1 2018-07-31 David Malcolm <dmalcolm@redhat.com>
3 * dump-context.h: Include "pretty-print.h".
4 (dump_context::refresh_dumps_are_enabled): New decl.
5 (dump_context::emit_item): New decl.
6 (class dump_context): Add fields "m_test_pp" and
8 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
9 (temp_dump_context::get_dumped_text): New decl.
10 (class temp_dump_context): Add field "m_pp".
11 * dumpfile.c (refresh_dumps_are_enabled): Convert to...
12 (dump_context::refresh_dumps_are_enabled): ...and add a test for
14 (set_dump_file): Update for above change.
15 (set_alt_dump_file): Likewise.
16 (dump_loc): New overload, taking a pretty_printer *.
17 (dump_context::dump_loc): Call end_any_optinfo. Dump the location
18 to any test pretty-printer.
19 (make_item_for_dump_gimple_stmt): New function, adapted from
20 optinfo::add_gimple_stmt.
21 (dump_context::dump_gimple_stmt): Call it, and use the result,
22 eliminating the direct usage of dump_file and alt_dump_file in
23 favor of indirectly using them via emit_item.
24 (make_item_for_dump_gimple_expr): New function, adapted from
25 optinfo::add_gimple_expr.
26 (dump_context::dump_gimple_expr): Call it, and use the result,
27 eliminating the direct usage of dump_file and alt_dump_file in
28 favor of indirectly using them via emit_item.
29 (make_item_for_dump_generic_expr): New function, adapted from
31 (dump_context::dump_generic_expr): Call it, and use the result,
32 eliminating the direct usage of dump_file and alt_dump_file in
33 favor of indirectly using them via emit_item.
34 (make_item_for_dump_printf_va): New function, adapted from
35 optinfo::add_printf_va.
36 (make_item_for_dump_printf): New function.
37 (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
38 and use the result, eliminating the direct usage of dump_file and
39 alt_dump_file in favor of indirectly using them via emit_item.
40 (make_item_for_dump_dec): New function.
41 (dump_context::dump_dec): Call it, and use the result,
42 eliminating the direct usage of dump_file and alt_dump_file in
43 favor of indirectly using them via emit_item.
44 (make_item_for_dump_symtab_node): New function, adapted from
45 optinfo::add_symtab_node.
46 (dump_context::dump_symtab_node): Call it, and use the result,
47 eliminating the direct usage of dump_file and alt_dump_file in
48 favor of indirectly using them via emit_item.
49 (dump_context::begin_scope): Reimplement, avoiding direct usage
50 of dump_file and alt_dump_file in favor of indirectly using them
52 (dump_context::emit_item): New member function.
53 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
54 Set up test pretty-printer on the underlying context. Call
55 refresh_dumps_are_enabled.
56 (temp_dump_context::~temp_dump_context): Call
57 refresh_dumps_are_enabled.
58 (temp_dump_context::get_dumped_text): New member function.
59 (selftest::verify_dumped_text): New function.
60 (ASSERT_DUMPED_TEXT_EQ): New macro.
61 (selftest::test_capture_of_dump_calls): Run all tests twice, with
62 and then without optinfo enabled. Add uses of
63 ASSERT_DUMPED_TEXT_EQ to all tests. Add test of nested scopes.
64 * dumpfile.h: Update comment for the dump_* API.
65 * optinfo-emit-json.cc
66 (selftest::test_building_json_from_dump_calls): Update for new
67 param for temp_dump_context ctor.
68 * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
70 (optinfo_item::~optinfo_item): Likewise.
71 (optinfo::add_item): New member function.
72 (optinfo::emit): Update comment.
73 (optinfo::add_string): Delete.
74 (optinfo::add_printf): Delete.
75 (optinfo::add_printf_va): Delete.
76 (optinfo::add_gimple_stmt): Delete.
77 (optinfo::add_gimple_expr): Delete.
78 (optinfo::add_tree): Delete.
79 (optinfo::add_symtab_node): Delete.
80 (optinfo::add_dec): Delete.
81 * optinfo.h (class dump_context): New forward decl.
82 (optinfo::add_item): New decl.
83 (optinfo::add_string): Delete.
84 (optinfo::add_printf): Delete.
85 (optinfo::add_printf_va): Delete.
86 (optinfo::add_gimple_stmt): Delete.
87 (optinfo::add_gimple_expr): Delete.
88 (optinfo::add_tree): Delete.
89 (optinfo::add_symtab_node): Delete.
90 (optinfo::add_dec): Delete.
91 (optinfo::add_poly_int): Delete.
92 (optinfo_item::optinfo_item): Remove "owned" param.
93 (class optinfo_item): Remove field "m_owned".
95 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
98 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
99 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
100 requested variable alignment.
101 (expand_one_ssa_partition): Likewise.
102 (expand_one_var): Likewise.
104 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
106 * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
107 to speculation_safe_value_not_needed.
109 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
111 * targhooks.h (speculation_safe_value_not_needed): New prototype.
112 * targhooks.c (speculation_safe_value_not_needed): New function.
113 * target.def (have_speculation_safe_value): Update documentation.
114 * doc/tm.texi: Regenerated.
116 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
118 * config/aarch64/iterators.md (ALLI_TI): New iterator.
119 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
121 (despeculate_copy<ALLI:mode>_insn): New insn.
122 (despeculate_copyti_insn): New insn.
123 (despeculate_simple<ALLI:mode>): New insn
124 (despeculate_simpleti): New insn.
125 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
127 (TARGET_SPECULATION_SAFE_VALUE): Redefine to
128 aarch64_speculation_safe_value.
129 (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
131 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
133 * config/aarch64/aarch64-speculation.cc: New file.
134 * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
135 before pass_reorder_blocks.
136 * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
138 * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
139 X14 and X15 when tracking speculation.
140 * config/aarch64/aarch64.md (register name constants): Add
141 SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
142 (unspec): Add UNSPEC_SPECULATION_TRACKER.
143 (speculation_barrier): New insn attribute.
144 (cmp<mode>): Allow SP in comparisons.
145 (speculation_tracker): New insn.
146 (speculation_barrier): Add speculation_barrier attribute.
147 * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
148 * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
149 * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
151 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
153 * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
154 aarch64_track_speculation is true.
155 (tb<optab><mode>1): Likewise.
156 * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
157 generate CB[N]Z when tracking speculation.
158 (aarch64_split_compare_and_swap): Likewise.
159 (aarch64_split_atomic_op): Likewise.
161 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
163 * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
165 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
167 * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
168 (speculation_barrier): New insn.
170 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
172 * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
173 * config/arm/arm.md (speculation_barrier): New expand.
174 (speculation_barrier_insn): New pattern.
176 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
178 * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
179 (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
180 (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
181 * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
183 * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
184 (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
185 (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
186 (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
187 (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
188 (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
189 (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
190 * builtins.c (expand_speculation_safe_value): New function.
191 (expand_builtin): Call it.
192 * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
193 * doc/extend.texi: Document __builtin_speculation_safe_value.
194 * doc/md.texi: Document "speculation_barrier" pattern.
195 * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
196 TARGET_HAVE_SPECULATION_SAFE_VALUE.
197 * doc/tm.texi: Regenerated.
198 * target.def (have_speculation_safe_value, speculation_safe_value): New
200 * targhooks.c (default_have_speculation_safe_value): New function.
201 (default_speculation_safe_value): New function.
202 * targhooks.h (default_have_speculation_safe_value): Add prototype.
203 (default_speculation_safe_value): Add prototype.
205 2018-07-31 David Malcolm <dmalcolm@redhat.com>
207 * dump-context.h (dump_context::dump_loc): New decl.
208 * dumpfile.c (dump_context::dump_loc): New member function.
209 (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
210 and dump_gimple_stmt.
211 (dump_context::dump_gimple_expr_loc): Likewise, using
213 (dump_context::dump_generic_expr_loc): Likewise, using
215 (dump_context::dump_printf_loc_va): Likewise, using
217 (dump_context::begin_scope): Explicitly using the global function
218 "dump_loc", rather than the member function.
220 2018-07-31 Martin Sebor <msebor@redhat.com>
222 PR tree-optimization/86741
223 * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
225 2018-07-31 Andreas Krebbel <krebbel@linux.ibm.com>
227 * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
228 depend on whether prefetch instructions will be emitted or not.
229 Use TARGET_SETMEM_PFD for checking whether prefetch instructions
230 will be emitted or not.
231 * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
232 (TARGET_SETMEM_PFD): New macros.
234 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
236 * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
237 (NULL_STMT_VEC_INFO): Delete.
238 (stmt_vec_info::operator*): Likewise.
239 (stmt_vec_info::operator gimple *): Likewise.
240 * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
241 of NULL_STMT_VEC_INFO.
242 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
243 (vect_reassociating_reduction_p): Likewise.
244 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
245 (vectorizable_store): Likewise.
246 * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
247 (vec_info::free_stmt_vec_infos): Likewise.
249 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
251 * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
252 * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
254 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
256 * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
257 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
258 (vec_info::free_stmt_vec_info): New private member functions.
259 (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
260 (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
261 * tree-parloops.c (gather_scalar_reductions): Remove calls to
262 set_stmt_vec_info_vec and free_stmt_vec_infos.
263 * tree-vect-loop.c (_loop_vec_info): Remove call to
264 set_stmt_vec_info_vec.
265 * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
266 (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
267 * tree-vectorizer.c (vec_info::new_stmt_vec_info)
268 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
269 (vec_info::free_stmt_vec_info): ...these new functions. Remove
270 assignments in {vec_info::,}new_stmt_vec_info that are redundant
271 with the clearing in the xcalloc.
272 (stmt_vec_info_vec): Delete.
273 (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
274 (vectorize_loops): Likewise.
275 (vec_info::~vec_info): Remove argument from call to
277 (vec_info::add_stmt): Remove vinfo argument from call to
280 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
282 * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
283 rather than a gimple stmt.
284 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. Don't free
285 information for pattern statements when passed the original
286 statement; instead wait to be passed the pattern statement itself.
287 Don't call set_vinfo_for_stmt here.
288 (free_stmt_vec_infos): Update call to free_stmt_vec_info.
289 * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
291 * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
292 * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
293 stmt_vec_infos entry.
295 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
297 * tree-vectorizer.h (vec_info::replace_stmt): Declare.
298 * tree-vectorizer.c (vec_info::replace_stmt): New function.
299 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
300 * tree-vect-stmts.c (vectorizable_call): Likewise.
301 (vectorizable_simd_clone_call): Likewise.
303 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
305 * tree-vectorizer.h (vec_info::remove_stmt): Declare.
306 * tree-vectorizer.c (vec_info::remove_stmt): New function.
307 * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
308 * tree-vect-loop.c (vect_transform_loop): Likewise.
309 * tree-vect-slp.c (vect_schedule_slp): Likewise.
310 * tree-vect-stmts.c (vect_remove_stores): Likewise.
312 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
314 * tree-vectorizer.h (vec_info::lookup_dr): New member function.
315 (vect_dr_stmt): Delete.
316 * tree-vectorizer.c (vec_info::lookup_dr): New function.
317 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
319 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
320 (vect_analyze_data_ref_dependence, vect_record_base_alignments)
321 (vect_verify_datarefs_alignment, vect_peeling_supportable)
322 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
323 (vect_analyze_data_refs): Likewise.
324 (vect_slp_analyze_data_ref_dependence): Likewise. Take a vec_info
326 (vect_find_same_alignment_drs): Likewise.
327 (vect_slp_analyze_node_dependences): Update calls accordingly.
328 (vect_analyze_data_refs_alignment): Likewise. Use vec_info::lookup_dr
329 instead of DR_VECT_AUX.
330 (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
331 of a vector data references. Use vec_info::lookup_dr instead of
333 (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
334 (vect_enhance_data_refs_alignment): Likewise. Use vec_info::lookup_dr
335 instead of DR_VECT_AUX.
337 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
339 * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
341 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
343 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
344 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
345 (vect_gen_prolog_loop_niters): Likewise.
347 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
349 * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
350 (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
351 (vect_known_alignment_in_bytes, vect_dr_behavior)
352 (vect_get_scalar_dr_size): Take references as dr_vec_infos
353 instead of data_references. Update calls to other routines for
354 which the same change has been made.
355 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
356 dr_vec_infos instead of stmt_vec_infos.
357 (vect_analyze_data_ref_dependence): Update call accordingly.
358 (vect_slp_analyze_data_ref_dependence)
359 (vect_record_base_alignments): Use DR_VECT_AUX.
360 (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
361 (vect_update_misalignment_for_peel, verify_data_ref_alignment)
362 (vector_alignment_reachable_p, vect_get_data_access_cost)
363 (vect_peeling_supportable, vect_analyze_group_access_1)
364 (vect_analyze_group_access, vect_analyze_data_ref_access)
365 (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
366 (vect_compile_time_alias, vect_small_gap_p)
367 (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
368 (vect_supportable_dr_alignment): Take references as dr_vec_infos
369 instead of data_references. Update calls to other routines for
370 which the same change has been made.
371 (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
372 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
373 (vect_slp_analyze_and_verify_node_alignment)
374 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
375 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
376 (vect_setup_realignment): Use dr_vec_infos. Update calls after
378 (_vect_peel_info::dr): Replace with...
379 (_vect_peel_info::dr_info): ...this new field.
380 (vect_peeling_hash_get_most_frequent)
381 (vect_peeling_hash_choose_best_peeling): Update accordingly.
382 (vect_peeling_hash_get_lowest_cost):
383 (vect_enhance_data_refs_alignment): Likewise. Update calls to other
384 routines for which the same change has been made.
385 (vect_peeling_hash_insert): Likewise. Take a dr_vec_info instead of a
387 * tree-vect-loop-manip.c (get_misalign_in_elems)
388 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after
390 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
391 * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
392 (vect_truncate_gather_scatter_offset, compare_step_with_zero)
393 (get_group_load_store_type, get_negative_load_store_type)
394 (vect_get_data_ptr_increment, vectorizable_store)
395 (vectorizable_load): Likewise.
396 (ensure_base_align): Take a dr_vec_info instead of a data_reference.
397 Update calls to other routines for which the same change has been made.
399 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
401 * tree-vectorizer.h (vec_info::move_dr): New member function.
402 (dataref_aux): Rename to...
403 (dr_vec_info): ...this and add "dr" and "stmt" fields.
404 (_stmt_vec_info::dr_aux): Update accordingly.
405 (_stmt_vec_info::data_ref_info): Delete.
406 (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
407 (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
408 (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
409 (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
411 (STMT_VINFO_DATA_REF): Likewise. Turn into an lvalue.
412 (STMT_VINFO_DR_INFO): New macro.
413 (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
414 (set_dr_misalignment): Update after rename of dataref_aux.
415 (vect_dr_stmt): Move earlier in file. Return dr_aux.stmt.
416 * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
417 initialization of STMT_VINFO_DATA_REF.
418 * tree-vectorizer.c (vec_info::move_dr): New function.
419 * tree-vect-patterns.c (vect_recog_bool_pattern)
420 (vect_recog_mask_conversion_pattern)
421 (vect_recog_gather_scatter_pattern): Use it.
422 * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
423 the "dr" and "stmt" fields of dr_vec_info instead of
426 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
428 * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
429 (is_pattern_stmt_p): Use it.
430 * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
431 on pattern statements.
433 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
435 * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
436 original stmt as a stmt_vec_info rather than a gimple stmt.
437 (vect_pattern_recog_1): Take the statement directly as a
438 stmt_vec_info, rather than via a gimple_stmt_iterator.
439 Update call to vect_mark_pattern_stmts.
440 (vect_pattern_recog): Update calls accordingly.
442 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
444 * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
445 (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
446 a vect_def_type for the first argument.
447 * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
448 (vect_get_vec_def_for_stmt_copy): Likewise. Return the original
449 operand if it isn't defined by a vectorized statement.
450 (vect_build_gather_load_calls): Remove the mask_dt argument and
451 update calls to vect_get_vec_def_for_stmt_copy.
452 (vectorizable_bswap): Likewise the dt argument.
453 (vectorizable_call): Update calls to vectorizable_bswap and
454 vect_get_vec_def_for_stmt_copy.
455 (vectorizable_simd_clone_call, vectorizable_assignment)
456 (vectorizable_shift, vectorizable_operation, vectorizable_condition)
457 (vectorizable_comparison): Update calls to
458 vect_get_vec_def_for_stmt_copy.
459 (vectorizable_store): Likewise. Remove now-unnecessary calls to
461 (vect_get_loop_based_defs): Remove dt argument and update call
462 to vect_get_vec_def_for_stmt_copy.
463 (vectorizable_conversion): Update calls to vect_get_loop_based_defs
464 and vect_get_vec_def_for_stmt_copy.
465 (vectorizable_load): Update calls to vect_build_gather_load_calls
466 and vect_get_vec_def_for_stmt_copy.
467 * tree-vect-loop.c (vect_create_epilog_for_reduction)
468 (vectorizable_reduction, vectorizable_live_operation): Update calls
469 to vect_get_vec_def_for_stmt_copy.
471 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
473 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
474 and gimple stmt arguments with a stmt_vec_info.
475 (vect_record_base_alignments): Update calls accordingly.
476 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
477 and gimple stmt arguments with a stmt_vec_info.
478 (vect_build_slp_tree_1): Remove vinfo argument and update call
479 to vect_record_max_nunits.
480 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
481 and vect_record_max_nunits.
483 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
485 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
486 file and take a stmt_vec_info instead of a gimple stmt.
487 (supportable_widening_operation, vect_finish_replace_stmt)
488 (vect_finish_stmt_generation, vect_get_store_rhs)
489 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
490 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
491 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
492 (vect_get_smallest_scalar_type, vect_check_gather_scatter)
493 (vect_create_data_ref_ptr, bump_vector_ptr)
494 (vect_permute_store_chain, vect_setup_realignment)
495 (vect_transform_grouped_load, vect_record_grouped_load_vectors)
496 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
497 (vectorizable_reduction, vectorizable_induction)
498 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
499 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
500 than gimple stmts as arguments.
501 * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
502 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
503 (can_group_stmts_p, vect_check_gather_scatter)
504 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
505 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
506 (vect_permute_load_chain, vect_shift_permute_load_chain)
507 (vect_transform_grouped_load)
508 (vect_record_grouped_load_vectors): Likewise.
509 * tree-vect-loop.c (vect_fixup_reduc_chain)
510 (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
511 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
512 (vectorizable_reduction, vectorizable_induction)
513 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
514 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
515 (vect_get_load_store_mask): Likewise.
516 * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
517 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
518 * tree-vect-stmts.c (vect_mark_relevant)
519 (is_simple_and_all_uses_invariant)
520 (exist_non_indexing_operands_for_use_p, process_use)
521 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
522 (vect_get_vec_def_for_operand, vect_get_vec_defs)
523 (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
524 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
525 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
526 (get_negative_load_store_type, get_load_store_type)
527 (vect_check_load_store_mask, vect_check_store_rhs)
528 (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
529 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
530 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
531 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
532 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
533 (vectorizable_load, vectorizable_condition, vectorizable_comparison)
534 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
535 (supportable_widening_operation): Likewise.
537 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
539 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
540 a stmt_vec_info instead of a gcall.
541 (vect_check_gather_scatter): Update call accordingly.
542 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
544 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
545 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
546 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
547 instead of a gimple stmt.
548 (vect_transform_loop): Update calls accordingly.
549 * tree-vect-slp.c (vect_split_slp_store_group): Take and return
550 stmt_vec_infos instead of gimple stmts.
551 (vect_analyze_slp_instance): Update use accordingly.
552 * tree-vect-stmts.c (read_vector_array, write_vector_array)
553 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
554 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
555 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
556 (vect_gen_widened_results_half, vect_get_loop_based_defs)
557 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
558 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
561 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
563 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
564 of the worklist from a vector of gimple stmts to a vector of
566 * tree-vect-stmts.c (vect_mark_relevant, process_use)
567 (vect_mark_stmts_to_be_vectorized): Likewise
569 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
571 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
572 statement before passing it to vect_analyze_stmt.
573 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
574 the chain of phi vector definitions. Track the exit phi via its
576 (vectorizable_reduction): Set cond_stmt_vinfo directly from the
577 STMT_VINFO_REDUC_DEF.
578 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
579 stmt_vec_infos to handle the statement chains.
580 (vect_get_slp_defs): Record the first statement in the node
581 using a stmt_vec_info.
582 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
583 statements here and pass their stmt_vec_info down to subroutines.
584 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
585 down to vect_finish_stmt_generation.
586 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
587 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
588 stmt_vec_infos to subroutines.
589 (vect_remove_stores): Use stmt_vec_infos to handle the statement
592 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
594 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
595 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
596 (vect_permute_store_chain, vect_setup_realignment)
597 (vect_permute_load_chain, vect_shift_permute_load_chain)
598 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
599 stmts internally, and when passing values to other vectorizer routines.
600 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
601 * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
602 (vect_analyze_loop_operations, get_initial_def_for_reduction)
603 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
604 (vectorizable_reduction, vectorizable_induction)
605 (vectorizable_live_operation, vect_transform_loop_stmt)
606 (vect_transform_loop): Likewise.
607 * tree-vect-patterns.c (vect_reassociating_reduction_p)
608 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
609 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
610 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
611 (vect_slp_analyze_node_operations_1): Likewise.
612 * tree-vect-stmts.c (vect_mark_relevant, process_use)
613 (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
614 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
615 (vect_finish_stmt_generation_1, get_group_load_store_type)
616 (get_load_store_type, vect_build_gather_load_calls)
617 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
618 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
619 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
620 (vectorizable_store, vectorizable_load, vectorizable_condition)
621 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
622 (supportable_widening_operation): Likewise.
623 (vect_get_vector_types_for_stmt): Likewise.
624 * tree-vectorizer.h (vect_dr_behavior): Likewise.
626 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
628 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
629 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
630 (vect_permute_store_chain, vect_permute_load_chain)
631 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
632 repeated stmt_vec_info lookups.
633 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
634 (vect_update_ivs_after_vectorizer): Likewise.
635 * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
636 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
637 * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
638 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
639 (vect_bb_slp_scalar_cost): Likewise.
640 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
642 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
644 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
645 gcall rather than the generic gimple stmt to gimple_call_internal_fn.
646 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
647 to get gassigns and gcalls, rather than operating on generc gimple
649 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
650 (vect_mark_stmts_to_be_vectorized, vectorizable_store)
651 (vectorizable_load, vect_analyze_stmt): Likewise.
652 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
654 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
656 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
657 return stmt_vec_infos rather than gimple stmts. Do not accept
659 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
661 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
662 Update use of get_later_stmt.
663 (vect_get_constant_vectors): Update call accordingly.
664 (vect_schedule_slp_instance): Likewise
665 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
666 (vect_slp_analyze_instance_dependence): Likewise.
667 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
669 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
671 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
672 (stmt_info_for_cost::stmt_info): ...this new field.
673 (add_stmt_costs): Update accordingly.
674 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
675 (vect_get_known_peeling_cost): Likewise.
676 (vect_estimate_min_profitable_iters): Likewise.
677 * tree-vect-stmts.c (record_stmt_cost): Likewise.
679 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
681 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
682 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
683 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
685 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
687 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
689 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
690 a gimple stmt to a stmt_vec_info.
691 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
693 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
695 * tree-vectorizer.h (vec_info::grouped_stores): Change from
696 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
697 (_loop_vec_info::reduction_chains): Likewise.
698 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
700 * tree-vect-slp.c (vect_analyze_slp): Likewise.
702 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
704 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
705 a gimple stmt to a stmt_vec_info.
706 (_stmt_vec_info::next_element): Likewise.
707 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
708 (vect_slp_analyze_and_verify_node_alignment)
709 (vect_analyze_group_access_1, vect_analyze_group_access)
710 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
711 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
712 (vect_supportable_dr_alignment): Update accordingly.
713 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
714 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
715 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
716 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
717 * tree-vect-slp.c (vect_build_slp_tree_1)
718 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
719 (vect_split_slp_store_group, vect_analyze_slp_instance)
720 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
721 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
722 (get_group_load_store_type, get_load_store_type)
723 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
724 (vect_transform_stmt, vect_remove_stores): Likewise.
726 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
728 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
730 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
731 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
732 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
733 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
734 (vector_alignment_reachable_p, vect_get_data_access_cost)
735 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
736 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
737 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
738 (vect_analyze_group_access_1, vect_analyze_group_access)
739 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
740 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
741 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
742 result of vect_dr_stmt and use the stmt_vec_info instead of
743 the associated gimple stmt.
744 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
745 (vect_gen_prolog_loop_niters): Likewise.
746 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
748 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
750 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
751 to a vec<stmt_vec_info>.
752 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
753 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
755 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
756 to a vec<stmt_vec_info>.
757 (bst_traits::value_type, bst_traits::value_type): Likewise.
758 (bst_traits::hash): Update accordingly.
759 (vect_get_and_check_slp_defs): Change the stmts parameter from
760 a vec<gimple *> to a vec<stmt_vec_info>.
761 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
762 (vect_build_slp_tree): Likewise.
763 (vect_build_slp_tree_2): Likewise. Update uses of
764 SLP_TREE_SCALAR_STMTS.
765 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
766 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
767 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
768 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
769 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
770 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
771 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
772 (vect_get_constant_vectors, vect_get_slp_defs)
773 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
774 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
775 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
776 instead of gimple stmts.
777 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
778 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
779 (vect_slp_analyze_instance_dependence): Update uses of
780 SLP_TREE_SCALAR_STMTS.
781 (vect_slp_analyze_and_verify_node_alignment): Likewise.
782 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
783 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
784 (get_initial_defs_for_reduction): Likewise.
785 (vect_create_epilog_for_reduction): Likewise.
786 (vectorize_fold_left_reduction): Likewise.
787 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
788 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
789 (can_vectorize_live_stmts): Likewise.
791 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
793 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
794 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
795 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
797 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
799 (gather_scalar_reductions): Update after above interface changes.
800 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
801 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
803 (vect_force_simple_reduction): Likewise.
804 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
805 LOOP_VINFO_REDUCTIONS.
806 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
808 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
810 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
811 a gimple stmt to a stmt_vec_info.
812 * tree-vect-loop.c (vect_active_double_reduction_p)
813 (vect_force_simple_reduction, vectorizable_reduction): Update
816 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
818 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
819 vec<gimple *> to a vec<stmt_vec_info>.
820 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
821 the reduction_phis argument from a vec<gimple *> to a
823 (vectorizable_reduction): Likewise the phis local variable that
824 is passed to vect_create_epilog_for_reduction. Update for new type
825 of SLP_TREE_VEC_STMTS.
826 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
827 (vectorizable_live_operation): Likewise.
828 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
829 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
831 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
833 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
834 a gimple stmt to a stmt_vec_info.
835 (vectorizable_condition, vectorizable_live_operation)
836 (vectorizable_reduction, vectorizable_induction): Pass back the
837 vectorized statement as a stmt_vec_info.
838 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
839 use of STMT_VINFO_VEC_STMT.
840 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
841 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
842 as stmt_vec_infos rather than gimple stmts.
843 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
845 (vectorizable_live_operation): Likewise.
846 (vectorizable_reduction, vectorizable_induction): Likewise,
847 updating use of STMT_VINFO_VEC_STMT.
848 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
849 of STMT_VINFO_VEC_STMT.
850 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
851 (vectorizable_simd_clone_call, vectorizable_conversion)
852 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
853 (vectorizable_store, vectorizable_load, vectorizable_condition)
854 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
855 from a gimple stmt to a stmt_vec_info.
856 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
857 pointer to a stmt_vec_info to the vectorizable_* routines.
859 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
861 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
862 a gimple stmt to a stmt_vec_info.
863 (is_pattern_stmt_p): Update accordingly.
864 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
865 (vect_record_grouped_load_vectors): Likewise.
866 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
867 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
868 (vect_model_reduction_cost): Likewise.
869 (vect_create_epilog_for_reduction): Likewise.
870 (vectorizable_reduction, vectorizable_induction): Likewise.
871 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
872 Return the stmt_vec_info for the pattern statement.
873 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
874 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
875 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
876 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
877 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
878 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
879 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
880 (free_stmt_vec_info, vect_is_simple_use): Likewise.
882 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
884 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
885 (vect_finish_stmt_generation): Likewise.
886 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
887 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
888 (vect_build_gather_load_calls): Use the return value of the above
889 functions instead of a separate call to vinfo_for_stmt. Use narrow
890 scopes for the input gimple stmt and wider scopes for the associated
891 stmt_vec_info. Use vec_info::lookup_def when setting these
892 stmt_vec_infos from an SSA_NAME definition.
893 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
894 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
895 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
896 (vectorizable_store, vectorizable_load, vectorizable_condition)
897 (vectorizable_comparison): Likewise.
898 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
899 (vectorizable_reduction): Likewise.
901 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
903 * tree-vectorizer.h (vect_is_simple_use): Add an optional
904 stmt_vec_info * parameter before the optional gimple **.
905 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
906 (process_use, vect_get_vec_def_for_operand_1): Update callers.
907 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
908 * tree-vect-loop.c (vectorizable_reduction): Likewise.
909 (vectorizable_live_operation): Likewise.
910 * tree-vect-patterns.c (type_conversion_p): Likewise.
911 (vect_look_through_possible_promotion): Likewise.
912 (vect_recog_rotate_pattern): Likewise.
913 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
915 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
917 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
918 a typedef to a wrapper class.
919 (NULL_STMT_VEC_INFO): New macro.
920 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
921 (stmt_vec_info::operator*): New function.
922 (stmt_vec_info::operator gimple *): Likewise.
923 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
924 (add_stmt_costs): Likewise.
925 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
926 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
927 (vect_get_known_peeling_cost): Likewise.
928 (vect_estimate_min_profitable_iters): Likewise.
929 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
930 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
931 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
932 (vectorizable_store, free_stmt_vec_infos): Likewise.
933 (new_stmt_vec_info): Change return type of xcalloc to
936 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
938 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
939 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
940 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
941 a single_imm_use-based sequence.
942 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
944 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
946 * tree-vectorizer.h (vec_info::lookup_def): Declare.
947 * tree-vectorizer.c (vec_info::lookup_def): New function.
948 * tree-vect-patterns.c (vect_get_internal_def): Use it.
949 (vect_widened_op_tree): Likewise.
950 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
951 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
952 (vectorizable_reduction): Likewise.
953 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
955 (vect_is_slp_reduction): Update calls accordingly. Use
956 vec_info::lookup_def.
957 (vect_is_simple_reduction): Likewise
958 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
960 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
962 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
963 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
964 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
966 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
967 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
968 (vect_update_vf_for_slp, vect_analyze_loop_operations)
969 (vect_is_slp_reduction, vectorizable_induction)
970 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
971 * tree-vect-patterns.c (vect_init_pattern_stmt):
972 (vect_determine_min_output_precision_1, vect_determine_precisions)
973 (vect_pattern_recog): Likewise.
974 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
975 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
976 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
977 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
978 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
979 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
980 info field from a loop to a loop_vec_info.
982 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
984 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
985 (vec_info::add_stmt): Declare.
986 * tree-vectorizer.c (vec_info::add_stmt): New function.
987 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
988 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
989 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
990 (vectorizable_induction): Likewise.
991 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
992 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
993 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
994 (vectorizable_load): Likewise.
995 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
996 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
997 (vect_recog_gather_scatter_pattern): Likewise.
998 (append_pattern_def_seq): Likewise. Remove a check that is
999 performed by add_stmt itself.
1001 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1003 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
1004 which make_ssa_name was called with new_stmt before new_stmt
1007 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1009 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
1011 (vect_is_slp_reduction): ...here...
1012 (vect_is_simple_reduction): ...and here. Remove repetition of tests
1013 that are already known to be false.
1015 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1017 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
1018 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
1019 STMT_VINFO_NUM_SLP_USES when it's true.
1020 (vect_free_slp_instance): Add a final_p parameter and pass it to
1022 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
1023 (vect_analyze_slp_instance): Likewise.
1024 (vect_slp_analyze_operations): Likewise.
1025 (vect_slp_analyze_bb_1): Likewise.
1026 * tree-vectorizer.c (vec_info): Likewise.
1027 * tree-vect-loop.c (vect_transform_loop): Likewise.
1029 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1031 * tree-vect-loop.c (vectorizable_reduction): Assert that the
1032 function is not called for second and subsequent members of
1035 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
1037 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
1038 cases for nested loops from here to ...
1039 (vect_create_epilog_for_reduction): ...here. Only call
1040 vect_is_simple_use for inner-loop reductions.
1042 2018-07-31 Martin Liska <mliska@suse.cz>
1044 PR gcov-profile/85338
1045 PR gcov-profile/85350
1046 PR gcov-profile/85372
1047 * profile.c (struct location_triplet): New.
1048 (struct location_triplet_hash): Likewise.
1049 (output_location): Do not output a BB that
1050 is already recorded for a line.
1051 (branch_prob): Use streamed_locations.
1053 2018-07-31 Martin Liska <mliska@suse.cz>
1055 PR gcov-profile/85370
1056 * coverage.c (coverage_begin_function): Do not mark target
1057 clones as artificial functions.
1059 2018-07-31 Martin Liska <mliska@suse.cz>
1061 PR gcov-profile/83813
1062 PR gcov-profile/84758
1063 PR gcov-profile/85217
1064 PR gcov-profile/85332
1065 * profile.c (branch_prob): Do not record GOTO expressions
1066 for GIMPLE statements which locations are already streamed.
1068 2018-07-31 Olivier Hainque <hainque@adacore.com>
1070 * gcc.c (handle_spec_function): Accept a soft_matched_part
1071 argument, as do_spec_1. Pass it down to ...
1072 (eval_spec_function): Accept a soft_matched_part argument,
1073 and pass it down to ...
1074 (do_spec_2): Accept a soft_matched_part argument, and pass
1075 it down to do_spec_1.
1076 (do_spec_1): Pass soft_matched_part to handle_spec_function.
1077 (handle_braces): Update call to handle_spec_function.
1078 (driver::set_up_specs): Update calls to do_spec_2.
1079 (compare_debug_dump_opt_spec_function): Likewise.
1080 (compare_debug_self_opt_spec_function): Likewise.
1082 2018-07-31 Olivier Hainque <hainque@adacore.com>
1084 * common.opt (nolibc): New option.
1085 * doc/invoke.texi (Link Options): Document it.
1086 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
1087 * config/alpha/linux.h: Likewise.
1088 * config/arc/elf.h: Likewise.
1089 * config/arm/uclinux-elf.h: Likewise.
1090 * config/arm/unknown-elf.h: Likewise.
1091 * config/avr/avrlibc.h: Likewise.
1092 * config/bfin/bfin.h: Likewise.
1093 * config/bfin/linux.h: Likewise.
1094 * config/bfin/uclinux.h: Likewise.
1095 * config/darwin.h: Likewise.
1096 * config/darwin10.h: Likewise.
1097 * config/darwin12.h: Likewise.
1098 * config/gnu-user.h: Likewise.
1099 * config/lm32/uclinux-elf.h: Likewise.
1100 * config/pa/pa-hpux11.h: Likewise.
1101 * config/pa/pa64-hpux.h: Likewise.
1102 * config/sparc/sparc.h: Likewise.
1104 2018-07-31 Olivier Hainque <hainque@adacore.com>
1106 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
1107 undefined variables.
1109 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
1112 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
1115 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
1117 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
1118 terminated string literal.
1120 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
1122 PR rtl-optimization/85160
1123 * combine.c (is_just_move): New function.
1124 (try_combine): Allow combining two instructions into two if neither of
1125 the original instructions was a move.
1127 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
1130 * doc/extend.texi (Global Register Variables): Discourage use of type
1132 (Local Register Variables): Likewise.
1134 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
1136 PR tree-optimization/86506
1137 * hwint.h (ceil_log2): Resync with hwint.c implementation.
1139 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
1142 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
1143 hard_regno, make sure no insn between `from` and `to` clobbers it.
1145 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
1146 Tom de Vries <tdevries@suse.de>
1148 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
1149 (PTX_DEFAULT_RUNTIME_DIM): ... this.
1150 (nvptx_goacc_validate_dims): Set default worker and gang dims to
1151 PTX_DEFAULT_RUNTIME_DIM.
1152 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
1154 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
1156 * config/pa/pa.c (pa_output_addr_vec): Align address table.
1157 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
1158 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
1160 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
1162 * config/rs6000/constraints.md (wG constraint): Delete, no longer
1164 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
1165 predicate to reflect toc fusion has been deleted.
1166 (toc_fusion_mem_raw): Delete, no longer used.
1167 (toc_fusion_mem_wrapped): Likewise.
1168 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
1170 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
1171 Delete, no longer used.
1172 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
1173 meant to be used for toc fusion.
1174 (rs6000_debug_print_mode): Delete toc fusion debugging.
1175 (rs6000_debug_reg_global): Likewise.
1176 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
1177 fusion and secondary reload support that were never used.
1178 (rs6000_option_override_internal): Delete TOC fusion, that was only
1179 partially defined, and it did not work unless you also used the
1181 (rs6000_legitimate_address_p): Delete TOC fusion support.
1182 (rs6000_opt_masks): Likewise.
1183 (fusion_wrap_memory_address): Delete function, no longer used.
1184 (fusion_split_address); Delete TOC fusion support.
1185 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
1186 longer used with toc fusion being deleted.
1187 (TARGET_TOC_FUSION_FP): Likewise.
1188 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
1190 (toc fusion spliter): Delete TOC fusion support.
1191 (toc_fusionload_<mode>): Likewise.
1192 (toc_fusionload_di): Likewise.
1193 (fusion_gpr_load_<mode>): Delete generator function, this insn no
1194 longer needs to be named. Rename predicate to delete TOC fusion.
1195 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
1196 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
1197 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
1198 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
1199 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
1201 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
1203 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
1204 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
1205 __int128 in built-in function prototypes.
1206 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
1207 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
1209 2018-07-27 Martin Sebor <msebor@redhat.com>
1211 PR tree-optimization/86696
1212 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
1213 types, including enums.
1214 (handle_char_store): Be prepared for the above function to fail.
1216 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
1218 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
1219 when optimization level is lower than 2 or optimize for size.
1221 2018-07-26 Martin Sebor <msebor@redhat.com>
1223 PR tree-optimization/86043
1224 PR tree-optimization/86042
1225 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
1226 (get_string_cst_length): Rename...
1227 (get_min_string_length): ...to this. Add argument.
1228 (handle_char_store): Extend to handle multi-character stores by
1230 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
1231 * tree.h (initializer_zerop): Add argument.
1233 2018-07-26 Jakub Jelinek <jakub@redhat.com>
1236 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
1237 declare target to variables if they have always,{to,from,tofrom} map
1240 2018-07-26 Martin Liska <mliska@suse.cz>
1243 * lto-wrapper.c: Add linker_output as prefix
1244 for ltrans_output_file.
1246 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
1248 PR rtl-optimization/85805
1249 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
1250 value for hard registers if that was written in the same mode.
1252 2018-07-26 Martin Liska <mliska@suse.cz>
1254 PR gcov-profile/86536
1255 * gcov.c (format_gcov): Use printf format %.*f directly
1256 and do not handle special values.
1258 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
1260 * common/config/arc/arc-common.c (arc_option_optimization_table):
1261 Update default optimizations for size.
1263 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
1265 * config/arc/arc.md (movsf_insn): Add short instruction selection.
1266 * config/arc/constraints.md (CfZ): New constraint.
1267 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
1268 (subsf3_fpu): Likewise.
1269 (cmpsf_fpu): Likewise.
1270 (cmpsf_fpu_uneq): Likewise.
1272 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
1274 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
1276 (arc_is_uncached_mem_p): uncached applies to both the variable and
1279 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
1281 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
1284 2018-07-25 David Malcolm <dmalcolm@redhat.com>
1286 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
1287 field "m_scopes" from vec to auto_vec.
1289 2018-07-25 Martin Liska <mliska@suse.cz>
1291 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
1294 2018-07-25 Richard Biener <rguenther@suse.de>
1297 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
1298 special wrt context_die late.
1299 (gen_subprogram_die): Re-use DIEs in local scope.
1301 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
1303 PR tree-optimization/86644
1304 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
1306 2018-07-25 Martin Liska <mliska@suse.cz>
1309 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
1310 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
1312 2018-07-25 Martin Liska <mliska@suse.cz>
1315 * params.def: Explain ASan abbreviation and provide
1316 a documentation link.
1318 2018-07-24 Martin Sebor <msebor@redhat.com>
1320 PR tree-optimization/86622
1321 PR tree-optimization/86532
1322 * builtins.h (string_length): Declare.
1323 * builtins.c (c_strlen): Correct handling of non-constant offsets.
1324 (check_access): Be prepared for non-constant length ranges.
1325 (string_length): Make extern.
1326 * expr.c (string_constant): Only handle the minor non-constant
1327 array index. Use string_constant to compute the length of
1328 a generic string constant.
1330 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
1332 PR tree-optimization/86618
1333 * tree-vect-stmts.c (vectorizable_call): Don't take the address
1334 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
1336 2018-07-24 David Malcolm <dmalcolm@redhat.com>
1338 PR tree-optimization/86636
1339 * json.cc (json::object::set): Fix comment. Add assertions.
1340 (json::array::append): Move here from json.h. Add comment and an
1342 (json::string::string): Likewise.
1343 * json.h (json::array::append): Move to json.cc.
1344 (json::string::string): Likewise.
1345 * optinfo-emit-json.cc
1346 (optrecord_json_writer::impl_location_to_json): Assert that we
1347 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
1348 wrapper around it. Expand the location once, rather than three
1350 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
1351 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
1353 (optrecord_json_writer::optinfo_to_json): Likewise, in four
1354 places. Fix some overlong lines.
1356 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
1358 * config/aarch64/aarch64-simd.md
1359 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
1360 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
1361 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
1362 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
1363 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
1364 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
1365 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
1366 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
1367 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
1369 2018-07-24 Jakub Jelinek <jakub@redhat.com>
1372 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
1373 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
1374 and abs_d == d, do the power of two handling if profitable.
1376 2018-07-24 Richard Biener <rguenther@suse.de>
1378 * match.pd: Add BIT_FIELD_REF canonicalizations.
1380 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
1383 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
1385 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
1387 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
1388 terminated STRING_CST object.
1390 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
1392 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
1394 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
1396 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
1397 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
1398 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
1399 the elements into a register.
1400 (rs6000_split_v4si_init_di_reg): Delete.
1401 (rs6000_split_v4si_init): Delete.
1402 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
1403 (vsx_init_v4si): Rewrite as a define_expand.
1405 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
1407 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
1408 zero_extend argument from memory): New.
1410 2018-07-22 Martin Sebor <msebor@redhat.com>
1413 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
1414 diagnosing calls with unknown arguments unless -Walloca-larger-than
1415 is restricted to less than PTRDIFF_MAX bytes.
1417 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
1419 * doc/gcov.texi (Invoking Gcov): Editorial changes.
1421 2018-07-20 David Malcolm <dmalcolm@redhat.com>
1423 * pretty-print.c (text_info::set_location): Remove redundant
1424 "line_table" parameter from call to rich_location::set_range.
1426 2018-07-20 Martin Sebor <msebor@redhat.com>
1429 * builtins.c (expand_builtin_alloca): Adjust.
1430 * calls.c (alloc_max_size): Simplify.
1431 * cgraphunit.c (cgraph_node::expand): Adjust.
1432 * common.opt (larger_than_size, warn_frame_larger_than): Remove
1434 (frame_larger_than_size): Same.
1435 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
1436 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
1438 * doc/invoke.texi (GCC Command Options): Document option arguments.
1439 Explain byte-size arguments and suffixes.
1440 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
1441 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
1442 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
1443 * doc/options.texi (UInteger): Expand.
1444 (Host_Wide_Int, ByteSize): Document new properties.
1445 * final.c (final_start_function_1): Include sizes in an error message.
1446 * function.c (frame_offset_overflow): Same.
1447 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
1448 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
1449 Diagnose unbounded alloca calls only for limits of less than
1451 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
1452 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
1454 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
1455 only for limits of less than PTRDIFF_MAX.
1456 * langhooks-def.h (lhd_handle_option): Change function argument
1458 * langhooks.c (lhd_handle_option): Same.
1459 * langhooks.h (handle_option): Same.
1460 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
1462 (var_type, var_type_struct): Same.
1463 (var_set): Handle ByteSize flag.
1464 * optc-gen.awk: Add comments to output to ease debugging. Make
1465 use of HOST_WIDE_INT where appropriate.
1466 * opts-gen-save.awk: Use %lx to format unsigned long.
1467 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
1468 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
1469 arguments. Parse bytes-size suffixes.
1470 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
1471 (enum_value_to_arg): Same.
1472 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
1473 (handle_option): Adjust.
1474 (generate_option): Change function argument to HOST_WIDE_INT.
1475 (cmdline_handle_error): Adjust.
1476 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
1477 (set_option): Change function argument to HOST_WIDE_INT.
1478 (option_enabled): Handle cl_host_wide_int.
1479 (get_option_state): Handle CLVC_SIZE.
1480 (control_warning_option): Same.
1481 * opts.c (common_handle_option): Change function argument to
1482 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
1483 OPT_Wvla_larger_than_.
1484 * opts.h (enum cl_var_type): Add an enumerator.
1485 * stor-layout.c (layout_decl): Print a more meaningful warning.
1486 * toplev.c (output_stack_usage): Adjust.
1488 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
1490 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
1491 call to inline_expand_builtin_string_cmp.
1492 (expand_builtin_strcmp): Likewise.
1493 (expand_builtin_strncmp): Likewise.
1494 (inline_string_cmp): Delete the last parameter, change char_type_node
1495 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
1497 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
1498 the inlining expansion on target where the type of the call has same or
1499 narrower precision than unsigned char.
1501 2018-07-20 David Malcolm <dmalcolm@redhat.com>
1503 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
1504 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
1505 * common.opt (fsave-optimization-record): New option.
1506 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
1507 * doc/invoke.texi (-fsave-optimization-record): New option.
1508 * dumpfile.c: Include "optinfo-emit-json.h".
1509 (struct kv_pair): Move to coretypes.h.
1510 (optgroup_options): Make non-static.
1511 (dump_context::end_scope): Call
1512 optimization_records_maybe_pop_dump_scope.
1513 * dumpfile.h (optgroup_options): New decl.
1514 * json.cc: New file.
1516 * optinfo-emit-json.cc: New file.
1517 * optinfo-emit-json.h: New file.
1518 * optinfo.cc: Include "optinfo-emit-json.h".
1519 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
1520 (optinfo_enabled_p): Check optimization_records_enabled_p.
1521 (optinfo_wants_inlining_info_p): Likewise.
1522 * optinfo.h: Update comment.
1523 * profile-count.c (profile_quality_as_string): New function.
1524 * profile-count.h (profile_quality_as_string): New decl.
1525 (profile_count::quality): New accessor.
1526 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
1527 and optinfo_emit_json_cc_tests.
1528 * selftest.h (selftest::json_cc_tests): New decl.
1529 (selftest::optinfo_emit_json_cc_tests): New decl.
1530 * toplev.c: Include "optinfo-emit-json.h".
1531 (compile_file): Call optimization_records_finish.
1532 (do_compile): Call optimization_records_start.
1533 * tree-ssa-live.c: Include optinfo.h.
1534 (remove_unused_scope_block_p): Retain inlining information if
1535 optinfo_wants_inlining_info_p returns true.
1537 2018-07-20 Richard Biener <rguenther@suse.de>
1540 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
1541 to cover -flto-partition=none.
1543 2018-07-20 Martin Liska <mliska@suse.cz>
1545 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
1546 (get_decl_source_range): Remove unused function.
1548 2018-07-20 Richard Biener <rguenther@suse.de>
1550 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
1551 (struct vn_phi_s): Likewise.
1552 (struct vn_reference_s): Likewise.
1553 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
1554 for searching the slot of an entry known to be in the hash itself.
1555 (vn_phi_hasher::equal): Likewise.
1556 (vn_reference_hasher::equal): Likewise.
1557 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
1559 (optimistic_info, current_info): Remove, keeping only valid_info.
1560 (vn_reference_lookup_1): Remove fallback lookup.
1561 (vn_reference_lookup_2): Likewise.
1562 (vn_nary_op_lookup_1): Likewise.
1563 (vn_phi_lookup): Likewise.
1564 (vn_nary_build_or_lookup_1): Make sure to not chain the built
1566 (vn_reference_insert): Adjust, chain the inserted hash element
1567 at last_inserted_ref.
1568 (vn_reference_insert_pieces): Likewise.
1569 (visit_reference_op_call): Likewise.
1570 (vn_nary_op_insert_into): Chain the inserted hash element at
1572 (vn_nary_op_insert_pieces): Adjust.
1573 (vn_nary_op_insert): Likewise.
1574 (vn_nary_op_insert_stmt): Likewise.
1575 (vn_phi_insert): Adjust, chain the inserted hash element at
1577 (process_scc): Remove clearing and copying the optimistic
1578 table. Instead remove elements inserted during an optimistic
1579 iteration from the single table we maintain.
1580 (init_scc_vn): Adjust.
1581 (free_scc_vn): Likewise.
1582 (sccvn_dom_walker::record_cond): Likewise.
1583 (sccvn_dom_walker::after_dom_children): Likewise.
1585 2018-07-19 Martin Sebor <msebor@redhat.com>
1587 PR tree-optimization/84047
1588 PR tree-optimization/83776
1589 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
1590 (check_array_bounds): Call it.
1592 2018-07-19 Martin Sebor <msebor@redhat.com>
1594 * align.h (align_flags): Use member initialization.
1596 2018-07-19 David Malcolm <dmalcolm@redhat.com>
1598 * Makefile.in (OBJS): Add optinfo.o.
1599 * coretypes.h (class symtab_node): New forward decl.
1600 (struct cgraph_node): New forward decl.
1601 (class varpool_node): New forward decl.
1602 * dump-context.h: New file.
1603 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
1605 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
1606 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
1607 (set_alt_dump_file): Likewise.
1608 (dump_context::~dump_context): New dtor.
1609 (dump_gimple_stmt): Move implementation to...
1610 (dump_context::dump_gimple_stmt): ...this new member function.
1611 Add the stmt to any pending optinfo, creating one if need be.
1612 (dump_gimple_stmt_loc): Move implementation to...
1613 (dump_context::dump_gimple_stmt_loc): ...this new member function.
1614 Start a new optinfo and add the stmt to it.
1615 (dump_gimple_expr): Move implementation to...
1616 (dump_context::dump_gimple_expr): ...this new member function.
1617 Add the stmt to any pending optinfo, creating one if need be.
1618 (dump_gimple_expr_loc): Move implementation to...
1619 (dump_context::dump_gimple_expr_loc): ...this new member function.
1620 Start a new optinfo and add the stmt to it.
1621 (dump_generic_expr): Move implementation to...
1622 (dump_context::dump_generic_expr): ...this new member function.
1623 Add the tree to any pending optinfo, creating one if need be.
1624 (dump_generic_expr_loc): Move implementation to...
1625 (dump_context::dump_generic_expr_loc): ...this new member
1626 function. Add the tree to any pending optinfo, creating one if
1628 (dump_printf): Move implementation to...
1629 (dump_context::dump_printf_va): ...this new member function. Add
1630 the text to any pending optinfo, creating one if need be.
1631 (dump_printf_loc): Move implementation to...
1632 (dump_context::dump_printf_loc_va): ...this new member function.
1633 Start a new optinfo and add the stmt to it.
1634 (dump_dec): Move implementation to...
1635 (dump_context::dump_dec): ...this new member function. Add the
1636 value to any pending optinfo, creating one if need be.
1637 (dump_context::dump_symtab_node): New member function.
1638 (dump_context::get_scope_depth): New member function.
1639 (dump_context::begin_scope): New member function.
1640 (dump_context::end_scope): New member function.
1641 (dump_context::ensure_pending_optinfo): New member function.
1642 (dump_context::begin_next_optinfo): New member function.
1643 (dump_context::end_any_optinfo): New member function.
1644 (dump_context::s_current): New global.
1645 (dump_context::s_default): New global.
1646 (dump_scope_depth): Delete global.
1647 (dumpfile_ensure_any_optinfo_are_flushed): New function.
1648 (dump_symtab_node): New function.
1649 (get_dump_scope_depth): Reimplement in terms of dump_context.
1650 (dump_begin_scope): Likewise.
1651 (dump_end_scope): Likewise.
1652 (selftest::temp_dump_context::temp_dump_context): New ctor.
1653 (selftest::temp_dump_context::~temp_dump_context): New dtor.
1654 (selftest::verify_item): New function.
1655 (ASSERT_IS_TEXT): New macro.
1656 (ASSERT_IS_TREE): New macro.
1657 (ASSERT_IS_GIMPLE): New macro.
1658 (selftest::test_capture_of_dump_calls): New test.
1659 (selftest::dumpfile_c_tests): Call it.
1660 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
1661 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
1662 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
1663 descriptive comment.
1664 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
1665 (dump_node, dump_bb): Move these unrelated decls.
1666 (class dump_manager): Add leading comment.
1667 * optinfo.cc: New file.
1668 * optinfo.h: New file.
1670 2018-07-19 Michael Collison <michael.collison@arm.com>
1671 Richard Henderson <rth@redhat.com>
1673 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
1674 (subti): Handle op1 zero.
1675 (subvti4, usub4ti4): New.
1676 (*sub<GPI>3_compare1_imm): New.
1677 (sub<GPI>3_carryinCV): New.
1678 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
1679 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
1681 2018-07-19 Michael Collison <michael.collison@arm.com>
1682 Richard Henderson <rth@redhat.com>
1684 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
1685 (addti3): Create simpler code if low part is already known to be 0.
1686 (addvti4, uaddvti4): New.
1687 (*add<GPI>3_compareC_cconly_imm): New.
1688 (*add<GPI>3_compareC_cconly): New.
1689 (*add<GPI>3_compareC_imm): New.
1690 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
1691 handle constants within this pattern..
1692 (*add<GPI>3_compareV_cconly_imm): New.
1693 (*add<GPI>3_compareV_cconly): New.
1694 (*add<GPI>3_compareV_imm): New.
1695 (add<GPI>3_compareV): New.
1696 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
1697 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
1698 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
1699 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
1700 with 'comparison' operator.
