial_defs_for_reduction): Replace the code argument with
11062 the neutral value calculated by neutral_op_for_slp_reduction.
11063 Use gimple_build_vector for constant-length vectors.
11064 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
11065 but the first group_size elements have a neutral value.
11066 Use duplicate_and_interleave otherwise.
11067 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
11068 Update call to get_initial_defs_for_reduction. Handle SLP
11069 reductions for variable-length vectors by creating one vector
11070 result for each scalar result, with the elements associated
11071 with other scalar results stubbed out with the neutral value.
11072 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
11073 Require IFN_VEC_SHL_INSERT for double reductions on
11074 variable-length vectors, or SLP reductions that have
11075 a neutral value. Require can_duplicate_and_interleave_p
11076 support for variable-length unchained SLP reductions if there
11077 is no neutral value, such as for MIN/MAX reductions. Also require
11078 the number of vector elements to be a multiple of the number of
11079 SLP statements when doing variable-length unchained SLP reductions.
11080 Update call to vect_create_epilog_for_reduction.
11081 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
11082 and remove initial values.
11083 (duplicate_and_interleave): Make public.
11084 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
11085 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
11087 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11088 Alan Hayward <alan.hayward@arm.com>
11089 David Sherwood <david.sherwood@arm.com>
11091 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
11092 (can_duplicate_and_interleave_p): New function.
11093 (vect_get_and_check_slp_defs): Take the vector of statements
11094 rather than just the current one. Remove excess parentheses.
11095 Restriction rejectinon of vect_constant_def and vect_external_def
11096 for variable-length vectors to boolean types, or types for which
11097 can_duplicate_and_interleave_p is false.
11098 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
11099 (duplicate_and_interleave): New function.
11100 (vect_get_constant_vectors): Use gimple_build_vector for
11101 constant-length vectors and suitable variable-length constant
11102 vectors. Use duplicate_and_interleave for other variable-length
11103 vectors. Don't defer the update when inserting new statements.
11105 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11106 Alan Hayward <alan.hayward@arm.com>
11107 David Sherwood <david.sherwood@arm.com>
11109 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
11110 min_profitable_iters doesn't go negative.
11112 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11113 Alan Hayward <alan.hayward@arm.com>
11114 David Sherwood <david.sherwood@arm.com>
11116 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
11117 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
11118 * optabs.def (vec_mask_load_lanes_optab): New optab.
11119 (vec_mask_store_lanes_optab): Likewise.
11120 * internal-fn.def (MASK_LOAD_LANES): New internal function.
11121 (MASK_STORE_LANES): Likewise.
11122 * internal-fn.c (mask_load_lanes_direct): New macro.
11123 (mask_store_lanes_direct): Likewise.
11124 (expand_mask_load_optab_fn): Handle masked operations.
11125 (expand_mask_load_lanes_optab_fn): New macro.
11126 (expand_mask_store_optab_fn): Handle masked operations.
11127 (expand_mask_store_lanes_optab_fn): New macro.
11128 (direct_mask_load_lanes_optab_supported_p): Likewise.
11129 (direct_mask_store_lanes_optab_supported_p): Likewise.
11130 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
11131 parameter.
11132 (vect_load_lanes_supported): Likewise.
11133 * tree-vect-data-refs.c (strip_conversion): New function.
11134 (can_group_stmts_p): Likewise.
11135 (vect_analyze_data_ref_accesses): Use it instead of checking
11136 for a pair of assignments.
11137 (vect_store_lanes_supported): Take a masked_p parameter.
11138 (vect_load_lanes_supported): Likewise.
11139 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
11140 vect_store_lanes_supported and vect_load_lanes_supported.
11141 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11142 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
11143 parameter. Don't allow gaps for masked accesses.
11144 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
11145 and vect_load_lanes_supported.
11146 (get_load_store_type): Take a masked_p parameter and update
11147 call to get_group_load_store_type.
11148 (vectorizable_store): Update call to get_load_store_type.
11149 Handle IFN_MASK_STORE_LANES.
11150 (vectorizable_load): Update call to get_load_store_type.
11151 Handle IFN_MASK_LOAD_LANES.
11153 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11154 Alan Hayward <alan.hayward@arm.com>
11155 David Sherwood <david.sherwood@arm.com>
11157 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
11158 modes for SVE.
11159 * config/aarch64/aarch64-protos.h
11160 (aarch64_sve_struct_memory_operand_p): Declare.
11161 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
11162 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
11163 (VPRED, vpred): Handle SVE structure modes.
11164 * config/aarch64/constraints.md (Utx): New constraint.
11165 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
11166 (aarch64_sve_struct_nonimmediate_operand): New predicates.
11167 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
11168 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
11169 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
11170 structure modes. Split into pieces after RA.
11171 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
11172 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
11173 New patterns.
11174 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
11175 SVE structure modes.
11176 (aarch64_classify_address): Likewise.
11177 (sizetochar): Move earlier in file.
11178 (aarch64_print_operand): Handle SVE register lists.
11179 (aarch64_array_mode): New function.
11180 (aarch64_sve_struct_memory_operand_p): Likewise.
11181 (TARGET_ARRAY_MODE): Redefine.
11183 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11184 Alan Hayward <alan.hayward@arm.com>
11185 David Sherwood <david.sherwood@arm.com>
11187 * target.def (array_mode): New target hook.
11188 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
11189 * doc/tm.texi: Regenerate.
11190 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
11191 * hooks.c (hook_optmode_mode_uhwi_none): New function.
11192 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
11193 targetm.array_mode.
11194 * stor-layout.c (mode_for_array): Likewise. Support polynomial
11195 type sizes.
11197 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11198 Alan Hayward <alan.hayward@arm.com>
11199 David Sherwood <david.sherwood@arm.com>
11201 * fold-const.c (fold_binary_loc): Check the argument types
11202 rather than the result type when testing for a vector operation.
11204 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11206 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
11207 * doc/tm.texi: Regenerate.
11209 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11210 Alan Hayward <alan.hayward@arm.com>
11211 David Sherwood <david.sherwood@arm.com>
11213 * doc/invoke.texi (-msve-vector-bits=): Document new option.
11214 (sve): Document new AArch64 extension.
11215 * doc/md.texi (w): Extend the description of the AArch64
11216 constraint to include SVE vectors.
11217 (Upl, Upa): Document new AArch64 predicate constraints.
11218 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
11219 enum.
11220 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
11221 (msve-vector-bits=): New option.
11222 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
11223 SVE when these are disabled.
11224 (sve): New extension.
11225 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
11226 modes. Adjust their number of units based on aarch64_sve_vg.
11227 (MAX_BITSIZE_MODE_ANY_MODE): Define.
11228 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
11229 aarch64_addr_query_type.
11230 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
11231 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
11232 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
11233 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
11234 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
11235 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
11236 (aarch64_simd_imm_zero_p): Delete.
11237 (aarch64_check_zero_based_sve_index_immediate): Declare.
11238 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11239 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11240 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11241 (aarch64_sve_float_mul_immediate_p): Likewise.
11242 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11243 rather than an rtx.
11244 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
11245 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
11246 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
11247 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
11248 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
11249 (aarch64_regmode_natural_size): Likewise.
11250 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
11251 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
11252 left one place.
11253 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
11254 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
11255 for VG and the SVE predicate registers.
11256 (V_ALIASES): Add a "z"-prefixed alias.
11257 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
11258 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
11259 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
11260 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
11261 (REG_CLASS_NAMES): Add entries for them.
11262 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
11263 and the predicate registers.
11264 (aarch64_sve_vg): Declare.
11265 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
11266 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
11267 (REGMODE_NATURAL_SIZE): Define.
11268 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
11269 SVE macros.
11270 * config/aarch64/aarch64.c: Include cfgrtl.h.
11271 (simd_immediate_info): Add a constructor for series vectors,
11272 and an associated step field.
11273 (aarch64_sve_vg): New variable.
11274 (aarch64_dbx_register_number): Handle VG and the predicate registers.
11275 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
11276 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
11277 (VEC_ANY_DATA, VEC_STRUCT): New constants.
11278 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
11279 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
11280 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
11281 (aarch64_get_mask_mode): New functions.
11282 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
11283 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
11284 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
11285 predicate modes and predicate registers. Explicitly restrict
11286 GPRs to modes of 16 bytes or smaller. Only allow FP registers
11287 to store a vector mode if it is recognized by
11288 aarch64_classify_vector_mode.
11289 (aarch64_regmode_natural_size): New function.
11290 (aarch64_hard_regno_caller_save_mode): Return the original mode
11291 for predicates.
11292 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
11293 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
11294 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
11295 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
11296 functions.
11297 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
11298 does not overlap dest if the function is frame-related. Handle
11299 SVE constants.
11300 (aarch64_split_add_offset): New function.
11301 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
11302 them aarch64_add_offset.
11303 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
11304 and update call to aarch64_sub_sp.
11305 (aarch64_add_cfa_expression): New function.
11306 (aarch64_expand_prologue): Pass extra temporary registers to the
11307 functions above. Handle the case in which we need to emit new
11308 DW_CFA_expressions for registers that were originally saved
11309 relative to the stack pointer, but now have to be expressed
11310 relative to the frame pointer.
11311 (aarch64_output_mi_thunk): Pass extra temporary registers to the
11312 functions above.
11313 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
11314 IP0 and IP1 values for SVE frames.
11315 (aarch64_expand_vec_series): New function.
11316 (aarch64_expand_sve_widened_duplicate): Likewise.
11317 (aarch64_expand_sve_const_vector): Likewise.
11318 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
11319 Handle SVE constants. Use emit_move_insn to move a force_const_mem
11320 into the register, rather than emitting a SET directly.
11321 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
11322 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
11323 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
11324 (offset_9bit_signed_scaled_p): New functions.
11325 (aarch64_replicate_bitmask_imm): New function.
11326 (aarch64_bitmask_imm): Use it.
11327 (aarch64_cannot_force_const_mem): Reject expressions involving
11328 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
11329 (aarch64_classify_index): Handle SVE indices, by requiring
11330 a plain register index with a scale that matches the element size.
11331 (aarch64_classify_address): Handle SVE addresses. Assert that
11332 the mode of the address is VOIDmode or an integer mode.
11333 Update call to aarch64_classify_symbol.
11334 (aarch64_classify_symbolic_expression): Update call to
11335 aarch64_classify_symbol.
11336 (aarch64_const_vec_all_in_range_p): New function.
11337 (aarch64_print_vector_float_operand): Likewise.
11338 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
11339 "vN" for FP registers with SVE modes. Handle (const ...) vectors
11340 and the FP immediates 1.0 and 0.5.
11341 (aarch64_print_address_internal): Handle SVE addresses.
11342 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
11343 (aarch64_regno_regclass): Handle predicate registers.
11344 (aarch64_secondary_reload): Handle big-endian reloads of SVE
11345 data modes.
11346 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
11347 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
11348 (aarch64_convert_sve_vector_bits): New function.
11349 (aarch64_override_options): Use it to handle -msve-vector-bits=.
11350 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11351 rather than an rtx.
11352 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
11353 Handle SVE vector and predicate modes. Accept VL-based constants
11354 that need only one temporary register, and VL offsets that require
11355 no temporary registers.
11356 (aarch64_conditional_register_usage): Mark the predicate registers
11357 as fixed if SVE isn't available.
11358 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
11359 Return true for SVE vector and predicate modes.
11360 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
11361 rather than an unsigned int. Handle SVE modes.
11362 (aarch64_preferred_simd_mode): Update call accordingly. Handle
11363 SVE modes.
11364 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
11365 if SVE is enabled.
11366 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11367 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11368 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11369 (aarch64_sve_float_mul_immediate_p): New functions.
11370 (aarch64_sve_valid_immediate): New function.
11371 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
11372 Explicitly reject structure modes. Check for INDEX constants.
11373 Handle PTRUE and PFALSE constants.
11374 (aarch64_check_zero_based_sve_index_immediate): New function.
11375 (aarch64_simd_imm_zero_p): Delete.
11376 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
11377 vector modes. Accept constants in the range of CNT[BHWD].
11378 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
11379 ask for an Advanced SIMD mode.
11380 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
11381 (aarch64_simd_vector_alignment): Handle SVE predicates.
11382 (aarch64_vectorize_preferred_vector_alignment): New function.
11383 (aarch64_simd_vector_alignment_reachable): Use it instead of
11384 the vector size.
11385 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
11386 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
11387 functions.
11388 (MAX_VECT_LEN): Delete.
11389 (expand_vec_perm_d): Add a vec_flags field.
11390 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
11391 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
11392 (aarch64_evpc_ext): Don't apply a big-endian lane correction
11393 for SVE modes.
11394 (aarch64_evpc_rev): Rename to...
11395 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
11396 (aarch64_evpc_rev_global): New function.
11397 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
11398 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
11399 MAX_VECT_LEN.
11400 (aarch64_evpc_sve_tbl): New function.
11401 (aarch64_expand_vec_perm_const_1): Update after rename of
11402 aarch64_evpc_rev. Handle SVE permutes too, trying
11403 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
11404 than aarch64_evpc_tbl.
11405 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
11406 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
11407 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
11408 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
11409 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
11410 (aarch64_expand_sve_vcond): New functions.
11411 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
11412 of aarch64_vector_mode_p.
11413 (aarch64_dwarf_poly_indeterminate_value): New function.
11414 (aarch64_compute_pressure_classes): Likewise.
11415 (aarch64_can_change_mode_class): Likewise.
11416 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
11417 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
11418 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
11419 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
11420 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
11421 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
11422 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
11423 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
11424 constraints.
11425 (Dn, Dl, Dr): Accept const as well as const_vector.
11426 (Dz): Likewise. Compare against CONST0_RTX.
11427 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
11428 of "vector" where appropriate.
11429 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
11430 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
11431 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
11432 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
11433 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
11434 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
11435 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
11436 (v_int_equiv): Extend to SVE modes.
11437 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
11438 mode attributes.
11439 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
11440 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
11441 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
11442 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
11443 (SVE_COND_FP_CMP): New int iterators.
11444 (perm_hilo): Handle the new unpack unspecs.
11445 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
11446 attributes.
11447 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
11448 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
11449 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
11450 (aarch64_equality_operator, aarch64_constant_vector_operand)
11451 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
11452 (aarch64_sve_nonimmediate_operand): Likewise.
11453 (aarch64_sve_general_operand): Likewise.
11454 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
11455 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
11456 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
11457 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
11458 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
11459 (aarch64_sve_float_arith_immediate): Likewise.
11460 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
11461 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
11462 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
11463 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
11464 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
11465 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
11466 (aarch64_sve_float_arith_operand): Likewise.
11467 (aarch64_sve_float_arith_with_sub_operand): Likewise.
11468 (aarch64_sve_float_mul_operand): Likewise.
11469 (aarch64_sve_vec_perm_operand): Likewise.
11470 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
11471 (aarch64_mov_operand): Accept const_poly_int and const_vector.
11472 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
11473 as well as const_vector.