1701 (*add<GPI>3_compareV_cconly_imm): Ditto.
1702 (*add<GPI>3_compareV_cconly): Ditto.
1703 (*add<GPI>3_compareV_imm): Ditto.
1704 (add<GPI>3_compareV): Ditto.
1705 (add<mode>3_carryinC): Ditto.
1706 (*add<mode>3_carryinC_zero): Ditto.
1707 (*add<mode>3_carryinC): Ditto.
1708 (add<mode>3_carryinV): Ditto.
1709 (*add<mode>3_carryinV_zero): Ditto.
1710 (*add<mode>3_carryinV): Ditto.
1712 2018-07-19 Michael Collison <michael.collison@arm.com>
1713 Richard Henderson <rth@redhat.com>
1715 * config/aarch64/aarch64-modes.def (CC_V): New.
1716 * config/aarch64/aarch64-protos.h
1717 (aarch64_addti_scratch_regs): Declare
1718 (aarch64_subvti_scratch_regs): Declare.
1719 (aarch64_expand_subvti): Declare.
1720 (aarch64_gen_unlikely_cbranch): Declare
1721 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
1722 for signed overflow using CC_Vmode.
1723 (aarch64_get_condition_code_1): Handle CC_Vmode.
1724 (aarch64_gen_unlikely_cbranch): New function.
1725 (aarch64_addti_scratch_regs): New function.
1726 (aarch64_subvti_scratch_regs): New function.
1727 (aarch64_expand_subvti): New function.
1729 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
1731 * config/aarch64/aarch64-option-extensions.def: New entry for profile
1733 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
1734 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
1737 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
1740 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
1741 address check not strict.
1743 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
1745 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
1747 (store_pair_lanes<mode>): Likewise.
1748 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
1749 enum value 'ADDR_QUERY_LDP_STP_N'.
1750 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
1751 (aarch64_print_address_internal): Add declaration.
1752 (aarch64_print_ldpstp_address): Remove.
1753 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
1754 (aarch64_print_operand): Change printing of 'y'.
1755 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
1756 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
1757 'true' rather than '1'.
1758 * config/aarch64/constraints.md (Uml): Likewise.
1759 (Uml): Rename to Umn.
1762 2018-07-19 Richard Biener <rguenther@suse.de>
1764 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
1766 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
1767 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
1768 (vn_reference_hasher): Likewise.
1769 (struct vn_tables_s): Remove obstack and alloc-pool members.
1770 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
1771 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
1772 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
1773 (vn_reference_insert_pieces): Likewise.
1774 (alloc_vn_nary_op_noinit): Adjust.
1775 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
1776 (vn_phi_eq): Adjust.
1777 (shared_lookup_phiargs): Remove.
1778 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
1779 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
1780 (visit_reference_op_call): Likewise.
1781 (copy_nary, copy_phi, copy_reference): Remove.
1782 (process_scc): Rewind the obstack when iterating. Do not
1783 copy the elements to valid_info but just move them from one
1784 hashtable to the other.
1785 (allocate_vn_table): Adjust.
1786 (free_vn_table): Likewise.
1787 (init_scc_vn): Likewise.
1788 (free_scc_vn): Likewise.
1790 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
1793 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
1794 indirect_return as function type attribute.
1795 (ix86_attribute_table): Change indirect_return to function
1797 * doc/extend.texi: Update indirect_return attribute.
1799 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
1801 * wide-int.h (widest2_int): New.
1802 * gimple-fold.c (arith_overflowed_p): Use it.
1803 * tree.h (widest2_int_cst): New.
1804 * tree-vrp.c (wide_int_binop_overflow): Rename from
1805 vrp_int_const_binop.
1806 Rewrite to work on trees.
1807 (extract_range_from_multiplicative_op_1): Abstract code to...
1808 (wide_int_range_min_max): ...here.
1809 (wide_int_range_cross_product): ...and here.
1810 (extract_range_from_binary_expr_1): Abstract overflow code to...
1811 (wide_int_range_mult_wrapping): ...here.
1812 * tree-vrp.h (wide_int_range_cross_product): New.
1813 (wide_int_range_mult_wrapping): New.
1815 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
1816 Julia Koval <julia.koval@intel.com>
1818 * config/i386/x86-tune-costs.h (skylake_memcpy,
1819 skylake_memset): Replace rep_prefix with unrolling for size 512.
1821 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
1824 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle comparision with EQ_EXPR
1827 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
1829 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
1830 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
1831 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
1832 previously known as "PowerPC AltiVec Built-in Functions". Move
1833 some material to new subsubsections "PowerPC AltiVec Built-in
1834 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
1836 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
1837 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
1838 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
1839 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
1841 2018-07-18 Richard Biener <rguenther@suse.de>
1843 PR tree-optimization/86557
1844 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
1847 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
1849 * config/s390/s390.c (s390_function_profiler): Generate CFI.
1851 2018-07-17 Jeff Law <law@redhat.com>
1853 * config/arm/arm.c (get_label_padding): Update for recent
1854 changes to label_to_alignment.
1856 PR tree-optimization/86010
1857 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
1859 * config/mips/mips.c (vr4130_align_insns): Update for recent
1860 changes to label_to_alignment.
1862 * config/frv/frv.c (frv_label_align): Update for recent changes
1863 to label_to_alignment.
1865 * config/nios2/nios2.c (nios2_label_align): Update for recent
1866 changes which dropped ALIGN_LABELS_LOG.
1868 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
1870 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
1871 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
1873 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
1875 * config/arc/arc.c (arc_label_align): Use align_labels instead of
1876 deprecated align_labels_log.
1878 2018-07-17 Richard Biener <rguenther@suse.de>
1881 * dwarf2out.c (init_sections_and_labels): Always generate
1882 a debug_line_str_section for early LTO debug.
1883 (dwarf2out_finish): Reset debug_line_str_hash output early.
1884 Bump counter for extra dwarf5 .debug_loc labels to not conflict
1886 (dwarf2out_early_finish): Output debug_line_str.
1888 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
1890 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
1891 index register on z196 or later.
1893 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
1895 * config/s390/s390.c (s390_default_align): Set default function
1897 (s390_override_options_after_change): Call s390_default align.
1898 (s390_option_override_internal): Call s390_default_align.
1899 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
1901 2018-07-17 Jakub Jelinek <jakub@redhat.com>
1904 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
1905 to _looptemp_ clauses, other than the first two.
1907 2018-07-17 Martin Liska <mliska@suse.cz>
1909 * opts.c: Do not enable OPT_falign_* for -Os.
1911 2018-07-17 Martin Liska <mliska@suse.cz>
1913 * align.h (MAX_CODE_ALIGN): New.
1914 (MAX_CODE_ALIGN_VALUE): New.
1915 * common/config/i386/i386-common.c (ix86_handle_option):
1916 (MAX_CODE_ALIGN): Moved to align.h.
1917 * final.c (MAX_CODE_ALIGN): Likewise.
1918 * opts.c (parse_and_check_align_values):
1919 (MAX_CODE_ALIGN): Likewise.
1920 (MAX_CODE_ALIGN_VALUE): Likewise.
1922 2018-07-17 Martin Liska <mliska@suse.cz>
1924 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
1925 in order to fulfil coding style.
1926 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
1927 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1928 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
1929 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
1930 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
1931 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
1932 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
1933 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
1934 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1936 2018-07-17 Martin Liska <mliska@suse.cz>
1938 * align.h: New file.
1939 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions directly.
1940 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
1941 align_flags of label_to_alignment.
1942 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into align_flags
1944 * config/m68k/m68k.c: Do not use removed align_labels_value and
1946 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
1947 (LOOP_ALIGN): Likewise.
1948 (LABEL_ALIGN): Likewise.
1949 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
1950 Remove not used macro.
1951 (rs6000_loop_align): Change return type to align_flags.
1952 (rs6000_loop_align_max_skip): Remove.
1953 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
1954 Change return type to align_flags.
1955 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
1956 Remove not used macro.
1957 (rs6000_loop_align): Change return type to align_flags.
1958 (rs6000_loop_align_max_skip): Remove.
1959 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
1960 * config/rx/rx-protos.h (rx_align_for_label): Make it
1962 * config/rx/rx.c (rx_align_for_label): Change return type
1964 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
1966 into align_flags class.
1967 (LABEL_ALIGN): Likewise.
1968 (LOOP_ALIGN): Likewise.
1969 * config/s390/s390.c (s390_label_align): Use align_flags
1971 (s390_asm_output_function_label): Likewise.
1972 * config/sh/sh.c (sh_override_options_after_change):
1973 Use align_flags class directly without macros.
1974 (find_barrier): Likewise.
1975 (barrier_align): Likewise.
1976 (sh_loop_align): Likewise.
1977 * config/spu/spu.c (spu_option_override):
1978 Use align_flags_tuple::get_value instead of removed macros.
1979 (spu_sched_init): Likewise.
1980 * config/spu/spu.h (GTY): Likewise.
1981 * config/visium/visium.c (visium_option_override):
1982 Set "8" as default secondary alignment.
1983 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
1984 in order to guarantee secondary alignment of 8.
1985 * coretypes.h: Include align.h header file.
1986 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
1987 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
1988 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
1989 * doc/tm.texi.in: Likewise.
1990 * final.c (struct label_alignment): Remove not used structure.
1991 (LABEL_ALIGN): Change type to align_flags.
1992 (LOOP_ALIGN): Likewise.
1993 (JUMP_ALIGN): Likewise.
1994 (default_loop_align_max_skip): Remove.
1995 (default_label_align_max_skip): Likewise.
1996 (default_jump_align_max_skip): Likewise.
1997 (default_label_align_after_barrier_max_skip):
1998 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
1999 (LABEL_TO_MAX_SKIP): Remove.
2000 (label_to_alignment): Return align_flags type instead of integer.
2001 (label_to_max_skip): Remove.
2002 (align_fuzz): Use align_flags type.
2003 (compute_alignments): Use align_flags type and use align_flags::max
2004 to combine multiple alignments.
2005 (grow_label_align): Grow vec instead of C array.
2006 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
2007 (shorten_branches): Use align_flags type and use align_flags::max
2008 to combine multiple alignments.
2009 (final_scan_insn_1): Remove usage of secondary alignment that comes
2010 from label alignment, but instead use proper secondary alignment
2011 which is computed in grow_label_align.
2012 * flags.h (struct align_flags_tuple): Move to align.h.
2013 (struct align_flags): Likewise.
2014 (state_align_loops): Rename to align_loops.
2015 (state_align_jumps): Rename to align_jumps.
2016 (state_align_labels): Rename to align_labels.
2017 (state_align_functions): Rename to align_functions.
2018 (align_loops_log): Remove.
2019 (align_jumps_log): Remove.
2020 (align_labels_log): Remove.
2021 (align_functions_log): Remove.
2022 (align_loops_max_skip): Remove.
2023 (align_jumps_max_skip): Remove.
2024 (align_labels_max_skip): Remove.
2025 (align_functions_max_skip): Remove.
2026 (align_loops_value): Remove.
2027 (align_jumps_value): Remove.
2028 (align_labels_value): Remove.
2029 (align_functions_value): Remove.
2030 * output.h (label_to_alignment): Change return type to align_flags.
2031 (label_to_max_skip): Remove.
2032 * target.def: Remove loop_align_max_skip, label_align_max_skip,
2033 jump_align_max_skip macros.
2034 * targhooks.h (default_loop_align_max_skip): Remove.
2035 (default_label_align_max_skip): Likewise.
2036 (default_jump_align_max_skip): Likewise.
2037 (default_label_align_after_barrier_max_skip): Remove.
2038 * toplev.c (read_log_maxskip): Use ::normalize function.
2039 (parse_N_M): Remove not used argument and also call ::normalize.
2040 (parse_alignment_opts): Do not pass unused arguments.
2041 * varasm.c (assemble_start_function): Use directly align_functions
2042 instead of removed macros.
2043 * system.h: Do not poison removed macros.
2045 2018-07-17 Jakub Jelinek <jakub@redhat.com>
2048 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
2049 and cond temporaries don't have reference type if iterator has
2050 pointer type. For init use &for_pre_body instead of pre_p if
2051 for_pre_body is non-empty.
2053 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
2055 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
2056 double-double modes to SFmode directly directly.
2057 (trunc<mode>sf2_fprs): Delete.
2059 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
2061 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
2062 for conversions between IFmode and the decimal floating point modes.
2063 (init_float128_ieee): Use the correct names for conversions between
2064 KFmode and the decimal floating point modes.
2066 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
2068 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
2069 for the conversions between TDmode and IFmode.
2070 (init_float128_ieee): Use more correct names for the conversions
2071 between TDmode and KFmode.
2073 2018-07-16 Jakub Jelinek <jakub@redhat.com>
2075 PR tree-optimization/86526
2076 * builtins.c (expand_builtin_memcmp): Formatting fixes.
2077 (inline_expand_builtin_string_cmp): Likewise.
2078 (inline_string_cmp): Likewise. Use c_readstr instead of
2079 builtin_memcpy_read_str. Add unit_mode temporary.
2081 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
2084 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
2085 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
2087 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
2089 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
2090 Alphabetize prototypes of built-in functions, separating out
2091 built-in functions that are listed in this section but should be
2092 described elsewhere.
2094 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
2097 * expmed.c (emit_store_flag): Do not emit setcc followed by a
2098 conditional move when trapping comparison was split to a
2099 non-trapping one (and vice versa).
2101 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
2103 * config/s390/s390.c (s390_function_profiler): Generate nops
2104 instead of profiler call sequences.
2105 * config/s390/s390.opt: Add the new option.
2107 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
2109 * config/s390/s390.c (s390_function_profiler): Generate
2110 __mcount_loc section.
2111 * config/s390/s390.opt: Add the new option.
2113 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
2115 * common.opt: Add the new warning.
2116 * config/s390/s390.c (s390_function_profiler): Emit "brasl
2117 %r0,__fentry__" when -mfentry is specified.
2118 (s390_option_override_internal): Disallow -mfentry for 31-bit
2120 * config/s390/s390.opt: Add the new option.
2122 2018-07-16 Richard Biener <rguenther@suse.de>
2125 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
2126 for function-local FUNCTION_DECL and RESULT_DECL immediately.
2128 2018-07-16 Martin Liska <mliska@suse.cz>
2131 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
2134 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
2136 * config/arc/arcHS.md: Update ARCHS scheduling rules.
2138 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
2140 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
2142 * config/arc/arc-cpus.def (hs4x): New cpu.
2144 * config/arc/arc-tables.opt: Regenerate.
2145 * config/arc/arc.c (arc_sched_issue_rate): New function.
2146 (TARGET_SCHED_ISSUE_RATE): Define.
2147 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
2148 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
2150 (attr tune): Add ARCHS4x tune values.
2151 (attr tune_dspmpy): Define.
2152 (*tst): Correct instruction type.
2153 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
2154 * config/arc/arcHS4x.md: New file.
2155 * config/arc/fpu.md: Update instruction type attributes.
2156 * config/arc/t-multilib: Regenerate.
2158 2018-07-16 Tom de Vries <tdevries@suse.de>
2161 * var-tracking.c (vt_initialize): Fix pre_dec handling.
2163 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2165 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
2168 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
2170 PR tree-optimization/86514
2171 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
2172 conversion to a boolean type from a type with greater precision.
2174 2018-07-16 Tom de Vries <tdevries@suse.de>
2176 * var-tracking.c (vt_initialize): Print adjusted insn slim if
2177 dump_flags request TDF_SLIM.
2179 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
2181 * fold-const.c (int_const_binop_1): Abstract...
2182 (wide_int_binop): ...wide int code here.
2183 (poly_int_binop): ...poly int code here.
2184 Abstract the rest of int_const_binop_1 into int_const_binop.
2185 * fold-const.h (wide_int_binop): New.
2186 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
2187 Remove useless PLUS/MINUS_EXPR case.
2188 (zero_nonzero_bits_from_vr): Move wide int code...
2189 (zero_nonzero_bits_from_bounds): ...here.
2190 (extract_range_from_binary_expr_1): Move mask optimization code...
2191 (range_easy_mask_min_max): ...here.
2192 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
2193 (range_easy_mask_min_max): New.
2195 2018-07-15 Jeff Law <law@redhat.com>
2198 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
2201 2018-07-14 Jim Wilson <jimw@sifive.com>
2203 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
2205 2018-07-14 Paul Koning <ni1d@arrl.net>
2207 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
2209 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
2211 * lto-streamer-out.c (copy_function_or_variable): Dump info about
2214 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
2215 Steve Munroe <munroesj52@gmail.com>
2217 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
2218 (_mm_andnot_si128): Likewise.
2219 (_mm_or_si128): Likewise.
2220 (_mm_xor_si128): Likewise.
2222 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
2225 * builtins.c (expand_builtin_memcmp): Inline the calls first
2226 when result_eq is false.
2227 (expand_builtin_strcmp): Inline the calls first.
2228 (expand_builtin_strncmp): Likewise.
2229 (inline_string_cmp): New routine. Expand a string compare
2230 call by using a sequence of char comparison.
2231 (inline_expand_builtin_string_cmp): New routine. Inline expansion
2232 a call to str(n)cmp/memcmp.
2233 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
2235 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
2237 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
2239 * config/arm/driver-arm.c: Include arm-native.h.
2240 (host_detect_local_cpu): Use auto-generated data tables.
2241 (vendors, arm_cpu_table): Delete. Move part information to ...
2242 * config/arm/arm-cpus.in: ... here.
2243 * config/arm/parsecpu.awk (gen_native): New function.
2244 (vendor, part): New CPU fields.
2245 (END): Add support for building the native CPU detection tables.
2246 * config/arm/t-arm (arm-native.h): Add build rule.
2247 (driver-arm.o): Add dependency on arm-native.h.
2249 2018-07-13 Richard Biener <rguenther@suse.de>
2252 * match.pd (addr1 - addr2): Allow either of the operand to
2255 2018-07-13 Tom de Vries <tdevries@suse.de>
2257 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
2260 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
2262 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
2263 arrays instead of numbered variables.
2265 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
2267 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
2268 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
2270 2018-07-13 Richard Biener <rguenther@suse.de>
2273 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
2274 instead of get_context_die.
2276 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
2277 Richard Biener <rguenther@suse.de>
2280 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
2281 that the loop latch destination where phi is defined.
2283 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
2285 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
2286 (riscv_expand_epilogue): Add assertion to check interrupt mode.
2287 (riscv_set_current_function): Extract getting interrupt type to new
2289 (riscv_get_interrupt_type): New function.
2290 (riscv_merge_decl_attributes): New function, checking interrupt type is
2292 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
2294 2018-07-12 Paul Koning <ni1d@arrl.net>
2296 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
2299 2018-07-12 Paul Koning <ni1d@arrl.net>
2301 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
2302 zero reference, add doloop_end instead.
2303 * doc/md.texi (decrement_and_branch_until_zero): Remove.
2304 (Looping patterns): Remove decrement_and_branch_until_zero. Add
2305 detail for doloop_end.
2307 2018-07-12 Martin Sebor <msebor@redhat.com>
2310 * attribs.c (decl_attributes): Reject conflicting attributes before
2311 calling attribute handlers.
2313 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
2315 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
2317 (gcc::dump_manager::get_dump_file_name): likewise.
2318 (dump_begin): Likewise.
2319 * dumpfile.h (dump_begin): Update prototype.
2320 (gcc::dump_manager::get_dump_file_name,
2321 gcc::dump_manager::get_dump_file_name): Update prototype.
2323 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
2325 * internal-fn.h (vectorizable_internal_fn_p): New function.
2326 * tree-vect-slp.c (compatible_calls_p): Likewise.
2327 (vect_build_slp_tree_1): Remove nops argument. Handle calls
2328 to internal functions.
2329 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
2331 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
2333 * fold-const.h (inverse_conditions_p): Declare.
2334 * fold-const.c (inverse_conditions_p): New function.
2335 * match.pd: Use inverse_conditions_p. Add folds of view_converts
2336 that test the inverse condition of a conditional internal function.
2337 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
2338 * internal-fn.c (internal_fn_mask_index): Handle conditional
2340 (vectorized_internal_fn_supported_p): New function.
2341 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
2342 (any_pred_load_store): Replace with...
2343 (need_to_predicate): ...this new variable.
2344 (redundant_ssa_names): New variable.
2345 (ifcvt_can_use_mask_load_store): Move initial checks to...
2346 (ifcvt_can_predicate): ...this new function. Handle tree codes
2347 for which a conditional internal function exists.
2348 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
2349 instead of ifcvt_can_use_mask_load_store. Update after variable
2351 (predicate_load_or_store): New function, split out from
2352 predicate_mem_writes.
2353 (check_redundant_cond_expr): New function.
2354 (value_available_p): Likewise.
2355 (predicate_rhs_code): Likewise.
2356 (predicate_mem_writes): Rename to...
2357 (predicate_statements): ...this. Use predicate_load_or_store
2358 and predicate_rhs_code.
2359 (combine_blocks, tree_if_conversion): Update after above name changes.
2360 (ifcvt_local_dce): Handle redundant_ssa_names.
2361 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
2362 general conditional functions.
2363 * tree-vect-stmts.c (vectorizable_call): Likewise.
2365 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
2366 Alan Hayward <alan.hayward@arm.com>
2367 David Sherwood <david.sherwood@arm.com>
2369 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
2370 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
2371 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
2372 plus and minus and convert them into IFN_COND_FMA-based sequences.
2373 (convert_mult_to_fma): Handle conditional plus and minus.
2375 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
2377 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
2378 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
2379 (cond_fnms_optab): New optabs.
2380 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
2382 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
2383 * internal-fn.h (get_conditional_internal_fn): Declare.
2384 (get_unconditional_internal_fn): Likewise.
2385 * internal-fn.c (cond_ternary_direct): New macro.
2386 (expand_cond_ternary_optab_fn): Likewise.
2387 (direct_cond_ternary_optab_supported_p): Likewise.
2388 (FOR_EACH_COND_FN_PAIR): Likewise.
2389 (get_conditional_internal_fn): New function.
2390 (get_unconditional_internal_fn): Likewise.
2391 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
2392 (gimple_match_op::gimple_match_op): Add a new overload for 5
2394 (gimple_match_op::set_op): Likewise.
2395 (gimple_resimplify5): Declare.
2396 * genmatch.c (decision_tree::gen): Generate simplifications for
2398 * gimple-match-head.c (gimple_simplify): Define an overload for
2399 5 operands. Handle calls with 5 arguments in the top-level overload.
2400 (convert_conditional_op): Handle conversions from unconditional
2401 internal functions to conditional ones.
2402 (gimple_resimplify5): New function.
2403 (build_call_internal): Pass a fifth operand.
2404 (maybe_push_res_to_seq): Likewise.
2405 (try_conditional_simplification): Try converting conditional
2406 internal functions to unconditional internal functions.
2407 Handle 3-operand unconditional forms.
2408 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
2409 Define ternary equivalents of the current rules for binary conditional
2411 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
2413 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
2414 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
2415 (optab): Handle them.
2416 (SVE_COND_FP_TERNARY): New int iterator.
2417 (sve_fmla_op, sve_fmad_op): New int attributes.
2418 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
2419 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
2420 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
2422 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
2424 * target.def (preferred_else_value): New target hook.
2425 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
2426 * doc/tm.texi: Regenerate.
2427 * targhooks.h (default_preferred_else_value): Declare.
2428 * targhooks.c (default_preferred_else_value): New function.
2429 * internal-fn.h (conditional_internal_fn_code): Declare.
2430 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
2431 (get_conditional_internal_fn): Use it.
2432 (conditional_internal_fn_code): New function.
2433 * gimple-match.h (gimple_match_cond): New struct.
2434 (gimple_match_op): Add a cond member function.
2435 (gimple_match_op::gimple_match_op): Update all forms to take a
2437 * genmatch.c (expr::gen_transform): Use the same condition as res_op
2438 for the suboperation, but don't specify a particular else_value.
2439 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
2440 (visit_nary_op, visit_reference_op_load): Pass
2441 gimple_match_cond::UNCOND to the gimple_match_op constructor.
2442 * gimple-match-head.c: Include tree-eh.h
2443 (convert_conditional_op): New function.
2444 (maybe_resimplify_conditional_op): Likewise.
2445 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
2446 (gimple_resimplify2): Likewise.
2447 (gimple_resimplify3): Likewise.
2448 (gimple_resimplify4): Likewise.
2449 (maybe_push_res_to_seq): Return null for conditional operations.
2450 (try_conditional_simplification): New function.
2451 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
2453 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
2455 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
2457 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
2459 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
2461 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
2463 (hash_tree): Do not hash DECL_FCONTEXT
2464 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
2465 Do not stream DECL_FCONTEXT.
2466 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
2467 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
2469 2018-07-12 Richard Biener <rguenther@suse.de>
2472 * dwarf2out.c (gen_block_die): Only output blocks when they have
2473 at least one !DECL_IGNORED_P variable.
2475 2018-07-12 Richard Biener <rguenther@suse.de>
2478 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
2479 Remove -mieee-fp handling.
2481 2018-07-12 Richard Biener <rguenther@suse.de>
2483 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
2484 left-over from last patch.
2486 2018-07-12 Jakub Jelinek <jakub@redhat.com>
2488 PR tree-optimization/86492
2489 * gimple-ssa-store-merging.c
2490 (imm_store_chain_info::coalesce_immediate_stores): Call
2491 check_no_overlap even for the merge_overlapping case. Formatting fix.
2493 2018-07-12 Richard Biener <rguenther@suse.de>
2496 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
2497 move possibly trapping operations into the conditional.
2499 2018-07-12 Richard Biener <rguenther@suse.de>
2501 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
2502 (vn_lookup_simplify_result): Remove recursion limit applied
2504 (vn_nary_build_or_lookup_1): Adjust.
2505 (try_to_simplify): Likewise.
2506 * gimple-match-head.c (gimple_resimplify1): Instead apply one
2508 (gimple_resimplify2): Likewise.
2509 (gimple_resimplify3): Likewise.
2510 (gimple_resimplify4): Likewise.
2512 2018-07-11 Jakub Jelinek <jakub@redhat.com>
2514 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
2515 Use __mmask64 type instead of __mmask8 for __M argument.
2516 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
2517 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
2519 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
2520 __mmask16 for __M argument.
2521 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
2522 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
2523 to __mmask16 instead of __mmask8.
2524 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
2525 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
2526 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
2527 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
2528 instead of __mmask16 for __U argument.
2529 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
2530 __mmask16 instead of __mmask8 for __U argument.
2531 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
2533 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
2535 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
2537 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
2539 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
2541 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
2543 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
2545 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
2547 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
2549 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
2551 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
2552 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
2553 return type as well as __M argument type and all casts from __mmask8
2555 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
2556 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
2557 return type as well as __M argument type and all casts from __mmask8
2559 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
2560 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
2561 return type as well as __M argument type and all casts from __mmask8
2563 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
2564 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
2565 return type as well as __M argument type and all casts from __mmask8
2567 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
2568 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
2571 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
2573 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
2574 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
2577 2018-07-11 Paul Koning <ni1d@arrl.net>
2579 * doc/md.texi (define_subst): Document how multiple occurrences of
2580 the same argument in the replacement pattern are handled.
2582 2018-07-11 Paul Koning <ni1d@arrl.net>
2584 * doc/extend.texi (Common Variable Attributes): Move "mode" into
2586 (Common Type Attributes): Add "mode" attribute.
2588 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
2590 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
2591 stream DECL_ORIGINAL_TYPE.
2592 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
2593 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
2594 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
2595 Do not walk original type.
2596 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
2598 (write_ts_decl_non_common_tree_pointers): Do not stream
2600 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
2601 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
2603 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
2605 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
2606 thread twice from the same starting edge.
2608 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
2610 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
2611 * gimple.c (gimple_call_nonnull_result_p): ...here...
2612 (gimple_call_nonnull_arg): ...and here.
2613 * gimple.h (gimple_call_nonnull_result_p): New.
2614 (gimple_call_nonnull_arg): New.
2616 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
2618 * config/arm/arm-cpus.in: Move information from fpu field of each
2619 cpu definition to the isa field.
2620 * config/arm/parsecpu.awk (fpu): Delete match rule.
2621 (gen_comm_data): Don't add bits from the CPU's FPU entry.
2623 2018-07-11 Richard Biener <rguenther@suse.de>
2626 * dwarf2out.c (init_sections_and_labels): Use
2627 output_asm_line_debug_info consistently.
2628 (dwarf2out_early_finish): Likewise.
2629 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
2632 2018-07-11 Richard Biener <rguenther@suse.de>
2634 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
2635 Rework father_bb setting in a way to avoid propagating constants
2636 multiple times on a loop body.
2638 2018-07-10 Mark Wielaard <mark@klomp.org>
2641 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
2643 2018-07-10 Richard Biener <rguenther@suse.de>
2645 * hash-map.h (hash_map::iterator::operator*): Return
2646 references to key and value.
2648 2018-07-10 Jakub Jelinek <jakub@redhat.com>
2651 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
2652 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
2653 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
2654 (gimplify_omp_for): For composite loops, move outer
2655 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
2656 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
2657 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
2658 TREE_LIST for both the original class iterator and the "last" helper
2659 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
2660 loop, remember has_decl_expr from outer composite loops for the
2661 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
2663 2018-07-09 Martin Sebor <msebor@redhat.com>
2667 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
2668 accessing implicitly initialized array elements.
2669 * expr.c (string_constant): Handle string initializers of
2670 character arrays within aggregates.
2671 * gimple-fold.c (fold_array_ctor_reference): Add argument.
2672 Store element offset. As a special case, handle zero size.
2673 (fold_nonarray_ctor_reference): Same.
2674 (fold_ctor_reference): Add argument. Store subobject offset.
2675 * gimple-fold.h (fold_ctor_reference): Add argument.
2677 2018-07-09 Paul Koning <ni1d@arrl.net>
2679 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
2680 (pdp11_insn_cost): New function.
2681 (pdp11_md_asm_adjust): New function.
2682 (TARGET_INVALID_WITHIN_DOLOOP): Define.
2683 (pdp11_rtx_costs): Update to match machine better.
2684 (output_addr_const_pdp11): Correct format mismatch warnings.
2685 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
2686 * config/pdp11/pdp11.md: General change to add base_cost and/or
2687 length attributes for use by new pdp11_insn_cost function.
2688 (MIN_BRANCH): Correct definition.
2690 (doloop_end): Use standard pattern name for looping pattern.
2691 (doloop_end_nocc): New.
2692 (movsf): Add another constraint alternative.
2693 (zero_extendqihi2): Add constraint alternatives for not in place
2695 (zero_extendhisi2): Remove.
2696 (shift patterns): Add CC handling variants.
2700 (define_peephole2): New peephole to recognize mov that sets CC for
2703 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2705 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
2706 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
2709 2018-07-09 Jakub Jelinek <jakub@redhat.com>
2712 * real.c (real_nextafter): Return true if result is denormal.
2714 2018-07-09 Martin Liska <mliska@suse.cz>
2716 * common.opt: Add back wrongly removed attribute.
2718 2018-07-09 Richard Biener <rguenther@suse.de>
2721 * dwarf2out.c (gen_block_die): For an early generated DIE
2722 always output high/low PC attributes.
2724 2018-07-09 Tom de Vries <tdevries@suse.de>
2726 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
2727 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
2728 onto VAR_DECL with abstract origin.
2730 2018-07-07 Jim Wilson <jimw@sifive.com>
2732 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
2734 2018-07-07 Tom de Vries <tdevries@suse.de>
2736 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
2738 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
2740 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
2742 (combine_bound): Use wide-int overflow calculation instead of
2744 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
2746 * fold-const.c (int_const_binop_2): Same.
2747 (extract_muldiv_1): Same.
2748 (fold_div_compare): Same.
2749 (fold_abs_const): Same.
2751 * poly-int.h (add): Same.
2755 * predict.c (predict_iv_comparison): Same.
2756 * profile-count.c (slow_safe_scale_64bit): Same.
2757 * simplify-rtx.c (simplify_const_binary_operation): Same.
2758 * tree-chrec.c (tree_fold_binomial): Same.
2759 * tree-data-ref.c (split_constant_offset_1): Same.
2760 * tree-if-conv.c (idx_within_array_bound): Same.
2761 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
2762 * tree-ssa-phiopt.c (minmax_replacement): Same.
2763 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
2764 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
2765 * vr-values.c (vr_values::adjust_range_with_scev): Same.
2766 * wide-int.cc (wi::add_large): Same.
2767 (wi::mul_internal): Same.
2768 (wi::sub_large): Same.
2769 (wi::divmod_internal): Same.
2770 * wide-int.h: Change overflow type to overflow_type for neg, add,
2771 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
2772 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
2773 mul_internal, divmod_internal.
2774 (overflow_type): New enum.
2775 (accumulate_overflow): New.
2777 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
2779 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
2780 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
2782 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
2784 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
2785 argument is checked for zero before entering loop, avoid checking again.
2787 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
2789 * gimplify.h (generic_expr_could_trap_p): Set as global function.
2790 * gimplify.h (generic_expr_could_trap_p): Likwise.
2791 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
2793 2018-07-06 Jakub Jelinek <jakub@redhat.com>
2795 PR tree-optimization/86401
2796 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
2797 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
2798 (fold_bit_and_mask): ... here. New helper function for match.pd.
2799 * fold-const.h (fold_bit_and_mask): Declare.
2800 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
2802 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
2805 * target.def (translate_mode_attribute): New hook.
2806 * targhooks.h (default_translate_mode_attribute): Declare.
2807 * targhooks.c (default_translate_mode_attribute): New function.
2808 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
2809 * doc/tm.texi: Regenerate.
2810 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
2811 (rs6000_translate_mode_attribute): New function.
2813 2018-07-06 Paul Koning <ni1d@arrl.net>
2815 * doc/md.texi (define_split): Document DONE and FAIL.
2816 (define_peephole2): Ditto.
2818 2018-07-05 Jeff Law <law@redhat.com>
2820 PR tree-optimization/86010
2821 * tree-ssa-dse.c (compute_trims): More aggressively trim at
2822 both the head and tail of mem* and str* calls.
2824 2018-07-05 Jim Wilson <jimw@sifive.com>
2826 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
2828 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
2830 * config/aarch64/aarch64-simd.md: correct flags text for
2831 MIN_EXPR replacement.
2833 2018-07-05 James Clarke <jrtc27@jrtc27.com>
2835 * configure: Regenerated.
2837 2018-07-05 Carl Love <cel@us.ibm.com>
2839 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
2840 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
2841 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
2842 VSX_BUILTIN_DOUBLEL_V4SF.
2844 2018-07-05 Martin Sebor <msebor@redhat.com>
2847 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
2848 than its domain to compute its the upper bound of a char array.
2850 2018-07-05 Nathan Sidwell <nathan@acm.org>
2852 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
2853 * doc/cpp.texi: Update comment.
2854 * doc/tm.texi: Rebuilt.
2855 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
2856 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
2857 * doc/extend.texi (Backwards Compatibility): Clarify it is system
2858 headers affected by extern "C".
2859 * system.h: Poison NO_IMPLICIT_EXTERN_C.
2860 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
2861 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
2862 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
2863 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
2864 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
2865 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
2866 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
2867 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
2868 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
2869 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
2870 config/sparc/sp64-elf.h, config/spu/spu.h,
2871 config/stormy16/stormy16.h, config/v850/v850.h,
2872 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
2873 define NO_IMPLICIT_EXTERN_C.
2874 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
2876 2018-07-05 Tamar Christina <tamar.christina@arm.com>
2879 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
2880 instead of GET_MODE_SIZE when comparing Units.
2882 2018-07-05 Tamar Christina <tamar.christina@arm.com>
2885 * rtlanal.c (set_noop_p): Constrain on mode change,
2886 include hard-reg-set.h
2888 2018-07-05 Tamar Christina <tamar.christina@arm.com>
2890 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
2892 2018-07-05 Jakub Jelinek <jakub@redhat.com>
2895 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
2898 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
2900 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
2902 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
2905 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
2907 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
2909 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
2912 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
2914 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
2916 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
2917 neg_*_op* variables.
2919 2018-07-04 Martin Liska <mliska@suse.cz>
2921 * tree-switch-conversion.c: Define
2922 max_ratio_for_speed and max_ratio_for_size constants.
2924 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
2925 Martin Liska <mliska@suse.cz>
2930 * common.opt: Rename align options with 'str_' prefix.
2931 * common/config/i386/i386-common.c (set_malign_value): New
2933 (ix86_handle_option): Use it to set -falign-* options/
2934 * config/aarch64/aarch64-protos.h (struct tune_params): Change
2935 type from int to string.
2936 * config/aarch64/aarch64.c: Update default values from int
2938 * config/alpha/alpha.c (alpha_override_options_after_change):
2940 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
2941 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
2942 max skip conditionally.
2943 * config/i386/freebsd.h (SUBALIGN_LOG): New.
2944 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
2945 max skip conditionally.
2946 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
2947 max skip conditionally.
2948 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
2949 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
2950 max skip conditionally.
2951 * config/i386/i386.c (struct ptt): Change type from int to
2953 (ix86_default_align): Set default values.
2954 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
2955 max skip conditionally.
2956 * config/i386/iamcu.h (SUBALIGN_LOG): New.
2957 (ASM_OUTPUT_MAX_SKIP_ALIGN):
2958 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
2959 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
2960 max skip conditionally.
2961 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
2962 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
2963 * config/i386/x86-64.h (SUBALIGN_LOG): New.
2964 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
2965 max skip conditionally.
2966 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
2967 * config/ia64/ia64.c (ia64_option_override): Set default values
2968 for alignment options.
2969 * config/m68k/m68k.c: Handle new str_align_* options.
2970 * config/mips/mips.c (mips_set_compression_mode): Change
2972 (mips_option_override): Set default values for options.
2973 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
2975 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2977 * config/rx/rx.c (rx_option_override): Likewise.
2978 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
2979 (LABEL_ALIGN): Use align_labels_log.
2980 (LOOP_ALIGN): Use align_loops_align.
2981 * config/s390/s390.c (s390_asm_output_function_label): Use new
2983 * config/sh/sh.c (sh_override_options_after_change):
2984 Change type of constants.
2985 * config/spu/spu.c (spu_sched_init): Likewise.
2986 * config/sparc/sparc.c (sparc_option_override): Set default
2988 * config/visium/visium.c (visium_option_override): Likewise.
2989 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
2990 emit p2align format with last argument if it's not needed.
2991 * doc/invoke.texi: Document extended format of -falign-*.
2992 * final.c: Use align_labels alignment.
2993 * flags.h (struct target_flag_state): Change type to use
2995 (struct align_flags_tuple): New.
2996 (struct align_flags): Likewise.
2997 (align_loops_log): Redefine macro to use new types.
2998 (align_loops_max_skip): Redefine macro to use new types.
2999 (align_jumps_log): Redefine macro to use new types.
3000 (align_jumps_max_skip): Redefine macro to use new types.
3001 (align_labels_log): Redefine macro to use new types.
3002 (align_labels_max_skip): Redefine macro to use new types.
3003 (align_functions_log): Redefine macro to use new types.
3004 (align_loops): Redefine macro to use new types.
3005 (align_jumps): Redefine macro to use new types.
3006 (align_labels): Redefine macro to use new types.
3007 (align_functions): Redefine macro to use new types.
3008 (align_functions_max_skip): Redefine macro to use new types.
3009 (align_loops_value): New macro.
3010 (align_jumps_value): New macro.
3011 (align_labels_value): New macro.
3012 (align_functions_value): New macro.
3013 * function.c (invoke_set_current_function_hook): Propagate
3014 alignment values from flags to global variables default in
3016 * ipa-icf.c (sem_function::equals_wpa): Use
3017 cl_optimization_option_eq instead of memcmp.
3018 * lto-streamer.h (cl_optimization_stream_out): Support streaming
3020 (cl_optimization_stream_in): Likewise.
3021 * optc-save-gen.awk: Support strings in cl_optimization.
3022 * opth-gen.awk: Likewise.
3023 * opts.c (finish_options): Remove error checking of invalid
3025 (MAX_CODE_ALIGN): Remove.
3026 (MAX_CODE_ALIGN_VALUE): Likewise.
3027 (parse_and_check_align_values): New function.
3028 (check_alignment_argument): Likewise.
3029 (common_handle_option): Use check_alignment_argument.
3030 * opts.h (parse_and_check_align_values): Declare.
3031 * toplev.c (init_alignments): Remove.
3032 (read_log_maxskip): New.
3033 (parse_N_M): Likewise.
3034 (parse_alignment_opts): Likewise.
3035 (backend_init_target): Remove usage of init_alignments.
3036 * toplev.h (parse_alignment_opts): Declare.
3037 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
3039 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
3040 * tree.c (cl_option_hasher::equal): New.
3041 * varasm.c: Use new global macros.
3043 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
3045 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
3046 Use a simpler align directive also if MAXSKIP = ALIGN-1.
3047 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
3048 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
3049 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
3050 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
3051 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
3052 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
3053 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
3054 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
3055 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
3056 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
3057 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
3059 2018-07-04 Martin Liska <mliska@suse.cz>
3060 Jonathan Wakely <jwakely@redhat.com>
3062 * coverage.c: Use correct type.
3063 * doc/invoke.texi: Language correction.
3065 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
3068 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
3069 ENDBRANCH for non-tail call which may return via indirect branch.
3070 * doc/extend.texi: Document indirect_return attribute.
3072 2018-07-03 Martin Sebor <msebor@redhat.com>
3074 PR tree-optimization/86274
3075 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
3077 (format_floating): Correct handling of infinities and NaNs.
3079 2018-07-03 Martin Sebor <msebor@redhat.com>
3081 * print-tree.c (print_real_cst): New function.
3082 (print_node_brief): Call it.
3083 (print_node): Ditto.
3085 2018-07-03 Jeff Law <law@redhat.com>
3087 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
3088 into a single pattern.
3090 * config/h8300/h8300.md (ors code_iterator): New.
3091 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
3092 a single pattern and single splitter.
3093 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
3094 (iorqi3_1, xorqi3_1): Likewise.
3095 (iorqi3, xorqi3 expanders): Similarly.
3097 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
3098 (movmd_internal) into a single pattern using the P mode iterator.
3099 (movmd splitters): Similarly.
3100 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
3101 (movsd splitters): Similarly.
3103 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
3104 ADDB, ADDW and ADDL into a single ADD attribute which selects the
3105 right table based on the size of the operand.
3106 * config/h8300/h8300.md (length_table): Corresponding changes. All
3107 references to "addb", "addw" and "addl" changed to "add".
3108 (btst patterns): Merge two variants into a single pattern.
3109 (tstqi, tsthi): Likewise.
3110 (addhi3_incdec, addsi3_incdec): Likewise.
3111 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
3112 (mulhi3, mulsi3): Likewise.
3113 (udivhi3, udivsi3): Likewise.
3114 (divhi3, divsi3): Likewise.
3115 (andorqi3, andorhi3, andorsi3): Likewise.
3117 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
3120 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
3121 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
3123 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
3125 PR tree-optimization/85694
3126 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
3127 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
3129 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
3130 (<u>avg<mode>3_ceil): New patterns.
3132 2018-07-03 David Malcolm <dmalcolm@redhat.com>
3134 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
3135 scan-tree-dump directive.
3136 * gcc.dg/vect/slp-perm-2.c: Likewise.
3137 * gcc.dg/vect/slp-perm-3.c: Likewise.
3138 * gcc.dg/vect/slp-perm-5.c: Likewise.
3139 * gcc.dg/vect/slp-perm-6.c: Likewise.
3140 * gcc.dg/vect/slp-perm-7.c: Likewise.
3141 * gcc.dg/vect/slp-perm-8.c: Likewise.
3143 2018-07-03 Marek Polacek <polacek@redhat.com>
3146 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
3148 2018-07-03 Richard Biener <rguenther@suse.de>
3151 * tree-ssa-structalias.c (find_func_clobbers): Properly
3152 handle indirect calls.
3154 2018-07-03 Jeff Law <law@redhat.com>
3156 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
3157 (shifts): New code iterator.
3158 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
3159 expander. Fix HImode handling on H8/SX.
3160 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
3161 (subqi3, subhi3, subsi3 expanders): Likewise.
3162 (andqi3, andhi3, andsi3 expanders): Likewise.
3163 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
3164 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
3165 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
3166 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
3167 (zero_extendqihi2, zero_extendqisi2): Likewise.
3168 (extendqihi2, extendqisi2): Likewise.
3169 (rotlqi3, rotlhi3, rotlsi3): Likewise.
3170 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
3171 (rotlqi3_1, rotlhi3_1): Likewise.
3172 (logicalhi3_sn, logicalsi3_sn): Likewise.
3173 (logicalhi3, logicalsi3): Likewise.
3175 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
3177 * tree-vect-patterns.c (vect_recog_rotate_pattern)
3178 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
3179 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
3180 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
3181 type to append_pattern_def_seq instead of creating a stmt_vec_info
3183 (build_mask_conversion): Likewise. Remove vinfo argument.
3184 (vect_add_conversion_to_patterm): Likewise, renaming to...
3185 (vect_add_conversion_to_pattern): ...this.
3186 (vect_recog_mask_conversion_pattern): Update call to
3187 build_mask_conversion. Pass the vector type to
3188 append_pattern_def_seq here too.
3189 (vect_recog_gather_scatter_pattern): Update call to
3190 vect_add_conversion_to_pattern.
3192 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
3194 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
3195 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
3196 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
3197 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
3198 STMT_VINFO_PATTERN_DEF_SEQ to null here.
3199 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
3200 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
3201 append_pattern_def_seq instead of new_pattern_def_seq.
3202 (vect_recog_divmod_pattern): Do both of the above.
3203 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
3206 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
3208 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
3209 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
3210 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
3211 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
3212 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
3213 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
3214 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
3215 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
3216 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
3217 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
3218 parameter with a single stmt_vec_info.
3219 (vect_recog_func_ptr): Likewise.
3220 (vect_recog_gather_scatter_pattern): Likewise, folding in...
3221 (vect_try_gather_scatter_pattern): ...this.
3222 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
3223 the stmt_vec_info of the statement to be matched. Don't clear
3224 STMT_VINFO_RELATED_STMT.
3225 (vect_pattern_recog): Update call accordingly.
3227 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
3229 PR tree-optimization/85694
3230 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
3231 (uavgM3_ceil): Document new optabs.
3232 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
3233 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
3235 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
3236 (savg_ceil_optab): New optabs.
3237 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
3238 (vect_vect_recog_func_ptrs): Add it.
3239 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
3240 constant directly from the associated lhs.
3242 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
3244 * tree-vect-patterns.c (vect_split_statement): New function.
3245 (vect_convert_input): Use it to try to split an existing cast.
3247 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
3249 * poly-int.h (print_hex): New function.
3250 * dumpfile.h (dump_dec, dump_hex): Declare.
3251 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
3252 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
3253 min_input_precision, operation_precision and operation_sign.
3254 * tree-vect-patterns.c (vect_get_range_info): New function.
3255 (vect_same_loop_or_bb_p, vect_single_imm_use)
3256 (vect_operation_fits_smaller_type): Delete.
3257 (vect_look_through_possible_promotion): Add an optional
3258 single_use_p parameter.
3259 (vect_recog_over_widening_pattern): Rewrite to use new
3260 stmt_vec_info infomration. Handle one operation at a time.
3261 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
3262 (vect_truncatable_operation_p, vect_set_operation_type)
3263 (vect_set_min_input_precision): New functions.
3264 (vect_determine_min_output_precision_1): Likewise.
3265 (vect_determine_min_output_precision): Likewise.
3266 (vect_determine_precisions_from_range): Likewise.
3267 (vect_determine_precisions_from_users): Likewise.
3268 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
3269 (vect_vect_recog_func_ptrs): Put over_widening first.
3271 (vect_pattern_recog): Call vect_determine_precisions.
3273 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
3275 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
3276 statements that have been replaced by further pattern statements.
3277 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
3279 2018-07-03 Richard Biener <rguenther@suse.de>
3281 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
3282 always set *dt. Dump vectype in vectype overload.
3283 * dumpfile.h (dump_gimple_expr): New function.
3284 (dump_gimple_expr_loc): Likewise.
3285 * dumpfile.c (dump_gimple_expr): New function.
3286 (dump_gimple_expr_loc): Likewise.
3288 2018-07-02 Jeff Law <law@redhat.com>
3290 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
3291 the H8/300, H8/300H and H8/S variants into a single pattern.
3292 (movhi_h8300, movqi_h8300hs): Similarly.
3293 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
3294 (QHI mode iterator): New.
3296 * config/h8300/h8300.md: Remove trailing whitespace.
3298 2018-07-02 Jim Wilson <jimw@sifive.com>
3300 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
3301 instead of emit_insn for interrupt returns.
3302 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
3303 (riscv_sret, riscv_uret): Likewise.
3305 2018-07-02 David Malcolm <dmalcolm@redhat.com>
3307 * pretty-print.c (selftest::test_pp_format): Move save and restore
3308 of quotes to class auto_fix_quotes, and add an instance.
3309 * selftest.c: Include "intl.h".
3310 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
3311 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
3312 * selftest.h (selftest::auto_fix_quotes): New class.
3314 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
3316 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
3317 (aarch64_sve_prepare_conditional_op): Remove.
3318 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
3319 Allow aarch64_simd_reg_or_zero as select operand; remove
3320 the aarch64_sve_prepare_conditional_op call.
3321 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
3322 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
3323 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
3324 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
3325 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
3326 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
3327 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
3328 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
3329 and a splitters to match all of the *_any patterns.
3330 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
3332 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
3333 (SVE_COND_FP_BINARY_REV): Remove.
3334 (sve_int_op_rev, sve_fp_op_rev): New.
3335 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
3336 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
3337 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
3338 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
3339 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
3340 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
3341 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
3342 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
3343 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
3345 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
3346 Remove match_dup 1 from the inner unspec.
3347 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
3349 * config/aarch64/aarch64.md (movprfx): New attr.
3350 (length): Default movprfx to 8.
3351 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
3352 (*madd<SVE_I>, *msub<SVE_I): Likewise.
3353 (*<su>mul<SVE_I>3_highpart): Likewise.
3354 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
3355 (*v<ASHIFT><SVE_I>3): Likewise.
3356 (*<su><MAXMIN><SVE_I>3): Likewise.
3357 (*<su><MAXMIN><SVE_F>3): Likewise.
3358 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
3359 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
3360 (*div<SVE_F>4): Likewise.
3362 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
3364 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
3367 2018-07-02 Richard Biener <rguenther@suse.de>
3369 PR tree-optimization/86363
3370 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
3371 memset argument refers to a non-variable address.
3373 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
3375 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
3376 {PLUS,MINUS}_EXPR code to...
3377 (adjust_symbolic_bound): ...here,
3378 (combine_bound): ...here,
3379 (set_value_range_with_overflow): ...and here.
3381 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
3383 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
3385 (extract_range_from_abs_expr): ...here.
3387 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
3389 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
3390 -fno-omit-frame-pointer when not optimizing.
3392 2018-07-02 Martin Liska <mliska@suse.cz>
3395 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
3396 (propagate_nothrow): Likewise.
3398 2018-07-02 Martin Liska <mliska@suse.cz>
3401 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
3403 2018-07-02 David Malcolm <dmalcolm@redhat.com>
3405 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
3406 function in r262149, changing "loc" param from source_location to
3407 const dump_location_t &.
3408 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
3409 declaration, as above.
3411 2018-07-01 Paul Koning <ni1d@arrl.net>
3413 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
3414 -munit-asm, -mgnu-asm, -mdec-asm.
3415 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
3416 (pdp11_output_labelref): New.
3417 (pdp11_output_def): New.
3418 (pdp11_output_addr_vec_elt): New.
3419 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
3420 %# and %@ format codes.
3421 (pdp11_option_override): New.
3422 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
3423 (pdp11_output_ident): New.
3424 (pdp11_asm_named_section): New.
3425 (pdp11_asm_init_sections): New.
3426 (pdp11_file_start): New.
3427 (pdp11_file_end): New.
3428 (output_ascii): Use .ascii/.asciz for -mdec-asm.
3429 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
3430 %o, like %c but octal.
3431 (pdp11_option_override): New.
3432 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
3434 (DATA_SECTION_ASM_OP): Ditto.
3435 (READONLY_DATA_SECTION_ASM_OP): New.
3436 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
3437 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
3438 (ASM_OUTPUT_LABELREF): Ditto.
3439 (ASM_OUTPUT_DEF): Ditto.
3440 (ASM_OUTPUT_EXTERNAL): New.
3441 (ASM_OUTPUT_SOURCE_FILENAME): New.
3442 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
3443 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
3444 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
3445 %# and %@ format codes.
3446 * config/pdp11/pdp11.opt (mgnu-asm): New.
3447 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
3448 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
3449 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
3451 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
3453 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
3454 dereferencing path[] beyond its length.
3456 (debug_all_paths): New.
3457 (rewire_first_differing_edge): New.
3458 (adjust_paths_after_duplication): New.
3459 (duplicate_thread_path): Call adjust_paths_after_duplication.
3461 (thread_through_all_blocks): Add new argument to
3462 duplicate_thread_path.
3464 2018-06-30 Jim Wilson <jimw@sifive.com>
3466 * config/riscv/predicates.md (p2m1_shift_operand): New.
3467 (high_mask_shift_operand): New.
3468 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
3469 pattern using p2m1_shift_operand.
3470 (lshsi3_zero_extend_3+2): New combiner pattern using
3471 high_mask_shift_operand.
3473 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
3475 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
3477 (vect_recog_rotate_pattern): ...here.
3478 (vect_convert_input): Try to insert casts of invariants in the
3480 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
3481 preheader to be empty.
3483 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
3485 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
3486 vector type. If given, install it in the new statement's
3488 (vect_element_precision): New function.
3489 (vect_unpromoted_value): New struct.
3490 (vect_unpromoted_value::vect_unpromoted_value): New function.
3491 (vect_unpromoted_value::set_op): Likewise.
3492 (vect_look_through_possible_promotion): Likewise.
3493 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
3494 (vect_widened_op_tree, vect_convert_input): Likewise.
3495 (vect_convert_inputs, vect_convert_output): Likewise.