11474 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
11475 in file. Use CONST0_RTX and CONSTM1_RTX.
11476 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
11477 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
11478 Use aarch64_simd_imm_zero.
11479 * config/aarch64/aarch64-sve.md: New file.
11480 * config/aarch64/aarch64.md: Include it.
11481 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
11482 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
11483 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
11484 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
11485 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
11486 (sve): New attribute.
11487 (enabled): Disable instructions with the sve attribute unless
11488 TARGET_SVE.
11489 (movqi, movhi): Pass CONST_POLY_INT operaneds through
11490 aarch64_expand_mov_immediate.
11491 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
11492 CNT[BHSD] immediates.
11493 (movti): Split CONST_POLY_INT moves into two halves.
11494 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
11495 Split additions that need a temporary here if the destination
11496 is the stack pointer.
11497 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
11498 (*add<mode>3_poly_1): New instruction.
11499 (set_clobber_cc): New expander.
11501 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11503 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
11504 parameter and use it instead of GET_MODE_SIZE (innermode). Use
11505 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
11506 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
11507 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
11508 Change innermode from fixed_mode_size to machine_mode.
11509 (simplify_subreg): Update call accordingly. Handle a constant-sized
11510 subreg of a variable-length CONST_VECTOR.
11512 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11513 Alan Hayward <alan.hayward@arm.com>
11514 David Sherwood <david.sherwood@arm.com>
11516 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
11517 (add_offset_to_base): New function, split out from...
11518 (create_mem_ref): ...here. When handling a scale other than 1,
11519 check first whether the address is valid without the offset.
11520 Add it into the base if so, leaving the index and scale as-is.
11522 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11524 PR c++/83778
11525 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
11526 fold_for_warn before checking if arg2 is INTEGER_CST.
11528 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
11530 * config/rs6000/predicates.md (load_multiple_operation): Delete.
11531 (store_multiple_operation): Delete.
11532 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
11533 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
11534 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
11535 guarded by TARGET_STRING.
11536 (rs6000_output_load_multiple): Delete.
11537 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
11538 OPTION_MASK_STRING / TARGET_STRING handling.
11539 (print_operand) <'N', 'O'>: Add comment that these are unused now.
11540 (const rs6000_opt_masks) <"string">: Change mask to 0.
11541 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
11542 (MASK_STRING): Delete.
11543 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
11544 parts. Simplify.
11545 (load_multiple): Delete.
11546 (*ldmsi8): Delete.
11547 (*ldmsi7): Delete.
11548 (*ldmsi6): Delete.
11549 (*ldmsi5): Delete.
11550 (*ldmsi4): Delete.
11551 (*ldmsi3): Delete.
11552 (store_multiple): Delete.
11553 (*stmsi8): Delete.
11554 (*stmsi7): Delete.
11555 (*stmsi6): Delete.
11556 (*stmsi5): Delete.
11557 (*stmsi4): Delete.
11558 (*stmsi3): Delete.
11559 (movmemsi_8reg): Delete.
11560 (corresponding unnamed define_insn): Delete.
11561 (movmemsi_6reg): Delete.
11562 (corresponding unnamed define_insn): Delete.
11563 (movmemsi_4reg): Delete.
11564 (corresponding unnamed define_insn): Delete.
11565 (movmemsi_2reg): Delete.
11566 (corresponding unnamed define_insn): Delete.
11567 (movmemsi_1reg): Delete.
11568 (corresponding unnamed define_insn): Delete.
11569 * config/rs6000/rs6000.opt (mno-string): New.
11570 (mstring): Replace by deprecation warning stub.
11571 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
11573 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11575 * regrename.c (regrename_do_replace): If replacing the same
11576 reg multiple times, try to reuse last created gen_raw_REG.
11578 PR debug/81155
11579 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
11580 main to workaround a bug in GDB.
11582 2018-01-12 Tom de Vries <tom@codesourcery.com>
11584 PR target/83737
11585 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
11587 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
11589 PR rtl-optimization/80481
11590 * ira-color.c (get_cap_member): New function.
11591 (allocnos_conflict_by_live_ranges_p): Use it.
11592 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
11593 (setup_slot_coalesced_allocno_live_ranges): Ditto.
11595 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
11597 PR target/83628
11598 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
11599 (*saddl_se_1): Ditto.
11600 (*ssubsi_1): Ditto.
11601 (*ssubl_se_1): Ditto.
11603 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11605 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
11606 rather than wi::to_widest for DR_INITs.
11607 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
11608 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
11609 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
11610 INTEGER_CSTs.
11611 (vect_analyze_group_access_1): Note that here.
11613 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11615 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
11616 polynomial type sizes.
11618 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
11620 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
11621 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
11622 (gimple_add_tmp_var): Likewise.
11624 2018-01-12 Martin Liska <mliska@suse.cz>
11626 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
11627 (gimple_alloc_sizes): Likewise.
11628 (dump_gimple_statistics): Use PRIu64 in printf format.
11629 * gimple.h: Change uint64_t to int.
11631 2018-01-12 Martin Liska <mliska@suse.cz>
11633 * tree-core.h: Use uint64_t instead of int.
11634 * tree.c (tree_node_counts): Likewise.
11635 (tree_node_sizes): Likewise.
11636 (dump_tree_statistics): Use PRIu64 in printf format.
11638 2018-01-12 Martin Liska <mliska@suse.cz>
11640 * Makefile.in: As qsort_chk is implemented in vec.c, add
11641 vec.o to linkage of gencfn-macros.
11642 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
11643 passing the info to record_node_allocation_statistics.
11644 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
11645 and pass the info.
11646 * ggc-common.c (struct ggc_usage): Add operator== and use
11647 it in operator< and compare function.
11648 * mem-stats.h (struct mem_usage): Likewise.
11649 * vec.c (struct vec_usage): Remove operator< and compare
11650 function. Can be simply inherited.
11652 2018-01-12 Martin Jambor <mjambor@suse.cz>
11654 PR target/81616
11655 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
11656 * tree-ssa-math-opts.c: Include domwalk.h.
11657 (convert_mult_to_fma_1): New function.
11658 (fma_transformation_info): New type.
11659 (fma_deferring_state): Likewise.
11660 (cancel_fma_deferring): New function.
11661 (result_of_phi): Likewise.
11662 (last_fma_candidate_feeds_initial_phi): Likewise.
11663 (convert_mult_to_fma): Added deferring logic, split actual
11664 transformation to convert_mult_to_fma_1.
11665 (math_opts_dom_walker): New type.
11666 (math_opts_dom_walker::after_dom_children): New method, body moved
11667 here from pass_optimize_widening_mul::execute, added deferring logic
11668 bits.
11669 (pass_optimize_widening_mul::execute): Moved most of code to
11670 math_opts_dom_walker::after_dom_children.
11671 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
11672 * config/i386/i386.c (ix86_option_override_internal): Added
11673 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
11675 2018-01-12 Richard Biener <rguenther@suse.de>
11677 PR debug/83157
11678 * dwarf2out.c (gen_variable_die): Do not reset old_die for
11679 inline instance vars.
11681 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
11683 PR target/81819
11684 * config/rx/rx.c (rx_is_restricted_memory_address):
11685 Handle SUBREG case.
11687 2018-01-12 Richard Biener <rguenther@suse.de>
11689 PR tree-optimization/80846
11690 * target.def (split_reduction): New target hook.
11691 * targhooks.c (default_split_reduction): New function.
11692 * targhooks.h (default_split_reduction): Declare.
11693 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
11694 target requests first reduce vectors by combining low and high
11695 parts.
11696 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
11697 (get_vectype_for_scalar_type_and_size): Export.
11698 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
11699 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
11700 * doc/tm.texi: Regenerate.
11701 * config/i386/i386.c (ix86_split_reduction): Implement
11702 TARGET_VECTORIZE_SPLIT_REDUCTION.
11704 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11706 PR target/83368
11707 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
11708 in PIC mode except for TARGET_VXWORKS_RTP.
11709 * config/sparc/sparc.c: Include cfgrtl.h.
11710 (TARGET_INIT_PIC_REG): Define.
11711 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
11712 (sparc_pic_register_p): New predicate.
11713 (sparc_legitimate_address_p): Use it.
11714 (sparc_legitimize_pic_address): Likewise.
11715 (sparc_delegitimize_address): Likewise.
11716 (sparc_mode_dependent_address_p): Likewise.
11717 (gen_load_pcrel_sym): Remove 4th parameter.
11718 (load_got_register): Adjust call to above. Remove obsolete stuff.
11719 (sparc_expand_prologue): Do not call load_got_register here.
11720 (sparc_flat_expand_prologue): Likewise.
11721 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
11722 (sparc_use_pseudo_pic_reg): New function.
11723 (sparc_init_pic_reg): Likewise.
11724 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11725 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
11727 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
11729 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
11730 Add item for branch_cost.
11732 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11734 PR rtl-optimization/83565
11735 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
11736 not extend the result to a larger mode for rotate operations.
11737 (num_sign_bit_copies1): Likewise.
11739 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11741 PR target/40411
11742 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
11743 -symbolic.
11744 Use values-Xc.o for -pedantic.
11745 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
11747 2018-01-12 Martin Liska <mliska@suse.cz>
11749 PR ipa/83054
11750 * ipa-devirt.c (final_warning_record::grow_type_warnings):
11751 New function.
11752 (possible_polymorphic_call_targets): Use it.
11753 (ipa_devirt): Likewise.
11755 2018-01-12 Martin Liska <mliska@suse.cz>
11757 * profile-count.h (enum profile_quality): Use 0 as invalid
11758 enum value of profile_quality.
11760 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
11762 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
11763 -mext-string options.
11765 2018-01-12 Richard Biener <rguenther@suse.de>
11767 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
11768 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
11769 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
11770 Likewise.
11771 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
11773 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
11775 * configure.ac (--with-long-double-format): Add support for the
11776 configuration option to change the default long double format on
11777 PowerPC systems.
11778 * config.gcc (powerpc*-linux*-*): Likewise.
11779 * configure: Regenerate.
11780 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
11781 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
11782 used without modification.
11784 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11786 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
11787 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
11788 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
11789 MISC_BUILTIN_SPEC_BARRIER.
11790 (rs6000_init_builtins): Likewise.
11791 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
11792 enum value.
11793 (speculation_barrier): New define_insn.
11794 * doc/extend.texi: Document __builtin_speculation_barrier.
11796 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11798 PR target/83203
11799 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
11800 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
11801 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
11802 iterators.
11803 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
11804 integral modes instead of "ss" and "sd".
11805 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
11806 vectors with 32-bit and 64-bit elements.
11807 (vecdupssescalarmodesuffix): New mode attribute.
11808 (vec_dup<mode>): Use it.
11810 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
11812 PR target/83330
11813 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
11814 frame if argument is passed on stack.
11816 2018-01-11 Jakub Jelinek <jakub@redhat.com>
11818 PR target/82682
11819 * ree.c (combine_reaching_defs): Optimize also
11820 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
11821 reg2=any_extend(exp); reg1=reg2;, formatting fix.
11823 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11825 PR middle-end/83189
11826 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
11828 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
11830 PR middle-end/83718
11831 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
11832 after they are computed.
11834 2018-01-11 Bin Cheng <bin.cheng@arm.com>
11836 PR tree-optimization/83695
11837 * gimple-loop-linterchange.cc
11838 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
11839 reset cached scev information after interchange.
11840 (pass_linterchange::execute): Remove call to scev_reset_htab.
11842 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11844 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
11845 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
11846 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
11847 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
11848 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
11849 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
11850 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
11851 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
11852 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
11853 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
11854 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
11855 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
11856 (V_lane_reg): Likewise.
11857 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
11858 New define_expand.
11859 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
11860 (vfmal_lane_low<mode>_intrinsic,
11861 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
11862 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
11863 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
11864 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
11865 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
11866 vfmsl_lane_high<mode>_intrinsic): New define_insns.
11868 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11870 * config/arm/arm-cpus.in (fp16fml): New feature.
11871 (ALL_SIMD): Add fp16fml.
11872 (armv8.2-a): Add fp16fml as an option.
11873 (armv8.3-a): Likewise.
11874 (armv8.4-a): Add fp16fml as part of fp16.
11875 * config/arm/arm.h (TARGET_FP16FML): Define.
11876 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
11877 when appropriate.
11878 * config/arm/arm-modes.def (V2HF): Define.
11879 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
11880 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
11881 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
11882 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
11883 vfmsl_low, vfmsl_high): New set of builtins.
11884 * config/arm/iterators.md (PLUSMINUS): New code iterator.
11885 (vfml_op): New code attribute.
11886 (VFMLHALVES): New int iterator.
11887 (VFML, VFMLSEL): New mode attributes.
11888 (V_reg): Define mapping for V2HF.
11889 (V_hi, V_lo): New mode attributes.
11890 (VF_constraint): Likewise.
11891 (vfml_half, vfml_half_selector): New int attributes.
11892 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
11893 define_expand.
11894 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
11895 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
11896 New define_insn.
11897 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
11898 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
11899 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
11900 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
11901 documentation.
11902 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
11903 Document new effective target and option set.
11905 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11907 * config/arm/arm-cpus.in (armv8_4): New feature.
11908 (ARMv8_4a): New fgroup.
11909 (armv8.4-a): New arch.
11910 * config/arm/arm-tables.opt: Regenerate.
11911 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
11912 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
11913 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
11914 Add matching rules for -march=armv8.4-a and extensions.
11915 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
11917 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
11919 PR target/81821
11920 * config/rx/rx.md (BW): New mode attribute.
11921 (sync_lock_test_and_setsi): Add mode suffix to insn output.
11923 2018-01-11 Richard Biener <rguenther@suse.de>
11925 PR tree-optimization/83435
11926 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
11927 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
11928 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
11930 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11931 Alan Hayward <alan.hayward@arm.com>
11932 David Sherwood <david.sherwood@arm.com>
11934 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
11935 field.
11936 (aarch64_classify_address): Initialize it. Track polynomial offsets.
11937 (aarch64_print_address_internal): Use it to check for a zero offset.
11939 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
11940 Alan Hayward <alan.hayward@arm.com>
11941 David Sherwood <david.sherwood@arm.com>
11943 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
11944 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
11945 Return a poly_int64 rather than a HOST_WIDE_INT.
11946 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
11947 rather than a HOST_WIDE_INT.
11948 * config/aarch64/aarch64.h (aarch64_frame): Protect with
11949 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
11950 hard_fp_offset, frame_size, initial_adjust, callee_offset and
11951 final_offset from HOST_WIDE_INT to poly_int64.
11952 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
11953 to_constant when getting the number of units in an Advanced SIMD
11954 mode.
11955 (aarch64_builtin_vectorized_function): Check for a constant number
11956 of units.
11957 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
11958 GET_MODE_SIZE.
11959 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
11960 attribute instead of GET_MODE_NUNITS.