3496 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
3497 to handle the optional cast of the multiplication result and
3498 vect_widened_op_tree to detect the widened multiplication itself.
3499 Do not require the input and output of promotion casts to have
3500 the same sign, but base the signedness of the operation on the
3501 input rather than the result. If the pattern includes two
3502 promotions, check that those promotions have the same sign.
3503 Do not restrict the MULT_EXPR handling to a double-width result;
3504 handle quadruple-width results and wider. Use vect_convert_inputs
3505 to convert the inputs to the common type.
3506 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
3507 to handle the optional cast of the ABS result. Also allow a sign
3508 change or a sign extension between the ABS and MINUS.
3509 Use vect_widened_op_tree to detect the widened subtraction and use
3510 vect_convert_inputs to convert the inputs to the common type.
3511 (vect_handle_widen_op_by_const): Delete.
3512 (vect_recog_widen_op_pattern): New function.
3513 (vect_recog_widen_mult_pattern): Use it.
3514 (vect_recog_widen_shift_pattern): Likewise.
3515 (vect_recog_widen_sum_pattern): Use
3516 vect_look_through_possible_promotion to handle the promoted
3519 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
3521 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
3522 the containing gimple_seq *.
3523 * gimple-iterator.h (gsi_for_stmt): Declare it.
3524 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
3525 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
3526 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
3527 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
3528 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
3530 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
3532 (vect_mark_pattern_stmts): ...here. Handle cases in which the
3533 statement being replaced is part of an existing pattern
3534 definition sequence, inserting the new pattern statements before
3536 (vect_pattern_recog_1): Don't return a bool. If the statement
3537 is already part of a pattern, instead apply pattern matching
3538 to the pattern definition statements. Don't clear the
3539 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
3540 (vect_pattern_recog): Don't break after the first match;
3541 continue processing the pattern definition statements instead.
3542 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
3544 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
3546 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
3547 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
3548 (vect_recog_widen_sum_pattern): Use it.
3550 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
3552 * tree-vect-loop.c (vectorizable_reduction): Assert that the
3553 phi is not a pattern statement and has not been replaced by
3554 a pattern statement.
3555 * tree-vect-patterns.c (type_conversion_p): Don't check
3556 STMT_VINFO_IN_PATTERN_P.
3557 (vect_recog_vector_vector_shift_pattern): Likewise.
3558 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
3559 the pattern statement rather than the original statement; check
3560 directly for a WIDEN_MULT_EXPR here.
3561 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
3562 vect_is_simple_use to return the pattern statement rather
3563 than the original statement; use is_pattern_stmt_p to check
3564 for such a pattern statement.
3565 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
3566 to return the pattern statement rather than the original statement;
3567 don't do the same transformation here.
3568 (vect_is_simple_use): If the defining statement has been replaced
3569 by a pattern statement, return the pattern statement instead.
3570 Remove the corresponding (local) transformation from the vectype
3573 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
3575 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
3576 end and default to null.
3577 * tree-vect-loop.c (vect_create_epilog_for_reduction)
3578 (vectorizable_reduction): Update calls accordingly, dropping the
3579 gimple ** argument if the passed-back statement isn't needed.
3580 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
3581 (vect_recog_rotate_pattern): Likewise.
3582 (vect_recog_mask_conversion_pattern): Likewise.
3583 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
3584 (vect_mask_constant_operand_p): Likewise.
3585 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
3586 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
3587 (get_group_load_store_type, get_load_store_type): Likewise.
3588 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
3589 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
3590 (vectorizable_conversion, vectorizable_assignment): Likewise.
3591 (vectorizable_shift, vectorizable_operation): Likewise.
3592 (vectorizable_store, vect_is_simple_cond): Likewise.
3593 (vectorizable_condition, vectorizable_comparison): Likewise.
3594 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
3595 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
3596 and move it to the end. Cope with null def_stmt_outs.
3598 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
3600 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
3602 2018-06-29 Jeff Law <law@redhat.com>
3604 * config/v850/v850.c (v850_legitimate_address_p): Handle large
3605 displacements for TARGET_V850E2V3 and newer.
3606 (TARGET_LRA_P): Remove. Defaults to LRA now.
3607 * config/v850/v850.md (sign23byte_load): Remove.
3608 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
3609 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
3611 2018-06-29 Martin Liska <mliska@suse.cz>
3614 * coverage.c (coverage_init): Mangle full path name.
3615 * doc/invoke.texi: Document the change.
3616 * gcov-io.c (mangle_path): New.
3617 * gcov-io.h (mangle_path): Likewise.
3618 * gcov.c (mangle_name): Use mangle_path for path mangling.
3620 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3622 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
3623 if starting source register is not even.
3625 2018-06-29 Martin Liska <mliska@suse.cz>
3627 PR tree-optimization/86263
3628 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
3629 Make edge redirection.
3631 2018-06-29 David Malcolm <dmalcolm@redhat.com>
3633 * dumpfile.c (dump_loc): Add indentation based on scope depth.
3634 (dump_scope_depth): New variable.
3635 (get_dump_scope_depth): New function.
3636 (dump_begin_scope): New function.
3637 (dump_end_scope): New function.
3638 * dumpfile.h (get_dump_scope_depth): New declaration.
3639 (dump_begin_scope): New declaration.
3640 (dump_end_scope): New declaration.
3641 (class auto_dump_scope): New class.
3642 (AUTO_DUMP_SCOPE): New macro.
3643 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
3646 2018-06-29 Richard Biener <rguenther@suse.de>
3648 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
3649 compute_all_dependences succeeds.
3650 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
3651 exceed --param loop-max-datarefs-for-datadeps.
3653 2018-06-29 Jakub Jelinek <jakub@redhat.com>
3655 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
3657 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
3660 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
3661 alternative 0 in preferred_for_speed attribute.
3663 2018-06-28 Paul Koning <ni1d@arrl.net>
3665 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
3666 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
3667 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
3668 * config/pdp11/pdp11.md: Correct "length" attribute calculation
3669 for shift insn patterns.
3671 2018-06-28 David Malcolm <dmalcolm@redhat.com>
3673 * cgraph.c (cgraph_node::get_body): Replace assignments to
3674 "dump_file" with calls to set_dump_file.
3675 * dumpfile.c (alt_dump_file): Make static, and group with...
3676 (alt_flags): ...this definition.
3677 (dumps_are_enabled): New variable.
3678 (refresh_dumps_are_enabled): New function.
3679 (set_dump_file): New function.
3680 (set_alt_dump_file): New function.
3681 (gcc::dump_manager::dump_start): Replace assignments to
3682 "dump_file" and "alt_dump_file" with calls to set_dump_file and
3684 (gcc::dump_manager::dump_finish): Likewise.
3685 * dumpfile.h (alt_dump_file): Delete decl.
3686 (dumps_are_enabled): New variable decl.
3687 (set_dump_file): New function decl.
3688 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
3690 * tree-nested.c (lower_nested_functions): Replace assignments to
3691 "dump_file" with calls to set_dump_file.
3693 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
3695 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
3696 goto_locus of each outgoing edge of each basic block.
3698 2018-06-28 Richard Biener <rguenther@suse.de>
3700 * dwarf2out.c (decl_scope_table): Remove.
3701 (push_decl_scope): Likewise.
3702 (pop_decl_scope): Likewise.
3703 (gen_type_die_for_member): Do not call push/pop_decl_scope.
3704 (gen_struct_or_union_type_die): Likewise.
3705 (gen_tagged_type_die): Likewise.
3706 (dwarf2out_init): Do not initialize decl_scope_table.
3707 (dwarf2out_c_finalize): Do not free it.
3709 2018-06-28 Richard Biener <rguenther@suse.de>
3711 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
3712 deciding whether to not re-use a DIE.
3714 2018-06-28 Richard Biener <rguenther@suse.de>
3716 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
3717 DW_AT_abstract_origin attribute.
3719 2018-06-28 Martin Liska <mliska@suse.cz>
3721 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
3722 Use newly introduced constants.
3723 * tree-switch-conversion.h (struct jump_table_cluster):
3724 Define max_ratio_for_size and max_ratio_for_speed.
3726 2018-06-28 Martin Liska <mliska@suse.cz>
3728 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
3729 Add new checking assert to catch invalid state.
3730 (jump_table_cluster::can_be_handled): Handle single case
3732 (jump_table_cluster::is_beneficial): Bail out for such case.
3733 (bit_test_cluster::find_bit_tests):
3734 Add new checking assert to catch invalid state.
3735 (bit_test_cluster::can_be_handled): Handle single case
3737 (bit_test_cluster::is_beneficial): Bail out for such case.
3738 (switch_decision_tree::analyze_switch_statement):
3741 2018-06-28 Martin Liska <mliska@suse.cz>
3743 * common.opt: Introduce -completion option.
3744 * gcc.c (driver_handle_option): Handle it.
3745 (driver::main): Print completions if completion
3747 * opt-suggestions.c (option_proposer::get_completions):
3749 (option_proposer::suggest_completion): Likewise.
3750 (option_proposer::find_param_completions): Likewise.
3751 (verify_autocompletions): Likewise.
3752 (test_completion_valid_options): Likewise.
3753 (test_completion_valid_params): Likewise.
3754 (in_completion_p): Likewise.
3755 (empty_completion_p): Likewise.
3756 (test_completion_partial_match): Likewise.
3757 (test_completion_garbage): Likewise.
3758 (opt_proposer_c_tests): Likewise.
3759 * opt-suggestions.h: Declare new functions.
3760 * opts.c (common_handle_option): Handle OPT__completion_.
3761 * selftest-run-tests.c (selftest::run_tests): Add
3762 opt_proposer_c_tests.
3763 * selftest.c (assert_str_startswith): New.
3764 * selftest.h (assert_str_startswith): Likewise.
3765 (opt_proposer_c_tests): New.
3766 (ASSERT_STR_STARTSWITH): Likewise.
3768 2018-06-28 Martin Liska <mliska@suse.cz>
3770 * Makefile.in: Add opt-suggestions.o.
3771 * gcc-main.c: Include opt-suggestions.h.
3772 * gcc.c (driver::driver): Likewise.
3773 (driver::~driver): Remove m_option_suggestions.
3774 (driver::build_option_suggestions): Moved to option_proposer.
3775 (driver::suggest_option): Likewise.
3776 (driver::handle_unrecognized_options): Use option_proposer.
3777 * gcc.h (class driver): Add new memver m_option_proposer.
3778 * opt-suggestions.c: New file.
3779 * opt-suggestions.h: New file.
3781 2018-06-28 Martin Liska <mliska@suse.cz>
3783 * vec.h (class auto_string_vec): New (moved from auto_argvec).
3784 (auto_string_vec::~auto_string_vec): Likewise.
3786 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
3788 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
3789 prevent_decl_creation_for_types fields up and add reset_location field.
3790 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
3791 statement if id->reset_location is true.
3792 (copy_edges_for_bb): Do not set goto_locus on the new edges if
3793 id->reset_location is true.
3794 (copy_phis_for_bb): Force input_location on the arguments if
3795 id->reset_location is true.
3796 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
3797 is set on the function to be inlined.
3799 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
3801 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
3803 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
3805 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
3806 registers for Pmode.
3807 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
3808 hard registers for the clobbered pseudo.
3810 2018-06-27 Paul Koning <ni1d@arrl.net>
3812 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
3813 mutually exclusive options.
3814 * config/pdp11/constraints.md (h): New constraint.
3815 (O): Update definition to match shift code generation.
3816 (D): New constraint.
3817 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
3819 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
3821 (output_jump): Change arguments.
3822 (pdp11_fixed_cc_regs): New function.
3823 (pdp11_cc_mode): Ditto.
3824 (pdp11_expand_shift): Ditto.
3825 (pdp11_assemble_shift): Ditto.
3826 (pdp11_small_shift): Ditto.
3827 (pdp11_branch_cost): Remove.
3828 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
3830 (pdp11_register_move_cost): Update for CC registers.
3831 (pdp11_rtx_costs): Add case for LSHIFTRT.
3832 (pdp11_output_jump): Add CCNZ mode conditional branches.
3833 (notice_update_cc_on_set): Remove.
3834 (pdp11_cc_mode): New function.
3835 (simple_memory_operand): Correct pre/post decrement case.
3836 (no_side_effect_operand): New function.
3837 (pdp11_regno_reg_class): Add CC_REGS class.
3838 (pdp11_fixed_cc_regs): New function.
3839 (pdp11_small_shift): New function.
3840 (pdp11_expand_shift): New function to expand shift insns.
3841 (pdp11_assemble_shift): New function to output shifts.
3842 (pdp11_branch_cost): Remove.
3843 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
3844 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
3845 (WCHAR_TYPE): Ditto.
3846 (PTRDIFF_TYPE): Ditto.
3847 (ADJUST_INSN_LENGTH): New macro.
3848 (FIXED_REGISTERS): Add CC registers.
3849 (CALL_USED_REGISTERS): Ditto.
3851 (REG_CLASS_NAMES): Ditto.
3852 (REG_CLASS_CONTENTS): Ditto.
3853 (SELECT_CC_MODE): Use new function.
3854 (TARGET_FLAGS_REGNUM): New macro.
3855 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
3856 (cc0_reg_rtx): Remove.
3857 (CC_STATUS_MDEP): Remove.
3858 (CC_STATUS_MDEFP_INIT): Remove.
3859 (CC_IN_FPU): Remove.
3860 (NOTICE_UPDATE_CC): Remove.
3861 (REGISTER_NAMES): Add CC registers.
3862 (BRANCH_COST): Change to constant 1.
3863 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
3865 * config/pdp11/pdp11.opt (mbcopy): Remove.
3866 (mbcopy-builtin): Remove.
3867 (mbranch-cheap): Remove.
3868 (mbranch-expensive): Remove.
3869 * config/pdp11/predicates.md (expand_shift_operand): Update to
3870 match shift code generation.
3871 (ccnz_operator): New predicate.
3872 * doc/invoke.texi (PDP-11 Options): Remove deleted options
3873 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
3874 Remove non-existent option -mabshi, -mno-abshi. Document mutually
3876 * doc/md.texi (PDP-11): Document new D and h constraints. Update
3877 description of O constraint.
3879 2018-06-27 Jeff Law <law@redhat.com>
3880 Austin Law <austinklaw@gmail.com>
3882 * config/v850/v850.md (addsi3_set_flags): New pattern.
3883 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
3884 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
3885 (zero_extendhisi2_v850_set_flags): Likewise.
3886 (zero_extendqisi2_v850_set_flags): Likewise.
3887 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
3888 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
3889 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
3891 * config/v850/v850-protos.h (notice_update_cc): Remove.
3892 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
3893 (v850_print_operand): Handle 'D' and "d".
3894 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
3895 Add handling of arithmetic/logical operations compared against zero.
3896 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
3897 Do not look at v850_compare_op, instead get mode from last argument.
3898 (v850_gen_compare): Remove
3899 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
3900 after reload for prologue insns.
3901 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
3903 (construct_save_jarl): Likewise.
3904 (TARGET_FLAGS_REGNUM): Define.
3905 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
3906 (NOTICE_UPDATE_CC): Remove.
3907 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
3908 than cc0. Conditionalize on reload_completed.
3909 (cmpsi_insn, setfcc_insn): Likewise.
3910 (tst1 splitter): Turn into define_and_split which sets the flags
3912 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
3913 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
3914 (cstoresf4, cstoredf4): Clobber the flags.
3915 (cmpsi, cmpsf, cmpdf): Remove expanders.
3916 (setf_insn): Remove pattern.
3917 (addsi3): Turn into define_and_split which clobbers the flags after
3918 reload and a suitable pattern (addsi3_clobber_flags) for use after
3920 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
3921 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
3922 (ashrsi3, ashrsi3_v850e2): Likewise.
3923 (bins): Clobber the flags.
3924 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
3925 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
3926 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
3927 (call_value_internal_short, call_value_internal_long): Likewise.
3928 (callt_save_interrupt, callt_return_interrupt): Likewise.
3929 (save_interrupt, return_interrupt): Likewise.
3930 (callt_save_all_interrupt, save_all_interrupt): Likewise.
3931 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
3932 (restore_all_interrupt, _restore_all_interrupt): Likewise.
3933 (All FP comparisons): Only allow after reload has completed.
3935 (divh, divhu): Tweak output template.
3936 (branch_z_normal, branch_z_invert): Remove
3937 (branch_nz_normal, branch_nz_invert): Likewise.
3938 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
3940 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
3941 * config/v850/v850.c (notice_update_cc): Remove.
3942 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
3943 (CC_NO_CARRY): Likewise.
3944 (NOTICE_UPDATE_CC): Define to nothing.
3945 * config/v850/v850.md: Remove block comment on cc0 handling
3946 Remove "cc" attribute from all patterns. Remove cc_status handling
3947 from all patterns. Minor formatting fixes.
3949 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3951 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
3952 (cortex-a76.cortex-a55): Likewise.
3953 * config/aarch64/aarch64-tune.md: Regenerate.
3954 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
3955 cortex-a76.cortex-a55.
3957 2018-06-27 Jeff Law <law@redhat.com>
3959 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
3960 (MULTILIB_DIRNAMES): Similarly.
3962 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
3964 * gimple.h (gimple_return_retbnd): Delete.
3965 (gimple_return_set_retbnd): Likewise.
3966 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
3967 gimple_return_set_retbnd.
3968 * gimple-pretty-print.c (dump_gimple_return): Remove call to
3969 gimple_return_retbnd and adjust.
3970 * tree-inline.h (struct copy_body_data): Remove retbnd field.
3971 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
3972 Explicitly return NULL in a couple more cases. Move assertion
3973 on debug statements and remove unreachable code.
3974 (reset_debug_binding): Do not test id->retbnd.
3975 (expand_call_inline): Do not set it.
3977 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
3979 * configure.ac: Add --disable-gcov option.
3980 * configure: Regenerate.
3981 * Makefile.in: Honour @enable_gcov@.
3982 * doc/install.texi: Document --disable-gcov.
3984 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3986 * config/arm/arm-cpus.in (cortex-a76): New entry.
3987 (cortex-a76.cortex-a55): Likewise.
3988 * config/arm/arm-tables.opt: Regenerate.
3989 * config/arm/arm-tune.md: Likewise.
3990 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
3991 * doc/invoke.texi (ARM Options): Document cortex-a76 and
3992 cortex-a76.cortex-a55.
3994 2018-06-27 Tamar Christina <tamar.christina@arm.com>
3997 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
3999 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
4001 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
4003 (EPILOGUE_USES): Likewise.
4005 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
4007 * tree-inline.c (remap_location): New function extracted from...
4008 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
4009 (copy_phis_for_bb): ...here. Call remap_location.
4010 (copy_cfg_body): Adjust call to copy_edges_for_bb.
4012 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
4014 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
4015 unaligned vsx for 16B memset.
4017 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
4020 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
4021 ieee128_float_type_node to long_double_type_node unless
4022 TARGET_LONG_DOUBLE_128 is set.
4024 2018-06-26 David Malcolm <dmalcolm@redhat.com>
4026 * cfgloop.c (get_loop_location): Convert return type from
4027 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
4028 by implicit construction from rtx_insn *, and using
4029 dump_user_location_t::from_function_decl for the fallback case.
4030 * cfgloop.h (get_loop_location): Convert return type from
4031 location_t to dump_user_location_t.
4032 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
4033 dump_printf_loc to pass in a dump_location_t rather than a
4034 location_t, via the gimple stmt.
4035 * coverage.c (get_coverage_counts): Update calls to
4036 dump_printf_loc to pass in dump_location_t rather than a
4038 * doc/optinfo.texi (Dump types): Convert example of
4039 dump_printf_loc from taking "locus" to taking "insn". Update
4040 description of the "_loc" calls to cover dump_location_t.
4041 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
4043 (dump_user_location_t::dump_user_location_t): New constructors,
4044 from gimple *stmt and rtx_insn *.
4045 (dump_user_location_t::from_function_decl): New function.
4046 (dump_loc): Make static.
4047 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
4048 const dump_location_t &.
4049 (dump_generic_expr_loc): Delete.
4050 (dump_printf_loc): Convert param "loc" from location_t to
4051 const dump_location_t &.
4052 (selftest::test_impl_location): New function.
4053 (selftest::dumpfile_c_tests): New function.
4054 * dumpfile.h: Include "profile-count.h".
4055 (class dump_user_location_t): New class.
4056 (struct dump_impl_location_t): New struct.
4057 (class dump_location_t): New class.
4058 (dump_printf_loc): Convert 2nd param from source_location to
4059 const dump_location_t &.
4060 (dump_generic_expr_loc): Delete.
4061 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
4062 const dump_location_t &.
4063 * gimple-fold.c (fold_gimple_assign): Update call to
4064 dump_printf_loc to pass in a dump_location_t rather than a
4065 location_t, via the gimple stmt.
4066 (gimple_fold_call): Likewise.
4067 * gimple-loop-interchange.cc
4068 (loop_cand::analyze_iloop_reduction_var): Update for change to
4069 check_reduction_path.
4070 (tree_loop_interchange::interchange): Update for change to
4072 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
4073 change in return-type of find_loop_location.
4074 (graphite_regenerate_ast_isl): Likewise.
4075 * graphite-optimize-isl.c (optimize_isl): Likewise.
4076 * graphite.c (graphite_transform_loops): Likewise.
4077 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
4078 pass in a dump_location_t rather than a location_t, via the
4080 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
4081 * ipa.c (walk_polymorphic_call_targets): Likewise.
4082 * loop-unroll.c (report_unroll): Convert "locus" param from
4083 location_t to dump_location_t.
4084 (decide_unrolling): Update for change to get_loop_location's
4086 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
4087 location_t to dump_user_location_t.
4088 (grid_find_single_omp_among_assignments_1): Updates calls to
4089 dump_printf_loc to pass in a dump_location_t rather than a
4090 location_t, via the gimple stmt.
4091 (grid_parallel_clauses_gridifiable): Convert "tloc" from
4092 location_t to dump_location_t. Updates calls to dump_printf_loc
4093 to pass in a dump_location_t rather than a location_t, via the
4095 (grid_inner_loop_gridifiable_p): Likewise.
4096 (grid_dist_follows_simple_pattern): Likewise.
4097 (grid_gfor_follows_tiling_pattern): Likewise.
4098 (grid_target_follows_gridifiable_pattern): Likewise.
4099 (grid_attempt_target_gridification): Convert initialization
4100 of local "grid" from memset to zero-initialization; FIXME: does
4101 this require C++11? Update call to dump_printf_loc to pass in a
4102 optinfo_location rather than a location_t, via the gimple stmt.
4103 * profile.c (read_profile_edge_counts): Updates call to
4104 dump_printf_loc to pass in a dump_location_t rather than a
4106 (compute_branch_probabilities): Likewise.
4107 * selftest-run-tests.c (selftest::run_tests): Call
4109 * selftest.h (dumpfile_c_tests): New decl.
4110 * tree-loop-distribution.c (pass_loop_distribution::execute):
4111 Update for change in return type of find_loop_location.
4112 * tree-parloops.c (parallelize_loops): Likewise.
4113 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
4114 "locus" from location_t to dump_user_location_t.
4115 (canonicalize_loop_induction_variables): Likewise.
4116 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
4117 for change in return type of find_loop_location.
4118 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
4119 to dump_printf_loc to pass in a dump_location_t rather than a
4120 location_t, via the stmt.
4121 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
4123 * tree-vect-loop-manip.c (find_loop_location): Convert return
4124 type from source_location to dump_user_location_t.
4125 (vect_do_peeling): Update for above change.
4126 (vect_loop_versioning): Update for change in type of
4128 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
4129 from location_t to dump_user_location_t.
4130 (vect_estimate_min_profitable_iters): Update for change in type
4132 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
4133 location_t to dump_location_t.
4134 (vect_slp_bb): Update for change in type of vect_location.
4135 * tree-vectorizer.c (vect_location): Convert from source_location
4136 to dump_user_location_t.
4137 (try_vectorize_loop_1): Update for change in vect_location's type.
4138 (vectorize_loops): Likewise.
4139 (increase_alignment): Likewise.
4140 * tree-vectorizer.h (vect_location): Convert from source_location
4141 to dump_user_location_t.
4142 (find_loop_location): Convert return type from source_location to
4143 dump_user_location_t.
4144 (check_reduction_path): Convert 1st param from location_t to
4145 dump_user_location_t.
4146 * value-prof.c (check_counter): Update call to dump_printf_loc to
4147 pass in a dump_user_location_t rather than a location_t; update
4148 call to error_at for change in type of "locus".
4149 (check_ic_target): Update call to dump_printf_loc to
4150 pass in a dump_user_location_t rather than a location_t, via the
4153 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
4155 * config/s390/s390.h (enum processor_flags): Do not use
4156 default tune parameter when -march was specified.
4158 2018-06-26 Jakub Jelinek <jakub@redhat.com>
4161 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
4162 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
4165 2018-06-26 Richard Biener <rguenther@suse.de>
4167 PR tree-optimization/86287
4169 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
4170 (vect_analyze_loop): Initialize n_stmts.
4172 2018-06-26 Richard Biener <rguenther@suse.de>
4175 * fold-const.c (fold_convertible_p): Pointer extension
4178 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
4181 * dwarf2out.c (loc_list_has_views): Adjust comments.
4182 (dw_loc_list): Split single cross-partition range with
4185 2018-06-25 Jeff Law <law@redhat.com>
4187 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
4188 on -mbig-switch by default.
4190 * config/v850/predicates.md (const_float_1_operand): Fix match_code
4192 (const_float_0_operand): Remove unused predicate.
4193 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
4194 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
4195 (recipsf2): New expander. Original pattern now called
4197 (recipdf2, recipdf2_insn): Similarly.
4198 (rsqrtsf2, rsqrtsf2_insn): Similarly
4199 (rsqrtdf2, rsqrtdf2_insn): Similarly
4201 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
4203 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
4204 Simplify logic for FreeBSD (twice).
4206 2018-06-25 Martin Sebor <msebor@redhat.com>
4208 PR tree-optimization/86204
4209 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
4210 a strnlen result if it's less than the length of the string.
4212 2018-06-25 Martin Sebor <msebor@redhat.com>
4214 PR tree-optimization/85700
4215 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
4216 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
4217 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
4219 2018-06-25 Martin Sebor <msebor@redhat.com>
4221 * doc/extend.texi (Zero-length arrays): Update and clarify.
4223 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
4225 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
4226 added IEEE/IBM long double multilib support on PowerPC little
4227 endian Linux systems.
4228 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
4229 (MULTILIB_DEFAULTS): Likewise.
4230 * config/rs6000/rs6000.c (rs6000_option_override_internal):
4232 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
4233 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
4234 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
4236 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
4239 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
4240 (REORDER_45): Likewise.
4242 2018-06-25 Jeff Law <law@redhat.com>
4244 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
4245 dividend to 32 bits. Adjust length.
4246 (udivmodhi4): Cleanup output template. Fix length.
4248 2018-06-25 Carl Love <cel@us.ibm.com>
4250 * config/rs6000/vsx.md: Change word selector to prefered location.
4252 2018-06-25 Richard Biener <rguenther@suse.de>
4254 PR tree-optimization/86304
4255 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
4256 epilogue-if-converted loops as well.
4258 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
4260 * lto-section-out.c (lto_begin_section): Do not print section
4261 name for noaddr and unnumbered dumps.
4263 2018-06-25 Richard Biener <rguenther@suse.de>
4265 * tree-vectorizer.h (struct vec_info_shared): New structure
4266 with parts split out from struct vec_info and loop_nest from
4267 struct _loop_vec_info.
4268 (struct vec_info): Adjust accordingly.
4269 (struct _loop_vec_info): Likewise.
4270 (LOOP_VINFO_LOOP_NEST): Adjust.
4271 (LOOP_VINFO_DATAREFS): Likewise.
4272 (LOOP_VINFO_DDRS): Likewise.
4273 (struct _bb_vec_info): Likewise.
4274 (BB_VINFO_DATAREFS): Likewise.
4275 (BB_VINFO_DDRS): Likewise.
4276 (struct _stmt_vec_info): Add dr_aux member.
4277 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
4278 (DR_MISALIGNMENT_UNINITIALIZED): New.
4279 (set_dr_misalignment): Adjust.
4280 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
4281 (vect_analyze_loop): Adjust prototype.
4282 (vect_analyze_loop_form): Likewise.
4283 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
4284 Compute dependences lazily.
4285 (vect_record_base_alignments): Use shared datarefs/ddrs.
4286 (vect_verify_datarefs_alignment): Likewise.
4287 (vect_analyze_data_refs_alignment): Likewise.
4288 (vect_analyze_data_ref_accesses): Likewise.
4289 (vect_analyze_data_refs): Likewise.
4290 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
4291 constructor parameter for shared part.
4292 (vect_analyze_loop_form): Pass in shared part and adjust.
4293 (vect_analyze_loop_2): Pass in storage for the number of
4294 stmts. Move loop nest finding to the caller. Compute
4296 (vect_analyze_loop): Pass in shared part.
4297 (vect_transform_loop): Verify shared datarefs are unchanged.
4298 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
4299 constructor parameter for shared part.
4300 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
4301 (vect_slp_bb): Verify shared datarefs are unchanged before
4303 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
4305 (new_stmt_vec_info): Initialize DR_AUX misalignment to
4306 DR_MISALIGNMENT_UNINITIALIZED.
4307 * tree-vectorizer.c (vec_info::vec_info): Add constructor
4308 parameter for shared part.
4309 (vec_info::~vec_info): Adjust.
4310 (vec_info_shared::vec_info_shared): New.
4311 (vec_info_shared::~vec_info_shared): Likewise.
4312 (vec_info_shared::save_datarefs): Likewise.
4313 (vec_info_shared::check_datarefs): Likewise.
4314 (try_vectorize_loop_1): Construct shared part live for analyses
4315 of a single loop for multiple vector sizes.
4316 * tree-parloops.c (gather_scalar_reductions): Adjust.
4318 2018-06-25 Richard Biener <rguenther@suse.de>
4320 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
4321 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
4322 (vect_analyze_data_refs): Remove similar code from here and
4323 simplify accordingly.
4325 2018-06-25 Richard Biener <rguenther@suse.de>
4327 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
4328 for reverse storage order accesses rather than asserting
4329 they cannot happen here.
4331 2018-06-25 Tom de Vries <tdevries@suse.de>
4334 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
4335 Use data16 instead of .byte for insn prefix.
4337 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
4340 * parser.c (make_char_string_pack): Pass this literal chars
4341 through cpp_interpret_string.
4342 (cp_parser_userdef_numeric_literal): Check the result of
4343 make_char_string_pack.
4345 2018-06-24 Maya Rashish <coypu@sdf.org>
4347 * ginclude/stddef.h: Simplify conditions around avoiding
4348 re-definition of __size_t.
4350 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
4352 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
4353 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
4355 2018-06-22 Maya Rashish <coypu@sdf.org>
4357 * doc/invoke.texi (mno-fancy-math-387): Update for changes
4358 made to OpenBSD and NetBSD through the years.
4360 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4362 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
4363 behavior of vec_pack (vector double, vector double) to match
4364 behavior of vec_float2 (vector double, vector double).
4366 2018-06-22 Olivier Hainque <hainque@adacore.com>
4368 * gimplify.c (gimplify_function_tree): Prevent creation
4369 of a trampoline for the address of the current function
4370 passed to entry/exit instrumentation hooks.
4372 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
4375 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
4378 2018-06-22 Martin Liska <mliska@suse.cz>
4380 PR tree-optimization/86263
4381 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
4382 Bail out if is_enabled is false.
4383 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
4385 (jump_table_cluster::is_enabled): New function.
4387 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
4389 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
4390 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
4391 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
4392 (lto_input_ts_binfo_tree_pointers): Likewise.
4393 * tree-streamer-out.c (streamer_write_tree_bitfields,
4394 write_ts_binfo_tree_pointers): Likewise.
4395 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
4397 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
4399 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
4401 2018-06-22 Martin Liska <mliska@suse.cz>
4403 * symbol-summary.h (get): Make it pure and inline move
4404 functionality from ::get function.
4405 (get): Remove and inline into ::get and ::get_create.
4406 (get_create): Move code from ::get function.
4408 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4411 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
4412 -x assembler-with-cpp.
4414 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4416 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
4417 _FILE_OFFSET_BITS=64 for C++.
4419 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
4421 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
4422 conversion insn that shows up when pr85657-3.c is compiled using
4423 IEEE 128-bit long double.
4424 (neg<mode>2_internal): Use the correct mode to check whether the
4425 mode is IBM extended.
4426 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
4427 multiply and divide external functions from being created more
4430 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
4432 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
4434 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
4435 the edge can be forwarded.
4436 (cfg_layout_merge_blocks): Likewise.
4438 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
4440 * except.c (finish_eh_generation): Commit edge insertions only after
4441 the EH edges have been redirected from post-landing to landing pads.
4443 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
4445 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
4446 create_tmp_var_for to create the FRAME decl.
4447 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
4449 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
4451 * tree-inline.c (copy_edges_for_bb): Minor tweak.
4452 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
4453 debug statement when resetting its value.
4454 (expand_call_inline): Copy the locus of the call onto the assignment
4455 of the return value, if any. Use local variable in more cases.
4457 2018-06-21 Martin Liska <mliska@suse.cz>
4459 * ipa-pure-const.c (propagate_nothrow): Use
4460 funct_state_summaries->get.
4461 (dump_malloc_lattice): Likewise.
4462 (propagate_malloc): Likewise.
4464 2018-06-21 Richard Biener <rguenther@suse.de>
4466 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
4467 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
4468 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
4470 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
4471 BLOCK_ABSTRACT_ORIGIN unconditionally.
4473 2018-06-21 David Malcolm <dmalcolm@redhat.com>
4475 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
4477 * ipa-reference.c (ipa_reference_c_finalize): Delete
4478 ipa_ref_opt_sum_summaries and set it to NULL.
4480 2018-06-21 Tom de Vries <tdevries@suse.de>
4482 PR tree-optimization/85859
4483 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
4484 test with comment from bb_no_side_effects_p.
4486 2018-06-21 Richard Biener <rguenther@suse.de>
4488 PR tree-optimization/86232
4489 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
4490 max for constant niter.
4492 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
4494 * config/aarch64/aarch64-simd.md
4495 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
4497 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
4499 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
4500 Make opernads of the unspec commutative.
4502 2018-06-21 Richard Biener <rguenther@suse.de>
4504 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
4505 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
4506 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
4507 (vect_analyze_data_ref_dependence): Re-order checks to deal with
4509 (vect_record_base_alignments): Do not record base alignment
4510 for gathers or scatters.
4511 (vect_compute_data_ref_alignment): Drop return value that is always
4512 true. Bail out early for gathers or scatters.
4513 (vect_enhance_data_refs_alignment): Bail out early for gathers
4515 (vect_find_same_alignment_drs): Likewise.
4516 (vect_analyze_data_refs_alignment): Remove dead code.
4517 (vect_slp_analyze_and_verify_node_alignment): Likewise.
4518 (vect_analyze_data_refs): For possible gathers or scatters do
4519 not create an alternate DR, just check their possible validity
4520 and mark them. Adjust DECL_NONALIASED handling to not rely
4522 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
4523 update inits of gathers or scatters.
4524 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
4525 Also copy gather/scatter flag to pattern vinfo.
4527 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
4529 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
4530 behavior of vec_packsu (vector unsigned long long, vector unsigned
4531 long long) to match behavior of vec_packs with same signature.
4533 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
4534 Thomas Schwinge <thomas@codesourcery.com>
4535 Cesar Philippidis <cesar@codesourcery.com>
4537 * gimplify.c (gimplify_scan_omp_clauses): Add support for
4538 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
4539 (gimplify_adjust_omp_clauses): Likewise.
4540 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
4541 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
4542 (gimplify_omp_target_update): Update handling of acc update and
4544 * omp-low.c (install_var_field): Remove unused parameter
4545 base_pointers_restrict.
4546 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
4547 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
4549 (omp_target_base_pointers_restrict_p): Delete.
4550 (scan_omp_target): Update call to scan_sharing_clauses.
4551 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
4553 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
4554 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
4555 (convert_local_omp_clauses): Likewise.
4556 * tree-pretty-print.c (dump_omp_clause): Likewise.
4557 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
4559 (omp_clause_code_name): Likewise.
4561 2018-06-20 Jakub Jelinek <jakub@redhat.com>
4564 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
4567 PR tree-optimization/86231
4568 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
4569 anti-range don't overwrite *vr0min before using it to compute *vr0max.
4571 2018-06-20 Tom de Vries <tdevries@suse.de>
4573 PR tree-optimization/86097
4574 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
4575 iv type if signedness of iv type is not the same as that of *nit.
4577 2018-06-20 Jakub Jelinek <jakub@redhat.com>
4579 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
4580 EDGE_EH edges, verify they are all EDGE_EH.
4582 2018-06-20 Maya Rashish <coypu@sdf.org>
4584 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
4586 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4588 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
4589 * config/aarch64/aarch64.c (xgene1_tunings): Add
4590 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
4591 (aarch64_mode_valid_for_sched_fusion_p):
4592 Allow 16-byte modes.
4593 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
4594 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
4596 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
4598 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
4599 * config/aarch64/iterators.md (VQ2): New mode iterator.
4601 2018-06-20 Martin Liska <mliska@suse.cz>
4603 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
4604 Change default ratio from 10 to 8.
4606 2018-06-20 Martin Liska <mliska@suse.cz>
4608 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
4610 (bit_test_cluster::find_bit_tests): Likewise.
4611 (switch_decision_tree::analyze_switch_statement): Find clusters.
4612 * tree-switch-conversion.h (struct jump_table_cluster): Document
4615 2018-06-20 Martin Liska <mliska@suse.cz>
4617 * tree-switch-conversion.c (switch_conversion::collect):
4618 Record m_uniq property.
4619 (switch_conversion::expand): Bail out for special conditions.
4620 (group_cluster::~group_cluster): New.
4621 (group_cluster::group_cluster): Likewise.
4622 (group_cluster::dump): Likewise.
4623 (jump_table_cluster::emit): New.
4624 (switch_decision_tree::fix_phi_operands_for_edges): New.
4625 (struct case_node): Remove struct.
4626 (jump_table_cluster::can_be_handled): New.
4627 (case_values_threshold): Moved to header.
4628 (reset_out_edges_aux): Likewise.
4629 (jump_table_cluster::is_beneficial): New.
4630 (bit_test_cluster::can_be_handled): Likewise.
4631 (add_case_node): Remove.
4632 (bit_test_cluster::is_beneficial): New.
4633 (case_bit_test::cmp): New.
4634 (bit_test_cluster::emit): New.
4635 (expand_switch_as_decision_tree_p): Remove.
4636 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
4637 (fix_phi_operands_for_edge): Likewise.
4638 (switch_decision_tree::analyze_switch_statement): New.
4639 (compute_cases_per_edge): Move ...
4640 (switch_decision_tree::compute_cases_per_edge): ... here.
4641 (try_switch_expansion): Likewise.
4642 (switch_decision_tree::try_switch_expansion): Likewise.
4643 (record_phi_operand_mapping): Likewise.
4644 (switch_decision_tree::record_phi_operand_mapping): Likewise.
4645 (emit_case_decision_tree): Likewise.
4646 (switch_decision_tree::emit): Likewise.
4647 (balance_case_nodes): Likewise.
4648 (switch_decision_tree::balance_case_nodes): Likewise.
4649 (dump_case_nodes): Likewise.
4650 (switch_decision_tree::dump_case_nodes): Likewise.
4651 (emit_jump): Likewise.
4652 (switch_decision_tree::emit_jump): Likewise.
4653 (emit_cmp_and_jump_insns): Likewise.
4654 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
4655 (emit_case_nodes): Likewise.
4656 (switch_decision_tree::emit_case_nodes): Likewise.
4657 (conditional_probability): Remove.
4658 * tree-switch-conversion.h (enum cluster_type): New.
4660 (struct cluster): Likewise.
4661 (cluster::cluster): Likewise.
4662 (struct simple_cluster): Likewise.
4663 (simple_cluster::simple_cluster): Likewise.
4664 (struct group_cluster): Likewise.
4665 (struct jump_table_cluster): Likewise.
4666 (struct bit_test_cluster): Likewise.
4667 (struct min_cluster_item): Likewise.
4668 (struct case_tree_node): Likewise.
4669 (case_tree_node::case_tree_node): Likewise.
4670 (jump_table_cluster::case_values_threshold): Likewise.
4671 (struct case_bit_test): Likewise.
4672 (struct switch_decision_tree): Likewise.
4673 (struct switch_conversion): Likewise.
4674 (switch_decision_tree::reset_out_edges_aux): Likewise.
4676 2018-06-20 Martin Liska <mliska@suse.cz>
4678 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
4679 (hoist_edge_and_branch_if_true): Likewise.
4680 (expand_switch_using_bit_tests_p): Likewise.
4681 (struct case_bit_test): Likewise.
4682 (case_bit_test_cmp): Likewise.
4683 (emit_case_bit_tests): Likewise.
4684 (switch_conversion::switch_conversion): New class.
4685 (struct switch_conv_info): Remove old struct.
4686 (collect_switch_conv_info): More to ...
4687 (switch_conversion::collect): ... this.
4688 (check_range): Likewise.
4689 (switch_conversion::check_range): Likewise.
4690 (check_all_empty_except_final): Likewise.
4691 (switch_conversion::check_all_empty_except_final): Likewise.
4692 (check_final_bb): Likewise.
4693 (switch_conversion::check_final_bb): Likewise.
4694 (create_temp_arrays): Likewise.
4695 (switch_conversion::create_temp_arrays): Likewise.
4696 (free_temp_arrays): Likewise.
4697 (gather_default_values): Likewise.
4698 (switch_conversion::gather_default_values): Likewise.
4699 (build_constructors): Likewise.
4700 (switch_conversion::build_constructors): Likewise.
4701 (constructor_contains_same_values_p): Likewise.
4702 (switch_conversion::contains_same_values_p): Likewise.
4703 (array_value_type): Likewise.
4704 (switch_conversion::array_value_type): Likewise.
4705 (build_one_array): Likewise.
4706 (switch_conversion::build_one_array): Likewise.
4707 (build_arrays): Likewise.
4708 (switch_conversion::build_arrays): Likewise.
4709 (gen_def_assigns): Likewise.
4710 (switch_conversion::gen_def_assigns): Likewise.
4711 (prune_bbs): Likewise.
4712 (switch_conversion::prune_bbs): Likewise.
4713 (fix_phi_nodes): Likewise.
4714 (switch_conversion::fix_phi_nodes): Likewise.
4715 (gen_inbound_check): Likewise.
4716 (switch_conversion::gen_inbound_check): Likewise.
4717 (process_switch): Use the newly created class.
4718 (switch_conversion::expand): New.
4719 (switch_conversion::~switch_conversion): New.
4720 * tree-switch-conversion.h: New file.
4722 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4724 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
4725 tree-vect-patterns.c.
4726 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
4727 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
4728 (vect_recog_sad_pattern): Likewise.
4729 (vect_recog_widen_sum_pattern): Likewise.
4730 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
4731 (vect_recog_widen_shift_pattern): Remove the type_in argument.
4732 (vect_recog_rotate_pattern): Likewise.
4733 (vect_recog_mult_pattern): Likewise.
4734 (vect_recog_vector_vector_shift_pattern): Likewise.
4735 (vect_recog_divmod_pattern): Likewise.
4736 (vect_recog_mixed_size_cond_pattern): Likewise.
4737 (vect_recog_bool_pattern): Likewise.
4738 (vect_recog_mask_conversion_pattern): Likewise.
4739 (vect_try_gather_scatter_pattern): Likewise.
4740 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
4741 (vect_recog_over_widening_pattern): Likewise.
4742 (vect_recog_gather_scatter_pattern): Likewise.
4743 (vect_recog_func_ptr): Move from tree-vectorizer.h
4744 (vect_vect_recog_func_ptrs): Move further down the file.
4745 (vect_recog_func): Likewise. Remove the third argument.
4746 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
4747 (vect_pattern_recog_1): Expect the pattern function to do any
4748 necessary target tests. Also expect it to provide a vector type.
4749 Remove the type_in handling.
4751 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4753 * tree-vect-patterns.c (vect_pattern_detected): New function.
4754 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
4755 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
4756 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
4757 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
4758 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
4759 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
4760 (vect_recog_mask_conversion_pattern)
4761 (vect_try_gather_scatter_pattern): Likewise.
4763 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4765 * tree-vect-patterns.c (vect_get_internal_def): New function.
4766 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
4767 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
4768 (search_type_for_mask_1): Use it.
4770 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4772 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
4773 redundant WIDEN_SUM_EXPR handling.
4774 (vect_recog_sad_pattern): Likewise.
4776 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4778 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
4779 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
4780 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
4781 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
4782 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
4784 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4786 * tree-vect-stmts.c (vectorizable_call): Make sure that we
4787 use the stmt_vec_info of the original bb statement for the
4788 new zero assignment, even if the call is part of a pattern.
4790 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4792 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
4793 that the sequence is attached to the original statement rather
4794 than the pattern statement.
4795 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
4796 PATTERN_DEF_SEQ from the original statement rather than
4797 the main pattern statement.
4798 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
4799 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
4800 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
4802 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4804 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
4805 definition statements before the early exit for statements that aren't
4807 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
4809 (vect_transform_loop): ...here. Process pattern definition
4810 statements without first checking whether the main pattern
4811 statement is live or relevant.
4813 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
4815 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
4816 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
4818 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
4820 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
4821 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
4822 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
4823 (expand_block_compare): Change select_block_compare_mode call.
4824 (expand_strncmp_align_check): Use new functions, fix comment.
4825 (emit_final_str_compare_gpr): New function.
4826 (expand_strn_compare): Refactor and clean up code.
4827 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
4829 2018-06-19 Tony Reix <tony.reix@atos.com>
4830 Damien Bergamini <damien.bergamini@atos.com>
4831 David Edelsohn <dje.gcc@gmail.com>
4833 * collect2.c (static_obj): New variable.
4834 (static_libs): New variable.
4835 (is_in_list): Uncomment declaration.
4836 (main): Track AIX libraries linked statically.
4837 (is_in_list): Uncomment definition.
4838 (scan_prog_file): Don't add AIX shared libraries initializer
4839 to constructor list if linking statically.
4841 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
4843 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
4845 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
4847 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
4849 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
4852 2018-06-19 Martin Liska <mliska@suse.cz>
4854 * config/i386/i386.c (ix86_can_inline_p): Do not use
4855 ipa_fn_summaries::get_create.
4856 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
4858 (devirtualization_time_bonus): Likewise.
4859 (ipcp_propagate_stage): Likewise.
4860 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
4861 (edge_set_predicate): Likewise.
4862 (evaluate_conditions_for_known_args): Likewise.
4863 (evaluate_properties_for_edge): Likewise.
4864 (ipa_call_summary::reset): Tranform to ...
4865 (ipa_call_summary::~ipa_call_summary): ... this.
4866 (ipa_fn_summary::reset): Transform to ...
4867 (ipa_fn_summary::~ipa_fn_summary): ... this.
4868 (ipa_fn_summary_t::remove): Rename to ...
4869 (ipa_fn_summary_t::remove_callees): ... this.
4870 (ipa_fn_summary_t::duplicate): Use placement new
4871 instead of memory copy.
4872 (ipa_call_summary_t::duplicate): Likewise.
4873 (ipa_call_summary_t::remove): Remove.
4874 (dump_ipa_call_summary): Change get_create to get.
4875 (ipa_dump_fn_summary): Dump only when summary exists.
4876 (analyze_function_body): Use symbol_summary::get instead
4878 (compute_fn_summary): Likewise.
4879 (estimate_edge_devirt_benefit): Likewise.
4880 (estimate_edge_size_and_time): Likewise.
4881 (inline_update_callee_summaries): Likewise.
4882 (remap_edge_change_prob): Likewise.
4883 (remap_edge_summaries): Likewise.
4884 (ipa_merge_fn_summary_after_inlining): Likewise.
4885 (write_ipa_call_summary): Likewise.
4886 (ipa_fn_summary_write): Likewise.
4887 (ipa_free_fn_summary): Likewise.
4888 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
4889 (struct ipa_call_summary): Likewise.
4890 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
4892 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
4893 (estimate_size_after_inlining): Likewise.
4894 (estimate_growth): Likewise.
4895 (growth_likely_positive): Likewise.
4896 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
4897 (inline_call): Likewise.
4898 * ipa-inline.c (caller_growth_limits): Likewise.
4899 (can_inline_edge_p): Likewise.
4900 (can_inline_edge_by_limits_p): Likewise.
4901 (compute_uninlined_call_time): Likewise.
4902 (compute_inlined_call_time): Likewise.
4903 (want_inline_small_function_p): Likewise.
4904 (edge_badness): Likewise.
4905 (update_caller_keys): Likewise.
4906 (update_callee_keys): Likewise.
4907 (inline_small_functions): Likewise.
4908 (inline_to_all_callers_1): Likewise.
4909 (dump_overall_stats): Likewise.
4910 (early_inline_small_functions): Likewise.
4911 (early_inliner): Likewise.
4912 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
4913 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
4914 * ipa-pure-const.c (malloc_candidate_p): Likewise.
4915 * ipa-split.c (execute_split_functions): Likewise.
4916 * symbol-summary.h: Likewise.
4917 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
4919 2018-06-19 Richard Biener <rguenther@suse.de>
4921 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
4922 (vectorize_loops): ... here. Fix dbgcnt handling.
4923 (try_vectorize_loop): Wrap try_vectorize_loop_1.
4925 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
4928 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
4929 ieee128 argument takes up only one (vector) register, not two (floating
4932 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
4934 * gimplify.c (gimplify_init_constructor): Really never clear for an
4935 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
4937 2018-06-19 Richard Biener <rguenther@suse.de>
4939 PR tree-optimization/86179
4940 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
4941 after failed recognition.
4943 2018-06-18 Martin Sebor <msebor@redhat.com>
4946 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
4947 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
4948 Handle integer subtraction.
4949 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
4950 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
4952 2018-06-18 David Malcolm <dmalcolm@redhat.com>
4954 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
4955 param from rtx to rtx_insn *.
4956 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
4958 (frv_ifcvt_modify_insn): Likwise.
4959 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
4960 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
4961 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
4962 as_a <rtx_insn *> cast to local "unprotected_region" once
4963 it's been established that it's not NULL or pc_rtx.
4964 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
4965 param "sethi" from rtx to rtx_insn *.
4966 (nds32_group_float_insns): Likewise for param "insn".
4967 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
4969 (vax_output_int_subtract): Likewise.
4970 * config/vax/vax.c (vax_output_int_add): Likewise for param
4972 (vax_output_int_subtract): Likewise.
4973 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
4974 (emit_pattern_after): Likewise for param "after".
4975 (emit_insn_after): Likewise.
4976 (emit_jump_insn_after): Likewise.
4977 (emit_call_insn_after): Likewise.
4978 (emit_debug_insn_after): Likewise.
4979 (emit_pattern_before): Likewise for param "before".
4980 (emit_insn_before): Likewise.
4981 (emit_jump_insn_before): Likewise.
4982 * final.c (get_insn_template): Likewise for param "insn", removing
4984 * output.h (get_insn_template): Likewise for 2nd param.
4985 * rtl.h (emit_insn_before): Likewise.
4986 (emit_jump_insn_before): Likewise.
4987 (emit_debug_insn_before_noloc): Likewise.
4988 (emit_insn_after): Likewise.
4989 (emit_jump_insn_after): Likewise.
4990 (emit_call_insn_after): Likewise.
4991 (emit_debug_insn_after): Likewise.
4992 (set_insn_deleted): Likewise for param.
4994 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
4997 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
4998 floating point modes, so that IFmode is numerically greater than
4999 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
5000 to declare the ordering. This prevents IFmode from being
5001 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
5002 machine. Include rs6000-modes.h to share the fractional values
5003 between genmodes* and the rest of the compiler.
5007 * config/rs6000/rs6000-modes.h: New file.
5008 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
5009 meaning of rs6000_long_double_size so that 126..128 selects an
5010 appropriate 128-bit floating point type.
5011 (rs6000_option_override_internal): Likewise.
5012 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
5013 (TARGET_LONG_DOUBLE_128): Change the meaning of
5014 rs6000_long_double_size so that 126..128 selects an appropriate
5015 128-bit floating point type.
5016 (LONG_DOUBLE_TYPE_SIZE): Update comment.
5017 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
5018 source and destination to match the standard usage.
5019 (truncifkf2): Likewise.
5020 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
5021 ISA 2.07 to use an explicit clobber, instead of passing in a
5023 (copysign<mode>3_soft): Likewise.
5025 2018-06-18 David Malcolm <dmalcolm@redhat.com>
5027 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
5028 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
5029 (vect_slp_analyze_instance_dependence): Likewise.
5030 (vect_enhance_data_refs_alignment): Likewise.
5031 (vect_analyze_data_refs_alignment): Likewise.
5032 (vect_slp_analyze_and_verify_instance_alignment
5033 (vect_analyze_data_ref_accesses): Likewise.
5034 (vect_prune_runtime_alias_test_list): Likewise.
5035 (vect_analyze_data_refs): Likewise.
5036 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
5037 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
5038 (vect_analyze_scalar_cycles_1): Likewise.
5039 (vect_get_loop_niters): Likewise.
5040 (vect_analyze_loop_form_1): Likewise.
5041 (vect_update_vf_for_slp): Likewise.
5042 (vect_analyze_loop_operations): Likewise.
5043 (vect_analyze_loop): Likewise.
5044 (vectorizable_induction): Likewise.
5045 (vect_transform_loop): Likewise.
5046 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
5047 * tree-vect-slp.c (vect_analyze_slp): Likewise.
5048 (vect_make_slp_decision): Likewise.
5049 (vect_detect_hybrid_slp): Likewise.
5050 (vect_slp_analyze_operations): Likewise.
5051 (vect_slp_bb): Likewise.
5052 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
5053 (vectorizable_bswap): Likewise.
5054 (vectorizable_call): Likewise.
5055 (vectorizable_simd_clone_call): Likewise.
5056 (vectorizable_conversion): Likewise.
5057 (vectorizable_assignment): Likewise.
5058 (vectorizable_shift): Likewise.
5059 (vectorizable_operation): Likewise.
5060 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
5062 2018-06-18 Martin Sebor <msebor@redhat.com>
5064 PR tree-optimization/81384
5065 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
5066 * builtins.c (expand_builtin_strnlen): New function.
5067 (expand_builtin): Call it.
5068 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
5069 * builtins.def (BUILT_IN_STRNLEN): New.
5070 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
5071 Warn for bounds in excess of maximum object size.
5072 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
5073 single-value ranges. Handle strnlen.
5074 (handle_builtin_strlen): Handle strnlen.
5075 (strlen_check_and_optimize_stmt): Same.
5076 * doc/extend.texi (Other Builtins): Document strnlen.
5078 2018-06-18 Maya Rashish <coypu@sdf.org>
5080 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
5081 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
5082 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
5084 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
5086 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
5088 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5090 * tree.c (escaped_string::escape): Replace cast to char * by
5091 const_cast<char *> (unescaped).
5093 2018-06-18 Nick Clifton <nickc@redhat.com>
5096 * tree.c (escaped_string): New class. Converts an unescaped
5097 string into its escaped equivalent.
5098 (warn_deprecated_use): Use the new class to convert the
5099 deprecation message, if present.
5100 (test_escaped_strings): New self test.
5101 (test_c_tests): Add test_escaped_strings.
5102 * doc/extend.texi (deprecated): Add a note that the
5103 deprecation message is affected by the -fmessage-length
5104 option, and that control characters will be escaped.
5105 (#pragma GCC error): Document this pragma.
5106 (#pragma GCC warning): Likewise.
5107 * doc/invoke.texi (-fmessage-length): Document this option's
5108 effect on the #warning and #error preprocessor directives and
5109 the deprecated attribute.
5111 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
5113 * tree.c (decl_value_expr_lookup): Revert latest change.
5114 (decl_value_expr_insert): Likewise.
5116 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
5118 * gimplify.c (nonlocal_vlas): Delete.
5119 (nonlocal_vla_vars): Likewise.
5120 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
5122 (gimplify_body): Do not create and destroy nonlocal_vlas.
5123 * tree-nested.c: Include diagnostic.h.
5124 (use_pointer_in_frame): Tweak.
5125 (lookup_field_for_decl): Add assertion and declare the transformation.
5126 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
5127 internal error when the reference is in a wrong context. Do not
5128 create a debug decl by default.
5129 (note_nonlocal_block_vlas): Delete.
5130 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
5131 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
5132 create a debug decl by default.
5133 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
5134 call is in a wrong context.
5135 (fixup_vla_decls): New function.
5136 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
5137 debug variables were created.
5138 * tree.c (decl_value_expr_lookup): Add checking assertion.
5139 (decl_value_expr_insert): Likewise.
5141 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
5144 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
5145 * tree-scalar-evolution.c (interpret_expr): Likewise.
5146 (expression_expensive_p): Likewise.
5147 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
5148 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
5149 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
5150 (ssa_defined_by_minus_one_stmt_p): New.
5152 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
5155 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
5156 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
5157 * dojump.c (do_jump): Likewise.
5158 * expr.c (expand_expr_real_2): Check operand type's sign.
5159 * fold-const.c (const_unop): Handle ABSU_EXPR.
5160 (fold_abs_const): Likewise.
5161 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
5162 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
5163 (strip_sign_op_1): Likesise.
5164 * match.pd: Add new pattern to generate ABSU_EXPR.
5165 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
5166 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
5167 * tree-eh.c (operation_could_trap_helper_p): Likewise.
5168 * tree-inline.c (estimate_operator_cost): Likewise.
5169 * tree-pretty-print.c (dump_generic_node): Likewise.
5170 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
5171 * tree.def (ABSU_EXPR): New.
5173 2018-06-16 Jakub Jelinek <jakub@redhat.com>
5176 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
5177 documented as preserved for backward compatibility only.
5178 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
5180 PR rtl-optimization/86108
5181 * bb-reorder.c (create_forwarder_block): Renamed to ...
5182 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
5183 jump from new landing pad to the second part.
5184 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
5187 2018-06-15 Jakub Jelinek <jakub@redhat.com>
5190 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
5191 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
5192 Only call store_expr for halves if the mode is the same.
5195 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
5196 Fix up comment formatting.
5198 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
5200 * typed-splay-tree.h (typed_splay_tree::remove): New function.
5201 (typed_splay_tree::closure,
5202 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
5203 (typed_splay_tree::typed_splay_tree,
5204 typed_splay_tree::operator =): Declared private.
5205 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
5206 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
5207 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
5208 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
5209 typed_splay_tree::splay_tree_splay,
5210 typed_splay_tree::splay_tree_foreach_helper,
5211 typed_splay_tree::splay_tree_insert,
5212 typed_splay_tree::splay_tree_remove,
5213 typed_splay_tree::splay_tree_lookup,
5214 typed_splay_tree::splay_tree_predecessor,
5215 typed_splay_tree::splay_tree_successor,
5216 typed_splay_tree::splay_tree_min,
5217 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
5218 (typed_splay_tree::root, typed_splay_tree::comp,
5219 typed_splay_tree::delete_key,
5220 typed_splay_tree::delete_value): New data members.
5221 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
5222 typed_splay_tree::remove.
5224 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
5226 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
5227 -mginv and -mno-ginv to the assembler.
5228 * config/mips/mips.opt (-mcrc): New option.
5230 * doc/invoke.text (-mcrc): Document.
5233 2018-06-15 Nick Clifton <nickc@redhat.com>
5236 * tree.c (escaped_string): New class. Converts an unescaped
5237 string into its escaped equivalent.
5238 (warn_deprecated_use): Use the new class to convert the
5239 deprecation message, if present.
5240 (test_escaped_strings): New self test.
5241 (test_c_tests): Add test_escaped_strings.
5242 * doc/extend.texi (deprecated): Add a note that the
5243 deprecation message is affected by the -fmessage-length
5244 option, and that control characters will be escaped.
5245 (#pragma GCC error): Document this pragma.
5246 (#pragma GCC warning): Likewise.
5247 * doc/invoke.texi (-fmessage-length): Document this option's
5248 effect on the #warning and #error preprocessor directives and
5249 the deprecated attribute.
5251 2018-06-15 Richard Biener <rguenther@suse.de>
5253 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
5254 here, also noting vector size used.
5255 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
5256 size used in MSG_OPTIMIZED_LOCATIONS dump.
5257 (pass_slp_vectorize::execute): Adjust.
5259 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
5262 * config/arc/arc.c (arc_return_address_register): Fix
5265 2018-06-15 Richard Biener <rguenther@suse.de>
5268 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
5269 leave useless conversion stripping to force_gimple_operand_gsi.
5270 (gimplify_build2): Likewise.
5271 (gimplify_build1): Likewise.
5273 2018-06-15 Richard Biener <rguenther@suse.de>
5276 * tree-cfg.c (move_stmt_op): unshare invariant addresses
5277 before adjusting their block.
5279 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
5281 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
5282 multilibs for *-*-rtems*.
5283 * config/riscv/t-rtems: New file.
5285 2018-06-14 Jakub Jelinek <jakub@redhat.com>
5288 * match.pd ((A +- CST1) +- CST2): Punt if last resort
5289 unsigned_type_for returns NULL.
5292 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
5293 subregs of multi-word pseudos unless the float mode has word size.
5295 2018-06-14 Richard Biener <rguenther@suse.de>
5298 * tree-vect-generic.c (build_word_mode_vector_type): Remove
5299 duplicate and harmful type_hash_canon.
5300 * tree.c (type_hash_canon): Assert we didn't find ourselves.
5302 2018-06-14 Richard Biener <rguenther@suse.de>
5305 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
5308 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
5310 * config/rtems.h (STDINT_LONG32): Define.
5312 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
5313 Prachi Godbole <prachi.godbole@imgtec.com>
5315 * config/mips/mips-cpus.def: Define P6600.
5316 * config/mips/mips-tables.opt: Regenerate.
5317 * config/mips/mips.c (mips_ucbranch_type): New enum.
5318 (mips_rtx_cost_data): Add support for P6600.
5319 (mips_issue_rate): Likewise.
5320 (mips_multipass_dfa_lookahead): Likewise.
5321 (mips_avoid_hazard): Likewise.
5322 (mips_reorg_process_insns): Likewise.
5323 (mips_classify_branch_p6600): New function.
5324 * config/mips/mips.h (TUNE_P6600): New define.
5325 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
5326 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
5327 * config/mips/mips.md: Include p6600.md.
5328 (processor): Add p6600.
5329 * config/mips/p6600.md: New file.
5330 * doc/invoke.texi: Add p6600 to supported architectures.
5332 2018-06-13 Martin Sebor <msebor@redhat.com>
5334 PR tree-optimization/86114
5335 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
5337 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
5339 2018-06-13 Richard Biener <rguenther@suse.de>
5341 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
5342 Properly set vector type of the intermediate stmt.
5343 * tree-vect-stmts.c (vectorizable_operation): The destination
5344 var always has vectype_out type.
5346 2018-06-13 Jeff Law <law@redhat.com>
5348 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
5349 integer 0 for argument to print_rtl_with_bb.
5350 (rl78_reorg): Likewise.
5352 2018-06-13 David Malcolm <dmalcolm@redhat.com>
5354 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
5355 from rtx to rtx_insn *.
5356 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
5358 (add_sched_insns_for_speculation): Likewise for local "target",
5359 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
5360 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
5361 from rtx_insn ** to rtx_code_label **.
5362 (reorg_emit_nops): Likewise.
5363 (c6x_reorg): Likewise for local "call_labels".
5364 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
5366 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
5367 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
5368 the loops over LABEL_REFS.
5369 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
5371 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
5372 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
5373 (split_branches): Strengthen local "olabel" from rtx to
5374 rtx_insn *, adding a safe_as_a cast.
5375 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
5377 (add_insn_after): Likewise for first two params.
5378 (add_insn_before): Likewise.
5379 (remove_insn): Likewise for param.
5380 (emit_pattern_before_noloc): Likewise for second and third params.
5381 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
5382 (emit_call_insn_before_noloc): Likewise.
5383 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
5385 (emit_barrier_before): Likewise.
5386 (emit_label_before): Strengthen "label" param from "rtx" to
5387 "rtx_code_label *". Strengthen "before" param from "rtx" to
5389 (emit_insn_after_1): Strengthen "after" param from "rtx" to
5391 (emit_pattern_after_noloc): Likewise.
5392 (emit_insn_after_noloc): Likewise.
5393 (emit_jump_insn_after_noloc): Likewise.
5394 (emit_call_insn_after_noloc): Likewise.
5395 (emit_debug_insn_after_noloc): Likewise.
5396 (emit_barrier_after): Likewise.
5397 (emit_label_after): Likewise for both params.
5398 (emit_pattern_after_setloc): Likewise for "after" param. Convert
5399 "loc" param from "int" to "location_t".
5400 (emit_insn_after_setloc): Likewise.
5401 (emit_jump_insn_after_setloc): Likewise.
5402 (emit_call_insn_after_setloc): Likewise.
5403 (emit_debug_insn_after_setloc): Likewise.
5404 (emit_pattern_before_setloc): Likewise for "before" param. Convert
5405 "loc" param from "int" to "location_t".
5406 (emit_pattern_before): Convert NULL_RTX to NULL.
5407 (emit_insn_before_setloc): Convert "loc" param from "int" to
5409 (emit_jump_insn_before_setloc): Likewise.
5410 (emit_call_insn_before_setloc): Likewise.
5411 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
5412 rtx_insn *. Convert "loc" param from "int" to "location_t".
5413 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
5414 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
5415 Convert 3rd param from "int" to "location_t".
5416 (emit_barrier_before, emit_barrier_after, next_real_insn):
5417 Strengthen param from rtx to rtx_insn *.
5418 (emit_label_before): Strengthen 1st param from "rtx" to
5419 "rtx_code_label *". Strengthen 2nd param from "rtx" to
5421 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
5422 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
5423 Strengthen 2nd param from "rtx" to "rtx_insn *".
5424 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
5425 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
5426 Likewise. Convert 3rd param from "int" to "location_t".
5427 (emit_label_after): Strengthen 1st param from "rtx" to
5429 (next_real_insn, remove_insn): Strengthen param from "rtx" to
5431 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
5432 from "rtx" to "rtx_insn *".
5434 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
5436 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
5437 bodies streamed in with -Q.
5438 * dumpfile.c (dump_files): Add lto-stream-out dump file.
5439 * dumpfile.h (tree_dump_index): Add lto_stream_out.
5440 * gimple-streamer-out.c: Include gimple-pretty-print.h
5441 (output_bb): Dump stmts streamed.
5442 * lto-section-out.c: Include print-tree.h
5443 (lto_begin_section): Dump sections created.
5444 (lto_output_decl_index): Dump decl encoded.
5445 * lto-streamer-out.c: Include print-tree.h
5446 (create_output_block): Dump output block created.
5447 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
5448 (output_function): Dump function output.
5449 (output_constructor): Dump constructor streamed.
5450 (write_global_stream): Output indexes encoded.
5451 (produce_asm_for_decls): Dump streams encoded.
5452 * lto-streamer.c (streamer_dump_file): New global var.
5453 * lto-streamer.h (streamer_dump_file): Declare.
5454 * passes.c (ipa_write_summaries): Initialize streamer dump.
5455 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
5458 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
5461 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
5462 offsets for register save directives. Emit a second batch of save
5463 directives, if need be, when the function accesses prior frames.
5465 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
5467 * config/arc/fpu.md (fmasf4): Force operand to register.
5468 (fnmasf4): Likewise.
5470 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
5472 * config/arc/arc-protos.h (arc_pad_return): Remove.
5473 * config/arc/arc.c (machine_function): Remove force_short_suffix
5475 (arc_print_operand): Adjust printing of '&'.
5476 (arc_verify_short): Remove conditional printing of short suffix.
5477 (arc_final_prescan_insn): Remove reference to size_reason.
5478 (pad_return): New function.
5479 (arc_reorg): Call pad_return.
5480 (arc_pad_return): Remove.
5481 (arc_init_machine_status): Remove reference to force_short_suffix.
5482 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
5483 (attr length): When attribute iscompact is true force to 2
5484 regardless; in the case of maybe check if we want to force the
5485 instruction to have 4 bytes length.
5486 (nopv): Change it to generate 4 byte long nop as well.
5487 (blockage): New pattern.
5488 (simple_return): Remove call to arc_pad_return.
5489 (p_return_i): Likewise.
5491 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
5493 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
5495 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
5497 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
5500 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
5502 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
5503 for ARC700 and ARCv2.
5505 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
5508 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
5509 operands[2] instead of operands[1].
5512 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
5514 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
5515 case, check whether the outer register overlaps an unallocatable
5516 register, not just whether it fits the required class.
5518 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
5520 * poly-int.h (can_div_trunc_p): Add new overload in which all values
5522 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
5523 (memrefs_conflict_p): Likewise.
5524 (init_alias_analysis): Likewise.
5525 * cfgexpand.c (expand_debug_expr): Likewise.
5526 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
5527 * cse.c (fold_rtx): Likewise.
5528 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
5529 * expr.c (emit_block_move_hints): Likewise.
5530 (clear_storage_hints, push_block, emit_push_insn): Likewise.
5531 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
5532 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
5533 (emit_group_store): Likewise.
5534 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
5535 to read the PRE/POST_MODIFY increment.
5536 * calls.c (store_one_arg): Use strip_offset.
5537 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
5539 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
5541 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
5542 (simplify_binary_operation_1): Extend CONST_INT handling to
5544 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
5545 than a HOST_WIDE_INT.
5546 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
5548 (adjust_mems, add_stores): Update accodingly.
5549 (vt_canonicalize_addr): Track polynomial offsets.
5550 (emit_note_insn_var_location): Likewise.
5551 (vt_add_function_parameter): Likewise.
5552 (vt_initialize): Likewise.
5554 2018-06-12 Jeff Law <law@redhat.com>
5556 * config.gcc (alpha*-*-freebsd*): Remove.
5557 * config/alpha/freebsd.h: Remove.
5559 2018-06-12 David Malcolm <dmalcolm@redhat.com>
5562 * spellcheck-tree.c (levenshtein_distance): Rename to...
5563 (get_edit_distance): ...this, and update for underlying renaming.
5564 * spellcheck-tree.h (levenshtein_distance): Rename to...
5565 (get_edit_distance): ...this.
5566 * spellcheck.c (levenshtein_distance): Rename to...
5567 (get_edit_distance): ...this. Convert from Levenshtein distance
5568 to Damerau-Levenshtein distance by supporting transpositions of
5569 adjacent characters. Rename "v1" to "v_next" and "v0" to
5571 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
5572 (selftest::test_edit_distance_unit_test_oneway): ...this, and
5573 update for underlying renaming.
5574 (selftest::levenshtein_distance_unit_test): Rename to...
5575 (selftest::test_get_edit_distance_unit): ...this, and update for
5576 underlying renaming.
5577 (selftest::test_find_closest_string): Add example from PR 69968
5578 where transposition helps
5579 (selftest::test_metric_conditions): Update for renaming.
5580 (selftest::test_metric_conditions): Likewise.
5581 (selftest::spellcheck_c_tests): Likewise.
5582 * spellcheck.h (levenshtein_distance): Rename both overloads to...
5583 (get_edit_distance): ...this.
5584 (best_match::consider): Update for renaming.
5586 2018-06-12 Martin Sebor <msebor@redhat.com>
5588 PR tree-optimization/85259
5589 * builtins.c (compute_objsize): Handle constant offsets.
5590 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
5591 true iff a warning has been issued.
5592 * gimple.h (gimple_nonartificial_location): New function.
5593 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
5594 gimple_nonartificial_location and handle -Wno-system-headers.
5595 (handle_builtin_stxncpy): Same.
5597 2018-06-12 Martin Sebor <msebor@redhat.com>
5600 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
5602 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
5604 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
5605 BUILTIN_VEC_XST entries for pointer to double and long long.
5607 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
5610 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
5612 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
5615 2018-06-12 Martin Liska <mliska@suse.cz>
5617 * doc/options.texi: Document IntegerRange.
5619 2018-06-12 Martin Liska <mliska@suse.cz>
5621 * config/i386/i386.opt: Make MPX-related options as Deprecated.
5622 * opt-functions.awk: Handle Deprecated flag.
5623 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
5625 (read_cmdline_option): Report warning for a deprecated option.
5626 * opts.h (struct cl_option): Add new field cl_deprecated.
5627 (CL_ERR_DEPRECATED): New.
5629 2018-06-12 Martin Liska <mliska@suse.cz>
5631 * doc/options.texi: Document Deprecated option flag.
5633 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
5635 * config/arc/arc-arch.h (arc_extras): New enum.
5636 (arc_cpu_t):Add field extra.
5637 (arc_cpu_types): Consider the extras.
5638 * config/arc/arc-cpus.def: Add extras info.
5639 * config/arc/arc-opts.h (processor_type): Consider extra field.
5640 * config/arc/arc.c (arc_override_options): Handle extra field.
5642 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
5644 * config/arc/arc-arch.h: Update ARC_OPTX macro.
5645 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
5647 * config/arc/arc.c (arc_init): Update pic warning.
5648 (irq_range): Update irq range parsing warnings.
5649 (arc_override_options): Update various warning messages.
5650 (arc_handle_aux_attribute): Likewise.
5652 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
5654 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
5656 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5658 * doc/sourcebuild.texi: Document usage of line number 0 in verify
5659 compiler messages directives.
5661 2018-06-12 Matthew Fortune <mfortune@gmail.com>
5663 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
5664 * config/mips/mips-tables.opt: Regenerate.
5665 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
5667 * doc/invoke.texi: Document -march=i6500.
5669 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
5671 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
5672 (i6400_gpmul): Add cpu_unit.
5673 (i6400_gpdiv): Likewise.
5674 (i6400_msa_add_d): Update reservations.
5675 (i6400_msa_int_add) Likewise.
5676 (i6400_msa_short_logic3) Likewise.
5677 (i6400_msa_short_logic2) Likewise.
5678 (i6400_msa_short_logic) Likewise.
5679 (i6400_msa_move) Likewise.
5680 (i6400_msa_cmp) Likewise.
5681 (i6400_msa_short_float2) Likewise.
5682 (i6400_msa_div_d) Likewise.
5683 (i6400_msa_long_logic1) Likewise.
5684 (i6400_msa_long_logic2) Likewise.
5685 (i6400_msa_mult) Likewise.
5686 (i6400_msa_long_float2) Likewise.
5687 (i6400_msa_long_float4) Likewise.
5688 (i6400_msa_long_float5) Likewise.
5689 (i6400_msa_long_float8) Likewise.
5690 (i6400_fpu_fadd): Include frint type.
5691 (i6400_fpu_store): New define_insn_reservation.
5692 (i6400_fpu_load): Likewise.
5693 (i6400_fpu_move): Likewise.
5694 (i6400_fpu_fcmp): Likewise.
5695 (i6400_fpu_fmadd): Likewise.
5696 (i6400_int_mult): Include imul3nc type and update reservation.
5697 (i6400_int_div): Include idiv3 type and update reservation.
5698 (i6400_int_load): Update to check type not move_type.
5699 (i6400_int_store): Likewise.
5700 (i6400_int_prefetch): Set zero latency.
5702 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
5704 * gcc.c: Document new %@{...} sequence.
5705 (LINK_COMMAND_SPEC): Use it for the -L switches.
5706 (cpp_unique_options): Use it for the -I switches.
5707 (at_file_argbuf): New global variable.
5708 (in_at_file): Likewise.
5709 (alloc_args): Create at_file_argbuf.
5710 (clear_args): Truncate at_file_argbuf.
5711 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
5712 (open_at_file): New function.
5713 (close_at_file): Likewise.
5714 (create_at_file): Delete.
5715 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
5718 (validate_switches_from_spec): Deal with %@{...} sequence.
5719 (validate_switches): Likewise.
5720 (driver::finalize): Call clear_args.
5722 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
5724 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
5726 2018-06-11 Martin Sebor <msebor@redhat.com>
5728 * doc/invoke.texi (-Wall): List -Wc++17-compat.
5729 (Wno-class-memaccess): Add @opindex.
5730 (Wno-templates, Wno-multiple-inheritance): Same.
5731 (Wno-virtual-inheritance, Wno-namespaces): Same.
5732 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
5733 (Wno-format-overflow, Wno-format-truncation): Same.
5734 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
5735 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
5736 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
5737 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
5738 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
5739 (Wno-misspelled-isr): Same.
5741 2018-06-11 Martin Sebor <msebor@redhat.com>
5743 * PR tree-optimization/86083
5744 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
5746 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
5748 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
5750 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
5753 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
5754 on the correct operand.
5755 (*movdi_internal64): Ditto.
5757 2018-06-11 Martin Liska <mliska@suse.cz>
5759 PR tree-optimization/86089
5760 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
5762 2018-06-11 Julia Koval <julia.koval@intel.com>
5764 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
5765 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
5766 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
5768 2018-06-11 Olivier Hainque <hainque@adacore.com>
5770 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
5771 for Ada with strict dwarf2.
5773 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
5776 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
5779 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
5781 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
5783 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
5785 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
5788 2018-06-08 Martin Liska <mliska@suse.cz>
5790 * tree-cfg.h (debug_function): Fix argument type to match
5793 2018-06-08 Martin Liska <mliska@suse.cz>
5795 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
5796 Remove usage of MPX-related (and removed) fields.
5797 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
5799 2018-06-08 David Malcolm <dmalcolm@redhat.com>
5801 * cfg.c (debug): Use TDF_NONE rather than 0.
5802 * cfghooks.c (debug): Likewise.
5803 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
5804 (struct dump_option_value_info): Convert to...
5805 (struct kv_pair): ...this template type.
5806 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
5808 (optinfo_verbosity_options): Likewise.
5809 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
5811 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
5812 than int for "optgroup_flags" param.
5813 (dump_generic_expr_loc): Use dump_flags_t rather than int for
5815 (dump_dec): Likewise.
5816 (dump_finish): Use TDF_NONE rather than 0.
5817 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
5818 rather than int for "optgroup_flags" param. Use TDF_NONE rather
5819 than 0. Update for change to option_ptr.
5820 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
5821 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
5822 0. Update for changes to optinfo_verbosity_options and
5824 (opt_info_switch_p): Convert optgroup_flags from int to
5826 (dump_basic_block): Use dump_flags_t rather than int
5827 for "dump_kind" param.
5828 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
5829 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
5830 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
5831 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
5832 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
5833 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
5834 TDF_NONE): Convert from macros to...
5835 (enum dump_flag): ...this new enum.
5836 (dump_flags_t): Update to use enum.
5837 (operator|, operator&, operator~, operator|=, operator&=):
5838 Implement for dump_flags_t.
5839 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
5840 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
5841 Convert from macros to...
5842 (enum optgroup_flag): ...this new enum.
5843 (optgroup_flags_t): New typedef.
5844 (operator|, operator|=): Implement for optgroup_flags_t.
5845 (struct dump_file_info): Convert field "alt_flags" to
5846 dump_flags_t. Convert field "optgroup_flags" to
5848 (dump_basic_block): Use dump_flags_t rather than int for param.
5849 (dump_generic_expr_loc): Likewise.
5850 (dump_dec): Likewise.
5851 (dump_register): Convert param "optgroup_flags" to
5853 (opt_info_enable_passes): Likewise.
5854 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
5856 * gimple-pretty-print.c (debug): Likewise.
5857 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
5858 (merged_store_group::apply_stores): Likewise.
5859 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
5860 * gimple.c (verify_gimple_pp): Likewise.
5861 * graphite-poly.c (print_pbb_body): Likewise.
5862 * passes.c (pass_manager::register_one_dump_file): Convert
5863 local "optgroup_flags" to optgroup_flags_t.
5864 * print-tree.c (print_node): Use TDF_NONE rather than 0.
5866 (debug_body): Likewise.
5867 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
5868 to optgroup_flags_t.
5869 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
5871 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
5872 (convert_mult_to_fma): Likewise.
5873 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
5874 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
5875 * tree-vect-data-refs.c (dump_lower_bound): Convert param
5876 "dump_kind" to dump_flags_t.
5878 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
5880 * config/rs6000/rs6000.c (min, max): Delete.
5882 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
5884 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
5885 -mabi=spe and -mabi=no-spe.
5887 2018-06-08 Martin Liska <mliska@suse.cz>
5889 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
5890 where we expect an existing summary.
5892 2018-06-08 Martin Liska <mliska@suse.cz>
5894 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
5895 * ipa-inline.h (estimate_edge_growth): Likewise.
5897 2018-06-08 Martin Liska <mliska@suse.cz>
5899 * cgraph.c (function_version_hasher::hash): Use
5900 cgraph_node::get_uid ().
5901 (function_version_hasher::equal):
5902 * cgraph.h (cgraph_node::get_uid): New method.
5903 * ipa-inline.c (update_caller_keys): Use
5904 cgraph_node::get_uid ().
5905 (update_callee_keys): Likewise.
5906 * ipa-utils.c (searchc): Likewise.
5907 (ipa_reduced_postorder): Likewise.
5908 * lto-cgraph.c (input_node): Likewise.
5909 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
5910 * symbol-summary.h (symtab_insertion): Likewise.
5911 (symtab_removal): Likewise.
5912 (symtab_duplication): Likewise.
5913 * tree-pretty-print.c (dump_function_header): Likewise.
5914 * tree-sra.c (convert_callers_for_node): Likewise.
5916 2018-06-08 Martin Liska <mliska@suse.cz>
5918 * cgraph.c (symbol_table::create_edge): Always assign a new
5920 (symbol_table::free_edge): Do not recycle numbers.
5921 * cgraph.h (cgraph_edge::get): New method.
5922 * symbol-summary.h (symtab_removal): Use it.
5923 (symtab_duplication): Likewise.
5924 (call_summary::hashable_uid): Remove.
5926 2018-06-08 Martin Liska <mliska@suse.cz>
5928 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
5929 (initialize_growth_caches): Remove.
5930 (free_growth_caches): Likewise.
5931 (do_estimate_edge_time): Use edge_growth_cache.
5932 (do_estimate_edge_size): Likewise.
5933 (do_estimate_edge_hints): Likewise.
5934 * ipa-inline.c (reset_edge_caches): Likewise.
5935 (recursive_inlining): Likewise.
5936 (inline_small_functions): Likewise.
5937 * ipa-inline.h (initialize_growth_caches): Remove.
5938 (estimate_edge_size): Likewise.
5939 (estimate_edge_time): Likewise.
5940 (estimate_edge_hints): Likewise.
5941 (reset_edge_growth_cache): Likewise.
5942 * symbol-summary.h (call_summary::remove): New method.
5944 2018-06-08 Martin Liska <mliska@suse.cz>
5946 * ipa-cp.c (class edge_clone_summary): New summary.
5947 (grow_edge_clone_vectors): Remove.
5948 (ipcp_edge_duplication_hook): Remove.
5949 (class edge_clone_summary_t): New call_summary class.
5950 (ipcp_edge_removal_hook): Remove.
5951 (edge_clone_summary_t::duplicate): New function.
5952 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
5953 (create_specialized_node): Likewise.
5954 (ipcp_driver): Initialize edge_clone_summaries and do not
5957 2018-06-08 Martin Liska <mliska@suse.cz>
5959 * symbol-summary.h (get): New function.
5960 (call_summary::m_initialize_when_cloning): New class member.
5962 2018-06-08 Martin Liska <mliska@suse.cz>
5964 * cgraph.c (cgraph_node::remove): Do not recycle uid.
5965 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
5966 (symbol_table::allocate_cgraph_symbol): Do not set uid.
5967 * passes.c (uid_hash_t): Record removed_nodes by their uids.
5968 (remove_cgraph_node_from_order): Use the removed_nodes set.
5969 (do_per_function_toporder): Likwise.
5970 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
5971 instead of summary_uid.
5972 (symtab_removal): Likewise.
5973 (symtab_duplication): Likewise.
5975 2018-06-08 Martin Liska <mliska@suse.cz>
5977 * ipa-cp.c (ipcp_store_bits_results): Use
5978 ipcp_transformation_sum.
5979 (ipcp_store_vr_results): Likewise.
5980 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
5982 (ipcp_transformation_initialize): ... this.
5983 (ipa_set_node_agg_value_chain):
5984 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
5985 (write_ipcp_transformation_info): Likewise.
5986 (read_ipcp_transformation_info): Likewise.
5987 (ipcp_update_bits): Likewise.
5988 (ipcp_update_vr): Likewise.
5989 (ipcp_transform_function): Likewise.
5990 * ipa-prop.h: Rename ipcp_transformation_summary to
5991 ipcp_transformation.
5992 (class ipcp_transformation_t): New function summary.
5993 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
5994 (ipa_get_agg_replacements_for_node): Likewise.
5996 2018-06-08 Martin Liska <mliska@suse.cz>
5998 * ipa-pure-const.c (struct funct_state_d): Do it class instead
6000 (class funct_state_summary_t): New function_summary class.
6001 (has_function_state): Remove.
6002 (get_function_state): Likewise.
6003 (set_function_state): Likewise.
6004 (add_new_function): Likewise.
6005 (funct_state_summary_t::insert): New function.
6006 (duplicate_node_data): Remove.
6007 (remove_node_data): Remove.
6008 (funct_state_summary_t::duplicate): New function.
6009 (register_hooks): Create new funct_state_summaries.
6010 (pure_const_generate_summary): Use it.
6011 (pure_const_write_summary): Likewise.
6012 (pure_const_read_summary): Likewise.
6013 (propagate_pure_const): Likewise.
6014 (propagate_nothrow): Likewise.
6015 (dump_malloc_lattice): Likewise.
6016 (propagate_malloc): Likewise.
6017 (execute): Do not register hooks, just remove summary
6019 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
6022 2018-06-08 Martin Liska <mliska@suse.cz>
6024 * ipa-reference.c (remove_node_data): Remove.
6025 (duplicate_node_data): Likewise.
6026 (class ipa_ref_var_info_summary_t): New class.
6027 (class ipa_ref_opt_summary_t): Likewise.
6028 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
6029 (get_reference_optimization_summary): Use
6030 ipa_ref_opt_sum_summaries.
6031 (set_reference_vars_info): Remove.
6032 (set_reference_optimization_summary): Likewise.
6033 (ipa_init): Create summaries.
6034 (init_function_info): Use function summary.
6035 (ipa_ref_opt_summary_t::duplicate): New function.
6036 (ipa_ref_opt_summary_t::remove): New function.
6037 (get_read_write_all_from_node): Fix GNU coding style.
6038 (propagate): Use function summary.
6039 (write_node_summary_p): Fix GNU coding style.
6040 (stream_out_bitmap): Likewise.
6041 (ipa_reference_read_optimization_summary): Use function summary.
6042 (ipa_reference_c_finalize): Do not release hooks.
6044 2018-06-08 Martin Liska <mliska@suse.cz>
6046 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
6047 (analyze_function_body): Extract multiple calls of get_create.
6048 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
6049 * ipa-inline.c (recursive_inlining): Use ::get method.
6050 * ipa-inline.h (estimate_edge_growth): Likewise.
6052 2018-06-08 Martin Liska <mliska@suse.cz>
6054 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
6056 (hsa_function_summary::hsa_function_summary): Use the new enum
6058 (hsa_gpu_implementation_p): Use hsa_summaries::get.
6059 * hsa-gen.c (hsa_get_host_function): Likewise.
6060 (get_brig_function_name): Likewise.
6061 * ipa-hsa.c (process_hsa_functions): Likewise.
6062 (ipa_hsa_write_summary): Likewise.
6063 * symbol-summary.h (symtab_duplication): Use ::get function/
6064 (get): New function.
6066 2018-06-08 Martin Liska <mliska@suse.cz>
6068 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
6070 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
6071 (hsa_register_kernel): Likewise.
6072 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
6073 * hsa-gen.c (hsa_get_host_function): Likewise.
6074 (get_brig_function_name): Likewise.
6075 (generate_hsa): Likewise.
6076 (pass_gen_hsail::execute): Likewise.
6077 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
6078 (devirtualization_time_bonus): Likewise.
6079 (ipcp_propagate_stage): Likewise.
6080 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
6081 (edge_set_predicate): Likewise.
6082 (evaluate_conditions_for_known_args): Likewise.
6083 (evaluate_properties_for_edge): Likewise.
6084 (ipa_fn_summary::reset): Likewise.
6085 (ipa_fn_summary_t::duplicate): Likewise.
6086 (dump_ipa_call_summary): Likewise.
6087 (ipa_dump_fn_summary): Likewise.
6088 (analyze_function_body): Likewise.
6089 (compute_fn_summary): Likewise.
6090 (estimate_edge_devirt_benefit): Likewise.
6091 (estimate_edge_size_and_time): Likewise.
6092 (estimate_calls_size_and_time): Likewise.
6093 (estimate_node_size_and_time): Likewise.
6094 (inline_update_callee_summaries): Likewise.
6095 (remap_edge_change_prob): Likewise.
6096 (remap_edge_summaries): Likewise.
6097 (ipa_merge_fn_summary_after_inlining): Likewise.
6098 (ipa_update_overall_fn_summary): Likewise.
6099 (read_ipa_call_summary): Likewise.
6100 (inline_read_section): Likewise.
6101 (write_ipa_call_summary): Likewise.
6102 (ipa_fn_summary_write): Likewise.
6103 (ipa_free_fn_summary): Likewise.
6104 * ipa-hsa.c (process_hsa_functions): Likewise.
6105 (ipa_hsa_write_summary): Likewise.
6106 (ipa_hsa_read_section): Likewise.
6107 * ipa-icf.c (sem_function::merge): Likewise.
6108 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
6109 (do_estimate_edge_time): Likewise.
6110 (estimate_size_after_inlining): Likewise.
6111 (estimate_growth): Likewise.
6112 (growth_likely_positive): Likewise.
6113 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
6114 (inline_call): Likewise.
6115 * ipa-inline.c (caller_growth_limits): Likewise.
6116 (can_inline_edge_p): Likewise.
6117 (can_inline_edge_by_limits_p): Likewise.
6118 (compute_uninlined_call_time): Likewise.
6119 (compute_inlined_call_time): Likewise.
6120 (want_inline_small_function_p): Likewise.
6121 (edge_badness): Likewise.
6122 (update_caller_keys): Likewise.
6123 (update_callee_keys): Likewise.
6124 (recursive_inlining): Likewise.
6125 (inline_small_functions): Likewise.
6126 (inline_to_all_callers_1): Likewise.
6127 (dump_overall_stats): Likewise.
6128 (early_inline_small_functions): Likewise.
6129 (early_inliner): Likewise.
6130 * ipa-inline.h (estimate_edge_growth): Likewise.
6131 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
6132 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
6133 * ipa-prop.h (IPA_NODE_REF): Likewise.
6134 (IPA_EDGE_REF): Likewise.
6135 * ipa-pure-const.c (malloc_candidate_p): Likewise.
6136 (propagate_malloc): Likewise.
6137 * ipa-split.c (execute_split_functions): Likewise.
6138 * symbol-summary.h: Rename get to get_create.
6140 (get_create): Likewise.
6141 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
6143 2018-06-08 Martin Liska <mliska@suse.cz>
6145 * symbol-summary.h (release): Move definition out of class
6147 (symtab_removal): Likewise.
6148 (symtab_duplication): Likewise.
6150 2018-06-08 Martin Liska <mliska@suse.cz>
6152 * symbol-summary.h (function_summary): Move constructor
6153 implementation out of class declaration.
6154 (release): Likewise.
6155 (symtab_insertion): Likewise.
6156 (symtab_removal): Likewise.
6157 (symtab_duplication): Likewise.
6160 2018-06-08 Martin Liska <mliska@suse.cz>
6162 * Makefile.in: Remove support for MPX (macros, related functions,
6163 fields in cgraph_node, ...).
6164 * builtin-types.def (BT_BND): Likewise.
6165 (BT_FN_BND_CONST_PTR): Likewise.
6166 (BT_FN_CONST_PTR_BND): Likewise.
6167 (BT_FN_VOID_PTR_BND): Likewise.
6168 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
6169 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
6170 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
6171 (expand_builtin_mempcpy_with_bounds): Likewise.
6172 (expand_builtin_memset_with_bounds): Likewise.
6173 (expand_builtin_memset_args): Likewise.
6174 (std_expand_builtin_va_start): Likewise.
6175 (expand_builtin): Likewise.
6176 (expand_builtin_with_bounds): Likewise.
6177 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
6178 (DEF_LIB_BUILTIN_CHKP): Likewise.
6179 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
6180 (DEF_CHKP_BUILTIN): Likewise.
6181 (BUILT_IN_MEMCPY): Likewise.
6182 (BUILT_IN_MEMMOVE): Likewise.
6183 (BUILT_IN_MEMPCPY): Likewise.
6184 (BUILT_IN_MEMSET): Likewise.
6185 (BUILT_IN_STPCPY): Likewise.
6186 (BUILT_IN_STRCAT): Likewise.
6187 (BUILT_IN_STRCHR): Likewise.
6188 (BUILT_IN_STRCPY): Likewise.
6189 (BUILT_IN_STRLEN): Likewise.
6190 (BUILT_IN_MEMCPY_CHK): Likewise.
6191 (BUILT_IN_MEMMOVE_CHK): Likewise.
6192 (BUILT_IN_MEMPCPY_CHK): Likewise.
6193 (BUILT_IN_MEMSET_CHK): Likewise.
6194 (BUILT_IN_STPCPY_CHK): Likewise.
6195 (BUILT_IN_STRCAT_CHK): Likewise.
6196 (BUILT_IN_STRCPY_CHK): Likewise.
6197 * calls.c (store_bounds): Likewise.
6198 (emit_call_1): Likewise.
6199 (special_function_p): Likewise.
6200 (maybe_warn_nonstring_arg): Likewise.
6201 (initialize_argument_information): Likewise.
6202 (finalize_must_preallocate): Likewise.
6203 (compute_argument_addresses): Likewise.
6204 (expand_call): Likewise.
6205 * cfgexpand.c (expand_call_stmt): Likewise.
6206 (expand_return): Likewise.
6207 (expand_gimple_stmt_1): Likewise.
6208 (pass_expand::execute): Likewise.
6209 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
6210 (cgraph_node::remove): Likewise.
6211 (cgraph_node::dump): Likewise.
6212 (cgraph_node::verify_node): Likewise.
6213 * cgraph.h (chkp_function_instrumented_p): Likewise.
6214 (symtab_node::get_alias_target): Likewise.
6215 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
6216 (cgraph_local_p): Likewise.
6217 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
6218 (cgraph_edge::rebuild_references): Likewise.
6219 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
6220 (walk_polymorphic_call_targets): Likewise.
6221 (cgraph_node::expand_thunk): Likewise.
6222 (symbol_table::output_weakrefs): Likewise.
6223 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
6224 (ix86_handle_option): Likewise.
6225 * config/i386/constraints.md: Likewise.
6226 * config/i386/i386-builtin-types.def (BND): Likewise.
6230 * config/i386/i386-builtin.def (BDESC_END): Likewise.
6231 (BDESC_FIRST): Likewise.
6233 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
6234 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
6235 * config/i386/i386.c (enum reg_class): Likewise.
6236 (ix86_target_string): Likewise.
6237 (ix86_option_override_internal): Likewise.
6238 (ix86_conditional_register_usage): Likewise.
6239 (ix86_valid_target_attribute_inner_p): Likewise.
6240 (ix86_set_indirect_branch_type): Likewise.
6241 (ix86_set_current_function): Likewise.
6242 (ix86_function_arg_regno_p): Likewise.
6243 (init_cumulative_args): Likewise.
6244 (ix86_function_arg_advance): Likewise.
6245 (ix86_function_arg): Likewise.
6246 (ix86_pass_by_reference): Likewise.
6247 (ix86_function_value_regno_p): Likewise.
6248 (ix86_function_value_1): Likewise.
6249 (ix86_function_value_bounds): Likewise.
6250 (ix86_return_in_memory): Likewise.
6251 (ix86_setup_incoming_vararg_bounds): Likewise.
6252 (ix86_va_start): Likewise.
6253 (indirect_thunk_need_prefix): Likewise.
6254 (print_reg): Likewise.
6255 (ix86_print_operand): Likewise.
6256 (ix86_expand_call): Likewise.
6257 (ix86_output_function_return): Likewise.
6258 (reg_encoded_number): Likewise.
6259 (BDESC_VERIFYS): Likewise.
6260 (ix86_init_mpx_builtins): Likewise.
6261 (ix86_init_builtins): Likewise.
6262 (ix86_emit_cmove): Likewise.
6263 (ix86_emit_move_max): Likewise.
6264 (ix86_expand_builtin): Likewise.
6265 (ix86_builtin_mpx_function): Likewise.
6266 (ix86_get_arg_address_for_bt): Likewise.
6267 (ix86_load_bounds): Likewise.
6268 (ix86_store_bounds): Likewise.
6269 (ix86_load_returned_bounds): Likewise.
6270 (ix86_store_returned_bounds): Likewise.
6271 (ix86_class_likely_spilled_p): Likewise.
6272 (ix86_hard_regno_mode_ok): Likewise.
6273 (x86_order_regs_for_local_alloc): Likewise.
6274 (ix86_mitigate_rop): Likewise.
6275 (ix86_bnd_prefixed_insn_p): Likewise.
6276 (ix86_mpx_bound_mode): Likewise.
6277 (ix86_make_bounds_constant): Likewise.
6278 (ix86_initialize_bounds): Likewise.
6279 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
6280 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
6281 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
6282 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
6283 (TARGET_CHKP_BOUND_MODE): Likewise.
6284 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
6285 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
6286 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
6287 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
6288 * config/i386/i386.h (TARGET_MPX): Likewise.
6289 (TARGET_MPX_P): Likewise.
6290 (VALID_BND_REG_MODE): Likewise.
6291 (FIRST_BND_REG): Likewise.
6292 (LAST_BND_REG): Likewise.
6293 (enum reg_class): Likewise.
6294 (BND_REG_P): Likewise.
6295 (BND_REGNO_P): Likewise.
6296 (BNDmode): Likewise.
6297 (ADJUST_INSN_LENGTH): Likewise.
6298 * config/i386/i386.md: Likewise.
6299 * config/i386/i386.opt: Likewise.
6300 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
6301 (defined): Likewise.
6302 (LINK_MPX): Likewise.
6303 (MPX_SPEC): Likewise.
6304 (LIBMPX_SPEC): Likewise.
6305 (LIBMPXWRAPPERS_SPEC): Likewise.
6306 (CHKP_SPEC): Likewise.
6307 * config/i386/predicates.md: Likewise.
6308 * dbxout.c (dbxout_type): Likewise.
6309 * doc/extend.texi: Likewise.
6310 * doc/invoke.texi: Likewise.
6311 * doc/md.texi: Likewise.
6312 * doc/tm.texi: Likewise.
6313 * doc/tm.texi.in: Likewise.
6314 * dwarf2out.c (is_base_type): Likewise.
6315 (gen_formal_types_die): Likewise.
6316 (gen_subprogram_die): Likewise.
6317 (gen_type_die_with_usage): Likewise.
6318 (gen_decl_die): Likewise.
6319 (dwarf2out_late_global_decl): Likewise.
6320 * expr.c (expand_assignment): Likewise.
6321 (emit_storent_insn): Likewise.
6322 (store_expr_with_bounds): Likewise.
6323 (store_expr): Likewise.
6324 (expand_expr_real_1): Likewise.
6325 * expr.h (store_expr_with_bounds): Likewise.
6326 * function.c (use_register_for_decl): Likewise.
6327 (struct bounds_parm_data): Likewise.
6328 (assign_parms_augmented_arg_list): Likewise.
6329 (assign_parm_find_entry_rtl): Likewise.
6330 (assign_parm_is_stack_parm): Likewise.
6331 (assign_parm_load_bounds): Likewise.
6332 (assign_bounds): Likewise.
6333 (assign_parms): Likewise.
6334 (expand_function_start): Likewise.
6335 * gcc.c (CHKP_SPEC): Likewise.
6336 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
6337 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
6338 (wrestrict_dom_walker::check_call): Likewise.
6339 * gimple.c (gimple_build_call_from_tree): Likewise.
6340 * gimple.h (enum gf_mask): Likewise.
6341 (gimple_call_with_bounds_p): Likewise.
6342 (gimple_call_set_with_bounds): Likewise.
6343 * gimplify.c (gimplify_init_constructor): Likewise.
6344 * ipa-cp.c (initialize_node_lattices): Likewise.
6345 (propagate_constants_across_call): Likewise.
6346 (find_more_scalar_values_for_callers_subset): Likewise.
6347 * ipa-hsa.c (process_hsa_functions): Likewise.
6348 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
6349 * ipa-icf.c (sem_function::merge): Likewise.
6350 * ipa-inline.c (early_inliner): Likewise.
6351 * ipa-pure-const.c (warn_function_noreturn): Likewise.
6352 (warn_function_cold): Likewise.
6353 (propagate_pure_const): Likewise.
6354 * ipa-ref.h (enum GTY): Likewise.
6355 * ipa-split.c (find_retbnd): Likewise.
6356 (consider_split): Likewise.
6357 (split_function): Likewise.
6358 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
6359 * ipa.c (walk_polymorphic_call_targets): Likewise.
6360 (symbol_table::remove_unreachable_nodes): Likewise.
6361 (process_references): Likewise.
6362 (cgraph_build_static_cdtor_1): Likewise.
6363 * lto-cgraph.c (lto_output_node): Likewise.
6364 (output_refs): Likewise.
6365 (compute_ltrans_boundary): Likewise.
6366 (input_overwrite_node): Likewise.
6367 (input_node): Likewise.
6368 (input_cgraph_1): Likewise.
6369 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
6370 * passes.c (pass_manager::execute_early_local_passes): Likewise.
6371 (class pass_chkp_instrumentation_passes): Likewise.
6372 (make_pass_chkp_instrumentation_passes): Likewise.
6373 * passes.def: Likewise.
6374 * rtl.h (struct GTY): Likewise.
6375 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
6376 * stor-layout.c (layout_type): Likewise.
6377 * symtab.c: Likewise.
6378 * target.def: Likewise.
6379 * targhooks.c (default_chkp_bound_type): Likewise.
6380 (default_chkp_bound_mode): Likewise.
6381 (default_builtin_chkp_function): Likewise.
6382 (default_chkp_function_value_bounds): Likewise.
6383 (default_chkp_make_bounds_constant): Likewise.
6384 (default_chkp_initialize_bounds): Likewise.
6385 * targhooks.h (default_chkp_bound_type): Likewise.
6386 (default_chkp_bound_mode): Likewise.
6387 (default_builtin_chkp_function): Likewise.
6388 (default_chkp_function_value_bounds): Likewise.
6389 (default_chkp_make_bounds_constant): Likewise.
6390 (default_chkp_initialize_bounds): Likewise.
6391 * toplev.c (compile_file): Likewise.
6392 (process_options): Likewise.
6393 * tree-core.h (DEF_BUILTIN): Likewise.
6394 (DEF_BUILTIN_CHKP): Likewise.
6395 * tree-inline.c (declare_return_variable): Likewise.
6396 (remap_gimple_stmt): Likewise.
6397 (copy_bb): Likewise.
6398 (initialize_inlined_parameters): Likewise.
6399 (expand_call_inline): Likewise.
6400 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
6401 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
6402 (make_pass_ipa_chkp_produce_thunks): Likewise.
6403 (make_pass_chkp): Likewise.
6404 (make_pass_chkp_opt): Likewise.
6405 (make_pass_chkp_instrumentation_passes): Likewise.
6406 * tree-pretty-print.c (dump_generic_node): Likewise.
6407 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
6408 * tree-ssa-dce.c (propagate_necessity): Likewise.
6409 (eliminate_unnecessary_stmts): Likewise.
6410 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
6411 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
6412 * tree-ssa-sccvn.h: Likewise.
6413 * tree-ssa-strlen.c (get_string_length): Likewise.
6414 (valid_builtin_call): Likewise.
6415 (adjust_last_stmt): Likewise.
6416 (handle_builtin_strchr): Likewise.
6417 (handle_builtin_strcpy): Likewise.
6418 (handle_builtin_stxncpy): Likewise.
6419 (handle_builtin_memcpy): Likewise.
6420 (handle_builtin_strcat): Likewise.
6421 (strlen_check_and_optimize_stmt): Likewise.
6422 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
6423 * tree-streamer-in.c: Likewise.
6424 * tree-streamer.c (record_common_node): Likewise.
6425 * tree.c (tree_code_size): Likewise.
6426 (wide_int_to_tree_1): Likewise.
6427 (type_contains_placeholder_1): Likewise.
6428 (build_common_tree_nodes): Likewise.
6429 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
6430 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
6431 (POINTER_BOUNDS_P): Likewise.
6432 (BOUNDED_TYPE_P): Likewise.
6433 (BOUNDED_P): Likewise.
6434 (CALL_WITH_BOUNDS_P): Likewise.
6435 (pointer_bounds_type_node): Likewise.
6436 * value-prof.c (gimple_ic): Likewise.
6437 * var-tracking.c (vt_add_function_parameters): Likewise.
6438 * varasm.c (make_decl_rtl): Likewise.
6439 (assemble_start_function): Likewise.
6440 (output_constant): Likewise.
6441 (maybe_assemble_visibility): Likewise.
6442 * varpool.c (ctor_for_folding): Likewise.
6443 * chkp-builtins.def: Remove.
6444 * ipa-chkp.c: Remove.
6445 * ipa-chkp.h: Remove.
6446 * rtl-chkp.c: Remove.
6447 * rtl-chkp.h: Remove.
6448 * tree-chkp-opt.c: Remove.
6449 * tree-chkp.c: Remove.
6450 * tree-chkp.h: Remove.
6452 2018-06-07 Carl Love <cel@us.ibm.com>
6454 * config/rs6000/vsx.md (vextract_fp_from_shorth,
6455 vextract_fp_from_shortl): Add BE support.
6457 2018-06-07 Paul Koning <ni1d@arrl.net>
6459 * compare-elim.c (try_merge_compare): Don't merge compare if
6460 address contains a side effect.
6461 (try_eliminate_compare): Likewise.
6463 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
6465 * config.gcc: Support "tremont".
6466 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
6467 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6469 * config/i386/i386.c (m_TREMONT): Define.
6470 (processor_target_table): Add "tremont".
6471 (PTA_TREMONT): Define.
6472 (ix86_lea_outperforms): Add TARGET_TREMONT.
6473 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
6474 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
6475 and M_INTEL_GOLDMONT_PLUS.
6476 (fold_builtin_cpu): Add "tremont".
6477 (ix86_add_stmt_cost): Add TARGET_TREMONT.
6478 (ix86_option_override_internal): Add "tremont".
6479 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
6480 (processor_type): Add PROCESSOR_TREMONT.
6481 * config/i386/x86-tune.def: Add m_TREMONT.
6482 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
6484 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
6486 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
6487 symbol defined for msp430i* devices to be lower case.
6489 2018-06-07 Richard Biener <rguenther@suse.de>
6491 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
6492 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
6493 Properly wrap signed arithmetic if overflow wraps.
6495 2018-06-07 Jakub Jelinek <jakub@redhat.com>
6497 PR tree-optimization/69615
6498 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
6499 of a cast from a same precision integral SSA_NAME in a bb dominated
6500 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
6501 cast to utype if rhs2 has already a compatible type.
6503 2018-06-07 Richard Biener <rguenther@suse.de>
6505 PR tree-optimization/85935
6506 * graphite-scop-detection.c (find_params_in_bb): Analyze
6507 condition operands with respect to the correct loop. Assert
6508 the analysis doesn't fail.
6510 2018-06-04 Carl Love <cel@us.ibm.com>
6512 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
6513 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
6514 as it is slightly cheaper.
6515 (first_match_or_eos_index_<mode>):
6516 Calculate index using natural element order.
6517 (first_match_index_<mode>):
6518 Calculate index using natural element order.
6519 (first_match_or_eos_index_<mode>):
6520 Calculate index using natural order.
6521 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
6522 for BE and LE modes.