11961 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
11962 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
11963 GET_MODE_SIZE for fixed-size registers.
11964 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
11965 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
11966 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
11967 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
11968 (aarch64_print_operand, aarch64_print_address_internal)
11969 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
11970 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
11971 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
11972 Handle polynomial GET_MODE_SIZE.
11973 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
11974 wider than SImode without modification.
11975 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
11976 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
11977 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
11978 passing and returning SVE modes.
11979 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
11980 rather than GEN_INT.
11981 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
11982 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
11983 (aarch64_allocate_and_probe_stack_space): Likewise.
11984 (aarch64_layout_frame): Cope with polynomial offsets.
11985 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
11986 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
11987 polynomial offsets.
11988 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
11989 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
11990 poly_int64 rather than a HOST_WIDE_INT.
11991 (aarch64_get_separate_components, aarch64_process_components)
11992 (aarch64_expand_prologue, aarch64_expand_epilogue)
11993 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
11994 (aarch64_anchor_offset): New function, split out from...
11995 (aarch64_legitimize_address): ...here.
11996 (aarch64_builtin_vectorization_cost): Handle polynomial
11997 TYPE_VECTOR_SUBPARTS.
11998 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
11999 GET_MODE_NUNITS.
12000 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
12001 number of elements from the PARALLEL rather than the mode.
12002 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
12003 rather than GET_MODE_BITSIZE.
12004 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
12005 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
12006 (aarch64_expand_vec_perm_const_1): Handle polynomial
12007 d->perm.length () and d->perm elements.
12008 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
12009 Apply to_constant to d->perm elements.
12010 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
12011 polynomial CONST_VECTOR_NUNITS.
12012 (aarch64_move_pointer): Take amount as a poly_int64 rather
12013 than an int.
12014 (aarch64_progress_pointer): Avoid temporary variable.
12015 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
12016 the mode attribute instead of GET_MODE.
12018 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12019 Alan Hayward <alan.hayward@arm.com>
12020 David Sherwood <david.sherwood@arm.com>
12022 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
12023 x exists before using it.
12024 (aarch64_add_constant_internal): Rename to...
12025 (aarch64_add_offset_1): ...this. Replace regnum with separate
12026 src and dest rtxes. Handle the case in which they're different,
12027 including when the offset is zero. Replace scratchreg with an rtx.
12028 Use 2 additions if there is no spare register into which we can
12029 move a 16-bit constant.
12030 (aarch64_add_constant): Delete.
12031 (aarch64_add_offset): Replace reg with separate src and dest
12032 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
12033 Use aarch64_add_offset_1.
12034 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
12035 an rtx rather than an int. Take the delta as a poly_int64
12036 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
12037 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
12038 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
12039 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
12040 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
12041 and aarch64_add_sp.
12042 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
12043 aarch64_add_constant.
12045 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12047 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
12048 Use scalar_float_mode.
12050 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12052 * config/aarch64/aarch64-simd.md
12053 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
12054 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
12055 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
12056 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
12057 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
12058 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
12059 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
12060 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
12061 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
12062 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
12064 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12066 PR target/83514
12067 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
12068 targ_options->x_arm_arch_string is non NULL.
12070 2018-01-11 Tamar Christina <tamar.christina@arm.com>
12072 * config/aarch64/aarch64.h
12073 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
12075 2018-01-11 Sudakshina Das <sudi.das@arm.com>
12077 PR target/82096
12078 * expmed.c (emit_store_flag_force): Swap if const op0
12079 and change VOIDmode to mode of op0.
12081 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12083 PR rtl-optimization/83761
12084 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
12085 than bytes to mode_for_size.
12087 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12089 PR middle-end/83189
12090 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
12091 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
12092 profile.
12094 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12096 PR middle-end/83575
12097 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
12098 when in layout mode.
12099 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
12100 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
12101 partition fixup.
12103 2018-01-10 Michael Collison <michael.collison@arm.com>
12105 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
12106 * config/aarch64/aarch64-option-extension.def: Add
12107 AARCH64_OPT_EXTENSION of 'fp16fml'.
12108 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12109 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
12110 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
12111 * config/aarch64/constraints.md (Ui7): New constraint.
12112 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
12113 (VFMLA_SEL_W): Ditto.
12114 (f16quad): Ditto.
12115 (f16mac1): Ditto.
12116 (VFMLA16_LOW): New int iterator.
12117 (VFMLA16_HIGH): Ditto.
12118 (UNSPEC_FMLAL): New unspec.
12119 (UNSPEC_FMLSL): Ditto.
12120 (UNSPEC_FMLAL2): Ditto.
12121 (UNSPEC_FMLSL2): Ditto.
12122 (f16mac): New code attribute.
12123 * config/aarch64/aarch64-simd-builtins.def
12124 (aarch64_fmlal_lowv2sf): Ditto.
12125 (aarch64_fmlsl_lowv2sf): Ditto.
12126 (aarch64_fmlalq_lowv4sf): Ditto.
12127 (aarch64_fmlslq_lowv4sf): Ditto.
12128 (aarch64_fmlal_highv2sf): Ditto.
12129 (aarch64_fmlsl_highv2sf): Ditto.
12130 (aarch64_fmlalq_highv4sf): Ditto.
12131 (aarch64_fmlslq_highv4sf): Ditto.
12132 (aarch64_fmlal_lane_lowv2sf): Ditto.
12133 (aarch64_fmlsl_lane_lowv2sf): Ditto.
12134 (aarch64_fmlal_laneq_lowv2sf): Ditto.
12135 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
12136 (aarch64_fmlalq_lane_lowv4sf): Ditto.
12137 (aarch64_fmlsl_lane_lowv4sf): Ditto.
12138 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
12139 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
12140 (aarch64_fmlal_lane_highv2sf): Ditto.
12141 (aarch64_fmlsl_lane_highv2sf): Ditto.
12142 (aarch64_fmlal_laneq_highv2sf): Ditto.
12143 (aarch64_fmlsl_laneq_highv2sf): Ditto.
12144 (aarch64_fmlalq_lane_highv4sf): Ditto.
12145 (aarch64_fmlsl_lane_highv4sf): Ditto.
12146 (aarch64_fmlalq_laneq_highv4sf): Ditto.
12147 (aarch64_fmlsl_laneq_highv4sf): Ditto.
12148 * config/aarch64/aarch64-simd.md:
12149 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
12150 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12151 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
12152 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12153 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
12154 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
12155 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
12156 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
12157 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
12158 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
12159 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
12160 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
12161 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
12162 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
12163 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
12164 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
12165 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
12166 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
12167 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
12168 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
12169 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
12170 (vfmlsl_low_u32): Ditto.
12171 (vfmlalq_low_u32): Ditto.
12172 (vfmlslq_low_u32): Ditto.
12173 (vfmlal_high_u32): Ditto.
12174 (vfmlsl_high_u32): Ditto.
12175 (vfmlalq_high_u32): Ditto.
12176 (vfmlslq_high_u32): Ditto.
12177 (vfmlal_lane_low_u32): Ditto.
12178 (vfmlsl_lane_low_u32): Ditto.
12179 (vfmlal_laneq_low_u32): Ditto.
12180 (vfmlsl_laneq_low_u32): Ditto.
12181 (vfmlalq_lane_low_u32): Ditto.
12182 (vfmlslq_lane_low_u32): Ditto.
12183 (vfmlalq_laneq_low_u32): Ditto.
12184 (vfmlslq_laneq_low_u32): Ditto.
12185 (vfmlal_lane_high_u32): Ditto.
12186 (vfmlsl_lane_high_u32): Ditto.
12187 (vfmlal_laneq_high_u32): Ditto.
12188 (vfmlsl_laneq_high_u32): Ditto.
12189 (vfmlalq_lane_high_u32): Ditto.
12190 (vfmlslq_lane_high_u32): Ditto.
12191 (vfmlalq_laneq_high_u32): Ditto.
12192 (vfmlslq_laneq_high_u32): Ditto.
12193 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
12194 (AARCH64_FL_FOR_ARCH8_4): New.
12195 (AARCH64_ISA_F16FML): New ISA flag.
12196 (TARGET_F16FML): New feature flag for fp16fml.
12197 (doc/invoke.texi): Document new fp16fml option.
12199 2018-01-10 Michael Collison <michael.collison@arm.com>
12201 * config/aarch64/aarch64-builtins.c:
12202 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
12203 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12204 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
12205 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
12206 (AARCH64_ISA_SHA3): New ISA flag.
12207 (TARGET_SHA3): New feature flag for sha3.
12208 * config/aarch64/iterators.md (sha512_op): New int attribute.
12209 (CRYPTO_SHA512): New int iterator.
12210 (UNSPEC_SHA512H): New unspec.
12211 (UNSPEC_SHA512H2): Ditto.
12212 (UNSPEC_SHA512SU0): Ditto.
12213 (UNSPEC_SHA512SU1): Ditto.
12214 * config/aarch64/aarch64-simd-builtins.def
12215 (aarch64_crypto_sha512hqv2di): New builtin.
12216 (aarch64_crypto_sha512h2qv2di): Ditto.
12217 (aarch64_crypto_sha512su0qv2di): Ditto.
12218 (aarch64_crypto_sha512su1qv2di): Ditto.
12219 (aarch64_eor3qv8hi): Ditto.
12220 (aarch64_rax1qv2di): Ditto.
12221 (aarch64_xarqv2di): Ditto.
12222 (aarch64_bcaxqv8hi): Ditto.
12223 * config/aarch64/aarch64-simd.md:
12224 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
12225 (aarch64_crypto_sha512su0qv2di): Ditto.
12226 (aarch64_crypto_sha512su1qv2di): Ditto.
12227 (aarch64_eor3qv8hi): Ditto.
12228 (aarch64_rax1qv2di): Ditto.
12229 (aarch64_xarqv2di): Ditto.
12230 (aarch64_bcaxqv8hi): Ditto.
12231 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
12232 (vsha512h2q_u64): Ditto.
12233 (vsha512su0q_u64): Ditto.
12234 (vsha512su1q_u64): Ditto.
12235 (veor3q_u16): Ditto.
12236 (vrax1q_u64): Ditto.
12237 (vxarq_u64): Ditto.
12238 (vbcaxq_u16): Ditto.
12239 * config/arm/types.md (crypto_sha512): New type attribute.
12240 (crypto_sha3): Ditto.
12241 (doc/invoke.texi): Document new sha3 option.
12243 2018-01-10 Michael Collison <michael.collison@arm.com>
12245 * config/aarch64/aarch64-builtins.c:
12246 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
12247 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12248 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
12249 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
12250 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
12251 (AARCH64_ISA_SM4): New ISA flag.
12252 (TARGET_SM4): New feature flag for sm4.
12253 * config/aarch64/aarch64-simd-builtins.def
12254 (aarch64_sm3ss1qv4si): Ditto.
12255 (aarch64_sm3tt1aq4si): Ditto.
12256 (aarch64_sm3tt1bq4si): Ditto.
12257 (aarch64_sm3tt2aq4si): Ditto.
12258 (aarch64_sm3tt2bq4si): Ditto.
12259 (aarch64_sm3partw1qv4si): Ditto.
12260 (aarch64_sm3partw2qv4si): Ditto.
12261 (aarch64_sm4eqv4si): Ditto.
12262 (aarch64_sm4ekeyqv4si): Ditto.
12263 * config/aarch64/aarch64-simd.md:
12264 (aarch64_sm3ss1qv4si): Ditto.
12265 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
12266 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
12267 (aarch64_sm4eqv4si): Ditto.
12268 (aarch64_sm4ekeyqv4si): Ditto.
12269 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
12270 (sm3part_op): Ditto.
12271 (CRYPTO_SM3TT): Ditto.
12272 (CRYPTO_SM3PART): Ditto.
12273 (UNSPEC_SM3SS1): New unspec.
12274 (UNSPEC_SM3TT1A): Ditto.
12275 (UNSPEC_SM3TT1B): Ditto.
12276 (UNSPEC_SM3TT2A): Ditto.
12277 (UNSPEC_SM3TT2B): Ditto.
12278 (UNSPEC_SM3PARTW1): Ditto.
12279 (UNSPEC_SM3PARTW2): Ditto.
12280 (UNSPEC_SM4E): Ditto.
12281 (UNSPEC_SM4EKEY): Ditto.
12282 * config/aarch64/constraints.md (Ui2): New constraint.
12283 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
12284 * config/arm/types.md (crypto_sm3): New type attribute.
12285 (crypto_sm4): Ditto.
12286 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
12287 (vsm3tt1aq_u32): Ditto.
12288 (vsm3tt1bq_u32): Ditto.
12289 (vsm3tt2aq_u32): Ditto.
12290 (vsm3tt2bq_u32): Ditto.
12291 (vsm3partw1q_u32): Ditto.
12292 (vsm3partw2q_u32): Ditto.
12293 (vsm4eq_u32): Ditto.
12294 (vsm4ekeyq_u32): Ditto.
12295 (doc/invoke.texi): Document new sm4 option.
12297 2018-01-10 Michael Collison <michael.collison@arm.com>
12299 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
12300 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
12301 (AARCH64_FL_FOR_ARCH8_4): New.
12302 (AARCH64_FL_V8_4): New flag.
12303 (doc/invoke.texi): Document new armv8.4-a option.
12305 2018-01-10 Michael Collison <michael.collison@arm.com>
12307 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12308 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
12309 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
12310 * config/aarch64/aarch64-option-extension.def: Add
12311 AARCH64_OPT_EXTENSION of 'sha2'.
12312 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
12313 (crypto): Disable sha2 and aes if crypto disabled.
12314 (crypto): Enable aes and sha2 if enabled.
12315 (simd): Disable sha2 and aes if simd disabled.
12316 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
12317 New flags.
12318 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
12319 (TARGET_SHA2): New feature flag for sha2.
12320 (TARGET_AES): New feature flag for aes.
12321 * config/aarch64/aarch64-simd.md:
12322 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
12323 conditional on TARGET_AES.
12324 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
12325 (aarch64_crypto_sha1hsi): Make pattern conditional
12326 on TARGET_SHA2.
12327 (aarch64_crypto_sha1hv4si): Ditto.
12328 (aarch64_be_crypto_sha1hv4si): Ditto.
12329 (aarch64_crypto_sha1su1v4si): Ditto.
12330 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
12331 (aarch64_crypto_sha1su0v4si): Ditto.
12332 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
12333 (aarch64_crypto_sha256su0v4si): Ditto.
12334 (aarch64_crypto_sha256su1v4si): Ditto.
12335 (doc/invoke.texi): Document new aes and sha2 options.
12337 2018-01-10 Martin Sebor <msebor@redhat.com>
12339 PR tree-optimization/83781
12340 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
12341 as string arrays.
12343 2018-01-11 Martin Sebor <msebor@gmail.com>
12344 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12346 PR tree-optimization/83501
12347 PR tree-optimization/81703
12349 * tree-ssa-strlen.c (get_string_cst): Rename...
12350 (get_string_len): ...to this. Handle global constants.
12351 (handle_char_store): Adjust.
12353 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
12354 Jim Wilson <jimw@sifive.com>
12356 * config/riscv/riscv-protos.h (riscv_output_return): New.
12357 * config/riscv/riscv.c (struct machine_function): New naked_p field.
12358 (riscv_attribute_table, riscv_output_return),
12359 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
12360 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
12361 (riscv_compute_frame_info): Only compute frame->mask if not a naked
12362 function.
12363 (riscv_expand_prologue): Add early return for naked function.
12364 (riscv_expand_epilogue): Likewise.
12365 (riscv_function_ok_for_sibcall): Return false for naked function.
12366 (riscv_set_current_function): New.
12367 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
12368 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
12369 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
12370 * doc/extend.texi (RISC-V Function Attributes): New.
12372 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
12374 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
12375 check for 128-bit long double before checking TCmode.
12376 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
12377 128-bit long doubles before checking TFmode or TCmode.
12378 (FLOAT128_IBM_P): Likewise.
12380 2018-01-10 Martin Sebor <msebor@redhat.com>
12382 PR tree-optimization/83671
12383 * builtins.c (c_strlen): Unconditionally return zero for the empty
12384 string.
12385 Use -Warray-bounds for warnings.
12386 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
12387 for non-constant array indices with COMPONENT_REF, arrays of
12388 arrays, and pointers to arrays.
12389 (gimple_fold_builtin_strlen): Determine and set length range for
12390 non-constant character arrays.
12392 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
12394 PR middle-end/81897
12395 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
12396 empty blocks.
12398 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
12400 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
12402 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12404 PR target/83399
12405 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
12406 VECTOR_MEM_ALTIVEC_OR_VSX_P.
12407 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
12408 indexed_or_indirect_operand predicate.
12409 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
12410 (*vsx_le_perm_load_v8hi): Likewise.
12411 (*vsx_le_perm_load_v16qi): Likewise.
12412 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
12413 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
12414 (*vsx_le_perm_store_v8hi): Likewise.
12415 (*vsx_le_perm_store_v16qi): Likewise.
12416 (eight unnamed splitters): Likewise.
12418 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12420 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
12421 * config/rs6000/emmintrin.h: Likewise.
12422 * config/rs6000/mmintrin.h: Likewise.
12423 * config/rs6000/xmmintrin.h: Likewise.
12425 2018-01-10 David Malcolm <dmalcolm@redhat.com>
12427 PR c++/43486
12428 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
12429 "public_flag".
12430 * tree.c (tree_nop_conversion): Return true for location wrapper
12431 nodes.
12432 (maybe_wrap_with_location): New function.
12433 (selftest::check_strip_nops): New function.
12434 (selftest::test_location_wrappers): New function.
12435 (selftest::tree_c_tests): Call it.
12436 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
12437 (maybe_wrap_with_location): New decl.
12438 (EXPR_LOCATION_WRAPPER_P): New macro.
12439 (location_wrapper_p): New inline function.
12440 (tree_strip_any_location_wrapper): New inline function.
12442 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
12444 PR target/83735
12445 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
12446 stack_realign_offset for the largest alignment of stack slot
12447 actually used.
12448 (ix86_find_max_used_stack_alignment): New function.
12449 (ix86_finalize_stack_frame_flags): Use it. Set
12450 max_used_stack_alignment if we don't realign stack.
12451 * config/i386/i386.h (machine_function): Add
12452 max_used_stack_alignment.
12454 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
12456 * config/arm/arm.opt (-mbranch-cost): New option.
12457 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
12458 account.
12460 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
12462 PR target/83629
12463 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
12464 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
12466 2018-01-10 Richard Biener <rguenther@suse.de>
12468 PR debug/83765
12469 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
12470 early out so it also covers the case where we have a non-NULL
12471 origin.
12473 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12475 PR tree-optimization/83753
12476 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
12477 for non-strided grouped accesses if the number of elements is 1.
12479 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12481 PR target/81616
12482 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
12483 * i386.h (TARGET_USE_GATHER): Define.
12484 * x86-tune.def (X86_TUNE_USE_GATHER): New.
12486 2018-01-10 Martin Liska <mliska@suse.cz>
12488 PR bootstrap/82831
12489 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
12490 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
12491 partitioning.
12492 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
12493 CLEANUP_NO_PARTITIONING is not set.
12495 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12497 * doc/rtl.texi: Remove documentation of (const ...) wrappers
12498 for vectors, as a partial revert of r254296.
12499 * rtl.h (const_vec_p): Delete.
12500 (const_vec_duplicate_p): Don't test for vector CONSTs.
12501 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
12502 * expmed.c (make_tree): Likewise.
12504 Revert:
12505 * common.md (E, F): Use CONSTANT_P instead of checking for
12506 CONST_VECTOR.
12507 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
12508 checking for CONST_VECTOR.
12510 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12512 PR middle-end/83575
12513 * predict.c (force_edge_cold): Handle in more sane way edges
12514 with no prediction.
12516 2018-01-09 Carl Love <cel@us.ibm.com>
12518 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
12519 V4SI, V4SF types.
12520 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
12521 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
12522 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
12523 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
12524 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
12525 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
12526 * config/rs6000/rs6000-protos.h: Add extern defition for
12527 rs6000_generate_float2_double_code.
12528 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
12529 function.
12530 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
12531 (float2_v2df): Add define_expand.
12533 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
12535 PR target/83628
12536 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
12537 op_mode in the force_to_mode call.
12539 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12541 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
12542 instead of checking each element individually.
12543 (aarch64_evpc_uzp): Likewise.
12544 (aarch64_evpc_zip): Likewise.
12545 (aarch64_evpc_ext): Likewise.
12546 (aarch64_evpc_rev): Likewise.
12547 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
12548 instead of checking each element individually. Return true without
12549 generating rtl if
12550 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
12551 whether all selected elements come from the same input, instead of
12552 checking each element individually. Remove calls to gen_rtx_REG,
12553 start_sequence and end_sequence and instead assert that no rtl is
12554 generated.
12556 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12558 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
12559 order of HIGH and CONST checks.
12561 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12563 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
12564 if the destination isn't an SSA_NAME.
12566 2018-01-09 Richard Biener <rguenther@suse.de>
12568 PR tree-optimization/83668
12569 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
12570 move prologue...
12571 (canonicalize_loop_form): ... here, renamed from ...
12572 (canonicalize_loop_closed_ssa_form): ... this and amended to
12573 swap successor edges for loop exit blocks to make us use
12574 the RPO order we need for initial schedule generation.
12576 2018-01-09 Joseph Myers <joseph@codesourcery.com>
12578 PR tree-optimization/64811
12579 * match.pd: When optimizing comparisons with Inf, avoid
12580 introducing or losing exceptions from comparisons with NaN.
12582 2018-01-09 Martin Liska <mliska@suse.cz>
12584 PR sanitizer/82517
12585 * asan.c (shadow_mem_size): Add gcc_assert.
12587 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
12589 Don't save registers in main().
12591 PR target/83738
12592 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
12593 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
12594 * config/avr/avr.c (avr_set_current_function): Don't error if
12595 naked, OS_task or OS_main are specified at the same time.
12596 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
12597 OS_main.
12598 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
12599 attribute.
12600 * common/config/avr/avr-common.c (avr_option_optimization_table):
12601 Switch on -mmain-is-OS_task for optimizing compilations.
12603 2018-01-09 Richard Biener <rguenther@suse.de>
12605 PR tree-optimization/83572
12606 * graphite.c: Include cfganal.h.
12607 (graphite_transform_loops): Connect infinite loops to exit
12608 and remove fake edges at the end.
12610 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12612 * ipa-inline.c (edge_badness): Revert accidental checkin.
12614 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12616 PR ipa/80763
12617 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
12618 symbols; not inline clones.
12620 2018-01-09 Jakub Jelinek <jakub@redhat.com>
12622 PR target/83507
12623 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
12624 hard registers. Formatting fixes.
12626 PR preprocessor/83722
12627 * gcc.c (try_generate_repro): Pass
12628 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
12629 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
12630 do_report_bug.
12632 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
12633 Kito Cheng <kito.cheng@gmail.com>
12635 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
12636 (riscv_leaf_function_p): Delete.
12637 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
12639 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12641 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
12642 function.
12643 (do_ifelse): New function.
12644 (do_isel): New function.
12645 (do_sub3): New function.
12646 (do_add3): New function.
12647 (do_load_mask_compare): New function.
12648 (do_overlap_load_compare): New function.
12649 (expand_compare_loop): New function.
12650 (expand_block_compare): Call expand_compare_loop() when appropriate.
12651 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
12652 option description.
12653 (-mblock-compare-inline-loop-limit): New option.
12655 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12657 PR target/83677
12658 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
12659 Reverse order of second and third operands in first alternative.
12660 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
12661 of first and second elements in UNSPEC_VPERMR vector.
12662 (altivec_expand_vec_perm_le): Likewise.
12664 2018-01-08 Jeff Law <law@redhat.com>
12666 PR rtl-optimizatin/81308
12667 * tree-switch-conversion.c (cfg_altered): New file scoped static.
12668 (process_switch): If group_case_labels makes a change, then set
12669 cfg_altered.
12670 (pass_convert_switch::execute): If a switch is converted, then
12671 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
12673 PR rtl-optimization/81308
12674 * recog.c (split_all_insns): Conditionally cleanup the CFG after
12675 splitting insns.
12677 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
12679 PR target/83663 - Revert r255946
12680 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
12681 generation for cases where splatting a value is not useful.
12682 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
12683 across a vec_duplicate and a paradoxical subreg forming a vector
12684 mode to a vec_concat.
12686 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12688 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
12689 -march=armv8.3-a variants.
12690 * config/arm/t-multilib: Likewise.
12691 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
12693 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12695 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
12696 to generate rtl.
12697 (cceq_ior_compare_complement): Give it a name so I can use it, and
12698 change boolean_or_operator predicate to boolean_operator so it can
12699 be used to generate a crand.
12700 (eqne): New code iterator.
12701 (bd/bd_neg): New code_attrs.
12702 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
12703 a single define_insn.
12704 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
12705 decrement (bdnzt/bdnzf/bdzt/bdzf).
12706 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
12707 with the new names of the branch decrement patterns, and added the
12708 names of the branch decrement conditional patterns.
12710 2018-01-08 Richard Biener <rguenther@suse.de>
12712 PR tree-optimization/83563
12713 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
12714 cache.
12716 2018-01-08 Richard Biener <rguenther@suse.de>
12718 PR middle-end/83713
12719 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
12721 2018-01-08 Richard Biener <rguenther@suse.de>
12723 PR tree-optimization/83685
12724 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
12725 references to abnormals.
12727 2018-01-08 Richard Biener <rguenther@suse.de>
12729 PR lto/83719
12730 * dwarf2out.c (output_indirect_strings): Handle empty
12731 skeleton_debug_str_hash.
12732 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
12734 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12736 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
12737 (emit_store_direct): Likewise.
12738 (arc_trampoline_adjust_address): Likewise.
12739 (arc_asm_trampoline_template): New function.
12740 (arc_initialize_trampoline): Use asm_trampoline_template.
12741 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
12742 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
12743 * config/arc/arc.md (flush_icache): Delete pattern.
12745 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
12747 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
12748 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
12749 munaligned-access.
12751 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12753 PR target/83681
12754 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
12755 by not USED_FOR_TARGET.
12756 (make_pass_resolve_sw_modes): Likewise.
12758 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
12760 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
12761 USED_FOR_TARGET.
12763 2018-01-08 Richard Biener <rguenther@suse.de>
12765 PR middle-end/83580
12766 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
12768 2018-01-08 Richard Biener <rguenther@suse.de>
12770 PR middle-end/83517
12771 * match.pd ((t * 2) / 2) -> t): Add missing :c.
12773 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
12775 PR middle-end/81897
12776 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
12777 basic blocks with a small number of successors.
12778 (convert_control_dep_chain_into_preds): Improve handling of
12779 forwarder blocks.
12780 (dump_predicates): Split apart into...
12781 (dump_pred_chain): ...here...
12782 (dump_pred_info): ...and here.
12783 (can_one_predicate_be_invalidated_p): Add debugging printfs.
12784 (can_chain_union_be_invalidated_p): Improve check for invalidation
12785 of paths.
12786 (uninit_uses_cannot_happen): Avoid unnecessary if
12787 convert_control_dep_chain_into_preds yielded nothing.
12789 2018-01-06 Martin Sebor <msebor@redhat.com>
12791 PR tree-optimization/83640
12792 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
12793 subtracting negative offset from size.
12794 (builtin_access::overlap): Adjust offset bounds of the access to fall
12795 within the size of the object if possible.
12797 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
12799 PR rtl-optimization/83699
12800 * expmed.c (extract_bit_field_1): Restrict the vector usage of
12801 extract_bit_field_as_subreg to cases in which the extracted
12802 value is also a vector.
12804 * lra-constraints.c (process_alt_operands): Test for the equivalence
12805 substitutions when detecting a possible reload cycle.
12807 2018-01-06 Jakub Jelinek <jakub@redhat.com>
12809 PR debug/83480
12810 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
12811 by default if flag_selective_schedling{,2}. Formatting fixes.
12813 PR rtl-optimization/83682
12814 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
12815 if it has non-VECTOR_MODE element mode.
12816 (vec_duplicate_p): Likewise.
12818 PR middle-end/83694
12819 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
12820 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
12822 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12824 PR target/83604
12825 * config/i386/i386-builtin.def
12826 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
12827 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
12828 Require also OPTION_MASK_ISA_AVX512F in addition to
12829 OPTION_MASK_ISA_GFNI.
12830 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
12831 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
12832 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
12833 to OPTION_MASK_ISA_GFNI.
12834 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
12835 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
12836 OPTION_MASK_ISA_AVX512BW.
12837 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
12838 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
12839 addition to OPTION_MASK_ISA_GFNI.
12840 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
12841 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
12842 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
12843 to OPTION_MASK_ISA_GFNI.
12844 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
12845 a requirement for all ISAs rather than any of them with a few
12846 exceptions.
12847 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
12848 processing.
12849 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
12850 bitmasks to be enabled with 3 exceptions, instead of requiring any
12851 enabled ISA with lots of exceptions.
12852 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
12853 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
12854 Change avx512bw in isa attribute to avx512f.
12855 * config/i386/sgxintrin.h: Add license boilerplate.
12856 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
12857 to __AVX512F__ and __AVX512VL to __AVX512VL__.
12858 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
12859 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
12860 defined.
12861 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
12862 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
12863 temporarily sse2 rather than sse if not enabled already.
12865 PR target/83604
12866 * config/i386/sse.md (VI248_VLBW): Rename to ...
12867 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
12868 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
12869 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
12870 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
12871 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
12872 mode iterator instead of VI248_VLBW.
12874 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
12876 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
12877 (record_modified): Skip clobbers; add debug output.
12878 (param_change_prob): Use sreal frequencies.
12880 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12882 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
12883 punt for user-aligned variables.
12885 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
12887 * tree-chrec.c (chrec_contains_symbols): Return true for
12888 POLY_INT_CST.
12890 2018-01-05 Sudakshina Das <sudi.das@arm.com>
12892 PR target/82439
12893 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
12894 of (x|y) == x for BICS pattern.