6523 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
6524 P9V_BUILTIN_VCLZLSBB_V16QI.
6525 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
6528 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
6530 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
6531 indentation and line wrap for many prototypes. Add missing
6532 @smallexample directives around block of prototypes for vec_xl and
6535 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
6537 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
6538 track if we pass or return IEEE 128-bit floating point.
6539 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
6540 C++ mangling that is compatible with GCC 8.1.
6541 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
6542 (init_cumulative_args): Note if we pass or return IEEE 128-bit
6543 floating point types.
6544 (rs6000_function_arg_advance_1): Likewise.
6545 (rs6000_mangle_type): Optionally generate mangled names that match
6546 what GCC 8.1 generated for IEEE 128-bit floating point types.
6547 (rs6000_globalize_decl_name): If we have an external function that
6548 passes or returns IEEE 128-bit types, generate a weak reference
6549 from the mangled name used in GCC 8.1 to the current mangled
6551 (rs6000_init_builtins): Make __ibm128 use the long double type if
6552 long double is IBM extended double. Make __float128 use the long
6553 double type if long double is IEEE 128-bit.
6556 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
6557 macro for __ibm128 built-in functions.
6558 (PACK_IF): Add __ibm128 pack/unpack functions.
6559 (UNPACK_IF): Likewise.
6560 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
6561 enable long double built-in functions if long double is IEEE
6562 128-bit floating point.
6563 (rs6000_invalid_builtin): Update long double built-in function
6565 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
6566 functions, adjust the built-in function to use the long double
6567 built-in function if __ibm128 and long double are the same type.
6568 * doc/extend.texi (PowerPC builtins): Update documention for
6569 __builtin_{,un}pack_longdouble. Add documentation for
6570 __builtin_{,un}pack_ibm128.
6572 2018-06-06 Jim Wilson <jimw@sifive.com>
6574 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
6575 (struct machine_function): New field interrupt_mode.
6576 (riscv_handle_type_attribute): New function. Add forward declaration.
6577 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
6578 (riscv_expand_epilogue): Check interrupt_mode field.
6579 (riscv_set_current_function): Check interrupt attribute args and
6580 set interrupt_mode field.
6581 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
6582 (riscv_sret, riscv_uret): New.
6583 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
6584 new arguments to interrupt attribute.
6586 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
6589 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
6590 Don't handle -mcpu=power8 if -mpower9-vector is also used.
6592 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
6594 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
6595 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
6596 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
6597 several redundant entries.
6599 2018-06-06 David Malcolm <dmalcolm@redhat.com>
6601 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
6602 type from "rtx" to "rtx_insn *".
6603 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
6604 for local "call_insn", removing cast.
6605 (ix86_expand_call): Likewise, introducing a "call_insn" local.
6607 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
6609 PR tree-optimization/86066
6610 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
6611 for BIT_INSERT_EXPR stores.
6613 2018-06-06 Richard Biener <rguenther@suse.de>
6615 PR tree-optimization/86062
6616 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
6617 component refs ontop
6618 of to be offsetted base.
6620 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
6622 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
6623 to be static and remove check on interrupt attribute name.
6625 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
6627 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
6628 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
6630 2018-06-05 Steve Ellcey <sellcey@cavium.com>
6633 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
6635 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
6636 Remove second argument, change how error is called.
6637 (aarch64_layout_arg): Remove second argument from
6638 aarch64_err_no_fpadvsimd call.
6639 (aarch64_init_cumulative_args): Ditto.
6640 (aarch64_gimplify_va_arg_expr): Ditto.
6641 * config/aarch64/aarch64.md (mov<mode>): Ditto.
6643 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
6645 * config/i386/i386.md (simple_return_indirect_internal): New expander.
6646 (*simple_return_indirect_internal<mode>): Rename from
6647 simple_return_indirect_internal. Use W mode iterator.
6648 (rstorssp): New expander.
6649 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
6650 (clrssbsy): New expander.
6651 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
6653 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
6655 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
6657 (cmse_check_pointed_object): Likewise.
6659 2018-06-05 Martin Liska <mliska@suse.cz>
6661 PR gcov-profile/47618
6662 * doc/invoke.texi: Document how -fprofile-dir format
6665 2018-06-05 Richard Biener <rguenther@suse.de>
6667 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
6668 removal pretend DOM info isn't available so we do not update
6669 it and only remove edges, not dominated blocks. Actually free
6670 DOM info in case we removed something. Remove unreachable blocks.
6671 (mfb_keep_latches): Work with either DOM info or marked backedges.
6672 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
6673 first. Mark backedges if DOM info isn't available.
6674 (Re-)compute DOM info after cleanup_control_flow_pre.
6676 2018-06-05 Richard Biener <rguenther@suse.de>
6678 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
6679 (locus_discrim_hasher::hash): Adjust.
6680 (locus_discrim_hasher::equal): Likewise.
6681 (next_discriminator_for_locus): Work on line directly.
6682 (same_line_p): Pass in expanded locus1 as well.
6683 (assign_discriminators): Avoid redundant location expansions.
6685 2018-06-05 Richard Biener <rguenther@suse.de>
6687 PR tree-optimization/86046
6688 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
6689 if required after clearing TREE_ADDRESSABLE.
6691 2018-06-05 Richard Biener <rguenther@suse.de>
6693 PR tree-optimization/86047
6694 * tree-ssa-loop.c (for_each_index): Glob handling of all
6695 decls and constants and really handle all of them.
6697 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6700 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
6701 qualifier_void_pointer and qualifier_const_void_pointer.
6702 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
6703 (arm_init_builtins): Handle the above.
6704 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
6705 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
6706 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
6709 2018-06-05 Martin Liska <mliska@suse.cz>
6711 * auto-profile.c (read_autofdo_file): Do not use
6712 gcov_ctr_summary struct.
6713 (afdo_callsite_hot_enough_for_early_inline): Likewise.
6714 * coverage.c (struct counts_entry): Likewise.
6715 (read_counts_file): Read just single summary entry.
6716 (get_coverage_counts): Use gcov_summary struct.
6717 * coverage.h (get_coverage_counts): Likewise.
6718 * gcov-dump.c (dump_working_sets): Likewise.
6719 (tag_summary): Dump just single summary.
6720 * gcov-io.c (gcov_write_summary): Write just histogram
6722 (gcov_read_summary): Read just single summary.
6723 (compute_working_sets): Use gcov_summary struct.
6724 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
6725 of GCOV_COUNTERS_SUMMABLE.
6726 (GCOV_COUNTERS_SUMMABLE): Remove.
6727 (GCOV_FIRST_VALUE_COUNTER): Replace with
6728 GCOV_COUNTER_V_INTERVAL.
6729 (struct gcov_ctr_summary): Remove.
6730 (struct gcov_summary): Directly use fields of former
6732 (compute_working_sets): Use gcov_summary struct.
6733 * gcov.c (read_count_file): Do not use ctrs fields.
6734 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
6736 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
6738 * profile.c: Likewise.
6739 * profile.h: Likewise.
6741 2018-06-05 Martin Liska <mliska@suse.cz>
6743 PR gcov-profile/84846
6744 * gcov.c (output_lines): Print working directory only
6745 in intermediate format.
6747 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
6749 * config/s390/s390-builtin-types.def: Add void function type.
6750 * config/s390/s390-builtins.def: Use the function type for the
6753 2018-06-04 Jim Wilson <jimw@sifive.com>
6755 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
6757 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
6758 and maybe_eh_return. Change regno to unsigned int. Use new args to
6759 handle EH_RETURN_DATA_REGNO registers properly.
6760 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
6761 (riscv_expand_epilogue): Update comment. Change argument name and
6762 type. Update code to use new name and type. Pass new args to
6763 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
6765 * config/riscv/riscv.md (NORMAL_RETURN): New.
6766 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
6767 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
6768 (eh_return): Call gen_eh_return_internal and emit barrier.
6769 (eh_return_internal): Call riscv_expand_epilogue.
6771 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
6773 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
6774 bit_insertion field and declare can_be_merged_into method.
6775 (merged_store_group::can_be_merged_into): New method.
6776 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
6777 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
6778 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
6780 2018-06-04 Richard Biener <rguenther@suse.de>
6782 PR tree-optimization/85955
6783 * builtins.c (fold_builtin_sincos): Convert pointers to
6784 destination to appropriate type before dereferencing.
6786 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
6788 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
6790 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
6792 * expr.c (expand_expr_real_1): Force the operand into memory if
6793 its TYPE_MODE is BLKmode and if there is no integer mode for
6794 the number of bits being extracted.
6796 2018-06-04 Jakub Jelinek <jakub@redhat.com>
6800 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
6801 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
6803 2018-06-04 Richard Biener <rguenther@suse.de>
6805 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
6806 (cleanup_tree_cfg_noloop): ... single caller. Do
6807 start_recording_case_labels later.
6809 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
6811 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
6812 to _IMMINTRIN_H_INCLUDED.
6813 * config/i386/pconfigintrin.h: Ditto.
6814 * config/i386/waitpkgintrin.h: Ditto.
6815 * config/i386/immintrin.h: Add includes for sgxintrin.h,
6816 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
6817 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
6818 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
6819 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
6820 waitpkgintrin.h and cldemoteintrin.h.
6822 2018-06-04 Richard Biener <rguenther@suse.de>
6824 PR tree-optimization/86038
6825 * tracer.c (find_best_successor): Check probability for
6826 being initialized, bail out if not.
6828 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
6831 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
6832 of bits to ignore when comparing architectures.
6834 2018-06-04 Jakub Jelinek <jakub@redhat.com>
6836 PR tree-optimization/69615
6837 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
6838 maximum or minimum of the type, try to merge it also as if
6839 range1 is + [-, x - 1] or + [x + 1, -].
6842 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
6844 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
6846 PR tree-optimization/86034
6847 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
6848 the unsigned bitfield type in a bit insertion sequence if it does not
6849 have a larger precision than the bitfield size.
6850 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
6852 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
6854 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
6856 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
6858 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
6859 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
6860 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
6861 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
6863 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
6865 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
6866 Disable -fdelete-null-pointer-checks for ELF toolchain.
6868 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
6869 Kito Cheng <kito.cheng@gmail.com>
6871 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
6872 (nds32le-*-*, nds32be-*-*): Integrate checking process.
6873 (nds32*-*-*): Add glibc and uclibc conditions.
6874 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
6875 (TARGET_EXCEPT_UNWIND_INFO): Define.
6876 * config/nds32/elf.h: New file.
6877 * config/nds32/linux.h: New file.
6878 * config/nds32/nds32-elf.opt: New file.
6879 * config/nds32/nds32-linux.opt: New file.
6880 * config/nds32/nds32-fp-as-gp.c
6881 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
6882 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
6884 (nds32_asm_file_end): Ditto.
6885 (nds32_print_operand): Ditto.
6886 (nds32_insert_attributes): Ditto.
6887 (nds32_init_libfuncs): New function.
6888 (TARGET_HAVE_TLS): Define.
6889 (TARGET_INIT_LIBFUNCS): Define.
6890 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
6892 (TARGET_ELF): Apply different mcmodel setting.
6893 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
6894 been migrated into elf.h and linux.h files.
6895 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
6896 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
6897 (mcmodel): The content has been migrated into nds32-elf.opt and
6898 nds32-linux.opt files.
6899 * config/nds32/t-elf: New file.
6900 * config/nds32/t-linux: New file.
6902 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
6903 Shiva Chen <shiva0217@gmail.com>
6905 * config/nds32/constants.md (unspec_volatile_element): Add
6906 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
6907 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
6909 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
6910 (make_pass_nds32_fp_as_gp): Declare.
6911 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
6913 (nds32_asm_function_end_prologue): Remove unused asm output.
6914 (nds32_asm_function_begin_epilogue): Remove unused asm output.
6915 (nds32_asm_file_start): Output necessary fp_as_gp information.
6916 (nds32_option_override): Adjust register usage.
6917 (nds32_expand_prologue): Consider fp_as_gp situation.
6918 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
6919 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
6922 (simple_return): Ditto.
6923 (omit_fp_begin): Output special directive for fp_as_gp.
6924 (omit_fp_end): Output special directive for fp_as_gp.
6925 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
6926 mforbid-fp-as-gp): New options.
6928 2018-06-01 Mark Wielaard <mark@klomp.org>
6930 * dwarf2out.c (dwarf2out_finish): Remove generation of
6931 DW_AT_loclists_base.
6933 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
6935 * gimple-ssa-store-merging.c: Include gimple-fold.h.
6936 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
6937 (struct merged_store_group): Add bit_insertion field.
6938 (dump_char_array): Use standard hexadecimal format.
6939 (merged_store_group::merged_store_group): Set bit_insertion to false.
6940 (merged_store_group::apply_stores): Use optimal buffer size. Deal
6941 with BIT_INSERT_EXPR stores. Move up code updating the mask and
6942 also print the mask in the dump file.
6943 (pass_store_merging::gate): Minor tweak.
6944 (imm_store_chain_info::coalesce_immediate): Fix wrong association
6945 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
6946 stores with INTEGER_CST stores.
6947 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
6948 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
6949 and use it throughout. Generate bit insertion sequences if need be.
6950 (pass_store_merging::process_store): Remove redundant condition.
6951 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
6953 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
6955 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
6956 the 128-bit floating point types. Fix function comment.
6958 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6960 * config/aarch64/aarch64-simd.md
6961 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
6963 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
6966 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
6968 PR tree-optimization/85989
6969 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
6971 (backprop::intersect_uses): Check it when deciding whether this
6972 is a backedge reference.
6973 (backprop::process_block): Add each phi to m_visited_phis
6974 after visiting it, then clear it at the end.
6976 2018-06-01 Richard Biener <rguenther@suse.de>
6978 * tree-vectorizer.h (vect_dr_stmt): New function.
6979 (vect_get_load_cost): Adjust.
6980 (vect_get_store_cost): Likewise.
6981 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
6982 Use vect_dr_stmt instead of DR_SMTT.
6983 (vect_record_base_alignments): Likewise.
6984 (vect_calculate_target_alignment): Likewise.
6985 (vect_compute_data_ref_alignment): Likewise and make static.
6986 (vect_update_misalignment_for_peel): Likewise.
6987 (vect_verify_datarefs_alignment): Likewise.
6988 (vector_alignment_reachable_p): Likewise.
6989 (vect_get_data_access_cost): Likewise. Pass down
6990 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
6991 (vect_get_peeling_costs_all_drs): Likewise.
6992 (vect_peeling_hash_get_lowest_cost): Likewise.
6993 (vect_enhance_data_refs_alignment): Likewise.
6994 (vect_find_same_alignment_drs): Likewise.
6995 (vect_analyze_data_refs_alignment): Likewise.
6996 (vect_analyze_group_access_1): Likewise.
6997 (vect_analyze_group_access): Likewise.
6998 (vect_analyze_data_ref_access): Likewise.
6999 (vect_analyze_data_ref_accesses): Likewise.
7000 (vect_vfa_segment_size): Likewise.
7001 (vect_small_gap_p): Likewise.
7002 (vectorizable_with_step_bound_p): Likewise.
7003 (vect_prune_runtime_alias_test_list): Likewise.
7004 (vect_analyze_data_refs): Likewise.
7005 (vect_supportable_dr_alignment): Likewise.
7006 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
7007 (vect_gen_prolog_loop_niters): Likewise.
7008 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
7009 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
7011 (vect_recog_mask_conversion_pattern): Likewise.
7012 (vect_try_gather_scatter_pattern): Likewise.
7013 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
7014 to vect_get_store_cost.
7015 (vect_get_store_cost): Get stmt_info instead of DR.
7016 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
7017 (vect_get_load_cost): Get stmt_info instead of DR.
7019 2018-06-01 Richard Biener <rguenther@suse.de>
7022 * gimple-fold.c (var_decl_component_p): Also allow offsetted
7023 vars wrapped in MEM_REFs.
7025 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
7027 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
7028 Fix subreg tests so that we only return a choice between
7029 GENERAL_REGS and FP_REGS if the original classes included both.
7031 2018-06-01 Richard Biener <rguenther@suse.de>
7034 * tree-ssa-structalias.c (get_function_part_constraint):
7035 Handle NULL fi->decl.
7036 (find_func_aliases_for_call): Properly handle indirect
7037 fi from direct call.
7038 (find_func_clobbers): Likewise.
7039 (ipa_pta_execute): Likewise.
7040 (create_variable_info_for): For functions that are ifunc_resolver
7041 resolve to a varinfo that contains the result of the resolver call.
7042 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
7045 2018-05-31 Michael Collison <michael.collison@arm.com>
7047 * config/aarch64/aarch64.md:
7048 (*fix_to_zero_extenddfdi2): New pattern.
7049 * gcc.target/aarch64/fix_extend1.c: New testcase.
7051 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
7055 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
7056 and BUILT_IN_STRNCMP_EQ.
7057 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
7058 BUILT_IN_STRNCMP_EQ.
7059 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
7060 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
7061 (gimple_fold_builtin): Likewise.
7062 * tree-ssa-strlen.c (compute_string_length): New function.
7063 (determine_min_obsize): New function.
7064 (handle_builtin_string_cmp): New function to handle calls to
7065 string compare functions.
7066 (strlen_optimize_stmt): Add handling to builtin string compare
7068 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
7069 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
7070 * tree.c (build_common_builtin_nodes): Add new defines of
7071 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
7073 2018-05-31 Jakub Jelinek <jakub@redhat.com>
7076 * bb-reorder.c (pass_partition_blocks::gate): Return false for
7077 functions with naked attribute.
7079 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
7081 * config/i386/sse.md (avx_vec_concat<mode>):
7082 Substitute concat_tg_mode mode attribute with xtg_mode.
7083 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
7084 (concat_tg_mode): Remove mode attribute.
7086 2018-05-31 Martin Sebor <msebor@redhat.com>
7089 * calls.c (alloc_max_size): Correct a logic error/typo.
7090 Treat excessive arguments as infinite. Warn for invalid arguments.
7091 * doc/invoke.texi (-Walloc-size-larger-than): Update.
7093 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
7096 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
7097 and movx for Haswell.
7099 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
7100 Cesar Philippidis <cesar@codesourcery.com>
7103 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
7104 when emitting error on private/firstprivate reductions.
7105 * omp-low.c (lower_omp_target): Avoid reference-type processing
7106 on pointers for firstprivate clause.
7108 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
7110 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
7113 * config/aarch64/aarch64-simd.md
7114 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
7115 (aarch64_ld1_x3_<mode>): Likewise
7116 (aarch64_st1x2<VALLDIF:mode>): Likewise
7117 (aarch64_st1_x2_<mode>): Likewise
7118 (aarch64_st1x3<VALLDIF:mode>): Likewise
7119 (aarch64_st1_x3_<mode>): Likewise
7120 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
7121 (vld1_s8_x3): Likewise.
7122 (vld1_u16_x3): Likewise.
7123 (vld1_s16_x3): Likewise.
7124 (vld1_u32_x3): Likewise.
7125 (vld1_s32_x3): Likewise.
7126 (vld1_u64_x3): Likewise.
7127 (vld1_s64_x3): Likewise.
7128 (vld1_f16_x3): Likewise.
7129 (vld1_f32_x3): Likewise.
7130 (vld1_f64_x3): Likewise.
7131 (vld1_p8_x3): Likewise.
7132 (vld1_p16_x3): Likewise.
7133 (vld1_p64_x3): Likewise.
7134 (vld1q_u8_x3): Likewise.
7135 (vld1q_s8_x3): Likewise.
7136 (vld1q_u16_x3): Likewise.
7137 (vld1q_s16_x3): Likewise.
7138 (vld1q_u32_x3): Likewise.
7139 (vld1q_s32_x3): Likewise.
7140 (vld1q_u64_x3): Likewise.
7141 (vld1q_s64_x3): Likewise.
7142 (vld1q_f16_x3): Likewise.
7143 (vld1q_f32_x3): Likewise.
7144 (vld1q_f64_x3): Likewise.
7145 (vld1q_p8_x3): Likewise.
7146 (vld1q_p16_x3): Likewise.
7147 (vld1q_p64_x3): Likewise.
7148 (vst1_s64_x2): Likewise.
7149 (vst1_u64_x2): Likewise.
7150 (vst1_f64_x2): Likewise.
7151 (vst1_s8_x2): Likewise.
7152 (vst1_p8_x2): Likewise.
7153 (vst1_s16_x2): Likewise.
7154 (vst1_p16_x2): Likewise.
7155 (vst1_s32_x2): Likewise.
7156 (vst1_u8_x2): Likewise.
7157 (vst1_u16_x2): Likewise.
7158 (vst1_u32_x2): Likewise.
7159 (vst1_f16_x2): Likewise.
7160 (vst1_f32_x2): Likewise.
7161 (vst1_p64_x2): Likewise.
7162 (vst1q_s8_x2): Likewise.
7163 (vst1q_p8_x2): Likewise.
7164 (vst1q_s16_x2): Likewise.
7165 (vst1q_p16_x2): Likewise.
7166 (vst1q_s32_x2): Likewise.
7167 (vst1q_s64_x2): Likewise.
7168 (vst1q_u8_x2): Likewise.
7169 (vst1q_u16_x2): Likewise.
7170 (vst1q_u32_x2): Likewise.
7171 (vst1q_u64_x2): Likewise.
7172 (vst1q_f16_x2): Likewise.
7173 (vst1q_f32_x2): Likewise.
7174 (vst1q_f64_x2): Likewise.
7175 (vst1q_p64_x2): Likewise.
7176 (vst1_s64_x3): Likewise.
7177 (vst1_u64_x3): Likewise.
7178 (vst1_f64_x3): Likewise.
7179 (vst1_s8_x3): Likewise.
7180 (vst1_p8_x3): Likewise.
7181 (vst1_s16_x3): Likewise.
7182 (vst1_p16_x3): Likewise.
7183 (vst1_s32_x3): Likewise.
7184 (vst1_u8_x3): Likewise.
7185 (vst1_u16_x3): Likewise.
7186 (vst1_u32_x3): Likewise.
7187 (vst1_f16_x3): Likewise.
7188 (vst1_f32_x3): Likewise.
7189 (vst1_p64_x3): Likewise.
7190 (vst1q_s8_x3): Likewise.
7191 (vst1q_p8_x3): Likewise.
7192 (vst1q_s16_x3): Likewise.
7193 (vst1q_p16_x3): Likewise.
7194 (vst1q_s32_x3): Likewise.
7195 (vst1q_s64_x3): Likewise.
7196 (vst1q_u8_x3): Likewise.
7197 (vst1q_u16_x3): Likewise.
7198 (vst1q_u32_x3): Likewise.
7199 (vst1q_u64_x3): Likewise.
7200 (vst1q_f16_x3): Likewise.
7201 (vst1q_f32_x3): Likewise.
7202 (vst1q_f64_x3): Likewise.
7203 (vst1q_p64_x3): Likewise.
7205 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
7207 * config/msp430/msp430.c (msp430_output_labelref): Prepend
7208 user_label_prefix to name.
7210 * tree-core.h: Update comment about the format of NAME string
7211 passed to handler in attribute_spec.
7213 * config/msp430/msp430.md: Remove erroneous subreg expression from
7214 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
7215 zero_extend{q,h}isi2.
7217 2018-05-30 Borislav Petkov <bp@suse.de>
7219 * doc/extend.texi: Document some architecture specific
7220 constraints and sort entries.
7222 2018-05-30 Martin Sebor <msebor@redhat.com>
7225 * builtins.c (expand_builtin_stpcpy_1): New function.
7226 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
7227 only if the former succeeds.
7229 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
7231 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
7234 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
7236 * doc/invoke.texi (-flinker-output): Document
7238 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
7240 * passes.c (ipa_write_summaries): Only modify statements if body
7242 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
7243 incrementally linking.
7244 (ipa_passes): Likewise.
7245 * lto-cgraph.c (lto_output_node): When incrementally linking do not
7246 pass down resolution info.
7247 * common.opt (flag_incremental_link): Update info.
7248 * gcc.c (plugin specs): Turn flinker-output=* to
7249 -plugin-opt=-linker-output-known
7250 * toplev.c (compile_file): Also cut compilation when doing incremental
7252 * flag-types. (enum lto_partition_model): Add
7253 LTO_LINKER_OUTPUT_NOLTOREL.
7254 (invoke.texi): Add -flinker-output docs.
7255 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
7256 link same way as WPA; do not stream in dead initializers.
7258 * dwarf2out.c (dwarf2out_die_ref_for_decl,
7259 darf2out_register_external_decl): Support incremental link.
7261 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
7263 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
7265 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
7267 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
7268 it down to simple_object_copy_lto_debug_sections.
7269 (run_gcc): Determine incremental LTO link time and configure
7270 lto1 into non-wpa mode, disable renaming of debug sections.
7272 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
7274 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
7275 descriptions of various incorrectly documented functions.
7277 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
7280 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
7281 address check not strict.
7283 2018-05-30 Richard Biener <rguenther@suse.de>
7285 PR tree-optimization/85964
7286 * tracer.c (better_p): Drop initialized count check, we only
7287 call the function with initialized counts now.
7288 (find_best_successor): Do find a best edge if one
7289 has uninitialized count.
7290 (find_best_predecessor): Likewise. Do BB frequency check only
7291 if count is initialized.
7293 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
7295 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
7296 (aarch64_ldrstr_offset_compare): New.
7297 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
7298 load/store orderings.
7299 (aarch64_gen_adjusted_ldpstp): Likewise.
7301 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
7303 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
7304 Check for subset of GENERAL_REGS and FP_REGS.
7305 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
7308 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
7310 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
7311 and wi::to_poly_offset. Add the current offset and then check
7312 whether the sum fits, rather than using an unchecked addition of
7313 a checked term. Check for a shwi rather than a uhwi.
7314 * expr.c (get_bit_range): Use tree_to_poly_uint64.
7315 (store_constructor): Use poly_int_tree_p.
7316 (expand_expr_real_1): Likewise.
7317 * function.c (assign_temp): Likewise.
7318 * fold-const.c (const_binop): Use poly_int_tree_p and
7320 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
7322 * ipa-icf-gimple.c (func_checker::compare_operand): Use
7323 to_poly_offset for MEM offsets.
7324 * ipa-icf.c (sem_variable::equals): Likewise.
7325 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
7326 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
7327 wi::to_poly_offset for BIT_FIELD_REF offsets.
7328 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
7330 * var-tracking.c (emit_note_insn_var_location): Use
7331 tree_to_poly_uint64.
7333 2018-05-29 Jim Wilson <jimw@sifive.com>
7335 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
7337 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
7340 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
7341 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
7343 (sse4_1_round<mode>2): Use nonimmediate_operand
7344 for operand 1 predicate.
7346 2018-05-29 Martin Sebor <msebor@redhat.com>
7347 Richard Biener <rguenther@suse.de>
7350 * calls.c (get_size_range): Call determine_value_range instead
7351 of get_value_range..
7352 * tree-vrp.h (determine_value_range): Declared new function.
7353 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
7355 2018-05-29 Richard Biener <rguenther@suse.de>
7357 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
7358 sure to use non-pattern stmts for get_earlier_stmt arguments.
7359 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
7360 called on pattern stmts.
7361 (get_later_stmt): Likewise.
7363 2018-05-29 Martin Liska <mliska@suse.cz>
7365 PR gcov-profile/85759
7366 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
7369 2018-05-29 Jakub Jelinek <jakub@redhat.com>
7371 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
7373 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
7377 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
7378 VEC_PACK_FLOAT_EXPR): New tree codes.
7379 * tree-pretty-print.c (op_code_prio): Handle
7380 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
7381 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
7382 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
7383 * tree-inline.c (estimate_operator_cost): Likewise.
7384 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
7385 * fold-const.c (const_binop): Likewise.
7386 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
7387 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
7388 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
7389 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
7390 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
7391 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
7392 * expr.c (expand_expr_real_2): Likewise.
7393 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
7394 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
7395 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
7397 * optabs.c (expand_widen_pattern_expr): For
7398 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
7399 sign from result type rather than operand's type.
7400 (expand_binop_directly): For vec_packu_float_optab and
7401 vec_packs_float_optab allow result type to be different from operand's
7403 * optabs-tree.c (optab_for_tree_code): Handle
7404 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
7405 VEC_PACK_FLOAT_EXPR. Formatting fixes.
7406 * tree-vect-generic.c (expand_vector_operations_1): Handle
7407 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
7408 VEC_PACK_FLOAT_EXPR.
7409 * tree-vect-stmts.c (supportable_widening_operation): Handle
7411 (supportable_narrowing_operation): Handle FLOAT_EXPR.
7412 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
7413 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
7414 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
7415 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
7417 (vec_pack<floatprefix>_float_<mode>): New expander.
7418 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
7420 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
7421 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
7422 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
7423 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
7424 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
7426 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
7427 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
7428 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
7429 VEC_PACK_FLOAT_EXPR): Document.
7431 2018-05-29 Richard Biener <rguenther@suse.de>
7433 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
7435 (stmt_vec_info_vec): Make pointer.
7436 (init_stmt_vec_info_vec): Remove.
7437 (free_stmt_vec_info_vec): Likewise.
7438 (set_stmt_vec_info_vec): New function.
7439 (free_stmt_vec_infos): Likewise.
7440 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
7441 (set_vinfo_for_stmt): Likewise.
7442 (get_earlier_stmt): Likewise.
7443 (get_later_stmt): Likewise.
7444 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
7445 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
7446 (vec_info::~vec_info): Free stmt_vec_infos.
7447 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
7448 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
7449 (pass_slp_vectorize::execute): Likewise.
7450 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
7451 (free_stmt_vec_info_vec): Likewise.
7452 (set_stmt_vec_info_vec): New function.
7453 (free_stmt_vec_infos): Likewise.
7454 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
7455 the global stmt_vec_info_vec.
7456 * tree-parloops.c (gather_scalar_reductions): Use
7457 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
7460 2018-05-29 Richard Biener <rguenther@suse.de>
7462 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
7464 2018-05-29 Martin Liska <mliska@suse.cz>
7465 David Malcolm <dmalcolm@redhat.com>
7467 * vec.c (test_reverse): New.
7468 (vec_c_tests): Add new test.
7469 * vec.h (vl_ptr>::reverse): New function.
7471 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
7473 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
7475 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
7478 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
7480 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
7482 2018-05-28 Richard Biener <rguenther@suse.de>
7484 PR tree-optimization/85933
7485 * tree-vect-data-refs.c (vect_record_base_alignments): Only
7486 look at stmts marked as vectorizable.
7488 2018-05-28 Richard Biener <rguenther@suse.de>
7490 PR tree-optimization/85934
7491 * tree-vect-generic.c (expand_vector_operations_1): Hoist
7492 vector boolean check before scalar optimization.
7494 2018-05-28 Jakub Jelinek <jakub@redhat.com>
7496 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
7499 2018-05-28 Mark Wielaard <mark@klomp.org>
7501 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
7502 if it is an expression containing a minus sign.
7504 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
7506 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
7508 2018-05-27 Paul Koning <ni1d@arrl.net>
7510 * config/pdp11/pdp11.md (truncsihi2): Remove.
7512 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
7513 Chung-Ju Wu <jasonwucj@gmail.com>
7515 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
7517 (unaligned_store_dw): Ditto.
7518 * config/nds32/nds32-memory-manipulation.c
7519 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
7520 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
7521 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
7522 (emit_setmem_word_loop): Rename to ...
7523 (emit_setmem_doubleword_loop): ... this.
7524 (nds32_gen_dup_4_byte_to_word_value): New function.
7525 (nds32_gen_dup_8_byte_to_double_word_value): New function.
7526 (nds32_expand_setmem_loop): Refine implementation.
7527 (nds32_expand_setmem_loop_v3m): Ditto.
7528 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
7531 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
7533 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
7535 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
7537 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
7538 (nds32_init_machine_status): Initialize machine->attr_naked_p and
7539 machine->attr_no_prologue_p.
7540 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
7541 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
7542 (nds32_expand_epilogue): Consider attr_naked_p.
7543 (nds32_expand_epilogue_v3pop): Likewise.
7544 (nds32_can_use_return_insn): Likewise.
7545 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
7546 attr_no_prologue_p fields.
7547 * config/nds32/nds32.opt (mret-in-naked-func): New option.
7549 2018-05-27 Jakub Jelinek <jakub@redhat.com>
7552 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
7554 * config/i386/sse.md
7555 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
7557 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
7559 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
7561 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
7563 (*<floatsuffix>floatv2div2sf2): Rename to ...
7564 (*float<floatunssuffix>v2div2sf2): ... this.
7565 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
7566 (float<floatunssuffix>v2div2sf2_mask): ... this.
7567 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
7568 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
7569 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
7571 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
7573 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
7575 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
7577 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
7579 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
7581 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
7582 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
7583 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
7584 gen_ufix_truncv8dfv8si2.
7585 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
7586 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
7587 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
7588 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
7589 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
7590 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
7591 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
7592 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
7594 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
7598 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
7600 2018-05-25 Jim Wilson <jimw@sifive.com>
7602 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
7603 * config/riscv/riscv.c (struct machine_function): Add
7604 interrupt_handler_p and attribute_checked_p fields.
7605 (riscv_attribute_table): Add interrupt.
7606 (riscv_interrupt_type_p): New.
7607 (riscv_save_reg_p): Save extra regs for interrupt handler.
7608 (riscv_use_save_libcall): Return false for interrupt handler.
7609 (riscv_first_stack_step): Add forward declaration.
7610 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
7611 for interrupt handler with large frame. Use it for saved reg list.
7612 (riscv_expand_prologue): Move flag_stack_usage_info support to
7613 eliminate duplication.
7614 (riscv_expand_epilogue): Generate mret for interrupt handler.
7615 (riscv_epilogue_uses): New.
7616 (riscv_can_use_return_insn): Return false for interrupt handler.
7617 (riscv_function_ok_for_sibcall): Likewise.
7618 (riscv_set_current_function): Add interrupt handler support.
7619 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
7620 * config/riscv/riscv.md (UNSPECV_MRET): New.
7622 (riscv_frflags, riscv_fsflags): Use tab after opcode.
7624 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
7626 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
7628 PR tree-optimization/85712
7629 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
7630 this candidate has already been replaced in-situ by a copy.
7632 2018-05-25 Jason Merrill <jason@redhat.com>
7634 PR c++/80485 - inline function non-zero address.
7635 * symtab.c (nonzero_address): Check DECL_COMDAT.
7637 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
7640 * config/alpha/alpha.md (ashlsi3): New insn pattern.
7641 (*ashlsi_se): Rename from *ashldi_se. Define as sign
7642 extension of SImode operation. Use const123_operand predicate.
7643 (*saddsi_1): Remove.
7644 (*saddl_se_1): Ditto.
7646 (*ssubl_se_1): Ditto.
7647 * config/alpha/predicates.md (const123_operand): New predicate.
7648 * config/alpha/constraints.md (P): Use IN_RANGE.
7650 2018-05-25 Richard Biener <rguenther@suse.de>
7652 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
7654 (ref_maybe_used_by_stmt_p): Likewise.
7655 (stmt_may_clobber_ref_p): Likewise.
7656 (stmt_may_clobber_ref_p_1): Likewise.
7657 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
7659 (ref_maybe_used_by_stmt_p): Likewise.
7660 (stmt_may_clobber_ref_p): Likewise.
7661 (stmt_may_clobber_ref_p_1): Likewise.
7662 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
7663 the alias oracle to disambiguate DRs with stmts DR analysis
7665 (vect_analyze_data_refs): Do not give up on not analyzable
7666 DRs for BB vectorization. Remove code truncating the dataref
7669 2018-05-25 Jakub Jelinek <jakub@redhat.com>
7672 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
7673 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
7674 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
7676 2018-05-25 Richard Biener <rguenther@suse.de>
7678 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
7679 function, combining stmt data ref gathering and fatal analysis
7681 (vect_analyze_data_refs): Remove now redudnant code and simplify.
7682 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
7683 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
7684 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
7685 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
7687 2018-05-25 Bin Cheng <bin.cheng@arm.com>
7689 PR tree-optimization/85720
7690 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
7691 SCC if all partitions are builtins.
7692 (version_loop_by_alias_check): New parameter. Generate cancelable
7693 runtime alias check if all partitions are builtins.
7694 (distribute_loop): Update call to above function.
7696 2018-05-25 Bin Cheng <bin.cheng@arm.com>
7698 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
7699 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
7700 (parm_default_def_partition_arg): Ditto.
7701 (set_parm_default_def_partition): Ditto.
7702 (get_parm_default_def_partitions): Ditto and make it static.
7703 (get_undefined_value_partitions): Ditto and make it static.
7704 (remove_ssa_form): Refactor call to init_var_map here.
7705 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
7706 computation for loop region.
7707 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
7708 (register_default_def): Delete.
7709 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
7710 (parm_default_def_partition_arg): Ditto.
7711 (set_parm_default_def_partition): Ditto.
7712 (get_parm_default_def_partitions): Ditto and make it static.
7713 (get_undefined_value_partitions): Ditto and make it static.
7714 (coalesce_with_default, coalesce_with_default): Update comment.
7715 (create_coalesce_list_for_region): New func factored out from
7716 create_outofssa_var_map.
7717 (populate_coalesce_list_for_outofssa): New func factored out from
7718 create_outofssa_var_map and coalesce_ssa_name.
7719 (create_outofssa_var_map): Delete.
7720 (coalesce_ssa_name): Refactor to support live range computation.
7721 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
7722 (get_parm_default_def_partitions): Delete.
7723 (get_undefined_value_partitions): Ditto.
7724 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
7725 computation for loop region.
7726 (new_tree_live_info, loe_visit_block): Ditto.
7727 (live_worklist, set_var_live_on_entry): Ditto.
7728 (calculate_live_on_exit, verify_live_on_entry): Ditto.
7729 * tree-ssa-live.h (struct _var_map): New fields.
7730 (init_var_map): Change decl.
7731 (region_contains_p): New.
7733 2018-05-25 Bin Cheng <bin.cheng@arm.com>
7735 * tree-ssa-live.h (live_merge_and_clear): Delete.
7737 2018-05-25 Richard Biener <rguenther@suse.de>
7740 * tree-dump.c (dequeue_and_dump): Remove access to removed
7741 operand 2 of a SWITCH_EXPR.
7743 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
7745 * doc/sourcebuild.texi (vect_double_cond_arith): Include
7746 multiplication and division.
7747 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
7748 (cond_udiv@var{m}, cond_umod@var{m}): Document.
7749 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
7750 (cond_udiv_optab, cond_umod_optab): New optabs.
7751 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
7752 (IFN_COND_RDIV): New internal functions.
7753 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
7754 TRUNC_MOD_EXPR and RDIV_EXPR.
7755 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
7756 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
7758 (SVE_INT_BINARY): Include mult.
7759 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
7760 (optab, sve_int_op): Handle mult.
7761 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
7763 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
7764 for SVE_INT_BINARY_SD.
7766 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
7768 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
7769 (optab, sve_int_op): Handle div and udiv.
7770 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
7771 for SVE_INT_BINARY_SD.
7772 (*<optab><mode>3): New insn for the same.
7774 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
7776 * tree-vect-patterns.c: Include predict.h.
7777 (vect_recog_divmod_pattern): Restrict check for division support
7778 to when optimizing for size.
7780 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
7782 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
7783 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
7784 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
7785 (gimple_match_op::set_op): Likewise.
7786 (gimple_resimplify4): Declare.
7787 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
7788 (expr::gen_transform): Likewise.
7789 (decision_tree::gen): Generate a simplification routine for 4 operands.
7790 * gimple-match-head.c (gimple_simplify): Add an overload for
7791 4 operands. In the top-level function, handle up to 4 call
7792 arguments and call gimple_resimplify4.
7793 (gimple_resimplify4): New function.
7794 (build_call_internal): Pass a fourth operand.
7795 (maybe_push_to_seq): Likewise.
7796 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
7797 Fold VEC_COND_EXPRs of an operation and a default value into
7798 an IFN_COND_* function if possible.
7799 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
7801 (SVE_COND_FP_BINARY): Include them.
7802 (optab, sve_fp_op): Handle them.
7803 (SVE_INT_BINARY_REV): New code iterator.
7804 (SVE_COND_FP_BINARY_REV): New int iterator.
7805 (commutative): New int attribute.
7806 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
7808 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
7810 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
7811 (*cond_<optab><mode>): New patterns for reversed operands.
7813 2018-05-25 Richard Biener <rguenther@suse.de>
7815 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
7816 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
7817 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
7818 (STMT_VINFO_GROUPED_ACCESS): Adjust.
7819 * tree-vect-data-refs.c (everywhere): Adjust users.
7820 * tree-vect-loop.c (everywhere): Likewise.
7821 * tree-vect-slp.c (everywhere): Likewise.
7822 * tree-vect-stmts.c (everywhere): Likewise.
7823 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
7825 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7827 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
7829 (gcc_cv_as_section_exclude): ... this.
7830 Try Solaris as #exclude syntax.
7831 * configure: Regenerate.
7832 * config.in: Regenerate.
7833 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
7835 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
7836 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
7838 * varasm.c (default_elf_asm_named_section): Don't check if
7839 HAVE_GAS_SECTION_EXCLUDE is defined.
7841 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
7843 * doc/md.texi: Update the documentation of the cond_* optabs
7844 to mention the new final operand. Fix GET_MODE_NUNITS call.
7845 Describe the scalar case too.
7846 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
7847 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
7849 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
7850 (get_conditional_internal_fn): Update comment.
7851 * tree-vect-loop.c (vectorizable_reduction): Pass the original
7852 accumulator value as a final argument to conditional functions.
7853 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
7854 a define_expand and add an "else" operand. Assert for now that
7855 the else operand is equal to operand 2. Use SVE_INT_BINARY and
7856 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
7857 (*cond_<optab><mode>): New patterns.
7858 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
7859 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
7860 (UNSPEC_COND_EOR): Delete.
7861 (optab): Remove associated mappings.
7862 (SVE_INT_BINARY): New code iterator.
7863 (sve_int_op): Remove int attribute and add "minus" to the code
7865 (SVE_COND_INT_OP): Delete.
7866 (SVE_COND_FP_OP): Rename to...
7867 (SVE_COND_FP_BINARY): ...this.
7869 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
7871 * optabs.c (can_reuse_operands_p): New function.
7872 (maybe_legitimize_operands): Try to reuse the results for
7875 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
7877 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
7878 Add {q} suffix to insn mnemonic.
7880 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
7882 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
7883 (msp430_warn_func_return): New.
7885 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
7887 * fold-const.c (tree_nonzero_bits): New function.
7888 * fold-const.h (tree_nonzero_bits): Likewise.
7889 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
7890 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
7892 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
7896 * varasm.c (assemble_alias): Check ifunc_resolver only on
7899 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
7902 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
7903 when memory input operand is handled.
7905 2018-05-24 Luis Machado <luis.machado@linaro.org>
7907 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
7909 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
7911 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
7913 * match.pd: Delay FMA folds until after vectorization.
7915 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
7918 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
7919 address check not strict.
7921 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
7923 * gimple-match.h (gimple_match_op): New class.
7924 (mprts_hook): Replace parameters with a gimple_match_op *.
7925 (maybe_build_generic_op): Likewise.
7926 (gimple_simplified_result_is_gimple_val): Replace parameters with
7927 a const gimple_match_op *.
7928 (gimple_simplify): Replace code_helper * and tree * parameters with
7929 a gimple_match_op * parameter.
7930 (gimple_resimplify1): Replace code_helper *, tree and tree *
7931 parameters with a gimple_match_op * parameter.
7932 (gimple_resimplify2): Likewise.
7933 (gimple_resimplify3): Likewise.
7934 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
7935 parameters with a gimple_match_op * parameter.
7936 * gimple-match-head.c (gimple_simplify): Change prototypes of
7937 auto-generated functions to take a gimple_match_op * instead of
7938 separate code_helper * and tree * parameters. Make the same
7939 change in the top-level overload and update calls to the
7940 gimple_resimplify routines. Update calls to the auto-generated
7941 functions and to maybe_push_res_to_seq in the publicly-facing
7942 operation-specific gimple_simplify overloads.
7943 (gimple_match_op::MAX_NUM_OPS): Define.
7944 (gimple_resimplify1): Replace rcode and ops with a single res_op
7945 parameter. Update call to gimple_simplify.
7946 (gimple_resimplify2): Likewise.
7947 (gimple_resimplify3): Likewise.
7948 (mprts_hook): Replace parameters with a gimple_match_op *.
7949 (maybe_build_generic_op): Likewise.
7950 (build_call_internal): Replace type, nargs and ops with
7951 a gimple_match_op *.
7952 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
7953 with a single gimple_match_op *. Update calls to mprts_hook,
7954 build_call_internal and gimple_simplified_result_is_gimple_val.
7955 Factor out code that is common to the tree_code and combined_fn cases.
7956 * genmatch.c (expr::gen_transform): Replace tem_code and
7957 tem_ops with a gimple_match_op called tem_op. Update calls
7958 to the gimple_resimplify functions and maybe_push_res_to_seq.
7959 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
7960 res_ops. Update call to the gimple_resimplify functions.
7961 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
7962 (decision_tree::gen): Make the functions take a gimple_match_op *
7963 called res_op instead of separate res_code and res_ops parameters.
7964 Update call accordingly.
7965 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
7966 and ops with a single res_op parameter. Update calls to
7967 maybe_build_generic_op and maybe_push_res_to_seq.
7968 (fold_stmt_1): Update calls to gimple_simplify and
7969 replace_stmt_with_simplification.
7970 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
7971 and gimple_simplified_result_is_gimple_val.
7972 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
7974 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
7975 with a gimple_match_op *.
7976 (vn_nary_build_or_lookup): Likewise. Update call to
7977 vn_nary_build_or_lookup_1.
7978 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
7979 gimple_match_op *. Update calls to the gimple_resimplify routines
7980 and to gimple_simplified_result_is_gimple_val.
7981 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
7982 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
7983 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
7984 (visit_nary_op): Likewise.
7985 (visit_reference_op_load): Likewise.
7987 2018-05-23 Luis Machado <luis.machado@linaro.org>
7989 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
7990 modifier for printing the step amount.
7992 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
7995 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
7998 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
8000 * doc/sourcebuild.texi (Endianness): New subsubsection.
8002 2018-05-23 Luis Machado <luis.machado@linaro.org>
8004 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
8005 <prefetch_dynamic_strides>: New const bool field.
8006 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
8007 prefetch_dynamic_strides.
8008 (exynosm1_prefetch_tune): Likewise.
8009 (thunderxt88_prefetch_tune): Likewise.
8010 (thunderx_prefetch_tune): Likewise.
8011 (thunderx2t99_prefetch_tune): Likewise.
8012 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
8014 (aarch64_override_options_internal): Update to set
8015 PARAM_PREFETCH_DYNAMIC_STRIDES.
8016 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
8017 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
8018 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
8019 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
8020 prefetch-dynamic-strides setting.
8022 2018-05-23 Luis Machado <luis.machado@linaro.org>
8024 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
8025 <minimum_stride>: New const int field.
8026 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
8027 minimum_stride field defaulting to -1.
8028 (exynosm1_prefetch_tune): Likewise.
8029 (thunderxt88_prefetch_tune): Likewise.
8030 (thunderx_prefetch_tune): Likewise.
8031 (thunderx2t99_prefetch_tune): Likewise.
8032 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
8033 <default_opt_level>: Set to 3.
8034 (aarch64_override_options_internal): Update to set
8035 PARAM_PREFETCH_MINIMUM_STRIDE.
8036 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
8037 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
8038 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
8039 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
8040 stride is constant and is below the minimum stride threshold.
8042 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8044 * config/arm/arm-cpus.in (mode26): Delete.
8045 (armv4): Delete mode26 reference.
8046 * config/arm/arm.c (arm_configure_build_target): Delete use of
8049 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
8051 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
8053 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
8054 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
8056 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
8057 (floatunsdidf2): Ditto.
8059 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
8061 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
8062 (fixuns_trunc<mode>si2_avx512f): Ditto.
8063 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
8064 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
8065 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
8067 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
8069 PR rtl-optimization/79985
8070 * df-scan.c (df_insn_refs_collect): Remove special case for
8071 global registers and asm statements.
8073 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
8075 * extend.texi (Global Register Variables): Rewrite the bullet list.
8076 Note that the register is available for allocation. Note that access
8077 via inline asm must use constraints. Add note about async-signal
8078 handlers. Remove paragraph about automagic register selection.
8080 2018-05-23 Richard Biener <rguenther@suse.de>
8082 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
8083 of fixed offset from memset VN.
8085 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
8087 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
8089 (alloc_cand_and_find_basis): Initialize first_interp field.
8090 (slsr_process_mul): Modify first_interp field.
8091 (slsr_process_add): Likewise.
8092 (slsr_process_cast): Modify first_interp field for each new
8094 (slsr_process_copy): Likewise.
8095 (dump_candidate): Dump first_interp field.
8096 (replace_mult_candidate): Process all interpretations, not just
8098 (replace_rhs_if_not_dup): Likewise.
8099 (replace_one_candidate): Likewise.
8101 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
8103 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
8105 (aarch64_needs_frame_chain): New function.
8106 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
8108 2018-05-23 Sudakshina Das <sudi.das@arm.com>
8111 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
8112 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
8113 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
8114 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
8115 as true for strict-align.
8116 (aarch64_can_inline_p): Perform checks even when callee has no
8117 attributes to check for strict alignment.
8118 * doc/extend.texi (AArch64 Function Attributes): Document
8120 * doc/invoke.texi: (AArch64 Options): Likewise.
8122 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
8124 PR tree-optimization/85853
8125 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
8126 the handling of the root of the node to...
8127 (vect_slp_analyze_node_operations_1): ...this new function,
8128 and run the whole thing with the child nodes' def types
8129 set according to their SLP node's def type.
8131 2018-05-23 Richard Biener <rguenther@suse.de>
8134 * tree-data-ref.c (create_runtime_alias_checks): Defer
8135 and ignore overflow warnings.
8137 2018-05-23 Yury Gribov <tetra2005@gmail.com>
8139 PR tree-optimization/85822
8140 * tree-vrp.c (is_masked_range_test): Fix handling of negative
8143 2018-05-23 Richard Biener <rguenther@suse.de>
8145 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
8146 memset constants via native_interpret_expr.
8148 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
8151 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
8153 (cgraph_node::create_alias): Likewise.
8154 (cgraph_node::get_availability): Check ifunc_resolver instead
8155 of looking up ifunc attribute.
8156 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
8157 * varasm.c (do_assemble_alias): Likewise.
8158 (assemble_alias): Likewise.
8159 (default_binds_local_p_3): Likewise.
8160 * cgraph.h (cgraph_node): Add ifunc_resolver.
8161 (cgraph_node::only_called_directly_or_aliased_p): Return false
8163 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
8165 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
8166 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
8167 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
8168 instead of looking up ifunc attribute.
8170 2018-05-22 Luis Machado <luis.machado@linaro.org>
8172 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
8174 2018-05-22 Martin Sebor <msebor@redhat.com>
8177 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
8178 only when expasion succeeds.
8179 (expand_builtin_strcmp): Same.
8180 (expand_builtin_strncmp): Same.
8182 2018-05-22 Martin Sebor <msebor@redhat.com>
8184 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
8186 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
8187 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8189 * config/aarch64/aarch64-ldpstp.md: Replace uses of
8190 aarch64_mem_pair_operand with memory_operand and delete operand swapping
8192 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
8193 Add check for legitimate_address.
8194 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
8195 (aarch64_swap_ldrstr_operands): New.
8196 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
8199 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
8200 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8202 * config/aarch64/aarch64.md: New patterns to generate stp
8204 (store_pair_sw, store_pair_dw): New patterns to generate stp for
8205 single words and double words.
8206 (load_pair_sw, load_pair_dw): Likewise.
8207 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
8209 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
8211 * config/aarch64/aarch64-ldpstp.md: Modify peephole
8212 for different mode ldpstp and add peephole for merged zero stores.
8214 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
8216 (aarch64_gen_store_pair): Rename calls to match new patterns.
8217 (aarch64_gen_load_pair): Rename calls to match new patterns.
8218 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
8219 (load_pair<DREG:mode><DREG2:mode>): ... This.
8220 (store_pair<mode>): Rename to...
8221 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
8222 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
8224 (V_INT_EQUIV): Handle SImode.
8225 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
8228 2018-05-22 Martin Sebor <msebor@redhat.com>
8231 * calls.c (maybe_warn_nonstring_arg): Use string length to set
8232 or ajust the presumed bound on an operation to avoid unnecessary
8235 2018-05-22 Martin Sebor <msebor@redhat.com>
8237 PR tree-optimization/85826
8238 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
8239 assuming that a DECL necesarily has a constant size.
8241 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
8244 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
8246 2018-05-22 Richard Biener <rguenther@suse.de>
8248 PR tree-optimization/85834
8249 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
8250 non-constant and non-zero memset arguments.
8252 2018-05-22 Martin Liska <mliska@suse.cz>
8255 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
8257 2018-05-22 Richard Biener <rguenther@suse.de>
8259 PR tree-optimization/85863
8260 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
8261 comparisons when vectype is specified.
8262 (vectorizable_condition): Do not specify vectype for
8263 vect_is_simple_cond when SLP vectorizing.
8265 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
8268 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
8269 define __ibm128 as long double.
8270 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
8271 as a distinct type when IEEE 128-bit support is enabled.
8272 (init_float128_ieee): Fix up conversions between IFmode and IEEE
8273 128-bit types to use the correct functions.
8274 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
8275 convert between 128-bit floating point types that have different
8276 modes but the same representation, instead of using gen_lowpart to
8278 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
8280 (IFKF_reg): New attributes to give the register constraints for
8282 (extend<mode>tf2_internal): New insns to mark an explicit
8283 conversion between 128-bit floating point types that have a
8284 different mode but share the same representation.
8286 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
8288 PR tree-optimization/85814
8289 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
8290 a null return from get_strinfo when unsharing the next
8291 strinfo in the chain.
8293 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
8296 * varasm.c (weak_finish): Clean up weak_decls.
8298 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8300 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
8301 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
8302 UNSPEC_UADALP values.
8303 * config/aarch64/iterators.md (ABAL): New int iterator.
8306 (sur): Add mappings for the above.
8307 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
8309 (aarch64_<sur>abal<mode>_4): Likewise.
8310 (aarch64_<sur>adalp<mode>_3): Likewise.