12896 2018-01-05 Jakub Jelinek <jakub@redhat.com>
12898 PR tree-optimization/83605
12899 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
12900 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
12901 can throw.
12903 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
12905 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
12906 * config/epiphany/rtems.h: New file.
12908 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12909 Uros Bizjak <ubizjak@gmail.com>
12911 PR target/83554
12912 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
12913 QIreg_operand instead of register_operand predicate.
12914 * config/i386/i386.c (ix86_rop_should_change_byte_p,
12915 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
12916 comments instead of -fmitigate[-_]rop.
12918 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12920 PR bootstrap/81926
12921 * cgraphunit.c (symbol_table::compile): Switch to text_section
12922 before calling assembly_start debug hook.
12923 * run-rtl-passes.c (run_rtl_passes): Likewise.
12924 Include output.h.
12926 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12928 * tree-vrp.c (extract_range_from_binary_expr_1): Check
12929 range_int_cst_p rather than !symbolic_range_p before calling
12930 extract_range_from_multiplicative_op_1.
12932 2018-01-04 Jeff Law <law@redhat.com>
12934 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
12935 redundant test in assertion.
12937 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12939 * doc/rtl.texi: Document machine_mode wrapper classes.
12941 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12943 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
12944 using tree_to_uhwi.
12946 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12948 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
12949 the VEC_PERM_EXPR fold to fail.
12951 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12953 PR debug/83585
12954 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
12955 to switched_sections.
12957 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12959 PR target/83680
12960 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
12961 test for d.testing.
12963 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
12965 PR target/83387
12966 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
12967 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
12969 2018-01-04 Jakub Jelinek <jakub@redhat.com>
12971 PR debug/83666
12972 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
12973 is BLKmode and bitpos not zero or mode change is needed.
12975 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
12977 PR target/83675
12978 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
12979 TARGET_VIS2.
12981 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
12983 PR target/83628
12984 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
12985 instead of MULT rtx. Update all corresponding splitters.
12986 (*saddl_se): Ditto.
12987 (*ssub<modesuffix>): Ditto.
12988 (*ssubl_se): Ditto.
12989 (*cmp_sadd_di): Update split patterns.
12990 (*cmp_sadd_si): Ditto.
12991 (*cmp_sadd_sidi): Ditto.
12992 (*cmp_ssub_di): Ditto.
12993 (*cmp_ssub_si): Ditto.
12994 (*cmp_ssub_sidi): Ditto.
12995 * config/alpha/predicates.md (const23_operand): New predicate.
12996 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
12997 Look for ASHIFT, not MULT inner operand.
12998 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
13000 2018-01-04 Martin Liska <mliska@suse.cz>
13002 PR gcov-profile/83669
13003 * gcov.c (output_intermediate_file): Add version to intermediate
13004 gcov file.
13005 * doc/gcov.texi: Document new field 'version' in intermediate
13006 file format. Fix location of '-k' option of gcov command.
13008 2018-01-04 Martin Liska <mliska@suse.cz>
13010 PR ipa/82352
13011 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
13013 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13015 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
13017 2018-01-03 Martin Sebor <msebor@redhat.com>
13019 PR tree-optimization/83655
13020 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
13021 checking calls with invalid arguments.
13023 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13025 * tree-vect-stmts.c (vect_get_store_rhs): New function.
13026 (vectorizable_mask_load_store): Delete.
13027 (vectorizable_call): Return false for masked loads and stores.
13028 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
13029 instead of gimple_assign_rhs1.
13030 (vectorizable_load): Handle IFN_MASK_LOAD.
13031 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
13033 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13035 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
13036 split out from..,
13037 (vectorizable_mask_load_store): ...here.
13038 (vectorizable_load): ...and here.
13040 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13042 * tree-vect-stmts.c (vect_build_all_ones_mask)
13043 (vect_build_zero_merge_argument): New functions, split out from...
13044 (vectorizable_load): ...here.
13046 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13048 * tree-vect-stmts.c (vect_check_store_rhs): New function,
13049 split out from...
13050 (vectorizable_mask_load_store): ...here.
13051 (vectorizable_store): ...and here.
13053 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13055 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
13056 split out from...
13057 (vectorizable_mask_load_store): ...here.
13059 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13061 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
13062 (vect_model_store_cost): Take a vec_load_store_type instead of a
13063 vect_def_type.
13064 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
13065 (vect_model_store_cost): Take a vec_load_store_type instead of a
13066 vect_def_type.
13067 (vectorizable_mask_load_store): Update accordingly.
13068 (vectorizable_store): Likewise.
13069 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
13071 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13073 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
13074 IFN_MASK_LOAD calls here rather than...
13075 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
13077 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13078 Alan Hayward <alan.hayward@arm.com>
13079 David Sherwood <david.sherwood@arm.com>
13081 * expmed.c (extract_bit_field_1): For vector extracts,
13082 fall back to extract_bit_field_as_subreg if vec_extract
13083 isn't available.
13085 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13086 Alan Hayward <alan.hayward@arm.com>
13087 David Sherwood <david.sherwood@arm.com>
13089 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
13090 they are variable or constant sized.
13091 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
13092 slots for constant-sized data.
13094 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13095 Alan Hayward <alan.hayward@arm.com>
13096 David Sherwood <david.sherwood@arm.com>
13098 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
13099 handling COND_EXPRs with boolean comparisons, try to find a better
13100 basis for the mask type than the boolean itself.
13102 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13104 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
13105 is calculated and how it can be overridden.
13106 * genmodes.c (max_bitsize_mode_any_mode): New variable.
13107 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
13108 if defined.
13109 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
13110 if nonzero.
13112 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13113 Alan Hayward <alan.hayward@arm.com>
13114 David Sherwood <david.sherwood@arm.com>
13116 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
13117 Remove the mode argument.
13118 (aarch64_simd_valid_immediate): Remove the mode and inverse
13119 arguments.
13120 * config/aarch64/iterators.md (bitsize): New iterator.
13121 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
13122 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
13123 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
13124 aarch64_simd_valid_immediate.
13125 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
13126 (aarch64_reg_or_bic_imm): Likewise.
13127 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
13128 with an insn_type enum and msl with a modifier_type enum.
13129 Replace element_width with a scalar_mode. Change the shift
13130 to unsigned int. Add constructors for scalar_float_mode and
13131 scalar_int_mode elements.
13132 (aarch64_vect_float_const_representable_p): Delete.
13133 (aarch64_can_const_movi_rtx_p)
13134 (aarch64_simd_scalar_immediate_valid_for_move)
13135 (aarch64_simd_make_constant): Update call to
13136 aarch64_simd_valid_immediate.
13137 (aarch64_advsimd_valid_immediate_hs): New function.
13138 (aarch64_advsimd_valid_immediate): Likewise.
13139 (aarch64_simd_valid_immediate): Remove mode and inverse
13140 arguments. Rewrite to use the above. Use const_vec_duplicate_p
13141 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
13142 and aarch64_float_const_representable_p on the result.
13143 (aarch64_output_simd_mov_immediate): Remove mode argument.
13144 Update call to aarch64_simd_valid_immediate and use of
13145 simd_immediate_info.
13146 (aarch64_output_scalar_simd_mov_immediate): Update call
13147 accordingly.
13149 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13150 Alan Hayward <alan.hayward@arm.com>
13151 David Sherwood <david.sherwood@arm.com>
13153 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
13154 (mode_nunits): Likewise CONST_MODE_NUNITS.
13155 * machmode.def (ADJUST_NUNITS): Document.
13156 * genmodes.c (mode_data::need_nunits_adj): New field.
13157 (blank_mode): Update accordingly.
13158 (adj_nunits): New variable.
13159 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
13160 parameter.
13161 (emit_mode_size_inline): Set need_bytesize_adj for all modes
13162 listed in adj_nunits.
13163 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
13164 listed in adj_nunits. Don't emit case statements for such modes.
13165 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
13166 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
13167 nothing if adj_nunits is nonnull.
13168 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
13169 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
13170 (emit_mode_fbit): Update use of print_maybe_const_decl.
13171 (emit_move_size): Likewise. Treat the array as non-const
13172 if adj_nunits.
13173 (emit_mode_adjustments): Handle adj_nunits.
13175 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13177 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
13178 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
13179 (VECTOR_MODES): Use it.
13180 (make_vector_modes): Take the prefix as an argument.
13182 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13183 Alan Hayward <alan.hayward@arm.com>
13184 David Sherwood <david.sherwood@arm.com>
13186 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
13187 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
13188 for MODE_VECTOR_BOOL.
13189 * machmode.def (VECTOR_BOOL_MODE): Document.
13190 * genmodes.c (VECTOR_BOOL_MODE): New macro.
13191 (make_vector_bool_mode): New function.
13192 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
13193 MODE_VECTOR_BOOL.
13194 * lto-streamer-in.c (lto_input_mode_table): Likewise.
13195 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
13196 Likewise.
13197 * stor-layout.c (int_mode_for_mode): Likewise.
13198 * tree.c (build_vector_type_for_mode): Likewise.
13199 * varasm.c (output_constant_pool_2): Likewise.
13200 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
13201 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
13202 for MODE_VECTOR_BOOL.
13203 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
13204 of mode class checks.
13205 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
13206 instead of a list of mode class checks.
13207 (expand_vector_scalar_condition): Likewise.
13208 (type_for_widest_vector_mode): Handle BImode as an inner mode.
13210 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13211 Alan Hayward <alan.hayward@arm.com>
13212 David Sherwood <david.sherwood@arm.com>
13214 * machmode.h (mode_size): Change from unsigned short to
13215 poly_uint16_pod.
13216 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
13217 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13218 or if measurement_type is not polynomial.
13219 (fixed_size_mode::includes_p): Check for constant-sized modes.
13220 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
13221 return a poly_uint16 rather than an unsigned short.
13222 (emit_mode_size): Change the type of mode_size from unsigned short
13223 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
13224 (emit_mode_adjustments): Cope with polynomial vector sizes.
13225 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13226 for GET_MODE_SIZE.
13227 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13228 for GET_MODE_SIZE.
13229 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
13230 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
13231 * caller-save.c (setup_save_areas): Likewise.
13232 (replace_reg_with_saved_mem): Likewise.
13233 * calls.c (emit_library_call_value_1): Likewise.
13234 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
13235 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
13236 (gen_lowpart_for_combine): Likewise.
13237 * convert.c (convert_to_integer_1): Likewise.
13238 * cse.c (equiv_constant, cse_insn): Likewise.
13239 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
13240 (cselib_subst_to_values): Likewise.
13241 * dce.c (word_dce_process_block): Likewise.
13242 * df-problems.c (df_word_lr_mark_ref): Likewise.
13243 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
13244 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
13245 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
13246 (rtl_for_decl_location): Likewise.
13247 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
13248 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
13249 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
13250 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
13251 (expand_expr_real_1): Likewise.
13252 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
13253 (pad_below): Likewise.
13254 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13255 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
13256 * ira.c (get_subreg_tracking_sizes): Likewise.
13257 * ira-build.c (ira_create_allocno_objects): Likewise.
13258 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
13259 (ira_sort_regnos_for_alter_reg): Likewise.
13260 * ira-costs.c (record_operand_costs): Likewise.
13261 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
13262 (resolve_simple_move): Likewise.
13263 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
13264 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
13265 (lra_constraints): Likewise.
13266 (CONST_POOL_OK_P): Reject variable-sized modes.
13267 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
13268 (add_pseudo_to_slot, lra_spill): Likewise.
13269 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13270 * optabs-query.c (get_best_extraction_insn): Likewise.
13271 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13272 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
13273 (expand_mult_highpart, valid_multiword_target_p): Likewise.
13274 * recog.c (offsettable_address_addr_space_p): Likewise.
13275 * regcprop.c (maybe_mode_change): Likewise.
13276 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
13277 * regrename.c (build_def_use): Likewise.
13278 * regstat.c (dump_reg_info): Likewise.
13279 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
13280 (find_reloads, find_reloads_subreg_address): Likewise.
13281 * reload1.c (eliminate_regs_1): Likewise.
13282 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
13283 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
13284 (simplify_binary_operation_1, simplify_subreg): Likewise.
13285 * targhooks.c (default_function_arg_padding): Likewise.
13286 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
13287 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
13288 (verify_gimple_assign_ternary): Likewise.
13289 * tree-inline.c (estimate_move_cost): Likewise.
13290 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13291 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
13292 (get_address_cost_ainc): Likewise.
13293 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
13294 (vect_supportable_dr_alignment): Likewise.
13295 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13296 (vectorizable_reduction): Likewise.
13297 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
13298 (vectorizable_operation, vectorizable_load): Likewise.
13299 * tree.c (build_same_sized_truth_vector_type): Likewise.
13300 * valtrack.c (cleanup_auto_inc_dec): Likewise.
13301 * var-tracking.c (emit_note_insn_var_location): Likewise.
13302 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
13303 (ADDR_VEC_ALIGN): Likewise.
13305 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13306 Alan Hayward <alan.hayward@arm.com>
13307 David Sherwood <david.sherwood@arm.com>
13309 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
13310 unsigned short.
13311 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13312 or if measurement_type is polynomial.
13313 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
13314 * combine.c (make_extraction): Likewise.
13315 * dse.c (find_shift_sequence): Likewise.
13316 * dwarf2out.c (mem_loc_descriptor): Likewise.
13317 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
13318 (extract_bit_field, extract_low_bits): Likewise.
13319 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
13320 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
13321 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
13322 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
13323 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13324 * reload.c (find_reloads): Likewise.
13325 * reload1.c (alter_reg): Likewise.
13326 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
13327 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
13328 * tree-if-conv.c (predicate_mem_writes): Likewise.
13329 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
13330 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
13331 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
13332 * valtrack.c (dead_debug_insert_temp): Likewise.
13333 * varasm.c (mergeable_constant_section): Likewise.
13334 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
13336 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13337 Alan Hayward <alan.hayward@arm.com>
13338 David Sherwood <david.sherwood@arm.com>
13340 * expr.c (expand_assignment): Cope with polynomial mode sizes
13341 when assigning to a CONCAT.
13343 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13344 Alan Hayward <alan.hayward@arm.com>
13345 David Sherwood <david.sherwood@arm.com>
13347 * machmode.h (mode_precision): Change from unsigned short to
13348 poly_uint16_pod.
13349 (mode_to_precision): Return a poly_uint16 rather than an unsigned
13350 short.
13351 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
13352 or if measurement_type is not polynomial.
13353 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
13354 in which the mode is already known to be a scalar_int_mode.
13355 * genmodes.c (emit_mode_precision): Change the type of mode_precision
13356 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
13357 initializer.
13358 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13359 for GET_MODE_PRECISION.
13360 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13361 for GET_MODE_PRECISION.
13362 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
13363 as polynomial.
13364 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
13365 (expand_field_assignment, make_extraction): Likewise.
13366 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
13367 (get_last_value): Likewise.
13368 * convert.c (convert_to_integer_1): Likewise.
13369 * cse.c (cse_insn): Likewise.
13370 * expr.c (expand_expr_real_1): Likewise.
13371 * lra-constraints.c (simplify_operand_subreg): Likewise.
13372 * optabs-query.c (can_atomic_load_p): Likewise.
13373 * optabs.c (expand_atomic_load): Likewise.
13374 (expand_atomic_store): Likewise.
13375 * ree.c (combine_reaching_defs): Likewise.
13376 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
13377 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
13378 * tree.h (type_has_mode_precision_p): Likewise.
13379 * ubsan.c (instrument_si_overflow): Likewise.
13381 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13382 Alan Hayward <alan.hayward@arm.com>
13383 David Sherwood <david.sherwood@arm.com>
13385 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
13386 polynomial numbers of units.
13387 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
13388 (valid_vector_subparts_p): New function.
13389 (build_vector_type): Remove temporary shim and take the number
13390 of units as a poly_uint64 rather than an int.
13391 (build_opaque_vector_type): Take the number of units as a
13392 poly_uint64 rather than an int.
13393 * tree.c (build_vector_from_ctor): Handle polynomial
13394 TYPE_VECTOR_SUBPARTS.
13395 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
13396 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
13397 (build_vector_from_val): If the number of units is variable,
13398 use build_vec_duplicate_cst for constant operands and
13399 VEC_DUPLICATE_EXPR otherwise.
13400 (make_vector_type): Remove temporary is_constant ().
13401 (build_vector_type, build_opaque_vector_type): Take the number of
13402 units as a poly_uint64 rather than an int.
13403 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
13404 VECTOR_CST_NELTS.
13405 * cfgexpand.c (expand_debug_expr): Likewise.
13406 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
13407 (store_constructor, expand_expr_real_1): Likewise.
13408 (const_scalar_mask_from_tree): Likewise.
13409 * fold-const-call.c (fold_const_reduction): Likewise.
13410 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
13411 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
13412 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
13413 (fold_relational_const): Likewise.
13414 (native_interpret_vector): Likewise. Change the size from an
13415 int to an unsigned int.
13416 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
13417 TYPE_VECTOR_SUBPARTS.
13418 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
13419 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
13420 duplicating a non-constant operand into a variable-length vector.
13421 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
13422 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
13423 * ipa-icf.c (sem_variable::equals): Likewise.
13424 * match.pd: Likewise.
13425 * omp-simd-clone.c (simd_clone_subparts): Likewise.
13426 * print-tree.c (print_node): Likewise.
13427 * stor-layout.c (layout_type): Likewise.
13428 * targhooks.c (default_builtin_vectorization_cost): Likewise.
13429 * tree-cfg.c (verify_gimple_comparison): Likewise.
13430 (verify_gimple_assign_binary): Likewise.
13431 (verify_gimple_assign_ternary): Likewise.
13432 (verify_gimple_assign_single): Likewise.
13433 * tree-pretty-print.c (dump_generic_node): Likewise.
13434 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13435 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
13436 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
13437 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
13438 (vect_shift_permute_load_chain): Likewise.
13439 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
13440 (expand_vector_condition, optimize_vector_constructor): Likewise.
13441 (lower_vec_perm, get_compute_type): Likewise.
13442 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13443 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
13444 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
13445 (vect_recog_mask_conversion_pattern): Likewise.
13446 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
13447 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
13448 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13449 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
13450 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
13451 (vectorizable_shift, vectorizable_operation, vectorizable_store)
13452 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
13453 (supportable_widening_operation): Likewise.
13454 (supportable_narrowing_operation): Likewise.
13455 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
13456 Likewise.
13457 * varasm.c (output_constant): Likewise.
13459 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13460 Alan Hayward <alan.hayward@arm.com>
13461 David Sherwood <david.sherwood@arm.com>
13463 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
13464 so that both the length == 3 and length != 3 cases set up their
13465 own permute vectors. Add comments explaining why we know the
13466 number of elements is constant.
13467 (vect_permute_load_chain): Likewise.
13469 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13470 Alan Hayward <alan.hayward@arm.com>
13471 David Sherwood <david.sherwood@arm.com>
13473 * machmode.h (mode_nunits): Change from unsigned char to
13474 poly_uint16_pod.
13475 (ONLY_FIXED_SIZE_MODES): New macro.
13476 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
13477 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
13478 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
13479 New typedefs.
13480 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
13481 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
13482 or if measurement_type is not polynomial.
13483 * genmodes.c (ZERO_COEFFS): New macro.
13484 (emit_mode_nunits_inline): Make mode_nunits_inline return a
13485 poly_uint16.
13486 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
13487 Use ZERO_COEFFS when emitting initializers.
13488 * data-streamer.h (bp_pack_poly_value): New function.
13489 (bp_unpack_poly_value): Likewise.
13490 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13491 for GET_MODE_NUNITS.
13492 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13493 for GET_MODE_NUNITS.
13494 * tree.c (make_vector_type): Remove temporary shim and make
13495 the real function take the number of units as a poly_uint64
13496 rather than an int.
13497 (build_vector_type_for_mode): Handle polynomial nunits.
13498 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
13499 * emit-rtl.c (const_vec_series_p_1): Likewise.
13500 (gen_rtx_CONST_VECTOR): Likewise.
13501 * fold-const.c (test_vec_duplicate_folding): Likewise.
13502 * genrecog.c (validate_pattern): Likewise.
13503 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
13504 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13505 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
13506 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
13507 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
13508 * rtlanal.c (subreg_get_info): Likewise.
13509 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13510 (vect_grouped_load_supported): Likewise.
13511 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
13512 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
13513 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
13514 (simplify_const_unary_operation, simplify_binary_operation_1)
13515 (simplify_const_binary_operation, simplify_ternary_operation)
13516 (test_vector_ops_duplicate, test_vector_ops): Likewise.
13517 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
13518 instead of CONST_VECTOR_NUNITS.
13519 * varasm.c (output_constant_pool_2): Likewise.
13520 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
13521 explicit-encoded elements in the XVEC for variable-length vectors.
13523 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13525 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
13527 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13528 Alan Hayward <alan.hayward@arm.com>
13529 David Sherwood <david.sherwood@arm.com>
13531 * coretypes.h (fixed_size_mode): Declare.
13532 (fixed_size_mode_pod): New typedef.
13533 * builtins.h (target_builtins::x_apply_args_mode)
13534 (target_builtins::x_apply_result_mode): Change type to
13535 fixed_size_mode_pod.
13536 * builtins.c (apply_args_size, apply_result_size, result_vector)
13537 (expand_builtin_apply_args_1, expand_builtin_apply)
13538 (expand_builtin_return): Update accordingly.
13540 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13542 * cse.c (hash_rtx_cb): Hash only the encoded elements.
13543 * cselib.c (cselib_hash_rtx): Likewise.
13544 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
13545 CONST_VECTOR encoding.
13547 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13548 Jeff Law <law@redhat.com>
13550 PR target/83641
13551 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
13552 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
13553 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
13554 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
13556 PR target/83641
13557 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
13558 explicitly probe *sp in a noreturn function if there were any callee
13559 register saves or frame pointer is needed.
13561 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13563 PR debug/83621
13564 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
13565 BLKmode for ternary, binary or unary expressions.
13567 PR debug/83645
13568 * var-tracking.c (delete_vta_debug_insn): New inline function.
13569 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
13570 insns from get_insns () to NULL instead of each bb separately.
13571 Use delete_vta_debug_insn. No longer static.
13572 (vt_debug_insns_local, variable_tracking_main_1): Adjust
13573 delete_vta_debug_insns callers.
13574 * rtl.h (delete_vta_debug_insns): Declare.
13575 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
13576 instead of variable_tracking_main.
13578 2018-01-03 Martin Sebor <msebor@redhat.com>
13580 PR tree-optimization/83603
13581 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
13582 arguments past the endof the argument list in functions declared
13583 without a prototype.
13584 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
13585 Avoid checking when arguments are null.
13587 2018-01-03 Martin Sebor <msebor@redhat.com>
13589 PR c/83559
13590 * doc/extend.texi (attribute const): Fix a typo.
13591 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
13592 issuing -Wsuggest-attribute for void functions.
13594 2018-01-03 Martin Sebor <msebor@redhat.com>
13596 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
13597 offset_int::from instead of wide_int::to_shwi.
13598 (maybe_diag_overlap): Remove assertion.
13599 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
13600 * gimple-ssa-sprintf.c (format_directive): Same.
13601 (parse_directive): Same.
13602 (sprintf_dom_walker::compute_format_length): Same.
13603 (try_substitute_return_value): Same.
13605 2018-01-03 Jeff Law <law@redhat.com>
13607 PR middle-end/83654
13608 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
13609 non-constant residual for zero at runtime and avoid probing in
13610 that case. Reorganize code for trailing problem to mirror handling
13611 of the residual.
13613 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13615 PR tree-optimization/83501
13616 * tree-ssa-strlen.c (get_string_cst): New.
13617 (handle_char_store): Call get_string_cst.
13619 2018-01-03 Martin Liska <mliska@suse.cz>
13621 PR tree-optimization/83593
13622 * tree-ssa-strlen.c: Include tree-cfg.h.
13623 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
13624 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
13625 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
13626 to false.
13627 (strlen_dom_walker::before_dom_children): Call
13628 gimple_purge_dead_eh_edges. Dump tranformation with details
13629 dump flags.
13630 (strlen_dom_walker::before_dom_children): Update call by adding
13631 new argument cleanup_eh.
13632 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
13634 2018-01-03 Martin Liska <mliska@suse.cz>
13636 PR ipa/83549
13637 * cif-code.def (VARIADIC_THUNK): New enum value.
13638 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
13639 thunks.
13641 2018-01-03 Jan Beulich <jbeulich@suse.com>
13643 * sse.md (mov<mode>_internal): Tighten condition for when to use
13644 vmovdqu<ssescalarsize> for TI and OI modes.
13646 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13648 Update copyright years.
13650 2018-01-03 Martin Liska <mliska@suse.cz>
13652 PR ipa/83594
13653 * ipa-visibility.c (function_and_variable_visibility): Skip
13654 functions with noipa attribure.
13656 2018-01-03 Jakub Jelinek <jakub@redhat.com>
13658 * gcc.c (process_command): Update copyright notice dates.
13659 * gcov-dump.c (print_version): Ditto.
13660 * gcov.c (print_version): Ditto.
13661 * gcov-tool.c (print_version): Ditto.
13662 * gengtype.c (create_file): Ditto.
13663 * doc/cpp.texi: Bump @copying's copyright year.
13664 * doc/cppinternals.texi: Ditto.
13665 * doc/gcc.texi: Ditto.
13666 * doc/gccint.texi: Ditto.
13667 * doc/gcov.texi: Ditto.
13668 * doc/install.texi: Ditto.
13669 * doc/invoke.texi: Ditto.
13671 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13673 * vector-builder.h (vector_builder::m_full_nelts): Change from
13674 unsigned int to poly_uint64.
13675 (vector_builder::full_nelts): Update prototype accordingly.
13676 (vector_builder::new_vector): Likewise.
13677 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
13678 (vector_builder::operator ==): Likewise.
13679 (vector_builder::finalize): Likewise.
13680 * int-vector-builder.h (int_vector_builder::int_vector_builder):
13681 Take the number of elements as a poly_uint64 rather than an
13682 unsigned int.
13683 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
13684 from unsigned int to poly_uint64.
13685 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
13686 (vec_perm_indices::new_vector): Likewise.
13687 (vec_perm_indices::length): Likewise.
13688 (vec_perm_indices::nelts_per_input): Likewise.
13689 (vec_perm_indices::input_nelts): Likewise.
13690 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
13691 number of elements per input as a poly_uint64 rather than an
13692 unsigned int. Use the original encoding for variable-length
13693 vectors, rather than clamping each individual element.
13694 For the second and subsequent elements in each pattern,
13695 clamp the step and base before clamping their sum.
13696 (vec_perm_indices::series_p): Handle polynomial element counts.
13697 (vec_perm_indices::all_in_range_p): Likewise.
13698 (vec_perm_indices_to_tree): Likewise.
13699 (vec_perm_indices_to_rtx): Likewise.
13700 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
13701 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
13702 (tree_vector_builder::new_binary_operation): Handle polynomial
13703 element counts. Return false if we need to know the number
13704 of elements at compile time.
13705 * fold-const.c (fold_vec_perm): Punt if the number of elements
13706 isn't known at compile time.
13708 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13710 * vec-perm-indices.h (vec_perm_builder): Change element type
13711 from HOST_WIDE_INT to poly_int64.
13712 (vec_perm_indices::element_type): Update accordingly.
13713 (vec_perm_indices::clamp): Handle polynomial element_types.
13714 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
13715 (vec_perm_indices::all_in_range_p): Likewise.
13716 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
13717 than shwi trees.
13718 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
13719 polynomial vec_perm_indices element types.
13720 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
13721 * fold-const.c (fold_vec_perm): Likewise.
13722 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
13723 * tree-vect-generic.c (lower_vec_perm): Likewise.
13724 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
13725 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
13726 element type to HOST_WIDE_INT.
13728 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13729 Alan Hayward <alan.hayward@arm.com>
13730 David Sherwood <david.sherwood@arm.com>
13732 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
13733 rather than an int. Use plus_constant.
13734 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
13735 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
13737 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13738 Alan Hayward <alan.hayward@arm.com>
13739 David Sherwood <david.sherwood@arm.com>
13741 * calls.c (emit_call_1, expand_call): Change struct_value_size from
13742 a HOST_WIDE_INT to a poly_int64.
13744 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13745 Alan Hayward <alan.hayward@arm.com>
13746 David Sherwood <david.sherwood@arm.com>
13748 * calls.c (load_register_parameters): Cope with polynomial
13749 mode sizes. Require a constant size for BLKmode parameters
13750 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
13751 forces a parameter to be padded at the lsb end in order to
13752 fill a complete number of words, require the parameter size
13753 to be ordered wrt UNITS_PER_WORD.
13755 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13756 Alan Hayward <alan.hayward@arm.com>
13757 David Sherwood <david.sherwood@arm.com>
13759 * reload1.c (spill_stack_slot_width): Change element type
13760 from unsigned int to poly_uint64_pod.
13761 (alter_reg): Treat mode sizes as polynomial.
13763 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13764 Alan Hayward <alan.hayward@arm.com>
13765 David Sherwood <david.sherwood@arm.com>
13767 * reload.c (complex_word_subreg_p): New function.
13768 (reload_inner_reg_of_subreg, push_reload): Use it.
13770 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13771 Alan Hayward <alan.hayward@arm.com>
13772 David Sherwood <david.sherwood@arm.com>
13774 * lra-constraints.c (process_alt_operands): Reject matched
13775 operands whose sizes aren't ordered.
13776 (match_reload): Refer to this check here.