8311 (<sur>sadv16qi): New define_expand.
8313 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
8315 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
8316 (*movdf_internal): Ditto.
8317 (*rcpsf2_sse): Ditto.
8318 (*rsqrtsf2_sse): Ditto.
8319 (*sqrt<mode>2_sse): Ditto.
8321 2018-05-21 Tamar Christina <tamar.christina@arm.com>
8323 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
8325 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
8326 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
8327 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
8328 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
8330 * config/aarch64/arm_neon.h: Likewise.
8331 * config/aarch64/iterators.md (VQ_I): New.
8333 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
8335 * config.gcc: Add arc/t-multilib-linux to tmake_file for
8337 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
8340 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
8342 * config/nds32/constraints.md (S): New constraint.
8343 * config/nds32/nds32.md (call_internal): Use constraint S.
8344 (call_value_internal): Likewise.
8345 (sibcall_internal): Likewise.
8346 (sibcall_value_internal): Likewise.
8348 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
8349 Chung-Ju Wu <jasonwucj@gmail.com>
8351 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
8354 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
8355 Chung-Ju Wu <jasonwucj@gmail.com>
8357 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
8358 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
8359 (nds32_rtx_costs_impl): Simplify.
8360 (nds32_address_cost_impl): Simplify.
8361 (nds32_init_rtx_costs): New function.
8362 (nds32_rtx_costs_speed_prefer): Likewise.
8363 (nds32_rtx_costs_size_prefer): Likewise.
8364 (nds32_address_cost_speed_prefer): Likewise.
8365 (nds32_address_cost_speed_fwprop): Likewise.
8366 (nds32_address_cost_size_prefer): Likewise.
8367 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
8368 * config/nds32/nds32.c (nds32_option_override): Use
8369 nds32_init_rtx_costs function.
8371 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
8373 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
8374 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
8375 (TARGET_PIPELINE_N8): Likewise.
8376 (TARGET_PIPELINE_N10): Likewise.
8377 (TARGET_PIPELINE_N13): Likewise.
8378 (TARGET_PIPELINE_GRAYWOLF): Likewise.
8380 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
8382 * config/nds32/nds32-fpu.md: Update copyright year.
8384 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
8386 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
8388 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
8390 * config/nds32/nds32.c
8391 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
8392 * config/nds32/nds32.opt (minline-asm-r15): New option.
8394 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
8396 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
8398 * config/nds32/nds32.md (abssi2): New pattern.
8400 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
8402 * config/i386/i386.md (rex64namesuffix): New mode attribute.
8403 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
8404 Merge insn pattern from sse_cvtsi2ss<round_name> and
8405 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
8406 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
8407 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
8408 using SWI48 mode iterator.
8409 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
8410 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
8411 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
8412 pattern from sse_cvttss2si<round_saeonly_name>
8413 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
8414 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
8415 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
8416 using SWI48 mode iterator.
8417 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
8418 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
8419 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
8420 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
8421 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
8422 using SWI48 mode iterator.
8423 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
8424 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
8425 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
8426 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
8427 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
8428 SWI48 mode iterator.
8429 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
8430 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
8431 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
8432 pattern from sse_cvttsd2si<round_saeonly_name>
8433 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
8435 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
8437 * config/nds32/nds32-md-auxiliary.c
8438 (nds32_valid_smw_lwm_base_p): Refine.
8439 (nds32_output_smw_single_word): Refine.
8440 (nds32_output_smw_double_word): New.
8441 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
8443 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
8445 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
8446 (nds32_output_stack_pop): Refine.
8447 (nds32_expand_unaligned_load): Refine.
8448 (nds32_expand_unaligned_store): Refine.
8450 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
8451 Chung-Ju Wu <jasonwucj@gmail.com>
8453 * config/nds32/constants.md: Add TP_REGNUM constant.
8454 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
8455 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
8457 * config/nds32/nds32-doubleword.md: Consider flag_pic.
8458 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
8459 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
8460 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
8461 and PIC code generation.
8462 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
8464 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
8466 * config/nds32/nds32.md: Support TLS and PIC.
8467 * config/nds32/nds32.c: Support TLS and PIC.
8468 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
8469 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
8472 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
8474 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
8475 mode with E_ prefix.
8477 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
8478 Chung-Ju Wu <jasonwucj@gmail.com>
8480 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
8481 * config/nds32/nds32-md-auxiliary.c
8482 (symbolic_reference_mentioned_p): New.
8483 (nds32_legitimize_ict_address): New.
8484 (nds32_expand_ict_move): New.
8485 (nds32_indirect_call_referenced_p): New.
8486 (nds32_symbol_binds_local_p): Delete.
8487 (nds32_long_call_p): Modify.
8488 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
8489 * config/nds32/nds32-protos.h
8490 (symbolic_reference_mentioned_p): Declare.
8491 (nds32_legitimize_ict_address): Declare.
8492 (nds32_expand_ict_move): Declare.
8493 (nds32_indirect_call_referenced_p): Declare.
8494 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
8495 (nds32_relax_group): Use nds32_ict_const_p as condition.
8496 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
8497 (nds32_asm_file_start): Output ict_model directive in asm code.
8498 (nds32_legitimate_address_p): Consider indirect call.
8499 (nds32_print_operand): Consider indirect call.
8500 (nds32_print_operand_address): Consider indirect call.
8501 (nds32_insert_attributes): Handle "indirect_call" attribute.
8502 (TARGET_LEGITIMATE_ADDRESS_P): Define.
8503 (TARGET_LEGITIMATE_CONSTANT_P): Define.
8504 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
8505 (TARGET_DELEGITIMIZE_ADDRESS): Define.
8506 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
8507 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
8508 (TARGET_ICT_MODEL_SMALL): Define.
8509 (TARGET_ICT_MODEL_LARGE): Define.
8510 * config/nds32/nds32.md (movsi): Consider ict model.
8511 (call, call_value): Consider ict model.
8512 (sibcall, sibcall_value): Consider ict model.
8513 * config/nds32/nds32.opt (mict-model): New option.
8514 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
8517 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
8518 Monk Chiang <sh.chiang04@gmail.com>
8519 Jim Wilson <jimw@sifive.com>
8521 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
8522 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
8523 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
8524 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
8525 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
8526 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
8527 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
8528 compute save_libcall_adjustment properly.
8529 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
8530 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
8531 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
8532 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
8533 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
8534 (ABI_SPEC): Handle mabi=ilp32e.
8535 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
8536 (RVE): Add RVE mask.
8537 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
8538 <-march>: Add rv32e as an example.
8540 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
8543 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
8544 (intra_create_variable_infos): Handle C++ constructors.
8546 2018-05-18 Martin Liska <mliska@suse.cz>
8548 * passes.def: Remove a redundant pass.
8550 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
8553 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
8555 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8557 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
8559 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
8561 (armv2, armv2a, armv3, armv3m): Delete architectures.
8562 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
8563 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
8564 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
8566 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
8567 (*mulsidi3adddi): Likewise.
8568 (mulsidi3): Likewise.
8569 (*mulsidi3_nov6): Likewise.
8570 (umulsidi3): Likewise.
8571 (umulsidi3_nov6): Likewise.
8572 (umaddsidi4): Likewise.
8573 (*umulsidi3adddi): Likewise.
8574 (smulsi3_highpart): Likewise.
8575 (*smulsi3_highpart_nov6): Likewise.
8576 (umulsi3_highpart): Likewise.
8577 (*umulsi3_highpart_nov6): Likewise.
8578 * config/arm/arm.h (arm_arch3m): Delete.
8579 * config/arm/arm.c (arm_arch3m): Delete.
8580 (arm_option_override_internal): Update armv3-related comment.
8581 (arm_configure_build_target): Delete use of isa_bit_mode32.
8582 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
8583 (arm_rtx_costs_internal): Delete check of arm_arch3m.
8584 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
8586 (mulusa3): Likewise.
8587 * config/arm/arm-protos.h (arm_arch3m): Delete.
8588 * config/arm/arm-tables.opt: Regenerate.
8589 * config/arm/arm-tune.md: Likewise.
8590 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
8591 deleted architectures.
8593 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8595 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
8596 (armv5t, armv5te): New features.
8597 (ARMv5, ARMv5e): Delete fgroups.
8598 (ARMv5t, ARMv5te): Adjust for above changes.
8600 (armv5, armv5e): Delete arches.
8601 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
8603 (*call_reg_arm): Likewise.
8604 (*call_value_reg_armv5): Likewise.
8605 (*call_value_reg_arm): Likewise.
8606 (*call_symbol): Likewise.
8607 (*call_value_symbol): Likewise.
8608 (*sibcall_insn): Likewise.
8609 (*sibcall_value_insn): Likewise.
8611 (prefetch): Likewise.
8612 (define_split and define_peephole2 dependent on arm_arch5):
8614 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
8616 (TARGET_ARM_QBIT): Likewise.
8617 (TARGET_DSP_MULTIPLY): Likewise.
8618 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
8619 (arm_arch5, arm_arch5e): Delete.
8620 (arm_arch5t, arm_arch5te): Declare.
8621 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
8622 (arm_arch5t): Declare.
8623 (arm_option_reconfigure_globals): Update for the above.
8624 (arm_options_perform_arch_sanity_checks): Update comment, replace
8625 use of arm_arch5 with arm_arch5t.
8626 (use_return_insn): Likewise.
8627 (arm_emit_call_insn): Likewise.
8628 (output_return_instruction): Likewise.
8629 (arm_final_prescan_insn): Likewise.
8630 (arm_coproc_builtin_available): Likewise.
8631 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
8632 arm_arch5e with arm_arch5t and arm_arch5te.
8633 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
8634 (arm_arch5t, arm_arch5te): Declare.
8635 * config/arm/arm-tables.opt: Regenerate.
8636 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
8637 * config/arm/t-multilib: Likewise.
8638 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
8639 instead of arm_arch5.
8640 (*call_reg_thumb1): Likewise.
8641 (*call_value_reg_thumb1_v5): Likewise.
8642 (*call_value_reg_thumb1): Likewise.
8643 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
8645 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
8647 2018-05-18 Martin Liska <mliska@suse.cz>
8649 PR gcov-profile/84846
8650 * doc/gcov.texi: Document -t option of gcov tool.
8652 2018-05-18 Martin Liska <mliska@suse.cz>
8654 PR gcov-profile/84846
8655 * gcov.c (print_usage): Add new -t option.
8656 (process_args): Handle the option.
8657 (generate_results): Use stdout as output when requested by
8660 2018-05-18 Martin Liska <mliska@suse.cz>
8662 PR gcov-profile/84846
8663 * coverage.c (coverage_init): Write PWD to .gcno file.
8664 * doc/gcov.texi: Document how working directory is printed.
8665 * gcov-dump.c (dump_gcov_file): Print PWD.
8666 * gcov.c (output_intermediate_file): Likewise.
8667 (read_graph_file): Read PWD string.
8668 (output_lines): Print PWD.
8670 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8673 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
8674 for retval and return false if all args to phi are zero.
8676 2018-05-18 Richard Biener <rguenther@suse.de>
8678 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
8680 (evrp_dom_walker::before_dom_children): Call it.
8682 2018-05-18 Richard Biener <rguenther@suse.de>
8684 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
8685 results when processing array refs with variable index.
8687 2018-05-18 Toon Moene <toon@moene.org>
8689 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
8690 directly after that of -floop-interchange. Indicate that both
8691 options are enabled by default when specifying -O3.
8693 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8695 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
8696 iterator. Delete separate integer-mode vec_set<mode> expander.
8697 (aarch64_simd_vec_setv2di): Delete.
8698 (vec_setv2di): Delete.
8699 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
8700 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
8701 the "w, r" alternative.
8703 2018-05-18 Martin Liska <mliska@suse.cz>
8705 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
8706 * tree-pass.h (make_pass_lower_switch_O0): New function.
8707 * tree-switch-conversion.c (node_has_low_bound): Remove.
8708 (node_has_high_bound): Likewise.
8709 (node_is_bounded): Likewise.
8710 (class pass_lower_switch): Make it a template type and create
8712 (pass_lower_switch::execute): Add template argument.
8713 (make_pass_lower_switch): New function.
8714 (make_pass_lower_switch_O0): New function.
8715 (do_jump_if_equal): Remove.
8716 (emit_case_nodes): Simplify to just handle all 3 cases and leave
8717 all the hard work to tree optimization passes.
8719 2018-05-18 Martin Liska <mliska@suse.cz>
8721 * dbgcnt.c (limit_low): Renamed from limit.
8722 (limit_high): New variable.
8723 (dbg_cnt_is_enabled): Check for upper limit.
8724 (dbg_cnt): Adjust dumping.
8725 (dbg_cnt_set_limit_by_index): Add new argument for high
8727 (dbg_cnt_set_limit_by_name): Likewise.
8728 (dbg_cnt_process_single_pair): Parse new format.
8729 (dbg_cnt_process_opt): Use strtok.
8730 (dbg_cnt_list_all_counters): Remove 'value' and add
8732 * doc/invoke.texi: Document changes.
8734 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
8736 * doc/sourcebuild.texi (scalar_all_fma): Document.
8737 * tree.def (FMA_EXPR): Delete.
8738 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
8739 * internal-fn.c (ternary_direct): New macro.
8740 (expand_ternary_optab_fn): Likewise.
8741 (direct_ternary_optab_supported_p): Likewise.
8742 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
8743 * builtins.c (fold_builtin_fma): Delete.
8744 (fold_builtin_3): Don't call it.
8745 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
8746 * expr.c (expand_expr_real_2): Likewise.
8747 * fold-const.c (operand_equal_p): Likewise.
8748 (fold_ternary_loc): Likewise.
8749 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
8750 * gimple.c (DEFTREECODE): Likewise.
8751 * gimplify.c (gimplify_expr): Likewise.
8752 * optabs-tree.c (optab_for_tree_code): Likewise.
8753 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
8754 * tree-eh.c (operation_could_trap_p): Likewise.
8755 (stmt_could_throw_1_p): Likewise.
8756 * tree-inline.c (estimate_operator_cost): Likewise.
8757 * tree-pretty-print.c (dump_generic_node): Likewise.
8758 (op_code_prio): Likewise.
8759 * tree-ssa-loop-im.c (stmt_cost): Likewise.
8760 * tree-ssa-operands.c (get_expr_operands): Likewise.
8761 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
8762 * fold-const-call.h (fold_fma): Delete.
8763 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
8764 CFN_FNMA and CFN_FNMS.
8766 * genmatch.c (combined_fn): New enum.
8767 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
8768 (commutative_op): New function.
8769 (commutate): Use it. Handle more than 2 operands.
8770 (dt_operand::gen_gimple_expr): Use commutative_op.
8771 (parser::parse_expr): Allow :c to be used with non-binary
8772 operators if the commutative operand is known.
8773 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
8774 CFN_FMS, CFN_FNMA and CFN_FNMS.
8775 (backprop::process_assign_use): Remove FMA_EXPR handling.
8776 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
8777 (gen_hsa_fma): New function.
8778 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
8779 IFN_FNMA and IFN_FNMS.
8780 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
8781 * gimple-fold.h (follow_all_ssa_edges): Declare.
8782 * gimple-fold.c (follow_all_ssa_edges): New function.
8783 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
8784 gimple_build interface and use follow_all_ssa_edges to fold the result.
8785 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
8786 instead of checking for optabs directly.
8787 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
8788 rather than FMA_EXPRs.
8789 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
8790 call to IFN_FMA instead of an FMA_EXPR.
8792 2018-05-17 Jim Wilson <jimw@sifive.com>
8794 * expr.c (do_tablejump): When converting index to Pmode, if we have a
8795 sign extended promoted subreg, and the range does not have the sign bit
8796 set, then do a sign extend.
8798 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
8799 test, check for sign extended subreg and/or constant operands, and
8800 do a sign extend in that case.
8802 2018-05-17 Steve Ellcey <sellcey@cavium.com>
8804 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
8805 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
8807 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
8808 Change logics_shift_reg to logics_shift_imm.
8809 (thunderx2t99_fp_loadpair_basic): Delete.
8810 (thunderx2t99_fp_storepair_basic): Delete.
8811 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
8812 (thunderx2t99_asimd_polynomial): Delete.
8813 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
8814 and neon_fp_mul_d_scalar_q.
8815 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
8816 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
8817 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
8818 (thunderx2t99_asimd_lut): Add missing tbl types.
8819 (thunderx2t99_asimd_ext): Delete.
8820 (thunderx2t99_asimd_load1_1_mult): Delete.
8821 (thunderx2t99_asimd_load1_2_mult): Delete.
8822 (thunderx2t99_asimd_load1_ldp): New.
8823 (thunderx2t99_asimd_load1): New.
8824 (thunderx2t99_asimd_load2): Add missing *load2* types.
8825 (thunderx2t99_asimd_load3): New.
8826 (thunderx2t99_asimd_load4): New.
8827 (thunderx2t99_asimd_store1_1_mult): Delete.
8828 (thunderx2t99_asimd_store1_2_mult): Delete.
8829 (thunderx2t99_asimd_store2_mult): Delete.
8830 (thunderx2t99_asimd_store2_onelane): Delete.
8831 (thunderx2t99_asimd_store_stp): New.
8832 (thunderx2t99_asimd_store1): New.
8833 (thunderx2t99_asimd_store2): New.
8834 (thunderx2t99_asimd_store3): New.
8835 (thunderx2t99_asimd_store4): New.
8837 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
8839 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
8840 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
8842 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
8843 Segher Boessenkool <segher@kernel.crashing.org>
8846 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
8849 2018-05-17 Richard Biener <rguenther@suse.de>
8851 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
8852 for pruning loop and prune defs feeding only already visited PHIs.
8854 2018-05-17 Richard Biener <rguenther@suse.de>
8856 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
8858 2018-05-17 Bin Cheng <bin.cheng@arm.com>
8859 Richard Biener <rguenther@suse.de>
8861 PR tree-optimization/85793
8862 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
8863 for VMAT_ELEMENTWISE.
8865 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
8867 * internal-fn.h (lookup_internal_fn): Declare
8868 * internal-fn.c (lookup_internal_fn): New function.
8869 * gimple.c (gimple_build_call_from_tree): Handle calls to
8871 * gimple-pretty-print.c (dump_gimple_call): Print "." before
8872 internal function names.
8873 * tree-pretty-print.c (dump_generic_node): Likewise.
8874 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
8876 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
8878 * gimple-fold.h (gimple_build): Make the function forms take
8879 combined_fn rather than built_in_function.
8880 (gimple_simplify): Likewise.
8881 * gimple-match-head.c (gimple_simplify): Likewise.
8882 * gimple-fold.c (gimple_build): Likewise.
8883 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
8884 rather than gimple_build_call_internal.
8885 (get_initial_defs_for_reduction): Likewise.
8886 (vect_create_epilog_for_reduction): Likewise.
8887 (vectorizable_live_operation): Likewise.
8889 2018-05-17 Martin Liska <mliska@suse.cz>
8891 * gimple-ssa-sprintf.c (format_directive): Do not use
8892 space in between 'G_' and '('.
8894 2018-05-17 Jakub Jelinek <jakub@redhat.com>
8897 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
8898 even if the mask is not all ones.
8901 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
8903 (ix86_gimple_fold_builtin): Likewise.
8906 * config/i386/i386.c: Include tree-vector-builder.h.
8907 (ix86_vector_shift_count): New function.
8908 (ix86_fold_builtin): Fold shift builtins by scalar count.
8909 (ix86_gimple_fold_builtin): Likewise.
8911 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
8912 _mm512_setzero): New intrinsics.
8914 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
8915 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8917 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
8918 code generation for cases where splatting a value is not useful.
8919 * simplify-rtx.c (simplify_ternary_operation): Simplify
8920 vec_merge across a vec_duplicate and a paradoxical subreg forming
8921 a vector mode to a vec_concat.
8923 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
8925 * config.gcc: Support "goldmont-plus".
8926 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
8928 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
8929 PROCESSOR_GOLDMONT_PLUS.
8930 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
8931 (processor_target_table): Add "goldmont-plus".
8932 (PTA_GOLDMONT_PLUS): Define.
8933 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
8934 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
8935 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
8936 (fold_builtin_cpu): Add "goldmont-plus".
8937 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
8938 (ix86_option_override_internal): Add "goldmont-plus".
8939 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
8940 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
8941 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
8942 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
8944 2018-05-17 Richard Biener <rguenther@suse.de>
8946 PR tree-optimization/85757
8947 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
8948 remove defs that only feed that PHI from further processing.
8950 2018-05-16 Jim Wilson <jimw@sifive.com>
8952 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
8954 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
8956 2018-05-16 Mark Wielaard <mark@klomp.org>
8958 * dwarf2out.c (count_index_strings): New function.
8959 (output_indirect_strings): Call count_index_strings and generate
8960 header for dwarf_version >= 5.
8962 2018-05-16 Mark Wielaard <mark@klomp.org>
8964 * dwarf2out.c (dwarf_FORM): New function.
8965 (set_indirect_string): Use dwarf_FORM.
8966 (reset_indirect_string): Likewise.
8967 (size_of_die): Likewise.
8968 (value_format): Likewise.
8969 (output_die): Likewise.
8970 (add_skeleton_AT_string): Likewise.
8971 (output_macinfo_op): Likewise.
8972 (index_string): Likewise.
8973 (output_index_string_offset): Likewise.
8974 (output_index_string): Likewise.
8975 (count_index_strings): Likewise.
8977 2018-05-16 Carl Love <cel@us.ibm.com>
8979 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
8980 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
8982 2018-05-16 Martin Jambor <mjambor@suse.cz>
8984 * ipa-prop.c (ipa_free_all_edge_args): Remove.
8985 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
8987 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
8989 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
8990 (fnma<mode>4): Likewise.
8991 (fms<mode>4): Likewise.
8992 (fnms<mode>4): Likewise.
8993 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
8994 (aarch64_fnma<mode>4): Likewise.
8995 (aarch64_fms<mode>4): Likewise.
8996 (aarch64_fnms<mode>4): Likewise.
8997 (aarch64_fnmadd<mode>4): Likewise.
8999 2018-05-16 Jason Merrill <jason@redhat.com>
9001 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
9003 2018-05-16 Richard Biener <rguenther@suse.de>
9005 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
9006 (dump_stmt_cost): Declare.
9007 (add_stmt_cost): Dump cost we add.
9008 (add_stmt_costs): New function.
9009 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
9011 (vect_analyze_stmt): Adjust prototype.
9012 (vectorizable_condition): Likewise.
9013 (vectorizable_live_operation): Likewise.
9014 (vectorizable_reduction): Likewise.
9015 (vectorizable_induction): Likewise.
9016 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
9017 cost vector to pass to vectorizable_ and record afterwards.
9018 (vect_model_reduction_cost): Take cost vector argument and adjust.
9019 (vect_model_induction_cost): Likewise.
9020 (vectorizable_reduction): Likewise.
9021 (vectorizable_induction): Likewise.
9022 (vectorizable_live_operation): Likewise.
9023 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
9024 SLP_TREE_NUMBER_OF_VEC_STMTS.
9025 (vect_analyze_slp_cost_1): Remove.
9026 (vect_analyze_slp_cost): Likewise.
9027 (vect_slp_analyze_node_operations): Take visited args and
9028 a target cost vector. Avoid processing already visited stmt sets.
9029 (vect_slp_analyze_operations): Use a local cost vector to gather
9030 costs and register those of non-discarded instances.
9031 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
9032 (vect_schedule_slp_instance): Remove copying of
9033 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
9035 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
9036 adding cost. Record cost entry location.
9037 (vect_prologue_cost_for_slp_op): Function to compute cost of
9038 a constant or invariant generated for SLP vect in the prologue,
9039 split out from vect_analyze_slp_cost_1.
9040 (vect_model_simple_cost): Make static. Adjust for SLP costing.
9041 (vect_model_promotion_demotion_cost): Likewise.
9042 (vect_model_store_cost): Likewise, make static.
9043 (vect_model_load_cost): Likewise.
9044 (vectorizable_bswap): Add cost vector arg and adjust.
9045 (vectorizable_call): Likewise.
9046 (vectorizable_simd_clone_call): Likewise.
9047 (vectorizable_conversion): Likewise.
9048 (vectorizable_assignment): Likewise.
9049 (vectorizable_shift): Likewise.
9050 (vectorizable_operation): Likewise.
9051 (vectorizable_store): Likewise.
9052 (vectorizable_load): Likewise.
9053 (vectorizable_condition): Likewise.
9054 (vectorizable_comparison): Likewise.
9055 (can_vectorize_live_stmts): Likewise.
9056 (vect_analyze_stmt): Likewise.
9057 (vect_transform_stmt): Adjust calls to vectorizable_*.
9058 * tree-vectorizer.c: Include gimple-pretty-print.h.
9059 (dump_stmt_cost): New function.
9061 2018-05-16 Richard Biener <rguenther@suse.de>
9063 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
9064 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
9065 * tree-ssa-dse.c: Include tree-ssa-loop.h.
9066 (check_name): New callback.
9067 (dse_classify_store): Track cycles via a visited bitmap of PHI
9068 defs and simplify handling of in-loop and across loop dead stores
9069 and properly fail for loop-variant refs. Handle byte-tracking with
9070 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
9073 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
9075 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
9076 (vect_get_mask_type_for_stmt): Likewise.
9077 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
9079 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
9080 to determine the statement's vector type and the vector type that
9081 should be used for calculating nunits. Deal with cases in which
9082 the type has to be deferred.
9083 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
9084 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
9085 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
9086 (vect_determine_vf_for_stmt): New functions, split out from...
9087 (vect_determine_vectorization_factor): ...here.
9088 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
9089 (vect_get_mask_type_for_stmt): New functions, split out from
9090 vect_determine_vectorization_factor.
9092 2018-05-16 Richard Biener <rguenther@suse.de>
9094 * tree-cfg.c (verify_gimple_assign_ternary): Properly
9095 verify the [VEC_]COND_EXPR embedded comparison.
9097 2018-05-15 Martin Sebor <msebor@redhat.com>
9099 PR tree-optimization/85753
9100 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
9101 RECORD_TYPE in addition to ARRAY_TYPE.
9103 2018-05-15 Martin Sebor <msebor@redhat.com>
9106 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
9108 2018-05-15 Richard Biener <rguenther@suse.de>
9110 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
9111 add by_clobber_p one. Change algorithm to collect all defs
9112 representing uses we need to walk and try reducing them to
9113 a single one before failing.
9114 (dse_dom_walker::dse_optimize_stmt): Adjust.
9116 2018-05-13 Mark Wielaard <mark@klomp.org>
9118 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
9119 (size_of_loc_descr): Likewise.
9120 (output_loc_operands): Likewise.
9121 (output_loc_operands_raw): Likewise.
9122 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
9123 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
9124 (hash_loc_operands): Likewise.
9125 (compare_loc_operands): Likewise.
9127 2018-05-14 Mark Wielaard <mark@klomp.org>
9129 * dwarf2out.c (count_index_addrs): New function.
9130 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
9132 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9134 PR tree-optimization/83648
9135 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
9136 return value as malloc candidate.
9138 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9141 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
9142 param as true in call to suggest_attribute.
9144 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
9146 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
9147 -mreadonly-in-sdata.
9149 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9151 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
9153 (aarch64_crypto_aesd_fused): Likewise.
9155 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
9157 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
9158 (movsi_aarch64): Likewise.
9159 (load_pairsi): Likewise.
9160 (load_pairdi): Likewise.
9161 (store_pairsi): Likewise.
9162 (store_pairdi): Likewise.
9163 (load_pairsf): Likewise.
9164 (load_pairdf): Likewise.
9165 (store_pairsf): Likewise.
9166 (store_pairdf): Likewise.
9167 (zero_extend): Likewise.
9168 (trunc): Swap alternatives.
9169 (fcvt_target): Add '?' to prefer w over r.
9171 2018-05-14 Jakub Jelinek <jakub@redhat.com>
9174 * config/i386/i386.md: Disallow non-commutative arithmetics in
9175 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
9176 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
9177 in the peephole2 before it.
9179 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
9181 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
9182 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
9183 (ix86_handle_option): Handle -mcldemote.
9184 * config.gcc: New header.
9185 * config/i386/cldemoteintrin.h: New file.
9186 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
9187 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
9189 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
9190 OPTION_MASK_ISA_CLDEMOTE.
9191 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
9192 (ix86_valid_target_attribute_inner_p): Ditto.
9193 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
9194 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
9195 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
9196 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
9197 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
9199 * config/i386/i386.opt: Add -mcldemote.
9200 * config/i386/x86intrin.h: New header.
9201 * doc/invoke.texi: Add -mcldemote.
9203 2018-05-14 Richard Biener <rguenther@suse.de>
9205 * doc/match-and-simplify.texi: Adjust :s documentation.
9207 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
9209 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
9210 intended memcpy size.
9211 (REORDER_45): Likewise.
9213 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
9215 * sort.cc: New file.
9216 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
9217 * vec.c (qsort_chk): Use gcc_qsort.
9218 * Makefile.in (OBJS-libcommon): Add sort.o.
9219 (build/sort.o): New target. Use it...
9220 (BUILD_RTL): ... here, and...
9221 (build/gencfn-macros): ... here, and...
9222 (build/genmatch): ... here.
9224 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
9225 Chung-Ju Wu <jasonwucj@gmail.com>
9227 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
9228 * config/nds32/nds32-graywolf.md: New file.
9229 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
9230 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
9232 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
9233 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
9234 * config/nds32/nds32.md (pipeline_model): Add graywolf.
9235 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
9236 * config/nds32/pipelines.md: Include n15 settings.
9238 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
9239 Chung-Ju Wu <jasonwucj@gmail.com>
9241 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
9242 * config/nds32/nds32-n13.md: New file.
9243 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
9244 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
9246 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
9247 * config/nds32/nds32.md (pipeline_model): Add n13.
9248 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
9249 * config/nds32/pipelines.md: Include n13 settings.
9251 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
9252 Chung-Ju Wu <jasonwucj@gmail.com>
9254 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
9255 * config/nds32/nds32-n10.md: New file.
9256 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
9257 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
9259 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
9260 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
9261 * config/nds32/nds32.md (pipeline_model): Add n10.
9262 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
9263 * config/nds32/pipelines.md: Include n10 settings.
9265 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
9266 Kito Cheng <kito.cheng@gmail.com>
9267 Chung-Ju Wu <jasonwucj@gmail.com>
9269 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
9270 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
9271 Add enum values for DSP extension instructions.
9272 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
9274 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
9275 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
9277 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
9278 * config/nds32/nds32-dspext.md: New file for DSP implementation.
9279 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
9280 * config/nds32/nds32-intrinsic.md: Likewise.
9281 * config/nds32/nds32_intrinsic.h: Likewise.
9282 * config/nds32/nds32-md-auxiliary.c: Likewise.
9283 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
9284 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
9285 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
9286 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
9287 * config/nds32/nds32-protos.h: New declarations for DSP extension.
9288 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
9289 TYPE_DMAC in switch statement.
9290 * config/nds32/nds32.c: New checking and implementation for DSP
9291 extension instructions.
9292 * config/nds32/nds32.h: Likewise.
9293 * config/nds32/nds32.md: Likewise.
9294 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
9295 * config/nds32/predicates.md: Implement new predicates for DSP
9298 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
9300 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
9301 Reformat alternatives and attributes so it is easier to identify
9302 which constraints/attributes go with which instruction.
9303 (mov<mode>_hardfloat32, FMOVE64): Likewise.
9304 (mov<mode>_softfloat32, FMOVE64): Likewise.
9305 (mov<mode>_hardfloat64, FMOVE64): Likewise.
9306 (mov<mode>_softfloat64, FMOVE64): Likewise.
9308 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
9310 * doc/extend.texi (PowerPC Built-in Functions): Rename this
9312 (Basic PowerPC Built-in Functions): The new name of the
9313 subsection previously known as "PowerPC Built-in Functions".
9314 (Basic PowerPC Built-in Functions Available on all Configurations):
9316 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
9317 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
9318 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
9319 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
9321 2018-05-11 Martin Jambor <mjambor@suse.cz>
9324 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
9327 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
9330 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
9332 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
9334 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
9335 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
9336 (ix86_handle_option): Handle -mwaitpkg.
9337 * config.gcc: New header.
9338 * config/i386/cpuid.h (bit_WAITPKG): New bit.
9339 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
9340 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
9342 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
9343 OPTION_MASK_ISA_WAITPKG.
9344 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
9345 (ix86_option_override_internal): Add PTA_WAITPKG.
9346 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
9347 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
9348 IX86_BUILTIN_TPAUSE.
9349 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
9350 __builtin_ia32_umwait and __builtin_ia32_tpause.
9351 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
9352 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
9353 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
9354 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
9355 UNSPECV_TPAUSE): New.
9356 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
9357 * config/i386/i386.opt: Add -mwaitpkg.
9358 * config/i386/waitpkgintrin.h: New file.
9359 * config/i386/x86intrin.h: New header.
9360 * doc/invoke.texi: Add -mwaitpkg.
9362 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
9365 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
9367 (cortex-m0): Use armv6s-m isa.
9368 (cortex-m0plus): Likewise.
9369 (cortex-m1): Likewise.
9370 (cortex-m0.small-multiply): Likewise.
9371 (cortex-m0plus.small-multiply): Likewise.
9372 (cortex-m1.small-multiply): Likewise.
9374 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
9375 Jakub Jelinek <jakub@redhat.com>
9377 PR tree-optimization/85692
9378 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
9379 source permute as well.
9381 2018-05-11 Martin Liska <mliska@suse.cz>
9384 * doc/extend.texi: Document LLVM style format for no_sanitize
9387 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
9389 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
9390 mode_supports_vsx_dform_quad to mode_supports_dq_form.
9391 (mode_supports_vsx_dform_quad): Likewise.
9392 (mode_supports_vmx_dform): Move these functions to be next to the
9393 other mode_supports functions.
9394 (mode_supports_dq_form): Likewise.
9395 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
9396 mode_supports_dq_form.
9397 (reg_offset_addressing_ok_p): Likewise.
9398 (offsettable_ok_by_alignment): Likewise.
9399 (rs6000_legitimate_offset_address_p): Likewise.
9400 (legitimate_lo_sum_address_p): Likewise.
9401 (rs6000_legitimize_address): Likewise.
9402 (rs6000_legitimize_reload_address): Likewise.
9403 (rs6000_secondary_reload_inner): Likewise.
9404 (rs6000_preferred_reload_class): Likewise.
9405 (rs6000_output_move_128bit): Likewise.
9407 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
9409 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
9410 Generate SImode target register for null target.
9411 <case IX86_BUILTIN_XGETBV>: Ditto.
9412 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
9413 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
9415 2018-05-10 Carl Love <cel@us.ibm.com>
9417 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
9418 dcbtt and dcbtstt if operands[2] is 0.
9420 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
9423 * config/i386/sse.md (usadv64qi): New expander.
9425 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
9427 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
9428 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
9429 -maltivec=be support.
9430 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
9431 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
9432 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
9433 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
9434 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
9435 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
9436 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
9437 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
9438 altivec_vsumsws): Adjust.
9439 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
9440 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
9441 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
9443 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
9444 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
9445 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
9446 (altivec_lve<VI_char>x): Delete expand.
9447 (*altivec_lve<VI_char>x_internal): Rename to...
9448 (altivec_lve<VI_char>x): ... this.
9449 (altivec_lvxl_<mode>): Delete expand.
9450 (*altivec_lvxl_<mode>_internal): Rename to ...
9451 (altivec_lvxl_<mode>): ... this.
9452 (altivec_stvxl_<mode>): Delete expand.
9453 (*altivec_stvxl_<mode>_internal): Rename to ...
9454 (altivec_stvxl_<mode>): ... this.
9455 (altivec_stve<VI_char>x): Delete expand.
9456 (*altivec_stve<VI_char>x_internal): Rename to ...
9457 (altivec_stve<VI_char>x): ... this.
9458 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
9459 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
9460 reduc_plus_scal_<mode>): Adjust.
9461 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
9463 (rs6000_cpu_cpp_builtins): Adjust.
9464 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
9465 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
9466 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
9467 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
9468 -maltivec=be support.
9469 (rs6000_split_vec_extract_var): Adjust.
9470 (rs6000_split_v4si_init): Adjust.
9471 (swap_selector_for_mode): Delete.
9472 (altivec_expand_lvx_be, altivec_expand_stvx_be,
9473 altivec_expand_stvex_be): Delete.
9474 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
9475 -maltivec=be support.
9476 (rs6000_gimple_fold_builtin): Ditto.
9477 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
9479 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
9480 (TARGET_DIRECT_MOVE_64BIT): Adjust.
9481 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
9482 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
9483 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
9484 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
9485 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
9486 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
9487 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
9488 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
9489 anonymous split): Adjust.
9490 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
9491 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
9493 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
9495 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
9496 when --with-gxx-include-dir is also specified.
9497 * configure: Regenerate.
9499 2018-05-09 Jim Wilson <jimw@sifive.com>
9502 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
9503 * config/riscv/t-withmultilib: New.
9504 * config/riscv/withmultilib.h: New.
9505 * doc/install.texi: Document RISC-V --with-multilib-list support.
9507 2018-05-09 Richard Biener <rguenther@suse.de>
9509 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
9511 (vect_bb_vectorization_profitable_p): Adjust. Compute
9512 actual scalar cost using the cost vector and the add_stmt_cost
9515 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
9517 PR rtl-optimization/85645
9518 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
9519 in the REG_CFA_REGISTER note for LR, don't leave it empty.
9521 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
9523 PR rtl-optimization/85645
9524 * shrink-wrap.c (spread_components): Return a boolean saying if
9525 anything was changed.
9526 (try_shrink_wrapping_separate): Iterate spread_components until
9527 nothing changes anymore.
9529 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
9531 PR rtl-optimization/85645
9532 * regrename.c (build_def_use): Also kill the chains that include the
9533 destination of a REG_CFA_REGISTER note.
9535 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
9537 PR rtl-optimization/85645
9538 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
9539 insn that has a REG_CFA_REGISTER note.
9541 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
9543 * cfgexpand.c (expand_clobber): New function.
9544 (expand_gimple_stmt_1): Use it.
9545 * tree-vect-stmts.c (vect_clobber_variable): New function,
9547 (vectorizable_simd_clone_call): ...here.
9548 (vectorizable_store): Emit a clobber either side of an
9549 IFN_STORE_LANES sequence.
9550 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
9552 2018-05-09 Tom de Vries <tom@codesourcery.com>
9555 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
9556 (define_insn "trap_if_false"): Add exit after trap.
9558 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
9560 PR rtl-optimization/85638
9561 * bb-reorder.c: Include common/common-target.h.
9562 (create_forwarder_block): New function extracted from...
9563 (fix_up_crossing_landing_pad): ...here. Rename into...
9564 (dw2_fix_up_crossing_landing_pad): ...this.
9565 (sjlj_fix_up_crossing_landing_pad): New function.
9566 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
9567 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
9568 from both partitions and exit the loop after one iteration.
9570 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
9573 * doc/extend.texi (PowerPC Built-in Functions): Rename this
9575 (Basic PowerPC Built-in Functions): The new name of the
9576 subsection previously known as "PowerPC Built-in Functions".
9577 (Basic PowerPC Built-in Functions Available on all Configurations):
9579 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
9581 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
9583 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
9585 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
9588 2018-05-08 Jim Wilson <jimw@sifive.com>
9590 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
9591 (LD_EMUL_SUFFIX): New.
9592 (LINK_SPEC): Use it.
9594 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
9596 * doc/extend.texi (PowerPC Built-in Functions): Rename this
9598 (Basic PowerPC Built-in Functions): The new name of the
9599 subsection previously known as "PowerPC Built-in Functions".
9600 (Basic PowerPC Built-in Functions Available on all Configurations):
9602 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
9604 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
9606 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
9608 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
9611 2018-05-08 Jakub Jelinek <jakub@redhat.com>
9614 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
9615 after cmpelim optimization.
9617 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
9619 * config.gcc: Support "goldmont".
9620 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
9621 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
9623 * config/i386/i386.c (m_GOLDMONT): Define.
9624 (processor_target_table): Add "goldmont".
9625 (PTA_GOLDMONT): Define.
9626 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
9627 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
9628 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
9629 (fold_builtin_cpu): Add "goldmont".
9630 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
9631 (ix86_option_override_internal): Add "goldmont".
9632 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
9633 (processor_type): Add PROCESSOR_GOLDMONT.
9634 * config/i386/i386.md: Add CPU "glm".
9635 * config/i386/glm.md: New file.
9636 * config/i386/x86-tune.def: Add m_GOLDMONT.
9637 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
9639 2018-05-08 Jakub Jelinek <jakub@redhat.com>
9642 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
9644 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
9645 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
9646 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
9649 * config/i386/i386.c (ix86_fold_builtin): Handle
9650 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
9653 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
9654 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
9656 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
9659 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
9660 (check_arch): Likewise.
9661 (check_fpu): Return the result rather than printing it.
9662 (end arch): Fix operator precedence.
9663 (end cpu): Likewise.
9664 (END): Print the result from check_fpu.
9666 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
9667 Alan Hayward <alan.hayward@arm.com>
9668 David Sherwood <david.sherwood@arm.com>
9670 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
9671 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
9672 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
9673 (*fcmuo<mode>_and): New patterns.
9675 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
9677 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
9678 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
9679 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
9680 (cmp_op, sve_imm_con): New code attributes.
9681 (SVE_COND_INT_CMP, imm_con): Delete.
9682 (cmp_op): Remove above unspecs from int attribute.
9683 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
9685 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
9686 comparison-specific unspecs.
9687 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
9688 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
9689 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
9690 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
9691 (*vec_fcm<cmp_op><mode>): Rename to...
9692 (*fcm<cmp_op><mode>): ...this and adjust likewise.
9693 (*vec_fcmuo<mode>): Rename to...
9694 (*fcmuo<mode>): ...this and adjust likewise.
9695 (*pred_fcm<cmp_op><mode>): New pattern.
9696 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
9697 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
9699 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
9701 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
9702 (aarch64_emit_sve_predicated_cond): New function.
9703 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
9704 (aarch64_emit_unspec_cond_or): Replace with...
9705 (aarch64_emit_sve_or_conds): ...this new function. Use
9706 aarch64_emit_sve_ptrue_op for the individual comparisons and
9707 aarch64_emit_binop to OR them together.
9708 (aarch64_emit_inverted_unspec_cond): Replace with...
9709 (aarch64_emit_sve_inverted_cond): ...this new function. Use
9710 aarch64_emit_sve_ptrue_op for the comparison and
9711 aarch64_emit_unop to invert the result.
9712 (aarch64_expand_sve_vec_cmp_float): Update after the above
9713 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
9715 2018-05-07 Nathan Sidwell <nathan@acm.org>
9717 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
9718 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
9719 (Backwards Compatibility): Likewise.
9721 2018-05-07 Luis Machado <luis.machado@linaro.org>
9725 2018-05-07 Luis Machado <luis.machado@linaro.org>
9727 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
9728 <prefetch_dynamic_strides>: New const bool field.
9729 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
9730 prefetch_dynamic_strides.
9731 (exynosm1_prefetch_tune): Likewise.
9732 (thunderxt88_prefetch_tune): Likewise.
9733 (thunderx_prefetch_tune): Likewise.
9734 (thunderx2t99_prefetch_tune): Likewise.
9735 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
9737 (aarch64_override_options_internal): Update to set
9738 PARAM_PREFETCH_DYNAMIC_STRIDES.
9739 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
9740 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
9741 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
9742 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
9743 prefetch-dynamic-strides setting.
9745 2018-05-07 Luis Machado <luis.machado@linaro.org>
9747 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
9748 <minimum_stride>: New const int field.
9749 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
9750 minimum_stride field.
9751 (exynosm1_prefetch_tune): Likewise.
9752 (thunderxt88_prefetch_tune): Likewise.
9753 (thunderx_prefetch_tune): Likewise.
9754 (thunderx2t99_prefetch_tune): Likewise.
9755 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
9756 (aarch64_override_options_internal): Update to set
9757 PARAM_PREFETCH_MINIMUM_STRIDE.
9758 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
9759 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
9760 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
9761 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
9762 stride is constant and is below the minimum stride threshold.
9764 2018-05-07 Luis Machado <luis.machado@linaro.org>
9766 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
9769 2018-05-07 Luis Machado <luis.machado@linaro.org>
9771 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
9772 <prefetch_dynamic_strides>: New const bool field.
9773 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
9774 prefetch_dynamic_strides.
9775 (exynosm1_prefetch_tune): Likewise.
9776 (thunderxt88_prefetch_tune): Likewise.
9777 (thunderx_prefetch_tune): Likewise.
9778 (thunderx2t99_prefetch_tune): Likewise.
9779 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
9781 (aarch64_override_options_internal): Update to set
9782 PARAM_PREFETCH_DYNAMIC_STRIDES.
9783 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
9784 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
9785 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
9786 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
9787 prefetch-dynamic-strides setting.
9789 2018-05-07 Luis Machado <luis.machado@linaro.org>
9791 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
9792 <minimum_stride>: New const int field.
9793 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
9794 minimum_stride field.
9795 (exynosm1_prefetch_tune): Likewise.
9796 (thunderxt88_prefetch_tune): Likewise.
9797 (thunderx_prefetch_tune): Likewise.
9798 (thunderx2t99_prefetch_tune): Likewise.
9799 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
9800 (aarch64_override_options_internal): Update to set
9801 PARAM_PREFETCH_MINIMUM_STRIDE.
9802 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
9803 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
9804 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
9805 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
9806 stride is constant and is below the minimum stride threshold.
9808 2018-05-06 Jakub Jelinek <jakub@redhat.com>
9811 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
9812 the type is addressable. Don't force op into register if it has
9815 2018-05-05 Roland McGrath <mcgrathr@google.com>
9818 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
9819 any section for which we don't know a specific type it should have,
9820 regardless of name. Previously this was done only for the exact
9821 names ".init_array", ".fini_array", and ".preinit_array".
9822 (default_elf_asm_named_section): Add comment about
9823 relationship with default_section_type_flags and SECTION_NOTYPE.
9824 (get_section): Don't consider it a type conflict if one side has
9825 SECTION_NOTYPE and the other doesn't, as long as neither has the
9826 SECTION_BSS et al used in the default_section_type_flags logic.
9828 2018-05-05 Tom de Vries <tom@codesourcery.com>
9831 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
9832 (workaround_barsyncs): New function.
9833 (nvptx_reorg): Use workaround_barsyncs.
9834 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
9835 (define_expand "nvptx_membar_cta"): New define_expand.
9836 (define_insn "*nvptx_membar_cta"): New insn.
9838 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
9840 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
9841 To improve optimization opportunities.
9842 * builtin-types.def: The new needed builtin types for the above.
9844 2018-05-04 Richard Biener <rguenther@suse.de>
9846 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
9847 * gimple-ssa-store-merging.c
9848 (imm_store_chain_info::output_merged_store): Remove redundant create,
9849 release split_store vector contents on failure.
9850 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
9851 scalar stmt vector on cache hit.
9853 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
9855 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
9857 * config.gcc (powerpc-xilinx-eabi*): Remove.
9858 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
9860 (fusion_addis_mem_combo_load): Ditto.
9861 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
9863 (rs6000_cpu_cpp_builtins): Ditto.
9864 * config/rs6000/rs6000-linux.c
9865 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
9866 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
9867 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
9869 (rs6000_setup_reg_addr_masks): Ditto.
9870 (rs6000_init_hard_regno_mode_ok): Ditto.
9871 (rs6000_option_override_internal): Ditto.
9872 (legitimate_lo_sum_address_p): Ditto.
9873 (rs6000_legitimize_address): Ditto.
9874 (rs6000_legitimize_reload_address): Ditto.
9875 (rs6000_legitimate_address_p): Ditto.
9876 (abi_v4_pass_in_fpr): Ditto.
9877 (setup_incoming_varargs): Ditto.
9878 (rs6000_gimplify_va_arg): Ditto.
9879 (rs6000_split_multireg_move): Ditto.
9880 (rs6000_savres_strategy): Ditto.
9881 (rs6000_emit_prologue_components): Ditto.
9882 (rs6000_emit_epilogue_components): Ditto.
9883 (rs6000_emit_prologue): Ditto.
9884 (rs6000_emit_epilogue): Ditto.
9885 (rs6000_elf_file_end): Ditto.
9886 (rs6000_function_value): Ditto.
9887 (rs6000_libcall_value): Ditto.
9888 * config/rs6000/rs6000.h: Ditto.
9889 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
9890 (TARGET_MINMAX): ... this. New.
9891 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
9892 * config/rs6000/rs6000.md: Remove Xilinx FP support.
9893 (*movsi_internal1_single): Delete.
9894 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
9895 mfpu=, mxilinx-fpu): Delete.
9896 * config/rs6000/singlefp.h: Delete.
9897 * config/rs6000/sysv4.h: Remove Xilinx FP support.
9898 * config/rs6000/t-rs6000: Ditto.
9899 * config/rs6000/t-xilinx: Delete.
9900 * config/rs6000/titan.md: Adjust for fp_type removal.
9901 * config/rs6000/vsx.md: Remove Xilinx FP support.
9902 (VStype_simple): Delete.
9903 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
9904 * config/rs6000/xfpu.h: Delete.
9905 * config/rs6000/xfpu.md: Delete.
9906 * config/rs6000/xilinx.h: Delete.
9907 * config/rs6000/xilinx.opt: Delete.
9908 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
9909 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
9911 2018-05-04 Tom de Vries <tom@codesourcery.com>
9914 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
9917 2018-05-04 Richard Biener <rguenther@suse.de>
9920 * tree-complex.c (update_complex_assignment): We are always in SSA form.
9921 (expand_complex_div_wide): Likewise.
9922 (expand_complex_operations_1): Likewise.
9923 (expand_complex_libcall): Preserve EH info of the original stmt.
9924 (tree_lower_complex): Handle removed blocks.
9925 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
9926 on complex multiplication and division libcall builtins.
9928 2018-05-04 Richard Biener <rguenther@suse.de>
9931 * fold-const.c (negate_expr_p): Restrict negation of operand
9932 zero of a division to when we know that can happen without
9934 (fold_negate_expr_1): Likewise.
9936 2018-05-04 Jakub Jelinek <jakub@redhat.com>
9939 * real.h (real_nextafter): Declare.
9940 * real.c (real_nextafter): New function.
9941 * fold-const-call.c (fold_const_nextafter): New function.
9942 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
9943 CASE_CFN_NEXTTOWARD.
9944 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
9945 even when arg1_mode is different from arg0_mode.
9947 2018-05-03 Nathan Sidwell <nathan@acm.org>
9949 * doc/extend.texi (Deprecated Features): Remove
9951 (Backwards Compatibility): Likewise.
9952 * doc/invoke.texi (C++ Language Options): Likewise.
9953 (C++ Dialect Options): Likewise.
9955 2018-05-03 Jakub Jelinek <jakub@redhat.com>
9958 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
9959 _mm512_mask_mullox_epi64): New intrinsics.
9961 2018-05-03 Tom de Vries <tom@codesourcery.com>
9964 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
9965 dump files): Add offload-tree.
9967 2018-05-03 Richard Biener <rguenther@suse.de>
9969 PR tree-optimization/85615
9970 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
9971 to loops not nested in BBs loop father to avoid creating multi-entry
9974 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9976 PR tree-optimization/70291
9977 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
9978 arguments. Change return type to tree. Emit libcall as a new
9979 statement rather than replacing existing one when inplace_p is true.
9980 (expand_complex_multiplication_components): New function.
9981 (expand_complex_multiplication): Expand floating-point complex
9982 multiplication using the above.
9983 (expand_complex_division): Rename inner_type parameter to type.
9984 Update expand_complex_libcall call-site.
9985 (expand_complex_operations_1): Update expand_complex_multiplication
9986 and expand_complex_division call-sites.
9988 2018-05-02 Jakub Jelinek <jakub@redhat.com>
9991 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
9992 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
9993 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
9994 the highest significant bit of the shift count mask is clear. In
9995 check whether and[sq]i3 is needed verify that all significant bits
9996 of the shift count other than the highest are set.
9998 2018-05-02 Tom de Vries <tom@codesourcery.com>
10001 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
10002 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
10003 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
10004 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
10005 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
10006 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
10007 __builtin_goacc_parlevel_size.
10009 2018-05-02 Richard Biener <rguenther@suse.de>
10011 PR tree-optimization/85597
10012 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
10013 do not use split vect_get_vec_defs call but call vect_get_slp_defs
10016 2018-05-02 Tom de Vries <tom@codesourcery.com>
10019 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
10020 dump files): Add ltrans-tree.
10022 2018-05-02 Tom de Vries <tom@codesourcery.com>
10025 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
10026 dump files): Add wpa-ipa.
10028 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
10030 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
10031 powerpc*-*-linux*paired* target.
10032 * config/rs6000/750cl.h: Delete.
10033 * config/rs6000/paired.h: Delete.
10034 * config/rs6000/paired.md: Delete.
10035 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
10037 * config/rs6000/rs6000-builtin.def: Remove paired float support.
10038 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
10039 comment. Remove paired float support.
10040 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
10041 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
10043 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
10044 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
10046 * config/rs6000/rs6000.c: Remove paired float support.
10047 (paired_expand_vector_init, paired_expand_vector_move,
10048 paired_emit_vector_compare, paired_emit_vector_cond_expr,
10049 (paired_expand_lv_builtin, paired_expand_stv_builtin,
10050 paired_expand_builtin, paired_expand_predicate_builtin,
10051 paired_init_builtins): Delete.
10052 * config/rs6000/rs6000.h: Remove paired float support.
10053 * config/rs6000/rs6000.md: Remove paired float support.
10054 (move_from_CR_ov_bit): Delete.
10055 * config/rs6000/rs6000.opt (mpaired): Delete.
10056 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
10057 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
10059 2018-05-02 Richard Biener <rguenther@suse.de>
10061 PR middle-end/85567
10062 * gimplify.c (gimplify_save_expr): When in SSA form allow
10063 SAVE_EXPRs to compute to SSA vars.
10065 2018-05-02 Jakub Jelinek <jakub@redhat.com>
10068 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
10069 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
10070 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
10071 clobber operands[2], instead use a new pseudo. Formatting fixes.