13778 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13779 Alan Hayward <alan.hayward@arm.com>
13780 David Sherwood <david.sherwood@arm.com>
13782 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
13783 that the mode size is in the set {1, 2, 4, 8, 16}.
13785 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13786 Alan Hayward <alan.hayward@arm.com>
13787 David Sherwood <david.sherwood@arm.com>
13789 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
13790 Use plus_constant instead of gen_rtx_PLUS.
13792 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13793 Alan Hayward <alan.hayward@arm.com>
13794 David Sherwood <david.sherwood@arm.com>
13796 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
13797 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
13798 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
13799 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
13800 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
13801 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
13802 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
13803 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
13804 * config/i386/i386.c (ix86_push_rounding): ...this new function.
13805 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
13806 a poly_int64.
13807 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
13808 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
13809 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
13810 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
13811 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
13812 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
13813 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
13814 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
13815 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
13816 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
13817 function.
13818 * expr.c (emit_move_resolve_push): Treat the input and result
13819 of PUSH_ROUNDING as a poly_int64.
13820 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
13821 (emit_push_insn): Likewise.
13822 * lra-eliminations.c (mark_not_eliminable): Likewise.
13823 * recog.c (push_operand): Likewise.
13824 * reload1.c (elimination_effects): Likewise.
13825 * rtlanal.c (nonzero_bits1): Likewise.
13826 * calls.c (store_one_arg): Likewise. Require the padding to be
13827 known at compile time.
13829 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13830 Alan Hayward <alan.hayward@arm.com>
13831 David Sherwood <david.sherwood@arm.com>
13833 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
13834 Use plus_constant instead of gen_rtx_PLUS.
13836 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13837 Alan Hayward <alan.hayward@arm.com>
13838 David Sherwood <david.sherwood@arm.com>
13840 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
13841 rather than an int.
13843 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13844 Alan Hayward <alan.hayward@arm.com>
13845 David Sherwood <david.sherwood@arm.com>
13847 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
13848 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
13849 via stack temporaries. Treat the mode size as polynomial too.
13851 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13852 Alan Hayward <alan.hayward@arm.com>
13853 David Sherwood <david.sherwood@arm.com>
13855 * expr.c (expand_expr_real_2): When handling conversions involving
13856 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
13857 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
13858 as a poly_uint64 too.
13860 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13861 Alan Hayward <alan.hayward@arm.com>
13862 David Sherwood <david.sherwood@arm.com>
13864 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
13866 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13867 Alan Hayward <alan.hayward@arm.com>
13868 David Sherwood <david.sherwood@arm.com>
13870 * combine.c (can_change_dest_mode): Handle polynomial
13871 REGMODE_NATURAL_SIZE.
13872 * expmed.c (store_bit_field_1): Likewise.
13873 * expr.c (store_constructor): Likewise.
13874 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
13875 and polynomial REGMODE_NATURAL_SIZE.
13876 (gen_lowpart_common): Likewise.
13877 * reginfo.c (record_subregs_of_mode): Likewise.
13878 * rtlanal.c (read_modify_subreg_p): Likewise.
13880 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13881 Alan Hayward <alan.hayward@arm.com>
13882 David Sherwood <david.sherwood@arm.com>
13884 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
13885 numbers of elements.
13887 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13888 Alan Hayward <alan.hayward@arm.com>
13889 David Sherwood <david.sherwood@arm.com>
13891 * match.pd: Cope with polynomial numbers of vector elements.
13893 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13894 Alan Hayward <alan.hayward@arm.com>
13895 David Sherwood <david.sherwood@arm.com>
13897 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
13898 in a POINTER_PLUS_EXPR.
13900 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13901 Alan Hayward <alan.hayward@arm.com>
13902 David Sherwood <david.sherwood@arm.com>
13904 * omp-simd-clone.c (simd_clone_subparts): New function.
13905 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
13906 (ipa_simd_modify_function_body): Likewise.
13908 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13909 Alan Hayward <alan.hayward@arm.com>
13910 David Sherwood <david.sherwood@arm.com>
13912 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
13913 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
13914 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
13915 (expand_vector_condition, vector_element): Likewise.
13916 (subparts_gt): New function.
13917 (get_compute_type): Use subparts_gt.
13918 (count_type_subparts): Delete.
13919 (expand_vector_operations_1): Use subparts_gt instead of
13920 count_type_subparts.
13922 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13923 Alan Hayward <alan.hayward@arm.com>
13924 David Sherwood <david.sherwood@arm.com>
13926 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
13927 (vect_compile_time_alias): ...this new function. Do the calculation
13928 on poly_ints rather than trees.
13929 (vect_prune_runtime_alias_test_list): Update call accordingly.
13931 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13932 Alan Hayward <alan.hayward@arm.com>
13933 David Sherwood <david.sherwood@arm.com>
13935 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
13936 numbers of units.
13937 (vect_schedule_slp_instance): Likewise.
13939 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13940 Alan Hayward <alan.hayward@arm.com>
13941 David Sherwood <david.sherwood@arm.com>
13943 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
13944 constant and extern definitions for variable-length vectors.
13945 (vect_get_constant_vectors): Note that the number of units
13946 is known to be constant.
13948 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13949 Alan Hayward <alan.hayward@arm.com>
13950 David Sherwood <david.sherwood@arm.com>
13952 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
13953 of units as polynomial. Choose between WIDE and NARROW based
13954 on multiple_p.
13956 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13957 Alan Hayward <alan.hayward@arm.com>
13958 David Sherwood <david.sherwood@arm.com>
13960 * tree-vect-stmts.c (simd_clone_subparts): New function.
13961 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
13963 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13964 Alan Hayward <alan.hayward@arm.com>
13965 David Sherwood <david.sherwood@arm.com>
13967 * tree-vect-stmts.c (vectorizable_call): Treat the number of
13968 vectors as polynomial. Use build_index_vector for
13969 IFN_GOMP_SIMD_LANE.
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-stmts.c (get_load_store_type): Treat the number of
13976 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
13977 for variable-length vectors.
13978 (vectorizable_mask_load_store): Treat the number of units as
13979 polynomial, asserting that it is constant if the condition has
13980 already been enforced.
13981 (vectorizable_store, vectorizable_load): Likewise.
13983 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13984 Alan Hayward <alan.hayward@arm.com>
13985 David Sherwood <david.sherwood@arm.com>
13987 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
13988 of units as polynomial. Punt if we can't tell at compile time
13989 which vector contains the final result.
13991 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13992 Alan Hayward <alan.hayward@arm.com>
13993 David Sherwood <david.sherwood@arm.com>
13995 * tree-vect-loop.c (vectorizable_induction): Treat the number
13996 of units as polynomial. Punt on SLP inductions. Use an integer
13997 VEC_SERIES_EXPR for variable-length integer reductions. Use a
13998 cast of such a series for variable-length floating-point
13999 reductions.
14001 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14002 Alan Hayward <alan.hayward@arm.com>
14003 David Sherwood <david.sherwood@arm.com>
14005 * tree.h (build_index_vector): Declare.
14006 * tree.c (build_index_vector): New function.
14007 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
14008 of units as polynomial, forcibly converting it to a constant if
14009 vectorizable_reduction has already enforced the condition.
14010 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
14011 to create a {1,2,3,...} vector.
14012 (vectorizable_reduction): Treat the number of units as polynomial.
14013 Choose vectype_in based on the largest scalar element size rather
14014 than the smallest number of units. Enforce the restrictions
14015 relied on above.
14017 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14018 Alan Hayward <alan.hayward@arm.com>
14019 David Sherwood <david.sherwood@arm.com>
14021 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
14022 number of units as polynomial.
14024 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14025 Alan Hayward <alan.hayward@arm.com>
14026 David Sherwood <david.sherwood@arm.com>
14028 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
14029 * target.def (autovectorize_vector_sizes): Return the vector sizes
14030 by pointer, using vector_sizes rather than a bitmask.
14031 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
14032 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
14033 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
14034 Likewise.
14035 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
14036 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
14037 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
14038 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
14039 * omp-general.c (omp_max_vf): Likewise.
14040 * omp-low.c (omp_clause_aligned_alignment): Likewise.
14041 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
14042 * tree-vect-loop.c (vect_analyze_loop): Likewise.
14043 * tree-vect-slp.c (vect_slp_bb): Likewise.
14044 * doc/tm.texi: Regenerate.
14045 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
14046 to a poly_uint64.
14047 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
14048 the vector size as a poly_uint64 rather than an unsigned int.
14049 (current_vector_size): Change from an unsigned int to a poly_uint64.
14050 (get_vectype_for_scalar_type): Update accordingly.
14051 * tree.h (build_truth_vector_type): Take the size and number of
14052 units as a poly_uint64 rather than an unsigned int.
14053 (build_vector_type): Add a temporary overload that takes
14054 the number of units as a poly_uint64 rather than an unsigned int.
14055 * tree.c (make_vector_type): Likewise.
14056 (build_truth_vector_type): Take the number of units as a poly_uint64
14057 rather than an unsigned int.
14059 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14060 Alan Hayward <alan.hayward@arm.com>
14061 David Sherwood <david.sherwood@arm.com>
14063 * target.def (get_mask_mode): Take the number of units and length
14064 as poly_uint64s rather than unsigned ints.
14065 * targhooks.h (default_get_mask_mode): Update accordingly.
14066 * targhooks.c (default_get_mask_mode): Likewise.
14067 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
14068 * doc/tm.texi: Regenerate.
14070 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14071 Alan Hayward <alan.hayward@arm.com>
14072 David Sherwood <david.sherwood@arm.com>
14074 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
14075 * omp-general.c (omp_max_vf): Likewise.
14076 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
14077 (expand_omp_simd): Handle polynomial safelen.
14078 * omp-low.c (omplow_simd_context): Add a default constructor.
14079 (omplow_simd_context::max_vf): Change from int to poly_uint64.
14080 (lower_rec_simd_input_clauses): Update accordingly.
14081 (lower_rec_input_clauses): Likewise.
14083 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14084 Alan Hayward <alan.hayward@arm.com>
14085 David Sherwood <david.sherwood@arm.com>
14087 * tree-vectorizer.h (vect_nunits_for_cost): New function.
14088 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
14089 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
14090 (vect_analyze_slp_cost): Likewise.
14091 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
14092 (vect_model_load_cost): Likewise.
14094 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14095 Alan Hayward <alan.hayward@arm.com>
14096 David Sherwood <david.sherwood@arm.com>
14098 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
14099 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
14100 from an unsigned int * to a poly_uint64_pod *.
14101 (calculate_unrolling_factor): New function.
14102 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
14104 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14105 Alan Hayward <alan.hayward@arm.com>
14106 David Sherwood <david.sherwood@arm.com>
14108 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
14109 from an unsigned int to a poly_uint64.
14110 (_loop_vec_info::slp_unrolling_factor): Likewise.
14111 (_loop_vec_info::vectorization_factor): Change from an int
14112 to a poly_uint64.
14113 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
14114 (vect_get_num_vectors): New function.
14115 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
14116 (vect_get_num_copies): Use vect_get_num_vectors.
14117 (vect_analyze_data_ref_dependences): Change max_vf from an int *
14118 to an unsigned int *.
14119 (vect_analyze_data_refs): Change min_vf from an int * to a
14120 poly_uint64 *.
14121 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14122 than an unsigned HOST_WIDE_INT.
14123 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
14124 (vect_analyze_data_ref_dependence): Change max_vf from an int *
14125 to an unsigned int *.
14126 (vect_analyze_data_ref_dependences): Likewise.
14127 (vect_compute_data_ref_alignment): Handle polynomial vf.
14128 (vect_enhance_data_refs_alignment): Likewise.
14129 (vect_prune_runtime_alias_test_list): Likewise.
14130 (vect_shift_permute_load_chain): Likewise.
14131 (vect_supportable_dr_alignment): Likewise.
14132 (dependence_distance_ge_vf): Take the vectorization factor as a
14133 poly_uint64 rather than an unsigned HOST_WIDE_INT.
14134 (vect_analyze_data_refs): Change min_vf from an int * to a
14135 poly_uint64 *.
14136 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
14137 vfm1 as a poly_uint64 rather than an int. Make the same change
14138 for the returned bound_scalar.
14139 (vect_gen_vector_loop_niters): Handle polynomial vf.
14140 (vect_do_peeling): Likewise. Update call to
14141 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
14142 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
14143 be constant.
14144 * tree-vect-loop.c (vect_determine_vectorization_factor)
14145 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
14146 (vect_get_known_peeling_cost): Likewise.
14147 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
14148 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
14149 (vect_transform_loop): Likewise. Use the lowest possible VF when
14150 updating the upper bounds of the loop.
14151 (vect_min_worthwhile_factor): Make static. Return an unsigned int
14152 rather than an int.
14153 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
14154 polynomial unroll factors.
14155 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
14156 (vect_make_slp_decision): Likewise.
14157 (vect_supported_load_permutation_p): Likewise, and polynomial
14158 vf too.
14159 (vect_analyze_slp_cost): Handle polynomial vf.
14160 (vect_slp_analyze_node_operations): Likewise.
14161 (vect_slp_analyze_bb_1): Likewise.
14162 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14163 than an unsigned HOST_WIDE_INT.
14164 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
14165 (vectorizable_load): Handle polynomial vf.
14166 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
14167 a poly_uint64.
14168 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
14170 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14171 Alan Hayward <alan.hayward@arm.com>
14172 David Sherwood <david.sherwood@arm.com>
14174 * match.pd: Handle bit operations involving three constants
14175 and try to fold one pair.
14177 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14179 * tree-vect-loop-manip.c: Include gimple-fold.h.
14180 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
14181 niters_maybe_zero parameters. Handle other cases besides a step of 1.
14182 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
14183 Add a path that uses a step of VF instead of 1, but disable it
14184 for now.
14185 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
14186 and niters_no_overflow parameters. Update calls to
14187 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
14188 Create a new SSA name if the latter choses to use a ste other
14189 than zero, and return it via niters_vector_mult_vf_var.
14190 * tree-vect-loop.c (vect_transform_loop): Update calls to
14191 vect_do_peeling, vect_gen_vector_loop_niters and
14192 slpeel_make_loop_iterate_ntimes.
14193 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
14194 (vect_gen_vector_loop_niters): Update declarations after above changes.
14196 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
14198 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
14199 128-bit round to integer instructions.
14200 (ceil<mode>2): Likewise.
14201 (btrunc<mode>2): Likewise.
14202 (round<mode>2): Likewise.
14204 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14206 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
14207 unaligned VSX load/store on P8/P9.
14208 (expand_block_clear): Allow the use of unaligned VSX
14209 load/store on P8/P9.
14211 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14213 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
14214 New function.
14215 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
14216 swap associated with both a load and a store.
14218 2018-01-02 Andrew Waterman <andrew@sifive.com>
14220 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
14221 * config/riscv/riscv.md (clear_cache): Use it.
14223 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
14225 * web.c: Remove out-of-date comment.
14227 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14229 * expr.c (fixup_args_size_notes): Check that any existing
14230 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
14231 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
14232 (emit_single_push_insn): ...here.
14234 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14236 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
14237 (const_vector_encoded_nelts): New function.
14238 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
14239 (const_vector_int_elt, const_vector_elt): Declare.
14240 * emit-rtl.c (const_vector_int_elt_1): New function.
14241 (const_vector_elt): Likewise.
14242 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
14243 of CONST_VECTOR_ELT.
14245 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14247 * expr.c: Include rtx-vector-builder.h.
14248 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
14249 directly on the tree encoding.
14250 (const_vector_from_tree): Likewise.
14251 * optabs.c: Include rtx-vector-builder.h.
14252 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
14253 sequence of "u" values.
14254 * vec-perm-indices.c: Include rtx-vector-builder.h.
14255 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
14256 directly on the vec_perm_indices encoding.
14258 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14260 * doc/rtl.texi (const_vector): Describe new encoding scheme.
14261 * Makefile.in (OBJS): Add rtx-vector-builder.o.
14262 * rtx-vector-builder.h: New file.
14263 * rtx-vector-builder.c: Likewise.
14264 * rtl.h (rtx_def::u2): Add a const_vector field.
14265 (CONST_VECTOR_NPATTERNS): New macro.
14266 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
14267 (CONST_VECTOR_DUPLICATE_P): Likewise.
14268 (CONST_VECTOR_STEPPED_P): Likewise.
14269 (CONST_VECTOR_ENCODED_ELT): Likewise.
14270 (const_vec_duplicate_p): Check for a duplicated vector encoding.
14271 (unwrap_const_vec_duplicate): Likewise.
14272 (const_vec_series_p): Check for a non-duplicated vector encoding.
14273 Say that the function only returns true for integer vectors.
14274 * emit-rtl.c: Include rtx-vector-builder.h.
14275 (gen_const_vec_duplicate_1): Delete.
14276 (gen_const_vector): Call gen_const_vec_duplicate instead of
14277 gen_const_vec_duplicate_1.
14278 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
14279 (gen_const_vec_duplicate): Use rtx_vector_builder.
14280 (gen_const_vec_series): Likewise.
14281 (gen_rtx_CONST_VECTOR): Likewise.
14282 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
14283 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14284 Build a new vector rather than modifying a CONST_VECTOR in-place.
14285 (handle_special_swappables): Update call accordingly.
14286 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
14287 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14288 Build a new vector rather than modifying a CONST_VECTOR in-place.
14289 (handle_special_swappables): Update call accordingly.
14291 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14293 * simplify-rtx.c (simplify_const_binary_operation): Use
14294 CONST_VECTOR_ELT instead of XVECEXP.
14296 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14298 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
14299 the selector elements to be different from the data elements
14300 if the selector is a VECTOR_CST.
14301 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
14302 ssizetype for the selector.
14304 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14306 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
14307 before testing each element individually.
14308 * tree-vect-generic.c (lower_vec_perm): Likewise.
14310 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14312 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
14313 * selftest-run-tests.c (selftest::run_tests): Call it.
14314 * vector-builder.h (vector_builder::operator ==): New function.
14315 (vector_builder::operator !=): Likewise.
14316 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
14317 (vec_perm_indices::all_from_input_p): New function.
14318 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14319 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
14320 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
14321 instead of reading the VECTOR_CST directly. Detect whether both
14322 vector inputs are the same before constructing the vec_perm_indices,
14323 and update the number of inputs argument accordingly. Use the
14324 utility functions added above. Only construct sel2 if we need to.
14326 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14328 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
14329 the broadcast of the low byte.
14330 (expand_mult_highpart): Use an explicit encoding for the permutes.
14331 * optabs-query.c (can_mult_highpart_p): Likewise.
14332 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
14333 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14334 (vectorizable_bswap): Likewise.
14335 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
14336 explicit encoding for the power-of-2 permutes.
14337 (vect_permute_store_chain): Likewise.
14338 (vect_grouped_load_supported): Likewise.
14339 (vect_permute_load_chain): Likewise.
14341 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14343 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
14344 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
14345 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
14346 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14347 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
14348 (vect_gen_perm_mask_any): Likewise.
14350 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14352 * int-vector-builder.h: New file.
14353 * vec-perm-indices.h: Include int-vector-builder.h.
14354 (vec_perm_indices): Redefine as an int_vector_builder.
14355 (auto_vec_perm_indices): Delete.
14356 (vec_perm_builder): Redefine as a stand-alone class.
14357 (vec_perm_indices::vec_perm_indices): New function.
14358 (vec_perm_indices::clamp): Likewise.
14359 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
14360 (vec_perm_indices::new_vector): New function.
14361 (vec_perm_indices::new_expanded_vector): Update for new
14362 vec_perm_indices class.
14363 (vec_perm_indices::rotate_inputs): New function.
14364 (vec_perm_indices::all_in_range_p): Operate directly on the
14365 encoded form, without computing elided elements.
14366 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
14367 encoding. Update for new vec_perm_indices class.
14368 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
14369 the given vec_perm_builder.
14370 (expand_vec_perm_var): Update vec_perm_builder constructor.
14371 (expand_mult_highpart): Use vec_perm_builder instead of
14372 auto_vec_perm_indices.
14373 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
14374 vec_perm_indices instead of auto_vec_perm_indices. Use a single
14375 or double series encoding as appropriate.
14376 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
14377 vec_perm_indices instead of auto_vec_perm_indices.
14378 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14379 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14380 (vect_permute_store_chain): Likewise.
14381 (vect_grouped_load_supported): Likewise.
14382 (vect_permute_load_chain): Likewise.
14383 (vect_shift_permute_load_chain): Likewise.
14384 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14385 (vect_transform_slp_perm_load): Likewise.
14386 (vect_schedule_slp_instance): Likewise.
14387 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14388 (vectorizable_mask_load_store): Likewise.
14389 (vectorizable_bswap): Likewise.
14390 (vectorizable_store): Likewise.
14391 (vectorizable_load): Likewise.
14392 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
14393 vec_perm_indices instead of auto_vec_perm_indices. Use
14394 tree_to_vec_perm_builder to read the vector from a tree.
14395 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
14396 vec_perm_builder instead of a vec_perm_indices.
14397 (have_whole_vector_shift): Use vec_perm_builder and
14398 vec_perm_indices instead of auto_vec_perm_indices. Leave the
14399 truncation to calc_vec_perm_mask_for_shift.
14400 (vect_create_epilog_for_reduction): Likewise.
14401 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
14402 from auto_vec_perm_indices to vec_perm_indices.
14403 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14404 instead of changing individual elements.
14405 (aarch64_vectorize_vec_perm_const): Use new_vector to install
14406 the vector in d.perm.
14407 * config/arm/arm.c (expand_vec_perm_d::perm): Change
14408 from auto_vec_perm_indices to vec_perm_indices.
14409 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14410 instead of changing individual elements.
14411 (arm_vectorize_vec_perm_const): Use new_vector to install
14412 the vector in d.perm.
14413 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
14414 Update vec_perm_builder constructor.
14415 (rs6000_expand_interleave): Likewise.
14416 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
14417 (rs6000_expand_interleave): Likewise.
14419 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14421 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
14422 to qimode could truncate the indices.
14423 * optabs.c (expand_vec_perm_var): Likewise.
14425 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14427 * Makefile.in (OBJS): Add vec-perm-indices.o.
14428 * vec-perm-indices.h: New file.
14429 * vec-perm-indices.c: Likewise.
14430 * target.h (vec_perm_indices): Replace with a forward class
14431 declaration.
14432 (auto_vec_perm_indices): Move to vec-perm-indices.h.
14433 * optabs.h: Include vec-perm-indices.h.
14434 (expand_vec_perm): Delete.
14435 (selector_fits_mode_p, expand_vec_perm_var): Declare.
14436 (expand_vec_perm_const): Declare.
14437 * target.def (vec_perm_const_ok): Replace with...
14438 (vec_perm_const): ...this new hook.
14439 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
14440 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
14441 * doc/tm.texi: Regenerate.
14442 * optabs.def (vec_perm_const): Delete.
14443 * doc/md.texi (vec_perm_const): Likewise.
14444 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
14445 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
14446 expand_vec_perm for constant permutation vectors. Assert that
14447 the mode of variable permutation vectors is the integer equivalent
14448 of the mode that is being permuted.
14449 * optabs-query.h (selector_fits_mode_p): Declare.
14450 * optabs-query.c: Include vec-perm-indices.h.
14451 (selector_fits_mode_p): New function.
14452 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
14453 is defined, instead of checking whether the vec_perm_const_optab
14454 exists. Use targetm.vectorize.vec_perm_const instead of
14455 targetm.vectorize.vec_perm_const_ok. Check whether the indices
14456 fit in the vector mode before using a variable permute.
14457 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
14458 vec_perm_indices instead of an rtx.
14459 (expand_vec_perm): Replace with...
14460 (expand_vec_perm_const): ...this new function. Take the selector
14461 as a vec_perm_indices rather than an rtx. Also take the mode of
14462 the selector. Update call to shift_amt_for_vec_perm_mask.
14463 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
14464 Use vec_perm_indices::new_expanded_vector to expand the original
14465 selector into bytes. Check whether the indices fit in the vector
14466 mode before using a variable permute.
14467 (expand_vec_perm_var): Make global.
14468 (expand_mult_highpart): Use expand_vec_perm_const.
14469 * fold-const.c: Includes vec-perm-indices.h.
14470 * tree-ssa-forwprop.c: Likewise.
14471 * tree-vect-data-refs.c: Likewise.
14472 * tree-vect-generic.c: Likewise.
14473 * tree-vect-loop.c: Likewise.
14474 * tree-vect-slp.c: Likewise.
14475 * tree-vect-stmts.c: Likewise.
14476 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
14477 Delete.
14478 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
14479 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
14480 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
14481 (aarch64_vectorize_vec_perm_const): ...this new function.
14482 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14483 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14484 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
14485 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
14486 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14487 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14488 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
14489 into...
14490 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
14491 check for NEON modes.
14492 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
14493 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
14494 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
14495 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
14496 into...
14497 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
14498 the old VEC_PERM_CONST conditions.
14499 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
14500 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
14501 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
14502 (ia64_vectorize_vec_perm_const_ok): Merge into...
14503 (ia64_vectorize_vec_perm_const): ...this new function.
14504 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
14505 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
14506 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
14507 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
14508 * config/mips/mips.c (mips_expand_vec_perm_const)
14509 (mips_vectorize_vec_perm_const_ok): Merge into...
14510 (mips_vectorize_vec_perm_const): ...this new function.
14511 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
14512 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
14513 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
14514 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
14515 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
14516 (rs6000_expand_vec_perm_const): Delete.
14517 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
14518 Delete.
14519 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14520 (altivec_expand_vec_perm_const_le): Take each operand individually.
14521 Operate on constant selectors rather than rtxes.
14522 (altivec_expand_vec_perm_const): Likewise. Update call to
14523 altivec_expand_vec_perm_const_le.
14524 (rs6000_expand_vec_perm_const): Delete.
14525 (rs6000_vectorize_vec_perm_const_ok): Delete.
14526 (rs6000_vectorize_vec_perm_const): New function.
14527 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14528 an element count and rtx array.
14529 (rs6000_expand_extract_even): Update call accordingly.
14530 (rs6000_expand_interleave): Likewise.
14531 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
14532 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
14533 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
14534 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
14535 (rs6000_expand_vec_perm_const): Delete.
14536 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14537 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14538 (altivec_expand_vec_perm_const_le): Take each operand individually.
14539 Operate on constant selectors rather than rtxes.
14540 (altivec_expand_vec_perm_const): Likewise. Update call to
14541 altivec_expand_vec_perm_const_le.
14542 (rs6000_expand_vec_perm_const): Delete.
14543 (rs6000_vectorize_vec_perm_const_ok): Delete.
14544 (rs6000_vectorize_vec_perm_const): New function. Remove stray
14545 reference to the SPE evmerge intructions.
14546 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14547 an element count and rtx array.
14548 (rs6000_expand_extract_even): Update call accordingly.
14549 (rs6000_expand_interleave): Likewise.
14550 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
14551 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
14552 new function.
14553 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14555 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14557 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
14558 vector mode and that that mode matches the mode of the data
14559 being permuted.
14560 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
14561 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
14562 directly using expand_vec_perm_1 when forcing selectors into
14563 registers.
14564 (expand_vec_perm_var): New function, split out from expand_vec_perm.
14566 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14568 * optabs-query.h (can_vec_perm_p): Delete.
14569 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
14570 * optabs-query.c (can_vec_perm_p): Split into...
14571 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
14572 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
14573 particular selector is valid.
14574 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14575 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14576 (vect_grouped_load_supported): Likewise.
14577 (vect_shift_permute_load_chain): Likewise.
14578 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14579 (vect_transform_slp_perm_load): Likewise.
14580 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14581 (vectorizable_bswap): Likewise.
14582 (vect_gen_perm_mask_checked): Likewise.
14583 * fold-const.c (fold_ternary_loc): Likewise. Don't take
14584 implementations of variable permutation vectors into account
14585 when deciding which selector to use.
14586 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
14587 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
14588 with a false third argument.
14589 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
14590 to test whether the constant selector is valid and can_vec_perm_var_p
14591 to test whether a variable selector is valid.
14593 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14595 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
14596 * optabs-query.c (can_vec_perm_p): Likewise.
14597 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
14598 instead of vec_perm_indices.
14599 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
14600 (vect_gen_perm_mask_checked): Likewise,
14601 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
14602 (vect_gen_perm_mask_checked): Likewise,
14604 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14606 * optabs-query.h (qimode_for_vec_perm): Declare.
14607 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
14608 (qimode_for_vec_perm): ...this new function.
14609 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
14611 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14613 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
14614 does not have a conditional at the top.
14616 2018-01-02 Richard Biener <rguenther@suse.de>
14618 * ipa-inline.c (big_speedup_p): Fix expression.
14620 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14622 PR target/81616
14623 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
14624 for generic 4->6.
14626 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
14628 PR target/81616
14629 Generic tuning.
14630 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
14631 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
14632 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
14633 cond_taken_branch_cost 3->4.
14635 2018-01-01 Jakub Jelinek <jakub@redhat.com>
14637 PR tree-optimization/83581
14638 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
14639 TODO_cleanup_cfg if any changes have been made.
14641 PR middle-end/83608
14642 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
14643 convert_modes if target mode has the right side, but different mode
14644 class.
14646 PR middle-end/83609
14647 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
14648 last argument when extracting from CONCAT. If either from_real or
14649 from_imag is NULL, use expansion through memory. If result is not
14650 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
14651 the parts directly to inner mode, if even that fails, use expansion
14652 through memory.
14654 PR middle-end/83623
14655 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
14656 check for bswap in mode rather than HImode and use that in expand_unop
14657 too.
14658 \f
14659 Copyright (C) 2018 Free Software Foundation, Inc.
14661 Copying and distribution of this file, with or without modification,
14662 are permitted in any medium without royalty provided the copyright
14663 notice and this notice are preserved.