10073 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
10075 PR tree-optimization/85586
10076 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
10077 exit early for statements in the same group if the accesses are
10080 2018-05-02 Tom de Vries <tom@codesourcery.com>
10083 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
10086 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
10088 PR tree-optimization/85143
10089 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
10091 2018-05-01 Tom de Vries <tom@codesourcery.com>
10094 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
10095 not found" error message.
10097 2018-05-01 Tom de Vries <tom@codesourcery.com>
10100 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
10101 * vec.c (test_ordered_remove_if): New function.
10102 (vec_c_tests): Call test_ordered_remove_if.
10103 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
10104 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
10105 * tree-vect-patterns.c (vect_pattern_recog_1): Use
10106 VEC_ORDERED_REMOVE_IF.
10108 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10110 PR tree-optimization/82665
10111 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
10112 pointer subtraction where arguments come from a memchr call.
10114 2018-05-01 Jakub Jelinek <jakub@redhat.com>
10116 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
10117 --push-state --as-needed and --pop-state instead of --as-needed and
10118 --no-as-needed if ld supports it.
10119 * configure: Regenerated.
10122 * doc/install.texi2html: Replace _002d with - and _002a with * in
10123 generated html files using sed.
10125 2018-04-30 David Malcolm <dmalcolm@redhat.com>
10128 * gcc-rich-location.c (blank_line_before_p): New function.
10129 (use_new_line): New function.
10130 (gcc_rich_location::add_fixit_insert_formatted): New function.
10131 * gcc-rich-location.h
10132 (gcc_rich_location::add_fixit_insert_formatted): New function.
10134 2018-04-30 David Malcolm <dmalcolm@redhat.com>
10136 * selftest.c (assert_streq): Rename "expected" and "actual" to
10137 "val1" and "val2". Extend NULL-handling to cover both inputs
10138 symmetrically, while still requiring both to be non-NULL for a pass.
10139 * selftest.h (assert_streq): Rename "expected" and "actual" to
10141 (ASSERT_EQ): Likewise.
10142 (ASSERT_EQ_AT): Likewise.
10143 (ASSERT_KNOWN_EQ): Likewise.
10144 (ASSERT_KNOWN_EQ_AT): Likewise.
10145 (ASSERT_NE): Likewise.
10146 (ASSERT_MAYBE_NE): Likewise.
10147 (ASSERT_MAYBE_NE_AT): Likewise.
10148 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
10149 the assertion to pass.
10150 (ASSERT_STREQ_AT): Likewise.
10152 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
10154 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
10155 interaction with -pie.
10157 2018-04-30 David Malcolm <dmalcolm@redhat.com>
10159 * selftest.h: Fix alphabetization of per-source-file selftest
10162 2018-04-30 Jason Merrill <jason@redhat.com>
10164 PR c++/61982 - dead stores to destroyed objects.
10165 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
10168 2018-04-30 Jason Merrill <jason@redhat.com>
10170 * tree.c (build_clobber): New.
10171 * tree.h: Declare it.
10172 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
10174 2018-04-30 David Malcolm <dmalcolm@redhat.com>
10176 * diagnostic-show-locus.c (layout::layout): Update for
10177 location_get_source_line returning a char_span.
10178 (struct char_span): Move to input.h.
10179 (struct correction): Update for fields in char_span becoming
10181 (struct source_line): Update for location_get_source_line
10182 returning a char_span.
10183 (layout::print_line): Likewise.
10184 * edit-context.c (edited_file::print_content): Likewise.
10185 (edited_file::print_diff_hunk): Likewise.
10186 (edited_file::print_run_of_changed_lines): Likewise.
10187 (edited_file::get_num_lines): Likewise.
10188 (edited_line::edited_line): Likewise.
10189 * final.c (asm_show_source): Likewise.
10190 * input.c (location_get_source_line): Convert return type
10191 from const char * to char_span, losing the final "line_len"
10193 (dump_location_info): Update for the above.
10194 (get_substring_ranges_for_loc): Likewise. Use a char_span
10195 when handling the literal within the line.
10196 (test_reading_source_line): Update for location_get_source_line
10197 returning a char_span.
10198 * input.h (class char_span): Move here from
10199 diagnostic-show-locus.c, converting from a struct to a class.
10200 Make data members private.
10201 (char_span::operator bool): New.
10202 (char_span::length): New.
10203 (char_span::get_buffer): New.
10204 (char_span::operator[]): New.
10205 (char_span::subspan): Make const.
10206 (char_span::xstrdup): New.
10207 (location_get_source_line): Convert return type from const char *
10208 to char_span, losing the final "line_size" param.
10210 2018-04-30 Jan Hubicka <jh@suse.cz>
10212 * lto-wrapper.c (ltrans_priorities): New static var.
10213 (cmp_priority): New.
10214 (run_gcc): Read priorities and if doing parallel build order
10215 the Makefile by them.
10217 2018-04-30 David Malcolm <dmalcolm@redhat.com>
10219 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
10221 2018-04-30 Richard Biener <rguenther@suse.de>
10223 * tree-cfg.c (verify_address): Remove base argument, add
10224 flag whether to check TREE_ADDRESSABLE and do that.
10225 (verify_expr): Remove.
10226 (verify_types_in_gimple_reference): Add pieces from verify_expr.
10227 (verify_gimple_assign_single): Likewise.
10228 (verify_gimple_switch): Likewise.
10229 (verify_expr_location_1): Dereference tp once. Add (disabled)
10230 piece from verify_expr.
10231 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
10233 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
10235 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
10237 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
10239 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
10240 (small_data_pattern): Likewise.
10241 (arc_rewrite_small_data): Likewise.
10242 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
10243 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
10244 (get_symbol_alignment): New function.
10245 (legitimate_small_data_address_p): Likewise.
10246 (legitimate_scaled_address): Update, call
10247 legitimate_small_data_address_p.
10248 (output_sdata): New static variable.
10249 (arc_print_operand): Update how we handle small data operands.
10250 (arc_print_operand_address): Likewise.
10251 (arc_legitimate_address_p): Update, use
10252 legitimate_small_data_address_p.
10253 (arc_rewrite_small_data_p): Remove.
10254 (arc_rewrite_small_data_1): Likewise.
10255 (arc_rewrite_small_data): Likewise.
10256 (small_data_pattern): Likewise.
10257 (compact_sda_memory_operand): Update to use
10258 legitimate_small_data_address_p and get_symbol_alignment.
10259 (prepare_move_operands): Don't rewite sdata pattern.
10260 (prepare_extend_operands): Remove.
10261 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
10263 (zero_extendqisi2): Likewise.
10264 (zero_extendhisi2): Likewise.
10265 (extendqihi2): Likewise.
10266 (extendqisi2): Likewise.
10267 (extendhisi2): Likewise.
10268 (addsi3): Likewise.
10269 (subsi3): Likewise.
10270 (andsi3): Likewise.
10271 * config/arc/constraints.md (Usd): Change it to memory constraint.
10273 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
10275 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
10276 as source of std instructions.
10277 * config/arc/arc.md (movsi_insn): Update pattern predicate to
10278 allow 6-bit constants as source for store instructions.
10279 (movdi_insn): Update instruction pattern to allow 6-bit constants
10280 as source for store instructions.
10282 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
10284 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
10286 2018-04-30 Nathan Sidwell <nathan@acm.org>
10287 Sandra Loosemore <sandra@codesourcery.com>
10289 * dumpfile.c (dump_open): Allow '-' for stdout.
10290 * doc/invoke.texi (Developer Options): Document dump filename
10291 determination early. Document stdin/stdout selection.
10293 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
10295 Microblaze Target: PIC data text relative
10297 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
10298 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
10300 * config/microblaze/microblaze.h (microblaze_constant_address_p):
10301 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
10302 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
10303 New addressing mode for data-text relative position indepenedent code.
10304 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
10305 'ADDRESS_SYMBOLIC_TXT_REL'.
10306 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
10307 (microblaze_legitimate_pic_operand): Exclude function calls from
10308 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
10309 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
10311 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
10312 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
10313 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
10314 for 'address + offset'.
10315 (microblaze_expand_prologue): Add new function prologue call for
10317 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
10318 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
10319 table in case of TARGET_PIC_DATA_TEXT_REL.
10320 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
10321 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
10322 Add new macros 'UNSPEC_TEXT',
10323 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
10324 + exclude function calls from 'UNSPEC_PLT' in case of data text
10326 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
10327 new target hook for generating address diff vector tables in case of
10329 * doc/tm.texi : Regenerate.
10330 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
10331 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
10332 of addr diff vector generation.
10333 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
10334 target hook definition.
10335 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
10336 Add default function for generate_pic_addr_diff_vec -> flag_pic.
10337 * doc/invoke.texi (Add new pic option): Add new microblaze pic
10338 option for data text relative.
10340 2018-04-30 Richard Biener <rguenther@suse.de>
10342 * tree-chrec.h (evolution_function_is_constant_p): Remove
10344 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
10346 2018-04-30 Richard Biener <rguenther@suse.de>
10349 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
10351 2018-04-30 Richard Biener <rguenther@suse.de>
10353 PR tree-optimization/28364
10354 PR tree-optimization/85275
10355 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
10356 copying first exit test.
10358 2018-04-28 Mark Wielaard <mark@klomp.org>
10360 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
10361 dwarf_version >= 5.
10362 (dwarf_AT): Handle DW_AT_addr_base.
10363 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
10365 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
10368 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
10369 (*ashl<dwi>3_doubleword_mask_1): Ditto.
10370 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
10371 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
10373 2018-04-28 Richard Biener <rguenther@suse.de>
10375 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
10376 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
10377 to reflect use. Only add interesting stmts.
10379 2018-04-27 Martin Jambor <mjambor@suse.cz>
10382 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
10383 the jump function allows for passing through aggregate values.
10385 2018-04-27 David Malcolm <dmalcolm@redhat.com>
10387 * input.h (in_system_header_at): Convert from macro to inline
10389 (from_macro_expansion_at): Likewise.
10390 (from_macro_definition_at): Likewise.
10392 2018-04-27 Jeff Law <law@redhat.com>
10394 * config.gcc: Mark tile* targets as deprecated/obsolete.
10396 2018-04-27 Richard Biener <rguenther@suse.de>
10398 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
10401 2018-04-27 Richard Biener <rguenther@suse.de>
10403 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
10405 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
10407 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
10408 with Yd constraint. Set "preferred_for_speed" attribute from
10409 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
10410 with Yd constraint.
10411 (*movdi_internal): Ditto.
10412 (movti_interunit splitters): Remove
10413 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
10414 (movdi_interunit splitters): Ditto.
10415 * config/i386/constraints.md (Ye): Remove.
10416 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
10418 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10421 * config/aarch64/constraints.md (Usg): Limit to 31.
10422 (Usj): Limit to 63.
10424 2018-04-27 Jakub Jelinek <jakub@redhat.com>
10426 PR tree-optimization/85529
10427 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
10428 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
10429 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
10430 zero extension or masking of the MSB bit.
10431 (optimize_range_tests): Add FIRST_BB argument, pass it through
10432 to optimize_range_tests_var_bound.
10433 (maybe_optimize_range_tests, reassociate_bb): Adjust
10434 optimize_range_tests callers.
10436 2018-04-26 Richard Biener <rguenther@suse.de>
10437 Jakub Jelinek <jakub@redhat.com>
10439 * cgraph.h (symbol_table): Just declare debug method here.
10440 * symtab.c (symbol_table::debug): Define.
10442 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
10444 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
10446 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
10448 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
10449 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
10450 (*movdi_internal): Substitute Yi and Yj constraint with x
10451 and Ym and Yn constraint with y constraint. Update "isa"
10452 attribute and set "preferred_for_speed" attribute from
10453 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
10454 (*movsi_internal): Ditto.
10455 (*movdf_internal): Ditto.
10456 (*movsf_internal): Ditto.
10457 (*zero_extendsidi2): Ditto.
10458 * config/i386/sse.md (vec_set<mode>_0): Ditto.
10459 (sse2_loadld): Ditto.
10460 (*vec_extract<ssevecmodelower>_0): Ditto.
10461 (*vec_extractv4si_0_zext_sse4): Ditto.
10462 (vec_concatv2di): Ditto.
10463 (*vec_dup<mode>): Ditto.
10464 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
10465 * config/i386/constraints.md (Yi): Remove.
10470 2018-04-26 Nathan Sidwell <nathan@acm.org>
10472 * dumpfile.c (dump_open): New.
10473 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
10474 (dump_finish): Detect stdio/stderr by value not name.
10476 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
10478 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
10480 2018-04-26 Tom de Vries <tom@codesourcery.com>
10483 * config/nvptx/nvptx.c (verify_neutering_jumps)
10484 (verify_neutering_labels): New function
10485 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
10487 2018-04-26 Tom de Vries <tom@codesourcery.com>
10490 * config/nvptx/nvptx.c (needs_neutering_p): New function.
10491 (nvptx_single): Use needs_neutering_p to skip over insns that do not
10494 2018-04-26 Richard Biener <rguenther@suse.de>
10495 Tom de Vries <tom@codesourcery.com>
10498 * lto-streamer-out.c (output_function): Fixup loops if required to match
10499 discovery done in the reader.
10501 2018-04-26 Richard Biener <rguenther@suse.de>
10503 PR tree-optimization/85116
10504 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
10505 have a loop exit from the single latch predecessor. Remove
10506 case of header with just condition.
10507 (ch_base::copy_headers): Exclude infinite loops from any
10509 (pass_ch::execute): Record exits.
10511 2018-04-26 Richard Biener <rguenther@suse.de>
10513 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
10514 prologue cost vector and pass it to vect_get_load_cost.
10515 (vect_get_peeling_costs_all_drs): Likewise.
10516 (vect_peeling_hash_get_lowest_cost): Likewise.
10517 (vect_enhance_data_refs_alignment): Likewise.
10519 2018-04-26 Richard Biener <rguenther@suse.de>
10521 PR middle-end/85450
10522 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
10523 checking of integer<->pointer conversions.
10524 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
10525 sign-/zero-extending pointer types.
10526 (expand_omp_for_static_chunk): Likewise.
10528 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
10529 Jean Lee <xiaoyur347@gmail.com>
10531 * config/mips/mips.c (mips_asan_shadow_offset): New function.
10532 (TARGET_ASAN_SHADOW_OFFSET): Define.
10533 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
10534 true for -fsanitize=address.
10536 2018-04-25 Mark Wielaard <mark@klomp.org>
10538 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
10541 2018-04-25 Jakub Jelinek <jakub@redhat.com>
10543 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
10544 than "alu", remove explicit "memory" and "imm_disp" attributes.
10545 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
10547 PR middle-end/85414
10548 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
10549 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
10550 gen_lowpart_no_emit.
10552 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
10555 * config/i386/i386.c (ix86_expand_builtin): Change memory
10556 operand to XI, extend p0 to Pmode.
10557 * config/i386/i386.md: Change unspec volatile and operand
10558 1 mode to XI, change operand 0 mode to P.
10560 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
10562 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
10563 GET_MODE_MASK before any checking.
10564 (nds32_can_use_bset_p): Likewise.
10565 (nds32_can_use_btgl_p): Likewise.
10567 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
10569 * config/nds32/nds32-doubleword.md: New define_split pattern for
10570 illegal register number.
10572 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
10574 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
10576 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
10578 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
10580 2018-04-25 Richard Biener <rguenther@suse.de>
10582 * lto-streamer.h (LTO_major_version): Bump to 8.
10584 2018-04-25 Jakub Jelinek <jakub@redhat.com>
10586 * BASE-VER: Set to 9.0.0.
10588 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
10590 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
10591 in __abskf2 and __powikf2.
10593 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10596 * config/aarch64/constraints.md (Usg, Usj): New constraints.
10597 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
10598 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
10599 Use the above on operand 2. Reindent.
10600 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
10602 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
10605 * common/config/i386/i386-common.c (ix86_handle_option): Don't
10607 * config/i386/i386.opt (mcet): Removed.
10608 * doc/install.texi: Remove -mcet documentation.
10609 * doc/invoke.texi: Likewise.
10611 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
10614 * doc/install.texi: Remove -mcet from bootstrap-cet.
10616 2018-04-24 Jakub Jelinek <jakub@redhat.com>
10619 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
10620 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
10624 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
10625 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
10626 containing a CONST_VECTOR.
10628 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
10630 * doc/install.texi: Update newlib dependency for nvptx.
10632 2018-04-24 Jakub Jelinek <jakub@redhat.com>
10635 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
10636 instead of INTVAL when shifting x left.
10638 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
10640 PR tree-optimization/85478
10641 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
10642 vect_grouped_store_supported for single element vectors.
10644 2018-04-24 Richard Biener <rguenther@suse.de>
10647 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
10648 load cost increase to the case of non-constant step.
10650 2018-04-24 Jakub Jelinek <jakub@redhat.com>
10653 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
10654 destination if any_malformed_asm.
10656 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
10658 PR middle-end/85496
10659 * expr.c (store_field): In the bitfield case, if the value comes from
10660 a function call and is returned in registers by means of a PARALLEL,
10661 do not change the mode of the temporary unless BLKmode and VOIDmode.
10663 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
10665 PR rtl-optimization/85423
10666 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
10667 dependencies to debug insns when the previous insn is non-debug.
10669 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
10671 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
10672 enums into a single definition.
10673 (fls): Fix predicates and printing.
10676 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
10678 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
10679 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
10680 and short u6 immediate.
10681 (check_if_valid_sleep_operand): Remove.
10682 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
10684 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
10686 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
10687 flag_always_save_lp condition.
10688 * config/nds32/nds32.opt (malways-save-lp): New option.
10690 2018-04-22 Shiva Chen <shiva0217@gmail.com>
10692 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
10693 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
10694 * config/nds32/nds32.h
10695 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
10696 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
10698 2018-04-22 Shiva Chen <shiva0217@gmail.com>
10700 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
10701 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
10703 2018-04-22 Shiva Chen <shiva0217@gmail.com>
10704 Chung-Ju Wu <jasonwucj@gmail.com>
10706 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
10708 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
10709 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
10711 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
10713 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
10715 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
10717 * config/nds32/nds32-protos.h (nds32_data_alignment,
10718 nds32_local_alignment): Declare.
10719 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
10720 nds32_local_alignment): New functions.
10721 (TARGET_CONSTANT_ALIGNMENT): Define.
10722 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
10724 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
10726 * config/nds32/nds32.c
10727 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
10728 (TARGET_MODES_TIEABLE_P): Likewise.
10730 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
10732 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
10733 level Ofast and Og.
10735 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
10736 Chung-Ju Wu <jasonwucj@gmail.com>
10738 * config/nds32/constants.md (unspec_volatile_element): Add enum values
10739 for unaligned access.
10740 * config/nds32/nds32-intrinsic.c: Implementation of expanding
10742 * config/nds32/nds32-intrinsic.md: Likewise.
10743 * config/nds32/nds32_intrinsic.h: Likewise.
10744 * config/nds32/nds32.h (nds32_builtins): Likewise.
10745 * config/nds32/nds32.opt (munaligned-access): New option.
10746 * config/nds32/nds32.c (nds32_asm_file_start): Display
10747 flag_unaligned_access status.
10749 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
10751 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
10752 -mno-relax is present.
10753 * config/riscv/linux.h (LINK_SPEC): Ditto.
10755 2018-04-20 Martin Sebor <msebor@redhat.com>
10758 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
10760 (gimple_fold_builtin_stxcpy_chk): Same.
10761 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
10763 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
10766 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
10767 __powikf2 when long double is IEEE 128-bit.
10769 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
10771 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
10772 step to make sure stack always aligned.
10774 2018-04-20 Carl Love <cel@us.ibm.com>
10777 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
10778 size check for arg0.
10780 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
10781 Tom de Vries <tom@codesourcery.com>
10784 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
10785 Emit insns for calls too.
10786 (nvptx_find_par): Always look for worker-level predecessor insn.
10787 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
10789 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
10790 (nvptx_process_pars): Propagate frames for calls.
10792 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
10795 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
10797 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
10798 (ix86_handle_option): Don't handle OPT_mibt.
10799 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
10801 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
10803 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
10804 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
10805 (ix86_target_macros): Define __CET__ with flag_cf_protection
10806 for -fcf-protection.
10807 * config/i386/i386.c (isa2_opts): Remove -mibt.
10808 * config/i386/i386.h (TARGET_IBT): Removed.
10809 (TARGET_IBT_P): Likewise.
10810 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
10811 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
10812 * config/i386/i386.opt (mcet): Update help message.
10813 (mshstk): Likewise.
10815 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
10816 -mcet as an alias for -mshstk.
10818 2018-04-20 Richard Biener <rguenther@suse.de>
10820 PR middle-end/85475
10821 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
10822 complexity by forcing a single use of the multiply operand.
10824 2018-04-20 Martin Jambor <mjambor@suse.cz>
10827 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
10828 recursion dependency to only apply to non-clones.
10830 2018-04-20 Martin Jambor <mjambor@suse.cz>
10833 * ipa-cp.c (create_specialized_node): Check that clones of
10834 self-recursive edges exist during IPA-CP.
10836 2018-04-19 Toon Moene <toon@moene.org>
10838 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
10841 2018-04-19 Jakub Jelinek <jakub@redhat.com>
10843 PR tree-optimization/85467
10844 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
10845 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
10846 VECTOR_CST element to type.
10848 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
10851 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
10852 * config/i386/i386.md (builtin_setjmp_setup): Removed.
10853 (builtin_longjmp): Likewise.
10854 (save_stack_nonlocal): New pattern.
10855 (restore_stack_nonlocal): Likewise.
10857 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
10860 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
10861 Replace ASM_OUTPUT_LABEL with fprintf.
10863 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
10866 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
10867 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
10868 * config/i386/i386-c.c (ix86_target_macros_internal): Also
10869 define __IBT__ and __SHSTK__ for -fcf-protection.
10870 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
10872 (ix86_trampoline_init): Likewise.
10873 (x86_output_mi_thunk): Likewise.
10874 (ix86_notrack_prefixed_insn_p): Likewise.
10875 (ix86_option_override_internal): Don't disallow -fcf-protection.
10876 * config/i386/i386.md (rdssp<mode>): Also enable for
10878 (incssp<mode>): Likewise.
10879 (nop_endbr): Likewise.
10880 * config/i386/i386.opt (mcet): Change help message to built-in
10883 (mshstk): Likewise.
10884 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
10885 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
10886 enable CET built-in functions.
10888 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
10890 * common/config/i386/i386-common.c
10891 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
10892 OPTION_MASK_ISA_MOVDIRI_UNSET,
10893 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
10894 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
10895 * config.gcc (movdirintrin.h): New header.
10896 * config/i386/cpuid.h (bit_MOVDIRI,
10897 bit_MOVDIR64B): New bits.
10898 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
10900 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
10901 (VOID, PVOID, PCVOID)): New function types.
10902 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
10903 __builtin_ia32_directstoreu_u64,
10904 __builtin_ia32_movdir64b): New builtins.
10905 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
10906 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
10908 (ix86_valid_target_attribute_inner_p): Ditto.
10909 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
10910 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
10911 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
10912 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
10913 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
10914 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
10915 (movdiri<mode>, movdir64b_<mode>): New.
10916 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
10917 * config/i386/immintrin.h: Include movdirintrin.h.
10918 * config/i386/movdirintrin.h: New file.
10919 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
10921 2018-04-19 Richard Biener <rguenther@suse.de>
10923 PR middle-end/85455
10924 * cfg.c (clear_bb_flags): When loop state says we have
10925 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
10927 2018-04-19 Richard Biener <rguenther@suse.de>
10929 PR tree-optimization/84737
10930 * tree-vect-data-refs.c (vect_copy_ref_info): New function
10931 copying restrict info.
10932 (vect_setup_realignment): Use it.
10933 * tree-vectorizer.h (vect_copy_ref_info): Declare.
10934 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
10935 the first DR to all generated stores.
10936 (vectorizable_load): Likewise for loads.
10938 2018-04-19 Jakub Jelinek <jakub@redhat.com>
10940 PR tree-optimization/85446
10941 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
10942 the integral and pointer types to have the same precision.
10944 * doc/install.texi: Document --disable-cet being the default and
10947 2018-04-18 Martin Liska <mliska@suse.cz>
10949 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
10952 2018-04-18 Martin Liska <mliska@suse.cz>
10955 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
10958 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
10959 arguments if they are comparable.
10961 2018-04-18 Martin Liska <mliska@suse.cz>
10964 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
10967 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
10970 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
10973 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
10974 ENDBR after calling __morestack.
10976 2018-04-18 David Malcolm <dmalcolm@redhat.com>
10979 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
10980 by using gcc_base_ver to generate a gcc_driver_version, and use
10981 it when generating GCC_DRIVER_NAME.
10982 * configure: Regenerate.
10984 2018-04-18 Jakub Jelinek <jakub@redhat.com>
10987 * config.gcc: Obsolete powerpc*-*-*spe*.
10989 2018-04-17 Jakub Jelinek <jakub@redhat.com>
10992 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
10993 (stabstr_D): Change type of unum from unsigned int to
10994 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
10997 2018-04-17 Jim Wilson <jimw@sifive.com>
11000 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
11001 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
11002 Set arg_pointer_offset after using pretend_args_size.
11004 2018-04-17 Jakub Jelinek <jakub@redhat.com>
11006 PR rtl-optimization/85431
11007 * dse.c (record_store): Ignore zero width stores.
11010 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
11011 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
11012 __builtin_stack_restore rather than after it.
11013 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
11014 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
11015 argument instead of virtual_dynamic_stack_rtx.
11017 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
11019 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
11021 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
11022 Add note to error message to explain internal mapping of overloaded
11023 built-in function name to non-overloaded built-in function name.
11024 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
11027 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
11030 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
11031 where the inputs overlap with the output.
11033 2018-04-17 Jakub Jelinek <jakub@redhat.com>
11035 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
11036 (=v, v) alternative and explicit "memory" attribute.
11037 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
11038 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
11040 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
11041 "sselog1" type instead of "sselog".
11042 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
11043 "sselog". Remove explicit "memory" attribute.
11044 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
11045 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
11047 (vec_extract_hi_v32hi): Merge all alternatives into one, use
11048 "sselog1" type instead of "sselog". Remove explicit "memory"
11050 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
11051 use "sselog1" type instead of "sselog". Remove explicit "memory"
11053 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
11054 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
11056 (vec_extract_hi_v64qi): Merge all alternatives into one, use
11057 "sselog1" type instead of "sselog". Remove explicit "memory"
11059 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
11060 use "sselog1" type instead of "sselog". Remove explicit "memory"
11064 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
11066 PR middle-end/85414
11067 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
11070 2018-04-17 Martin Jambor <mjambor@suse.cz>
11073 * ipa-cp.c (create_specialized_node): Call
11074 expand_all_artificial_thunks if necessary.
11076 2018-04-17 Martin Liska <mliska@suse.cz>
11079 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
11080 in message, remote space in between '_G' and '('.
11082 2018-04-17 Jakub Jelinek <jakub@redhat.com>
11085 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
11086 avx512f_vmcmp<mode>3<round_saeonly_name>,
11087 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
11088 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
11089 avx512f_rndscale<mode><round_saeonly_name>,
11090 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
11091 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
11092 Use %<iptr>2 instead of %2 for -masm=intel.
11093 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
11094 avx512f_vcvttss2usi<round_saeonly_name>,
11095 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
11097 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
11098 avx512f_vcvttsd2usi<round_saeonly_name>,
11099 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
11100 Use %q1 instead of %1 for -masm=intel.
11101 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
11102 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
11103 of %3 for -masm=intel.
11104 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
11106 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
11108 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
11110 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
11111 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
11113 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
11115 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
11116 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
11117 %g1 and one with %0 and %1.
11118 (avx512er_vmrcp28<mode><round_saeonly_name>,
11119 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
11120 %1 for -masm=intel.
11121 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
11122 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
11123 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
11124 of %0 and %{%4%} for -masm=intel.
11125 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
11126 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
11127 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
11128 order of %0 and %{%5%}%{z%} for -masm=intel.
11130 2018-04-17 Jan Hubicka <jh@suse.cz>
11133 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
11135 2018-04-17 Martin Liska <mliska@suse.cz>
11138 * multiple_target.c (create_dispatcher_calls): Set apostrophes
11139 for target_clone error message. Make default implementation
11140 clone to be a local declaration.
11141 (separate_attrs): Add new argument and check for an empty
11143 (expand_target_clones): Handle it.
11144 (ipa_target_clone): Make redirection just for target_clones
11147 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
11148 Tom de Vries <tom@codesourcery.com>
11150 PR middle-end/84955
11151 * omp-expand.c (expand_oacc_for): Add dummy false branch for
11152 tiled basic blocks without omp continue statements.
11154 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
11157 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
11158 vec_extract expression as having side effects to make sure it gets
11161 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
11164 * config/i386/i386.c (get_builtin_code_for_version): Check
11167 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
11170 * config.gcc: Support "skylake".
11171 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
11173 * config/i386/i386.c (m_SKYLAKE): Define.
11174 (processor_target_table): Add "skylake".
11175 (ix86_option_override_internal): Add "skylake".
11176 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
11177 PROCESSOR_CANNONLAKE.
11178 (get_builtin_code_for_version): Fix priority for
11179 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
11180 PROCESSOR_SKYLAKE-AVX512.
11181 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
11182 (processor_type): Add PROCESSOR_SKYLAKE.
11184 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
11185 Jason Merrill <jason@redhat.com>
11188 * convert.c (convert_to_integer_1): Use direct recursion for
11189 enumeral types and types with a precision less than the number
11190 of bits in their mode.
11192 2018-04-16 Julia Koval <julia.koval@intel.com>
11195 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
11196 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
11198 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
11201 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
11202 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
11203 and -mno-direct-move.
11205 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
11208 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
11209 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
11210 Ensure negative shifts result in {0}.
11212 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
11214 PR rtl-optimization/79916
11215 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
11216 regs (if any) to define how to gnerate SD moves when LRA is in
11219 2018-04-13 Jakub Jelinek <jakub@redhat.com>
11221 PR rtl-optimization/85393
11222 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
11223 * except.c (expand_dw2_landing_pad_for_region): Make static.
11224 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
11225 a label and unconditional jump to old_bb, rather than
11226 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
11229 PR rtl-optimization/85376
11230 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
11231 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
11232 instead of a specific value.
11234 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
11235 Bin Cheng <bin.cheng@arm.com>
11237 PR tree-optimization/82965
11238 PR tree-optimization/83991
11239 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
11240 by_profile_only parameter.
11241 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
11242 information if the loop was predicted to iterate too many times.
11243 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
11245 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
11248 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
11251 2018-04-13 Martin Liska <mliska@suse.cz>
11252 Jakub Jelinek <jakub@redhat.com>
11254 PR middle-end/81657
11255 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
11256 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
11257 * builtins.c (expand_builtin_memory_copy_args): Use
11258 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
11259 handle dest_addr == pc_rtx.
11261 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
11264 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
11265 asked to not generate direct moves.
11266 (fix_trunc<mode>si2_stfiwx): Similar.
11267 (fix_trunc<mode>si2_internal): Similar.
11269 2018-04-12 Jakub Jelinek <jakub@redhat.com>
11272 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
11273 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
11274 lookup if dest in some wider mode is known to be const0_rtx and
11275 if so, record permanent equivalence for it to be ZERO_EXTEND of
11276 the narrower mode destination.
11278 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
11280 * lto-streamer-out.c (output_function): Revert 259346.
11281 * omp-expand.c (expand_oacc_for): Likewise.
11283 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
11285 PR rtl-optimization/85354
11286 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
11287 * sel-sched.c (sel_global_init): ... here.
11289 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
11292 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
11293 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
11294 mode for PE-COFF targets.
11295 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
11296 (i386_pe_asm_lto_end): Likewise.
11297 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
11298 (TARGET_ASM_LTO_END): Likewise.
11299 * config/i386/winnt.c (saved_debug_info_level): New static variable.
11300 (i386_pe_asm_lto_start): New function.
11301 (i386_pe_asm_lto_end): Likewise.
11303 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
11304 Richard Biener <rguenther@suse.de>
11306 PR middle-end/84955
11307 * lto-streamer-out.c (output_function): Fix CFG loop state before
11309 * omp-expand.c (expand_oacc_for): Handle calls to internal
11310 functions like regular functions.
11312 2018-04-12 Richard Biener <rguenther@suse.de>
11315 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
11316 for the early LTO debug to properly generate references to it
11317 during DIE emission. Do not re-use that for the skeleton for
11319 (dwarf2out_early_finish): Likewise.
11321 2018-04-12 Jakub Jelinek <jakub@redhat.com>
11324 * config/i386/sse.md
11325 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
11326 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
11327 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
11328 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
11329 and output is a reg, avoid creating invalid lowpart subreg, but
11330 instead split into a 512-bit move. Don't split if not AVX512VL,
11331 input is xmm16+ reg and output is a mem.
11332 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
11333 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
11334 xmm16+ reg and output is a mem.
11336 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11338 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
11339 also for flag_dwarf2_cfi_asm.
11341 2018-04-12 Jakub Jelinek <jakub@redhat.com>
11343 PR rtl-optimization/85342
11344 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
11345 a bool scalar var inside of the loop instead. Don't try to update
11346 recog_data.operand after failed apply_change_group.
11348 2018-04-12 Tom de Vries <tom@codesourcery.com>
11351 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
11352 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
11353 array with flexible array member as array without given dimension.
11354 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
11355 argument for undefined param to true.
11357 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
11360 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
11361 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
11362 from PowerPC section.
11363 * config/rs6000/sysv4.opt (mcall-): Improve help text.
11364 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
11365 help text that is too long.
11366 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
11367 help text that is too long.
11368 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
11369 help text that is too long.
11371 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
11373 * config/alpha/alpha.md (stack_probe_internal): Rename
11374 from "probe_stack". Update all callers.
11376 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
11378 PR rtl-optimization/84566
11379 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
11380 sched_macro_fuse_insns.
11382 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
11385 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
11386 (compute_block_dependences): ... from here.
11388 2018-04-11 Jakub Jelinek <jakub@redhat.com>
11390 PR tree-optimization/85331
11391 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
11392 from int to HOST_WIDE_INT.
11394 2018-04-11 Martin Jambor <mjambor@suse.cz>
11397 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
11398 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
11399 not the same as the source val.
11400 (cgraph_edge_brings_value_p): New parameter.
11401 (gather_edges_for_value): Pass destination value to
11402 cgraph_edge_brings_value_p.
11403 (perhaps_add_new_callers): Likewise.
11404 (get_info_about_necessary_edges): Likewise and exclude values brought
11405 only by self-recursive edges.
11406 (create_specialized_node): Redirect only clones of self-calling edges.
11407 (+self_recursive_pass_through_p): New function.
11408 (find_more_scalar_values_for_callers_subset): Use it.
11409 (find_aggregate_values_for_callers_subset): Likewise.
11410 (known_aggs_to_agg_replacement_list): Removed.
11411 (decide_whether_version_node): Re-calculate known constants for all
11412 remaining context clones.
11414 2018-04-11 Richard Biener <rguenther@suse.de>
11417 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
11418 from early DWARF output.
11419 (dwarf2out_early_finish): Output line info unconditionally into
11420 early DWARF and add reference to it.
11422 2018-04-11 Jakub Jelinek <jakub@redhat.com>
11425 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
11426 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
11427 other than V2DFmode using iptr mode attribute.
11428 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
11430 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
11432 PR rtl-optimization/84659
11433 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
11435 2018-04-11 Jakub Jelinek <jakub@redhat.com>
11438 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
11440 (output_loc_list): Pass address of a dummy size variable even in the
11441 locview handling loop.
11442 (index_location_lists): Add comment on why skip_loc_list_entry can't
11445 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
11448 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
11451 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
11454 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
11455 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
11456 and -mstring-compare-inline-limit.
11458 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
11461 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
11462 for stack clash protection in a register whenever we need it to be in
11465 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
11467 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
11468 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
11470 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
11473 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
11475 (mlong-double-): Ditto.
11476 * config/rs6000/sysv4.opt (msdata=): Ditto.
11477 (mtls-size=): Ditto.
11479 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
11481 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
11482 erroneous entries for
11483 "vector int vec_ldl (int, long int *)", and
11484 "vector unsigned int vec_ldl (int, unsigned long int *)".
11485 Add comments and entries for
11486 "vector bool char vec_ldl (int, bool char *)",
11487 "vector bool short vec_ldl (int, bool short *)",
11488 "vector bool int vec_ldl (int, bool int *)",
11489 "vector bool long long vec_ldl (int, bool long long *)",
11490 "vector pixel vec_ldl (int, pixel *)",
11491 "vector long long vec_ldl (int, long long *)",
11492 "vector unsigned long long vec_ldl (int, unsigned long long *)".
11493 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
11494 type tree bool_long_long_type_node and correct definition of
11495 bool_V2DI_type_node to make reference to this new type tree.
11496 (rs6000_mangle_type): Replace erroneous reference to
11497 bool_long_type_node with bool_long_long_type_node.
11498 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
11499 comments to emphasize sign distinctions for char and int types and
11500 replace RS6000_BTI_bool_long constant with
11501 RS6000_BTI_bool_long_long constant. Also add comment to restrict
11502 use of RS6000_BTI_pixel.
11503 (bool_long_type_node): Remove this macro definition.
11504 (bool_long_long_type_node): New macro definition
11506 2018-04-10 Jakub Jelinek <jakub@redhat.com>
11508 PR rtl-optimization/85300
11509 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
11510 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
11511 simplify_unary_operation fails.
11513 2018-04-10 Martin Liska <mliska@suse.cz>
11515 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
11516 cgraph_edge and ipa_ref.
11518 2018-04-10 Jakub Jelinek <jakub@redhat.com>
11522 * config/i386/sse.md
11523 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
11524 computation of the VEC_MERGE selector from mask.
11525 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
11526 Fix decoding of the VEC_MERGE selector into mask.
11528 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
11530 PR tree-optimization/85286
11531 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
11533 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
11535 * final.c (final_1): Set insn_last_address as well as
11536 insn_current_address.
11538 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11541 * explow.c (emit_stack_probe): Call validize_mem on memory location
11542 before passing it to gen_probe_stack. Create address operand and
11543 legitimize it for the probe_stack_address case.
11545 2018-04-09 Jan Hubicka <jh@suse.cz>
11548 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
11549 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
11550 * tree.c (free_lang_data_in_type): Fix handling of binfos;
11552 (free_lang_data): Rebuild type inheritance graph.
11554 2018-04-09 Martin Sebor <msebor@redhat.com>
11556 * invoke.texi (-finline-small-functions): Mention other optimization
11558 (-findirect-inlining, -fpartial-inlining): Same.
11559 (-finline-functions-called-once): Same.
11560 (-freorder-blocks-and-partition): Same.
11562 2018-04-09 Jan Hubicka <jh@suse.cz>
11565 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
11566 jumps; choose last target that matches the criteria (i.e.
11567 no partition changes for non-crossing jumps).
11568 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
11569 support for redirecting crossing jumps to non-crossing.
11571 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
11573 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
11574 also for naked functions.
11576 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
11578 * config/arc/arc.md (add_shift): New pattern.
11579 (add_shift2): Likewise.
11580 (sub_shift): Likewise.
11581 (sub_shift_cmp0_noout): Likewise.
11582 (compare_si_ashiftsi): Likewise.
11583 (xbfu_cmp0_noout): New combine pattern.
11584 (xbfu_cmp0"): Likewise.
11585 (movsi_set_cc_insn): Place the predicable variant first.
11586 (commutative_binary_cmp0_noout): Remove clobber.
11587 (commutative_binary_cmp0): New pattern.
11588 (noncommutative_binary_cmp0): Likewise.
11589 (noncommutative_binary_cmp0_noout): Likewise.
11590 (noncommutative_binary_comparison_result_used): Removed.
11591 (rsub_cmp0): New pattern.
11592 (rsub_cmp0_noout): Likewise.
11593 (extzvsi): Changed, keep only meaningful variants.
11594 (SQH, SEZ): New iterators.
11595 (SQH_postfix): New mode attribute.
11596 (SEZ_prefix): New code attribute.
11597 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
11598 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
11599 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
11600 of numerical value.
11601 (noncommutative_operator): Check the availability of barrel
11604 2018-04-09 Richard Biener <rguenther@suse.de>
11606 PR tree-optimization/85284
11607 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
11608 Only use the niter constraining form of simple_iv when the exit
11609 is always executed.
11611 2018-04-09 Tom de Vries <tom@codesourcery.com>
11614 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
11615 (define_expand "*memory_barrier"): New define_expand.
11616 (define_insn "memory_barrier"): New insn.
11618 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
11620 PR rtl-optimization/80463
11621 PR rtl-optimization/83972
11622 PR rtl-optimization/83480
11624 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
11625 correct producer for the insn.
11626 (tidy_control_flow): Fixup seqnos in case of debug insns.
11628 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
11630 PR rtl-optimization/83913
11632 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
11633 different sched-times when merging exprs.
11635 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
11637 PR rtl-optimization/83962
11639 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
11640 tidy_fallthru_edge and tidy_control_flow.
11642 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
11644 PR rtl-optimization/83530
11646 * sel-sched.c (force_next_insn): New global variable.
11647 (remove_insn_for_debug): When force_next_insn is true, also leave only
11648 next insn in the ready list.
11649 (sel_sched_region): When the region wasn't scheduled, make another pass
11650 over it with force_next_insn set to 1.
11652 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
11654 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
11656 * config/nds32/constants.md (unspec_volatile_element): Add enum values
11657 for interrupt control.
11658 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
11659 functions for interrupt control.
11660 * config/nds32/nds32-intrinsic.md: Likewise.
11661 * config/nds32/nds32_intrinsic.h: Likewise.
11662 * config/nds32/nds32.h (nds32_builtins): Likewise.
11664 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
11666 * config/nds32/nds32.c (nds32_init_machine_status,
11667 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
11668 strict_aligned_p field.
11669 (nds32_expand_to_rtl_hook): New function.
11670 (TARGET_EXPAND_TO_RTL_HOOK): Define.
11671 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
11673 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
11674 Chung-Ju Wu <jasonwucj@gmail.com>
11676 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
11677 * config/nds32/nds32-n7.md: New file.
11678 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
11679 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
11681 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
11682 * config/nds32/nds32.md (pipeline_model): Add n7.
11683 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
11684 * config/nds32/pipelines.md: Include n7 settings.
11686 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
11687 Chung-Ju Wu <jasonwucj@gmail.com>
11689 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
11690 * config/nds32/nds32-e8.md: New file.
11691 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
11692 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
11694 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
11695 * config/nds32/nds32.md (pipeline_model): Add e8.
11696 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
11697 * config/nds32/pipelines.md: Include e8 settings.
11699 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
11700 Chung-Ju Wu <jasonwucj@gmail.com>
11702 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
11703 * config/nds32/nds32-n8.md: New file.
11704 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
11705 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
11707 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
11708 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
11709 * config/nds32/nds32.md (pipeline_model): Add n8.
11710 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
11711 * config/nds32/pipelines.md: Include n8 settings.
11713 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
11714 Chung-Ju Wu <jasonwucj@gmail.com>
11716 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
11717 * config/nds32/nds32-n9-2r1w.md: New file.
11718 * config/nds32/nds32-n9-3r2w.md: New file.
11719 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
11720 nds32_register_ports): New or modify for cpu n9.
11721 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
11723 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
11724 * config/nds32/nds32-utils.c: New file.
11725 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
11726 TARGET_MUL_SLOW): Define.
11727 * config/nds32/nds32.md (pipeline_model): New attribute.
11728 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
11729 New options that support cpu n9.
11730 * config/nds32/pipelines.md: Include n9 settings.
11731 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
11733 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
11735 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
11736 information if necessary.
11737 (output_cond_branch_compare_zero): Likewise.
11738 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
11739 (nds32_target_alignment): Refine for alignment.
11740 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
11741 (FUNCTION_BOUNDARY): Modify.
11742 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
11744 * config/nds32/nds32.opt (malways-align, malign-functions): New.
11746 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
11748 * config/nds32/constants.md (unspec_volatile_element): Add values for
11749 TLB operation and data prefetch.
11750 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
11751 functions for TLB operation and data prefetch.
11752 * config/nds32/nds32-intrinsic.md: Likewise.
11753 * config/nds32/nds32_intrinsic.h: Likewise.
11754 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
11755 (nds32_print_operand): Likewise.
11756 * config/nds32/nds32.h (nds32_builtins): Likewise.
11758 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
11759 Andrew Pinski <pinsika@gcc.gnu.org>
11761 PR middle-end/82976
11762 * match.pd: Use constant_boolean_node of correct type instead of
11763 boolean_true_node or boolean_false_node for simplifying
11764 pointer comparisons to zero.
11766 2018-04-07 Jakub Jelinek <jakub@redhat.com>
11768 PR tree-optimization/80021
11769 * tree.c (verify_type_variant): Make error call in verify_variant_match
11770 translatable and remove final full stop.
11772 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
11774 * config/nds32/constants.md (unspec_volatile_element): Add
11775 UNSPEC_VOLATILE_EH_RETURN.
11776 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
11777 nds32_output_stack_pop): Support dwarf exception handling process.
11778 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
11779 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
11780 exception handling process.
11781 (nds32_compute_stack_frame): Likewise.
11782 (nds32_return_addr_rtx): Likewise.
11783 (nds32_initial_elimination_offset): Likewise.
11784 (nds32_expand_prologue): Likewise.
11785 (nds32_expand_epilogue): Likewise.
11786 (nds32_dynamic_chain_address): New function.
11787 * config/nds32/nds32.h (machine_function): Add fields for dwarf
11788 exception handling.
11789 (DYNAMIC_CHAIN_ADDRESS): Define.
11790 (EH_RETURN_DATA_REGNO): Define.
11791 (EH_RETURN_STACKADJ_RTX): Define.
11792 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
11793 patterns for dwarf exception handling.
11795 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
11797 * config/nds32/nds32.h: Clean up obsolete macros.
11799 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
11801 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
11802 Add enum values for particular instructions.
11803 * config/nds32/nds32-intrinsic.c: Implementation of expanding
11804 particular intrinsic functions.
11805 * config/nds32/nds32-intrinsic.md: Likewise.
11806 * config/nds32/nds32_intrinsic.h: Likewise.
11807 * config/nds32/nds32.h (nds32_builtins): Likewise.
11808 * config/nds32/nds32.md (type): Add pbsad and pbsada.
11809 (btst, ave): New patterns for particular instructions.
11811 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
11813 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
11814 Add enum values for atomic load/store and memory sync.
11815 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
11817 * config/nds32/nds32-intrinsic.md: Likewise.
11818 * config/nds32/nds32_intrinsic.h: Likewise.
11819 * config/nds32/nds32.h (nds32_builtins): Likewise.
11821 2018-04-07 Jakub Jelinek <jakub@redhat.com>
11823 PR tree-optimization/85257
11824 * fold-const.c (native_encode_vector): If not all elts could fit
11825 and off is -1, return 0 rather than offset.
11826 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
11827 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
11828 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
11829 adjust buffer in native_interpret_expr call.
11831 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
11833 * config/nds32/constants.md (unspec_volatile_element): Add cache
11834 control enum values.
11835 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
11836 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
11837 * config/nds32/nds32.c (nds32_cctl_names): New.
11838 (nds32_print_operand): Handle cache control register names.
11839 * config/nds32/nds32.h (nds32_builtins): New enum values.
11840 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
11842 * config/nds32/nds32.md (type): Add mmu.
11843 * config/nds32/pipelines.md (simple_insn): Add mmu.
11845 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
11847 * config/nds32/nds32.md (type): Remove call.
11848 * config/nds32/pipelines.md (simple_insn): Likewise.
11850 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
11852 * config/nds32/constants.md (unspec_volatile_element): Add
11853 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
11854 UNSPEC_VOLATILE_FMFCFG.
11855 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
11856 description for fmfcfg and fmfcsr.
11857 (bdesc_1arg): Add fmtcsr.
11858 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
11859 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
11860 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
11861 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
11862 unspec_fmfcfg): New patterns.
11863 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
11864 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
11865 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
11866 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
11867 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
11868 __nds32__fmfcfg): Define.
11870 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
11872 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
11873 intrinsic register names.
11874 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
11875 intrinsic register enum values and macros.
11877 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
11879 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
11880 for load/store addressing form.
11881 (nds32_print_operand_address): Likewise.
11883 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
11886 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
11887 based on LABEL_REF. Remove useless assertion.
11888 (pic_address_needs_scratch): Fix formatting.
11889 (sparc_legitimize_pic_address): Minor tweaks.
11890 (sparc_delegitimize_address): Adjust assertion accordingly.
11891 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
11892 into symbolic_operand.
11893 (movsi_high_pic_label_ref): Likewise.
11894 (movsi_lo_sum_pic_label_ref): Likewise.
11895 (movdi_pic_label_ref): Likewise.
11896 (movdi_high_pic_label_ref): Likewise.
11897 (movdi_lo_sum_pic_label_ref): Likewise.
11899 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
11901 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
11902 custom LIB_SPEC setup.
11904 2018-04-06 Ruslan Bukin <br@bsdpad.com>
11905 Kito Cheng <kito.cheng@gmail.com>
11907 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
11908 * config/riscv/freebsd.h: New.
11910 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
11912 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
11913 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
11916 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
11917 Kito Cheng <kito.cheng@gmail.com>
11919 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
11920 nds32_output_call, nds32_symbol_binds_local_p): New functions.
11921 * config/nds32/nds32-protos.h (nds32_output_call,
11922 nds32_output_return): Declare.
11923 * config/nds32/nds32.md: Refine all the call and return patterns.
11925 2018-04-06 Jakub Jelinek <jakub@redhat.com>
11928 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
11929 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
11931 PR rtl-optimization/84872
11932 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
11933 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
11934 EDGE_CROSSING edge.
11936 2018-04-06 Tamar Christina <tamar.christina@arm.com>
11938 * expr.c (copy_blkmode_to_reg): Revert 254862.
11939 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
11941 2018-04-06 Richard Biener <rguenther@suse.de>
11943 PR middle-end/85244
11944 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
11945 after seeing a component reference with an adjacent field. Treat
11946 refs to arrays at struct end of external decls similar to
11947 refs to unconstrained commons.
11949 2018-04-06 Jakub Jelinek <jakub@redhat.com>
11952 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
11953 look through SAVE_EXPRs with non-side-effects argument. Adjust
11955 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
11958 2018-04-06 Richard Biener <rguenther@suse.de>
11960 PR middle-end/85180
11961 * alias.c (find_base_term): New wrapper around find_base_term
11962 unwinding CSELIB_VAL_PTR changes.
11963 (find_base_term): Do not restore CSELIB_VAL_PTR during the
11966 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11968 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
11970 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
11971 constant definitions.
11972 ("nop"): lr 0,0 -> nopr r0
11973 ("nop_lr0", "nop_lr1"): New insn definitions.
11975 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
11977 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
11978 NDS32_V3PUSH_AVAILABLE_P macro.
11980 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
11981 Chung-Ju Wu <jasonwucj@gmail.com>
11983 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
11984 (nds32*-*-*): Add float and fpu_config into supported_defaults.
11985 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
11986 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
11987 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
11988 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
11989 * config/nds32/constraints.md: New constraints and checking for hard
11990 float configuration.
11991 * config/nds32/iterators.md: New mode iterator and attribute for hard
11992 float configuration.
11993 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
11995 * config/nds32/nds32-fpu.md: New file.
11996 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
11997 deal with hard float code generation.
11998 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
12000 (abi_type, float_reg_number): New enum type.
12001 * config/nds32/nds32-predicates.c: New predicates for hard float.
12002 * config/nds32/nds32-protos.h: Declare functions for hard float.
12003 * config/nds32/nds32.c: Implementation for hard float configuration.
12004 * config/nds32/nds32.h: Definitions for hard float configuration.
12005 * config/nds32/nds32.md: Include hard float machine description and
12006 modify patterns for hard float configuration.
12007 * config/nds32/nds32.opt: New options for hard float configuration.
12008 * config/nds32/predicates.md: New predicates for hard float
12011 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
12013 * common/config/nds32/nds32-common.c
12014 (nds32_option_optimization_table): Enable -mreleax-hint by default.
12016 2018-04-05 Jakub Jelinek <jakub@redhat.com>
12018 PR middle-end/85195
12019 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
12020 CONSTRUCTOR_ELT (ctor, ...)->value.
12022 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
12025 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
12027 2018-04-05 Tom de Vries <tom@codesourcery.com>
12030 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
12033 2018-04-05 Shiva Chen <shiva0217@gmail.com>
12034 Kito Cheng <kito.cheng@gmail.com>
12036 * config/nds32/constraints.md (U33): Fine-tune checking condition.
12037 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
12038 * config/nds32/nds32.h (nds32_16bit_address_type): Add
12039 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
12041 2018-04-05 Shiva Chen <shiva0217@gmail.com>
12042 Kito Cheng <kito.cheng@gmail.com>
12044 * config/nds32/constraints.md (Ufe): New memory constraint.
12045 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
12046 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
12047 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
12049 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
12050 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
12052 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
12054 * config/nds32/nds32.md: Use optimize_size in the condition for
12055 alu-shift instructions.
12057 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
12059 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
12061 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
12063 * config/nds32/nds32.md (negsi2): Refine pattern.
12065 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
12066 Chung-Ju Wu <jasonwucj@gmail.com>
12068 * config/nds32/iterators.md (shift_rotate): New code iterator.
12069 (shift): New code attribute.
12070 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
12071 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
12072 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
12073 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
12074 bit-wise operations.
12075 (andsi3, *andsi3): Ditto.
12076 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
12077 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
12078 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
12079 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
12080 nds32_ior_operand, nds32_xor_operand): New predicates.
12082 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
12084 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
12085 (addsi3, subsi3): ... this.
12087 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
12089 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
12091 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
12093 * config/nds32/nds32.md: Adjust indention.
12095 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
12097 * config/nds32/nds32.md (feature): New attribute.
12099 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
12101 * config/nds32/nds32.md (subtype): New attribute.
12103 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
12106 * config/arm/arm-builtins.c (arm_expand_builtin): Change
12107 expansion to perform a bitwise AND of the argument followed by a
12108 boolean negation of the result.
12110 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
12112 PR rtl-optimization/84878
12113 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
12114 the basic block. Assert the use reference is not artificial and that
12115 it has an associated insn.
12117 2018-04-04 Michael Matz <matz@suse.de>
12119 * builtins.c (compute_objsize): Pass correct operand
12120 to array_at_struct_end_p.
12122 2018-04-04 Richard Biener <rguenther@suse.de>
12125 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
12126 from contexts for DINFO_LEVEL_TERSE and below.
12128 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
12130 * config/nds32/nds32-doubleword.md (move_<mode>): Require
12131 resiter_operand condition.
12132 * config/nds32/nds32.md (*move<mode>): Ditto.
12134 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
12135 Monk Chiang <sh.chiang04@gmail.com>
12137 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
12139 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
12141 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
12143 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
12144 Kito Cheng <kito.cheng@gmail.com>
12146 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
12147 nds32_cond_code_str, output_cond_branch,
12148 output_cond_branch_compare_zero, nds32_expand_cbranch,
12149 nds32_expand_cstore, nds32_expand_movcc,
12150 nds32_output_cbranchsi4_equality_zero,
12151 nds32_output_cbranchsi4_equality_reg,
12152 nds32_output_cbranchsi4_equality_reg_or_const_int,
12153 nds32_output_cbranchsi4_greater_less_zero: New functions.
12154 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
12155 nds32_expand_cstore, nds32_expand_movcc,
12156 nds32_output_cbranchsi4_equality_zero,
12157 nds32_output_cbranchsi4_equality_reg,
12158 nds32_output_cbranchsi4_equality_reg_or_const_int,
12159 nds32_output_cbranchsi4_greater_less_zero): Declare.
12160 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
12161 nds32_rimm11s_operand): New predicates.
12162 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
12163 * config/nds32/nds32.md: Rewrite all the branch and conditional move
12166 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
12168 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
12169 * config/nds32/nds32.md: Ditto.
12170 * config/nds32/pipelines.md: Ditto.
12172 2018-04-04 Richard Biener <rguenther@suse.de>
12174 PR tree-optimization/85168
12175 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
12176 propagating abnormals.
12178 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
12180 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
12182 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
12183 Kito Cheng <kito.cheng@gmail.com>
12185 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
12186 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
12187 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
12188 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
12189 * config/nds32/nds32.md (sibcall_internal): New.
12190 (sibcall_register): Remove.
12191 (sibcall_immediate): Remove.
12192 (sibcall_value_internal): New.
12193 (sibcall_value_register): Remove.
12194 (sibcall_value_immediate): Remove.
12195 * config/nds32/predicates.md (nds32_general_register_operand): New.
12196 (nds32_call_address_operand): New.
12198 2018-04-03 Jakub Jelinek <jakub@redhat.com>
12200 PR rtl-optimization/85167
12201 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
12202 bb_defs if *split_p, instead preinitialize it to NULL.
12204 PR tree-optimization/85156
12205 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
12206 evaluating the argument multiple times.
12208 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
12210 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
12212 (_mm_cvtpd_ps): Likewise.
12213 (_mm_cvttpd_epi32): Likewise.
12214 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
12215 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
12216 vector, pixel, and bool following altivec.h include.
12218 2018-04-03 Martin Sebor <msebor@redhat.com>
12220 * doc/extend.texi (Common Function Attributes): Clarify.
12221 (const attribute): Likewise.
12222 (pure attribute): Likewise.
12224 2018-04-03 Jakub Jelinek <jakub@redhat.com>
12227 * config/i386/i386.c (ix86_expand_vector_set): Use
12228 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
12230 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
12232 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
12233 instructions when changing rounding bits to preserve precision bits
12234 in the x87 control word.
12236 2018-04-03 Martin Liska <mliska@suse.cz>
12238 PR tree-optimization/82491
12239 * rtl.h (strip_offset_and_add): Replace += suboffset with
12240 poly_uint64 () + suboffset.
12242 2018-03-29 Martin Liska <mliska@suse.cz>
12243 Martin Jambor <mjambor@suse.cz>
12246 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
12247 param_type is not an integral or pointer type.
12249 2018-04-03 Richard Biener <rguenther@suse.de>
12251 * sese.h (recompute_all_dominators): Remove.
12253 2018-04-02 Martin Sebor <msebor@redhat.com>
12255 * doc/invoke.texi (-Wrestrict): Fix typos.
12257 2018-04-02 Jim Wilson <jimw@sifive.com>
12259 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
12260 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
12261 (<optab>di3, <optab>si3_extend): Likewise.
12262 (<optab>si3_mask, <optab>si3_mask_1): New.
12263 (<optab>di3_mask, <optab>di3_mask_1): New.
12264 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
12265 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
12266 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
12268 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
12270 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
12273 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
12275 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
12276 (nds32_canonicalize_comparison): New function.
12278 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
12279 Kito Cheng <kito.cheng@gmail.com>
12280 Kuan-Lin Chen <kuanlinchentw@gmail.com>
12282 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
12283 * config/nds32/constants.md (unspec_volatile_element): Add
12284 UNSPEC_VOLATILE_RELAX_GROUP.
12285 * config/nds32/nds32-relax-opt.c: New file.
12286 * config/nds32/nds32-predicates.c
12287 (nds32_symbol_load_store_p): New function.
12288 * config/nds32/nds32-protos.h
12289 (nds32_symbol_load_store_p): Declare function.
12290 (make_pass_nds32_relax_opt): Declare new rtl pass function.
12291 * config/nds32/nds32.c
12292 (nds32_register_pass): New function to register pass.
12293 (nds32_register_passes): New function to register passes.
12294 * config/nds32/nds32.md (relax_group): New pattern.
12295 * config/nds32/nds32.opt (mrelax-hint): New option.
12296 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
12298 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
12300 * config/nds32/t-nds32: Modify files dependency.
12302 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
12304 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
12305 (PROFILE_HOOK): Define its implementation.
12307 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
12309 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
12310 type and 32-bit size.
12312 2018-04-01 Jakub Jelinek <jakub@redhat.com>
12314 PR middle-end/85090
12315 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
12316 (V_128_256): New mode iterator.
12317 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
12318 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
12319 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
12321 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
12322 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
12324 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
12327 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
12328 NaN inputs correctly.
12330 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
12333 * config/rs6000/vsx.md (??r): New mode attribute.
12334 (*vsx_mov<mode>_64bit): Use it.
12335 (*vsx_mov<mode>_32bit): Likewise.
12337 2018-03-30 Martin Sebor <msebor@redhat.com>
12339 PR tree-optimization/84818
12340 * builtins.c (check_access): Use warning_n.
12342 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12345 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
12347 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
12350 2018-03-30 Julia Koval <julia.koval@intel.com>
12353 * x86-tune.def (movx, partial_reg_dependency): Enable for
12356 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
12358 PR inline-asm/84985
12359 * lra-constraints.c (process_alt_operands): Move setting
12360 this_alternative_matches below.
12362 2018-03-29 Martin Liska <mliska@suse.cz>
12365 * doc/invoke.texi: Document how LTO works with debug info.
12366 Describe auto-load support of binutils. Mention 'x86-64'
12367 as valid option value of -march option.
12369 2018-03-29 Jakub Jelinek <jakub@redhat.com>
12371 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
12374 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
12375 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
12376 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
12379 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
12382 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
12383 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
12384 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
12385 for RS6000_BTM_POWERPC64.
12386 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
12387 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
12388 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
12391 (DIVDEU): Likewise.
12393 2018-03-28 Carl Love <cel@us.ibm.com>
12396 2017-09-27 Carl Love <cel@us.ibm.com>
12398 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
12399 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
12400 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
12403 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12405 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
12406 instead of __vector bool.
12407 (_mm_max_pu8): Likewise.
12408 (_mm_min_pi16): Likewise.
12410 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
12413 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
12414 (DIVWEUO): Likewise.
12415 (DIVDEO): Likewise.
12416 (DIVDEUO): Likewise.
12417 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
12418 DIVWEUO and DIVDEUO.
12419 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
12420 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
12421 (div_extend): Likewise.
12422 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
12424 (__builtin_divweuo): Likewise.
12425 (__builtin_divdeo): Likewise.
12426 (__builtin_divdeuo): Likewise.
12428 2018-03-28 Jakub Jelinek <jakub@redhat.com>
12431 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
12432 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
12434 PR tree-optimization/82004
12435 * gimple-match-head.c (optimize_pow_to_exp): New function.
12436 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
12437 Don't fold to exp if optimize_pow_to_exp is false.
12439 2018-03-28 Martin Liska <mliska@suse.cz>
12442 * calls.c (initialize_argument_information): Fix trailing space.
12443 * common.opt: Fix typo and provide better explanation for
12444 -fsanitize-coverage option.
12445 * config/i386/i386.opt: Fix typo.
12447 2018-03-28 Jakub Jelinek <jakub@redhat.com>
12448 Martin Liska <mliska@suse.cz>
12451 * gimplify.c (asan_poison_variable): Don't do the check for
12452 gimplify_omp_ctxp here.
12453 (gimplify_decl_expr): Do it here.
12454 (gimplify_target_expr): Likewise.
12456 2018-03-28 Martin Liska <mliska@suse.cz>
12459 * config/i386/i386.c (ix86_function_arg_advance): Do not call
12460 chkp_type_bounds_count if MPX is not enabled.
12462 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
12464 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
12466 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
12469 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
12470 function to create the function decl for complex long double
12471 multiply and divide for -mabi=ieeelongdouble.
12472 (init_float128_ieee): Call it.
12474 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
12477 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
12478 -fcf-protection=branch -mibt.
12479 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
12481 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
12484 * config/arm/arm.c (arm_valid_symbolic_address): Handle
12485 arm_word_relocations.
12487 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
12490 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
12491 extern array declarations.
12493 2018-03-27 Richard Biener <rguenther@suse.de>
12495 PR middle-end/84067
12496 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
12497 explicit single_use checks.
12499 2018-03-27 Richard Biener <rguenther@suse.de>
12501 PR tree-optimization/85082
12502 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
12505 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
12507 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
12508 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
12509 Turn on fasynchronous-unwind-tables and funwind-tables.
12511 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
12514 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
12515 (*bmi_blsr_<mode>_ccz): Ditto.
12517 2018-03-26 Tom de Vries <tom@codesourcery.com>
12519 PR tree-optimization/85063
12520 * omp-general.c (offloading_function_p): New function. Factor out
12522 * omp-offload.c (pass_omp_target_link::gate): ... here.
12523 * omp-general.h (offloading_function_p): Declare.
12524 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
12525 with attribute omp declare target for offloading functions.
12527 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
12529 PR tree-optimization/84005
12530 * tree-data-ref.h (get_base_for_alignment): Declare.
12531 * tree-data-ref.c (get_base_for_alignment_1): New function.
12532 (get_base_for_alignment): Likewise.
12533 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
12534 get_base_for_alignment to find a suitable base object, instead
12535 of always using drb->base_address.
12537 2018-03-23 Jakub Jelinek <jakub@redhat.com>
12539 PR inline-asm/85022
12540 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
12541 known size by default.
12543 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
12545 PR inline-asm/85030
12546 * lra-constraints.c (process_alt_operands): Don't match BLKmode
12547 and non BLKmode operands.
12549 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12552 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
12553 Clean up attributes.
12555 2018-03-23 Richard Biener <rguenther@suse.de>
12558 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
12559 we are going to emit early debug for LTO.
12561 2018-03-23 Jakub Jelinek <jakub@redhat.com>
12563 PR inline-asm/85034
12564 * function.c (match_asm_constraints_1): Don't optimize if input
12565 doesn't satisfy general_operand predicate for output's mode.
12567 PR inline-asm/85022
12568 * alias.c (write_dependence_p): Don't require for x_canonicalized
12569 non-VOIDmode if x has VOIDmode.
12572 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
12573 just don't try to optimize it rather than assert it never happens.
12575 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
12577 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
12578 macro expansions for definition of ST_INTERNAL_<mode> and
12579 LD_INTERNAL_<mode> builtins.
12580 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
12582 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
12584 (altivec_expand_st_builtin): Likewise.
12585 (altivec_expand_builtin): Remove calls to deleted functions.
12586 (rs6000_address_for_altivec): Delete this function.
12587 * config/rs6000/vector.md: Remove expands for
12588 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
12590 2018-03-22 Sudakshina Das <sudi.das@arm.com>
12593 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
12594 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
12595 re-computing once computed.
12596 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
12597 (arm_init_machine_status): Initialize
12598 machine->static_chain_stack_bytes.
12600 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
12603 * doc/extend.texi: Add four new prototypes for vec_ld.
12604 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
12605 definitions for more logical presentation.
12606 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
12607 entries for V1TI variants of __builtin_altivec_ld builtin.
12608 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
12609 handling of V1TI variant of LVX icode pattern.
12610 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
12611 (rs6000_gimple_fold_builtin): Likewise.
12612 (altivec_init_builtins): Add code to define
12613 __builtin_altivec_lvx_v1ti function.
12615 2018-03-22 Jakub Jelinek <jakub@redhat.com>
12617 PR inline-asm/84941
12618 * function.c (match_asm_constraints_1): Don't do the optimization
12619 if input isn't a REG, SUBREG, MEM or constant.
12621 2018-03-22 Tom de Vries <tom@codesourcery.com>
12623 PR tree-optimization/84956
12624 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
12625 bb_has_abnormal_pred.
12627 2018-03-22 Jakub Jelinek <jakub@redhat.com>
12630 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
12631 DECL_INITIAL (decl) to decl at the end.
12632 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
12633 adjust the comment.
12635 2018-03-21 Joseph Myers <joseph@codesourcery.com>
12637 * doc/extend.texi (__builtin_tgmath): Document when complex
12638 integer types are treated as _Complex _Float64.
12640 2018-03-21 Tom de Vries <tom@codesourcery.com>
12642 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
12644 2018-03-21 Jakub Jelinek <jakub@redhat.com>
12646 PR tree-optimization/84960
12647 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
12648 if it is ENTRY block, move them into single succ of ENTRY in that case.
12650 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
12652 PR tree-optimization/84811
12653 * poly-int.h (poly_span_traits): Remove the T3 parameter and
12654 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
12655 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
12656 (known_subrange_p): Update accordingly. Cast each value involved
12657 in the size comparison, rather than casting the result of the
12660 2018-03-21 Jakub Jelinek <jakub@redhat.com>
12662 PR tree-optimization/84982
12663 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
12664 by flipping the least significant bit rather than all bits from
12665 bitpos to bitpos + bitsize - 1.
12667 2018-03-21 Nathan Sidwell <nathan@acm.org>
12669 * doc/extend.texi (Deprecated Features): Remove mention of
12670 long-deleted deprecations.
12672 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12675 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
12676 * configure: Regenerate.
12678 2018-03-21 Tom de Vries <tom@codesourcery.com>
12680 PR tree-optimization/83126
12681 * tree-parloops.c (num_phis): New function.
12682 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
12684 2018-03-21 Nathan Sidwell <nathan@acm.org>
12686 * doc/extend.texi (Deprecated Features): Update deprecated flags,
12687 mention anon-struct/union members and trailing attributes.
12689 2018-03-21 Bin Cheng <bin.cheng@arm.com>
12691 PR tree-optimization/84969
12692 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
12693 builtin memset partitions if they set different rhs values.
12695 2018-03-21 Jakub Jelinek <jakub@redhat.com>
12697 PR rtl-optimization/84989
12698 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
12699 VEC_DUPLICATE with scalar result mode.
12701 2018-03-21 Martin Liska <mliska@suse.cz>
12704 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
12705 not intended return statement.
12707 2018-03-21 Martin Liska <mliska@suse.cz>
12710 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
12711 (chkp_find_bound_slots_1): Limit number of iterations.
12713 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
12716 * Minor grammar fixes for x86 options.
12718 2018-03-20 Jakub Jelinek <jakub@redhat.com>
12721 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
12722 holding REG_CFA_RESTORE notes, instead turn them into a USE.
12724 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
12727 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
12728 (altivec_lvx_<mode>_1op): Likewise.
12729 (altivec_stvx_<mode>_2op): Likewise.
12730 (altivec_stvx_<mode>_1op): Likewise.
12731 (altivec_lvx_<VM2:mode>): New define_expand.
12732 (altivec_stvx_<VM2:mode>): Likewise.
12733 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
12734 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
12735 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
12736 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
12737 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
12738 (rs6000_gen_lvx): Likewise.
12739 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
12740 (altivec_expand_stv_builtin): Likewise.
12741 (altivec_expand_builtin): Likewise.
12742 * config/rs6000/vector.md: Likewise.
12744 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12747 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
12750 2018-03-20 Richard Biener <rguenther@suse.de>
12753 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
12754 sign-conversions as zero, fall back to standard scalar_stmt
12757 2018-03-20 Martin Liska <mliska@suse.cz>
12760 * predict.c (rebuild_frequencies): Handle case when we have
12761 PROFILE_ABSENT, but flag_guess_branch_prob is false.
12763 2018-03-20 Jakub Jelinek <jakub@redhat.com>
12766 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
12767 flag_section_anchors.
12768 * varasm.c (use_blocks_for_decl_p): Remove hack for
12769 dw2_force_const_mem.
12772 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
12774 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
12775 be created, use lowpart_subreg of operands[0] rather than operands[0]
12777 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
12778 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
12779 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
12780 and n constraint instead of aarch64_shift_imm_di and Usd.
12781 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
12782 (*aarch64_<optab>_reg_minus<mode>3): ... this.
12784 2018-03-20 Sudakshina Das <sudi.das@arm.com>
12787 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
12788 to favor GPR over NEON registers.
12789 (<shift>di3_neon): Likewise.
12791 2018-03-20 Tom de Vries <tom@codesourcery.com>
12794 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
12795 (nvptx_process_pars): Emit bar.sync asap and alap.
12797 2018-03-20 Tom de Vries <tom@codesourcery.com>
12800 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
12801 seen_label if seen_label is already set.
12803 2018-03-20 Jakub Jelinek <jakub@redhat.com>
12806 * config/i386/i386.c (fold_builtin_cpu): For features above 31
12807 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
12808 Use 1U instead of 1. Formatting fixes.
12811 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
12812 instead of TREE_TYPE (s1) for the return value.
12814 2018-03-19 Jakub Jelinek <jakub@redhat.com>
12816 PR tree-optimization/84946
12817 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
12818 bitsize + bitsize in poly_uint64 rather than poly_int64.
12821 * dwarf2asm.c: Include fold-const.c.
12822 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
12823 of decl rather than decl itself.
12825 PR rtl-optimization/84643
12826 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
12828 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
12831 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
12832 calling assemble_variable.
12834 2018-03-19 Sudakshina Das <sudi.das@arm.com>
12837 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
12838 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
12840 2018-03-19 Jim Wilson <jimw@sifive.com>
12843 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
12844 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
12845 (riscv_first_stack_step): Likewise.
12846 (riscv_option_override): Use STACK_BOUNDARY instead of
12847 MIN_STACK_BOUNDARY.
12848 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
12849 MIN_STACK_BOUNDARY.
12850 (BIGGEST_ALIGNMENT): Set to 128.
12851 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
12852 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
12855 2018-03-19 Richard Biener <rguenther@suse.de>
12857 PR tree-optimization/84933
12858 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
12859 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
12861 2018-03-19 Richard Biener <rguenther@suse.de>
12863 PR tree-optimization/84859
12864 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
12865 (cond_if_else_store_replacement): Perform sinking operation on
12866 single-store BBs regardless of MAX_STORES_TO_SINK setting.
12867 Generalize what a BB with a single eligible store is.
12869 2018-03-19 Richard Biener <rguenther@suse.de>
12871 PR tree-optimization/84929
12872 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
12873 chrec_is_positive against non-chrec arg.
12875 2018-03-19 Tamar Christina <tamar.christina@arm.com>
12878 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
12880 2018-03-18 Martin Liska <mliska@suse.cz>
12882 PR rtl-optimization/84635
12883 * regrename.c (build_def_use): Use matches_mode only when
12886 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
12888 PR tree-optimization/84913
12889 * tree-vect-loop.c (vectorizable_reduction): Don't try to
12890 vectorize chains of COND_EXPRs.
12892 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
12894 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
12896 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
12898 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
12900 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
12902 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
12904 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
12905 Kito Cheng <kito.cheng@gmail.com>
12907 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
12908 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
12909 (nds32_adjust_reg_alloc_order): New function.
12910 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
12912 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
12914 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
12915 nds32_print_operand, nds32_print_operand_address): Use
12916 HOST_WIDE_INT_PRINT_DEC instead.
12918 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
12920 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
12922 2018-03-17 Jakub Jelinek <jakub@redhat.com>
12925 * config/i386/i386.c (initial_ix86_tune_features,
12926 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
12927 unsigned long long.
12928 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
12929 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
12930 rather than 1u << ix86_tune. Formatting fix.
12931 (ix86_option_override_internal): Change ix86_arch_mask from
12932 unsigned int to unsigned HOST_WIDE_INT, initialize to
12933 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
12934 (ix86_function_specific_restore): Likewise.
12936 2018-03-16 Jakub Jelinek <jakub@redhat.com>
12939 * postreload.c (reload_combine_recognize_pattern): Perform
12940 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
12941 truncate_int_for_mode the result for the destination's mode.
12944 * hsa-gen.c (mem_type_for_type): Fix comment typo.
12945 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
12947 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
12950 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
12953 * lra-assigns.c (lra_split_hard_reg_for): Don't use
12954 regno_allocno_class_array and sorted_pseudos.
12955 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
12956 insns where regno is used.
12958 2018-03-16 Martin Liska <mliska@suse.cz>
12961 * multiple_target.c (create_dispatcher_calls): Redirect
12962 reference in the symbol table.
12964 2018-03-16 Martin Liska <mliska@suse.cz>
12967 * multiple_target.c (create_dispatcher_calls): Redirect also
12970 2018-03-16 Jakub Jelinek <jakub@redhat.com>
12974 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
12975 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
12976 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
12978 2018-03-16 Julia Koval <julia.koval@intel.com>
12980 * doc/invoke.texi (Skylake Server): Add CLWB.
12981 Cannonlake): Remove CLWB.
12983 2018-03-16 Jakub Jelinek <jakub@redhat.com>
12985 PR tree-optimization/84841
12986 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
12988 (FLOAT_ONE_CONST_TYPE): Define.
12989 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
12990 (sort_by_operand_rank): Put entries with higher constant_type last
12991 rather than first to match comments.
12993 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
12995 * config/nios2/nios2.md (movsi_internal): Fix thinko in
12998 2018-03-15 Jakub Jelinek <jakub@redhat.com>
13001 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
13002 check and use address of target always.
13004 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
13007 * config/i386/i386.c (indirect_thunk_needed): Update comments.
13008 (indirect_thunk_bnd_needed): Likewise.
13009 (indirect_thunks_used): Likewise.
13010 (indirect_thunks_bnd_used): Likewise.
13011 (indirect_return_needed): New.
13012 (indirect_return_bnd_needed): Likewise.
13013 (output_indirect_thunk_function): Add a bool argument for
13015 (output_indirect_thunk_function): Don't generate alias for
13016 function return thunk.
13017 (ix86_code_end): Call output_indirect_thunk_function to generate
13018 function return thunks.
13019 (ix86_output_function_return): Set indirect_return_bnd_needed
13020 and indirect_return_needed instead of indirect_thunk_bnd_needed
13021 and indirect_thunk_needed.
13023 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
13025 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
13026 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
13027 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
13029 2018-03-15 David Malcolm <dmalcolm@redhat.com>
13030 Paul Hua <paul.hua.gm@gmail.com>
13033 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
13035 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
13037 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
13038 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
13039 resp. SFmode cases.
13041 2018-03-15 Tamar Christina <tamar.christina@arm.com>
13044 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
13045 instead of GET_MODE_SIZE when comparing Units.
13047 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
13050 * varasm.c (hash_section): Return an unchangeble hash value
13051 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
13052 Return !aarch64_can_use_per_function_literal_pools_p ().
13054 2018-03-15 Jakub Jelinek <jakub@redhat.com>
13057 * optabs.c (emit_conditional_move): Pass address of cmode's copy
13058 rather than address of cmode as last argument to prepare_cmp_insn.
13060 2018-03-15 Julia Koval <julia.koval@intel.com>
13062 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
13063 F_AVX512VNNI, F_AVX512BITALG): New.
13065 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
13068 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
13069 insn for floating-point loads and stores.
13071 2018-03-14 Carl Love <cel@us.ibm.com>
13073 * config/rs6000/rs6000-c.c: Add macro definitions for
13074 ALTIVEC_BUILTIN_VEC_PERMXOR.
13075 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
13076 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
13077 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
13078 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
13080 * config/doc/extend.texi: Add prototypes for vec_permxor.
13082 2018-03-14 David Malcolm <dmalcolm@redhat.com>
13085 * diagnostic-show-locus.c (class layout_point): Convert m_line
13086 from int to linenum_type.
13087 (line_span::comparator): Use linenum "compare" function when
13088 comparing line numbers.
13089 (test_line_span): New function.
13090 (layout_range::contains_point): Convert param "row" from int to
13092 (layout_range::intersects_line_p): Likewise.
13093 (layout::will_show_line_p): Likewise.
13094 (layout::print_source_line): Likewise.
13095 (layout::should_print_annotation_line_p): Likewise.
13096 (layout::print_annotation_line): Likewise.
13097 (layout::print_leading_fixits): Likewise.
13098 (layout::annotation_line_showed_range_p): Likewise.
13099 (struct line_corrections): Likewise for field m_row.
13100 (line_corrections::line_corrections): Likewise for param "row".
13101 (layout::print_trailing_fixits): Likewise.
13102 (layout::get_state_at_point): Likewise.
13103 (layout::get_x_bound_for_row): Likewise.
13104 (layout::print_line): Likewise.
13105 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
13106 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
13107 * input.c (selftest::test_linenum_comparisons): New function.
13108 (selftest::input_c_tests): Call it.
13109 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
13110 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
13111 * selftest.h (ASSERT_GT): New macro.
13112 (ASSERT_GT_AT): New macro.
13113 (ASSERT_LT): New macro.
13114 (ASSERT_LT_AT): New macro.
13116 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
13118 PR rtl-optimization/84780
13119 * combine.c (distribute_links): Don't make a link based on pc_rtx.
13121 2018-03-14 Martin Liska <mliska@suse.cz>
13123 * tree.c (record_node_allocation_statistics): Use
13124 get_stats_node_kind.
13125 (get_stats_node_kind): New function extracted from
13126 record_node_allocation_statistics.
13127 (free_node): Use get_stats_node_kind.
13129 2018-03-14 Richard Biener <rguenther@suse.de>
13131 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
13132 that the value-set of ANTIC_IN doesn't grow.
13135 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
13137 (BB_VISITED_WITH_VISITED_SUCCS): New define.
13138 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
13140 2018-03-14 Julia Koval <julia.koval@intel.com>
13142 * config.gcc (icelake-client, icelake-server): New.
13144 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
13145 (initial_ix86_arch_features): Ditto.
13146 (PTA_SKYLAKE): Add SGX.
13147 (PTA_ICELAKE): Remove.
13148 (PTA_ICELAKE_CLIENT): New.
13149 (PTA_ICELAKE_SERVER): New.
13150 (ix86_option_override_internal): Split up icelake on icelake client and
13152 (get_builtin_code_for_version): Ditto.
13153 (fold_builtin_cpu): Ditto.
13154 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
13155 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
13156 * config/i386/i386.h (processor_type): Ditto.
13157 * doc/invoke.texi: Ditto.
13159 2018-03-14 Jakub Jelinek <jakub@redhat.com>
13162 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
13163 INTEGER_CST offset, add it together with bitpos / 8 and
13164 sign extend based on POINTER_SIZE.
13168 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
13171 * config/i386/constraints.md (Yc): New register constraint.
13172 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
13173 Use Yc constraint for alternative 2 of operand 0. Remove
13174 preferred_for_speed attribute.
13176 2018-03-14 Richard Biener <rguenther@suse.de>
13178 PR tree-optimization/84830
13179 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
13180 with the old one to avoid oscillations.
13182 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
13185 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
13187 (assign_by_spills): Return a flag of reload assignment failure.
13188 Do not process the reload assignment failures. Do not spill other
13189 reload pseudos if they has the same reg class. Update n if
13191 (lra_assign): Add a return arg. Set up from the result of
13192 assign_by_spills call.
13193 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
13194 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
13195 usage_insns if it is not NULL.
13196 (spill_hard_reg_in_range): New function.
13197 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
13198 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
13199 function prototypes.
13200 (lra_assign): Change prototype.
13201 * lra.c (lra): Add code to deal with fails by splitting hard reg
13204 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
13206 * config/riscv/riscv.opt (mrelax): New option.
13207 * config/riscv/riscv.c (riscv_file_start): Emit ".option
13208 "norelax" when riscv_mrelax is disabled.
13209 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
13211 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13214 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
13215 reassociation for int modes.
13217 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
13219 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
13220 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
13222 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
13223 * config/aarch64/aarch64-sve.md
13224 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
13225 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
13226 (*extend<mode><Vwide>2): Rename to...
13227 (aarch64_sve_extend<mode><Vwide>2): ...this.
13228 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
13229 renaming the old pattern to...
13230 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
13232 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
13233 define_expand, renaming the old pattern to...
13234 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
13235 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
13236 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
13237 account when deciding which SVE instruction the optab should use.
13238 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
13240 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
13242 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
13243 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
13244 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
13245 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
13246 (tlsdesc_small_<mode>): Turn a define_expand and use
13247 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
13248 (tlsdesc_small_advsimd_<mode>): ...this.
13249 (tlsdesc_small_sve_<mode>): New pattern.
13251 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
13253 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
13254 (UNSPEC_UMUL_HIGHPART): New constants.
13255 (MUL_HIGHPART): New int iteraor.
13256 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
13257 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
13259 (*<su>mul<mode>3_highpart): New define_insn.
13261 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
13264 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
13267 2018-03-13 Martin Liska <mliska@suse.cz>
13270 * (sem_item_optimizer::sem_item_optimizer): Initialize new
13272 (sem_item_optimizer::~sem_item_optimizer): Release it.
13273 (sem_item_optimizer::merge_classes): Register variable aliases.
13274 (sem_item_optimizer::fixup_pt_set): New function.
13275 (sem_item_optimizer::fixup_points_to_sets): Likewise.
13276 * ipa-icf.h: Declare new variables and functions.
13278 2018-03-13 Jakub Jelinek <jakub@redhat.com>
13280 PR middle-end/84834
13281 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
13282 integer_pow2p@2 and test integer_pow2p in condition.
13283 (A < 0 ? C : 0): Similarly for @1.
13285 PR middle-end/84831
13286 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
13287 characters starting at p contain '\0' character, don't look beyond
13291 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
13292 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
13295 * reg-stack.c (change_stack): Change update_end var from int to
13296 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
13297 also call set_block_for_insn on the newly added insns and rescan.
13300 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
13301 on the last operand.
13304 * tree.c (stabilize_reference_1): Return save_expr (e) for
13305 STATEMENT_LIST even if it doesn't have side-effects.
13307 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
13309 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
13311 2018-03-12 Renlin Li <renlin.li@arm.com>
13313 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
13314 aarch64_output_scalar_simd_mov_immediate.
13316 2018-03-12 Martin Sebor <msebor@redhat.com>
13318 PR tree-optimization/83456
13319 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
13320 for perfectly overlapping calls to memcpy.
13321 (gimple_fold_builtin_memory_chk): Same.
13322 (gimple_fold_builtin_strcpy): Handle no-warning.
13323 (gimple_fold_builtin_stxcpy_chk): Same.
13324 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
13326 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
13328 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
13329 parameter. Use it for SFmode.
13330 (rs6000_function_arg_advance_1): Adjust.
13331 (rs6000_function_arg): Adjust.
13332 (rs6000_gimplify_va_arg): Pass false for that new parameter.
13334 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
13336 PR rtl-optimization/84169
13337 PR rtl-optimization/84780
13338 * combine.c (can_combine_p): Check for a 2-insn combination whether
13339 the destination register is used between the two insns, too.
13341 2018-03-12 Richard Biener <rguenther@suse.de>
13343 PR tree-optimization/84803
13344 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
13345 for refs DR analysis didn't process.
13347 2018-03-12 Richard Biener <rguenther@suse.de>
13349 PR tree-optimization/84777
13350 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
13351 force-vectorize loops ignore whether we are optimizing for size.
13353 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
13355 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
13356 (TARGET_MD_ASM_ADJUST): Define.
13358 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
13359 Kito Cheng <kito.cheng@gmail.com>
13360 Chung-Ju Wu <jasonwucj@gmail.com>
13362 * config/nds32/nds32.c (nds32_compute_stack_frame,
13363 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
13364 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
13365 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
13366 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
13367 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
13368 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
13369 * config/nds32/nds32.md (prologue, epilogue): Use macro
13370 NDS32_V3PUSH_AVAILABLE_P to do checking.
13372 2018-03-11 Jakub Jelinek <jakub@redhat.com>
13375 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
13376 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
13377 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
13378 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
13379 addition of most attributes on !orig_type_die or the attribute not
13380 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
13382 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
13383 Chung-Ju Wu <jasonwucj@gmail.com>
13385 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
13386 __NDS32_VH__ macro.
13387 * config/nds32/nds32.opt (mvh): New option.
13389 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
13390 Chung-Ju Wu <jasonwucj@gmail.com>
13392 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
13394 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
13395 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
13398 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
13399 Chung-Ju Wu <jasonwucj@gmail.com>
13401 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
13403 * config/nds32/nds32-multiple.md (strlensi): New pattern.
13404 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
13406 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
13407 Kito Cheng <kito.cheng@gmail.com>
13408 Chung-Ju Wu <jasonwucj@gmail.com>
13410 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
13411 UNSPEC_FFMISM and UNSPEC_FLMISM.
13412 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
13413 for ffb, ffmism and flmism.
13414 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
13415 (unspec_ffmism): Ditto.
13416 (unspec_flmism): Ditto.
13417 (nds32_expand_builtin_impl): Check if string extension is available.
13418 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
13419 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
13421 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
13424 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
13427 * lra-assigns.c (assign_by_spills): Return a flag of reload
13428 assignment failure. Do not process the reload assignment
13429 failures. Do not spill other reload pseudos if they has the same
13431 (lra_assign): Add a return arg. Set up from the result of
13432 assign_by_spills call.
13433 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
13434 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
13435 usage_insns if it is not NULL.
13436 (spill_hard_reg_in_range): New function.
13437 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
13438 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
13439 function prototypes.
13440 (lra_assign): Change prototype.
13441 * lra.c (lra): Add code to deal with fails by splitting hard reg
13444 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
13447 * config/i386/i386.opt: Replace Enforcment with Enforcement.
13449 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
13452 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
13453 (dw_val_node): Add val_symbolic_view.
13454 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
13455 (symview_upper_bound): New.
13456 (new_line_info_table): Initialize symviews_since_reset.
13457 (dwarf2out_source_line): Count symviews_since_reset and set
13458 symview_upper_bound.
13459 (dw_val_equal_p): Handle symview.
13460 (add_AT_symview): New.
13461 (print_dw_val): Handle symview.
13462 (attr_checksum, attr_checksum_ordered): Likewise.
13463 (same_dw_val_p, size_of_die): Likewise.
13464 (value_format, output_die): Likewise.
13465 (add_high_low_attributes): Use add_AT_symview for entry_view.
13466 (dwarf2out_finish): Reset symview_upper_bound, clear
13469 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
13472 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
13473 Add strict argument and use it.
13474 (rs6000_split_multireg_move): Update for new strict argument.
13475 (mem_operand_gpr): Disallow all non-offsettable addresses.
13476 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
13478 2018-03-09 Jakub Jelinek <jakub@redhat.com>
13481 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
13482 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
13483 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
13486 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
13487 decl, use remap_type if we want to use the type.
13489 2018-03-09 Martin Sebor <msebor@redhat.com>
13491 PR tree-optimization/84526
13492 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
13494 (builtin_access::generic_overlap): Be prepared to handle non-array
13497 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
13499 PR rtl-optimization/84682
13500 * lra-constraints.c (process_address_1): Check is_address flag
13501 for address constraints.
13502 (process_alt_operands): Likewise.
13503 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
13504 preprocess_constraints.
13505 * recog.h (preprocess_constraints): Add oploc parameter.
13507 * recog.c (preprocess_constraints): Test address_operand for
13508 CT_ADDRESS constraints.
13510 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
13513 * lra-assigns.c (assign_by_spills): Return a flag of reload
13514 assignment failure. Do not process the reload assignment
13515 failures. Do not spill other reload pseudos if they has the same
13517 (lra_assign): Add a return arg. Set up from the result of
13518 assign_by_spills call.
13519 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
13520 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
13521 usage_insns if it is not NULL.
13522 (spill_hard_reg_in_range): New function.
13523 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
13524 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
13525 function prototypes.
13526 (lra_assign): Change prototype.
13527 * lra.c (lra): Add code to deal with fails by splitting hard reg
13530 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13533 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
13534 Accept complain bool parameter. Only emit errors if it is true.
13535 (arm_parse_cpu_option_name): Likewise.
13536 (arm_target_thumb_only): Adjust callers of the above.
13537 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
13538 prototype to take a default true bool parameter.
13539 (arm_parse_arch_option_name): Likewise.
13541 2018-03-09 David Malcolm <dmalcolm@redhat.com>
13542 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
13546 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
13547 * configure: Regenerate.
13548 * configure.ac ("linker --version-script option"): New.
13549 ("linker soname option"): New.
13551 2018-03-09 Richard Biener <rguenther@suse.de>
13553 PR tree-optimization/84775
13554 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
13555 immediate uses of predicate stmts and mark them modified.
13558 PR tree-optimization/84178
13559 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
13561 (version_loop_for_if_conversion): Delay update_ssa call.
13562 (tree_if_conversion): Delay update_ssa until after predicate
13565 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
13568 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
13569 when the function accesses prior frames.
13571 2018-03-08 Jakub Jelinek <jakub@redhat.com>
13574 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
13575 gen_llsym, otherwise call maybe_gen_llsym.
13577 PR inline-asm/84742
13578 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
13579 has ',' character inside of it.
13581 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13584 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
13585 as clobbering CC_REGNUM.
13587 2018-03-08 Richard Biener <rguenther@suse.de>
13589 PR middle-end/84552
13590 * tree-scalar-evolution.c: Include tree-into-ssa.h.
13591 (follow_copies_to_constant): Do not follow SSA names registered
13594 2018-03-08 Richard Biener <rguenther@suse.de>
13596 PR tree-optimization/84178
13597 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
13599 (version_loop_for_if_conversion): Delay update_ssa call.
13600 (tree_if_conversion): Delay update_ssa until after predicate
13603 2018-03-08 David Malcolm <dmalcolm@redhat.com>
13605 PR tree-optimization/84178
13606 * tree-if-conv.c (release_bb_predicate): Remove the
13607 the assertion that the stmts have NULL use_ops.
13608 Discard the statements, asserting that they haven't
13609 yet been added to a BB.
13611 2018-03-08 Richard Biener <rguenther@suse.de>
13613 PR tree-optimization/84746
13614 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
13615 (phi_translate): Pass in destination ANTIC_OUT set.
13616 (phi_translate_1): Likewise. For a simplified result lookup
13617 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
13618 (phi_translate_set): Adjust.
13619 (do_pre_regular_insertion): Likewise.
13620 (do_pre_partial_partial_insertion): Likewise.
13622 2018-03-08 Martin Liska <mliska@suse.cz>
13624 PR gcov-profile/84735
13625 * doc/gcov.texi: Document usage of profile files.
13626 * gcov-io.h: Document changes in the format.
13628 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
13632 * dwarf2out.c (struct dw_line_info_table): Update comments for
13634 (FORCE_RESET_NEXT_VIEW): New.
13635 (FORCE_RESETTING_VIEW_P): New.
13636 (RESETTING_VIEW_P): Check for -1 too.
13637 (ZERO_VIEW_P): Likewise.
13638 (new_line_info_table): Force-reset next view.
13639 (dwarf2out_begin_function): Likewise.
13640 (dwarf2out_source_line): Simplify zero_view_p initialization.
13641 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
13642 view directly. Omit view when omitting .loc at line 0.
13644 2018-03-08 Jakub Jelinek <jakub@redhat.com>
13646 PR tree-optimization/84740
13647 * tree-switch-conversion.c (process_switch): Call build_constructors
13648 only if info.phi_count is non-zero.
13650 PR tree-optimization/84739
13651 * tree-tailcall.c (find_tail_calls): Check call arguments against
13652 DECL_ARGUMENTS (current_function_decl) rather than
13653 DECL_ARGUMENTS (func) when checking for tail recursion.
13655 2018-03-07 Jakub Jelinek <jakub@redhat.com>
13657 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
13658 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
13659 Volker Reichelt's entry and add entries for people that perform
13660 GCC fuzzy testing and report numerous bugs.
13662 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
13665 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
13666 readonly data in sdata, if that is disabled.
13667 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
13668 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
13669 -mreadonly-in-sdata option.
13671 2018-03-07 Martin Sebor <msebor@redhat.com>
13673 PR tree-optimization/84468
13674 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
13675 basic block when looking for nul assignment.
13677 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
13680 * except.h (output_function_exception_table): Adjust prototype.
13681 * except.c (output_function_exception_table): Remove FNNAME parameter
13682 and add SECTION parameter. Ouput one part of the table at a time.
13683 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
13684 the first part of the exception table and emit unwind directives.
13685 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
13686 (i386_pe_seh_cold_init): Likewise.
13687 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
13688 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
13689 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
13690 (ix86_output_call_insn): Emit a nop in one more case for SEH.
13691 * config/i386/winnt.c: Include except.h.
13692 (struct seh_frame_state): Add reg_offset, after_prologue and
13693 in_cold_section fields.
13694 (i386_pe_seh_end_prologue): Set seh->after_prologue.
13695 (i386_pe_seh_cold_init): New function.
13696 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
13697 to seh->in_cold_section.
13698 (seh_emit_push): Record the offset of the push.
13699 (seh_emit_save): Record the offet of the save.
13700 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
13701 Test seh->after_prologue to disregard the epilogue.
13702 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
13703 (i386_pe_end_cold_function): New function.
13705 2018-03-07 Jakub Jelinek <jakub@redhat.com>
13708 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
13709 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
13712 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
13714 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
13715 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
13717 PR middle-end/84723
13718 * multiple_target.c: Include tree-inline.h and intl.h.
13719 (expand_target_clones): Diagnose and fail if node->definition and
13720 !tree_versionable_function_p (node->decl).
13722 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
13724 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
13726 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
13727 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
13728 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
13730 2018-03-06 Jakub Jelinek <jakub@redhat.com>
13733 * combine.c (try_combine): Use reg_or_subregno instead of handling
13734 just paradoxical SUBREGs and REGs.
13736 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
13738 * config/arc/arc.c (arc_finalize_pic): Remove function.
13739 (arc_must_save_register): We use single base PIC register, remove
13740 checks to save/restore the PIC register.
13741 (arc_expand_prologue): Likewise.
13742 * config/arc/arc-protos.h (arc_set_default_type_attributes):
13744 (arc_verify_short): Likewise.
13745 (arc_attr_type): Likewise.
13746 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
13747 (walk_stores): Likewise.
13748 (arc_address_cost): Make it static.
13749 (arc_verify_short): Likewise.
13750 (branch_dest): Likewise.
13751 (arc_attr_type): Likewise.
13752 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
13753 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
13754 (arc_final_prescan_insn): Remove inserting the nops due to
13755 hardware hazards. It is done in reorg step.
13756 (insn_length_variant_t): Remove.
13757 (insn_length_parameters_t): Likewise.
13758 (arc_insn_length_parameters): Likewise.
13759 (arc_get_insn_variants): Likewise.
13760 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
13762 2018-03-06 Jakub Jelinek <jakub@redhat.com>
13764 PR inline-asm/84683
13765 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
13768 PR tree-optimization/84687
13769 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
13771 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
13773 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13775 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
13776 Rename to ppc_speculation_barrier.
13777 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
13778 __builtin_ppc_speculation_barrier.
13780 2018-03-05 Jakub Jelinek <jakub@redhat.com>
13783 * combine.c (combine_simplify_rtx): Don't try to simplify if
13784 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
13787 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
13789 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
13790 to 32 bytes when compiling for POWER9.
13792 2018-03-05 Jakub Jelinek <jakub@redhat.com>
13795 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
13796 regparm >= 3 with no arg reg available also for calls with
13797 flag_force_indirect_call. Pass decl to ix86_function_regparm.
13800 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
13802 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
13804 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
13807 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
13809 2018-03-05 Richard Biener <rguenther@suse.de>
13811 PR tree-optimization/84486
13812 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
13813 When inserting a __builtin_assume_aligned call set the LHS
13814 SSA name alignment info accordingly.
13816 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
13818 PR tree-optimization/84114
13819 * config/aarch64/aarch64.c (aarch64_reassociation_width)
13820 Avoid reassociation of FLOAT_MODE addition.
13822 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
13824 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
13825 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
13826 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
13827 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
13828 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
13829 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
13830 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
13832 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
13833 __builtin_ia32_wbinvd): New builtins.
13834 (SPECIAL_ARGS2): New.
13835 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
13836 (SPECIAL_ARGS2): New.
13837 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
13838 (ix86_valid_target_attribute_inner_p): Ditto.
13839 (ix86_init_mmx_sse_builtins): Add special_args2.
13840 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
13841 TARGET_WBNOINVD_P): New.
13842 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
13843 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
13844 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
13845 * config/i386/immintrin.h (_wbinvd): New intrinsic.
13846 * config/i386/pconfigintrin.h: New file.
13847 * config/i386/wbnoinvdintrin.h: Ditto.
13848 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
13850 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
13852 2018-03-05 Richard Biener <rguenther@suse.de>
13854 PR tree-optimization/84670
13855 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
13857 (BB_VISITED_WITH_VISITED_SUCCS): New define.
13858 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
13859 (compute_antic_aux): Only assert the number of values in ANTIC_IN
13860 doesn't grow if all successors (recursively) were visited at least
13863 2018-03-05 Richard Biener <rguenther@suse.de>
13865 PR tree-optimization/84650
13866 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
13867 if executed in the loop pipeline.
13869 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
13871 * doc/configfiles.texi (Configuration Files): Move info about
13872 conditionalizing $target-protos.h to...
13873 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
13874 differs from $target-protos.h.
13876 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
13877 Chung-Ju Wu <jasonwucj@gmail.com>
13879 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
13880 * config/nds32/nds32-multiple.md (setmemsi): Define.
13881 * config/nds32/nds32-memory-manipulation.c
13882 (nds32_gen_dup_4_byte_to_word_value): New.
13883 (emit_setmem_word_loop): New.
13884 (emit_setmem_byte_loop): New.
13885 (nds32_expand_setmem_loop): New.
13886 (nds32_expand_setmem_loop_v3m): New.
13887 (nds32_expand_setmem_unroll): New.
13888 (nds32_expand_setmem): New.
13890 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
13891 Chung-Ju Wu <jasonwucj@gmail.com>
13893 * config/nds32/nds32-memory-manipulation.c
13894 (nds32_emit_load_store): New.
13895 (nds32_emit_post_inc_load_store): New.
13896 (nds32_emit_mem_move): New.
13897 (nds32_emit_mem_move_block): New.
13898 (nds32_expand_movmemsi_loop_unknown_size): New.
13899 (nds32_expand_movmemsi_loop_known_size): New.
13900 (nds32_expand_movmemsi_loop): New.
13901 (nds32_expand_movmemsi_unroll): New.
13902 (nds32_expand_movmemqi): Rename ...
13903 (nds32_expand_movmemsi): ... to this.
13904 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
13905 (movmemsi): ... to this.
13906 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
13907 (nds32_expand_movmemsi): ... to this.
13909 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
13910 Monk Chiang <sh.chiang04@gmail.com>
13911 Chung-Ju Wu <jasonwucj@gmail.com>
13913 * config/nds32/nds32-protos.h
13914 (nds32_expand_load_multiple): New arguments.
13915 (nds32_expand_store_multiple): Ditto.
13916 (nds32_valid_multiple_load_store): Rename ...
13917 (nds32_valid_multiple_load_store_p): ... to this.
13918 * config/nds32/nds32-memory-manipulation.c
13919 (nds32_expand_load_multiple): Refine implementation.
13920 (nds32_expand_store_multiple): Ditto.
13921 * config/nds32/nds32-multiple.md
13922 (load_multiple): Update nds32_expand_load_multiple interface.
13923 (store_multiple): Update nds32_expand_store_multiple interface.
13924 * config/nds32/nds32-predicates.c
13925 (nds32_valid_multiple_load_store): Rename ...
13926 (nds32_valid_multiple_load_store_p): ... to this and refine
13928 * config/nds32/predicates.md
13929 (nds32_load_multiple_and_update_address_operation): New predicate.
13930 (nds32_store_multiple_and_update_address_operation): New predicate.
13932 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
13933 Chung-Ju Wu <jasonwucj@gmail.com>
13935 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
13936 (combo): New attribute.
13937 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
13939 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
13941 * config/nds32/nds32.opt: Change -mcmodel= default value.
13943 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
13944 Monk Chiang <sh.chiang04@gmail.com>
13945 Chung-Ju Wu <jasonwucj@gmail.com>
13947 * config/nds32/constants.md (unspec_element): New enum.
13948 * config/nds32/constraints.md (Umw): New constraint.
13949 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
13950 * config/nds32/nds32-intrinsic.md: Likewise.
13951 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
13952 (nds32_valid_smw_lwm_base_p): New.
13953 (nds32_output_smw_single_word): New.
13954 (nds32_output_lmw_single_word): New.
13955 (nds32_expand_unaligned_load): New.
13956 (nds32_expand_unaligned_store): New.
13957 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
13958 (nds32_output_smw_single_word): Declare.
13959 (nds32_output_lmw_single_word): Declare.
13960 (nds32_expand_unaligned_load): Declare.
13961 (nds32_expand_unaligned_store): Declare.
13962 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
13963 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
13964 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
13965 NDS32_BUILTIN_UASTORE_DW.
13966 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
13969 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
13970 Kito Cheng <kito.cheng@gmail.com>
13971 Chung-Ju Wu <jasonwucj@gmail.com>
13973 * config/nds32/nds32-intrinsic.c
13974 (nds32_expand_builtin_null_ftype_reg): Delete.
13975 (nds32_expand_builtin_reg_ftype_imm): Ditto.
13976 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
13977 (nds32_read_argument): New.
13978 (nds32_legitimize_target): Ditto.
13979 (nds32_legitimize_argument): Ditto.
13980 (nds32_check_constant_argument): Ditto.
13981 (nds32_expand_unop_builtin): Ditto.
13982 (nds32_expand_unopimm_builtin): Ditto.
13983 (nds32_expand_binop_builtin): Ditto.
13984 (nds32_builtin_decl_impl): Ditto.
13985 (builtin_description): Ditto.
13986 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
13987 (nds32_init_builtins_impl): Ditto.
13988 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
13989 (nds32_builtin_decl): New.
13990 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
13991 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
13993 2018-03-02 Jeff Law <law@redhat.com>
13995 * reorg.c (stop_search_p): Handle DEBUG_INSN.
13996 (redundant_insn, fill_simple_delay_slots): Likewise.
13997 (fill_slots_from_thread): Likewise.
13998 * resource.c (mark_referenced_resources): Likewise.
13999 (mark_set_resources, find_dead_or_set_registers): Likewise.
14001 2018-03-02 Jakub Jelinek <jakub@redhat.com>
14003 * substring-locations.h (format_warning_va): Formatting fix for
14004 ATTRIBUTE_GCC_DIAG.
14005 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
14007 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
14008 * substring-locations.c: Include intl.h.
14009 (format_warning_va): Turned into small wrapper around
14010 format_warning_n_va, renamed to ...
14011 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
14012 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
14014 (format_warning_at_substring_n): New function.
14015 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
14016 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
14017 format_warning_at_substring with just a shorter name instead of
14018 const function pointer.
14019 (fmtwarn_n): New function.
14020 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
14021 appropriate, get rid of all the fmtstr temporaries, move conditionals
14022 with G_() wrapped string literals directly into fmtwarn arguments,
14023 cast dir.len to (int), formatting fixes.
14025 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
14027 * doc/invoke.texi: Remove "Cilk Plus" references.
14029 2018-03-02 Jakub Jelinek <jakub@redhat.com>
14030 Richard Biener <rguenther@suse.de>
14033 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
14034 for error or warning attributes if CALL_FROM_THUNK_P is set.
14037 2018-03-02 Jakub Jelinek <jakub@redhat.com>
14040 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
14041 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
14044 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
14045 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
14047 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
14048 instead of -1U in last predictors element's probability member.
14050 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
14053 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
14054 arguments if they are comparable.
14056 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
14058 PR tree-optimization/84634
14059 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
14060 masks and masked_loop_p with a single loop_masks, making sure it's
14061 null for bb vectorization.
14063 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
14065 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
14066 (vect_analyze_data_ref_access): Use loop->safe_len rather than
14067 loop->force_vectorize to check whether there is no alias.
14069 2018-03-02 Jakub Jelinek <jakub@redhat.com>
14072 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
14074 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
14076 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
14077 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
14078 instead of a loop around prev_real_insn.
14079 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
14082 PR inline-asm/84625
14083 * config/i386/i386.c (ix86_print_operand): Use conditional
14084 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
14087 2018-03-02 Richard Biener <rguenther@suse.de>
14089 PR tree-optimization/84427
14090 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
14091 (bitmap_set_subtract_values): Rewrite to handle multiple
14094 (prune_clobbered_mems): Likewise.
14095 (phi_translate): Take edge instead of pred/phiblock.
14096 (phi_translate_1): Likewise.
14097 (phi_translate_set): Likewise. Insert all translated
14098 exprs for a value into the set, keeping possibly multiple
14099 expressions per value.
14100 (compute_antic_aux): Adjust for phi_translate changes.
14101 When intersecting union the expressions and prune those
14102 not in the final value set, keeping possibly multiple
14103 expressions per value. Do not use value-insertion
14104 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
14105 all expressions. Add verification that the value-sets
14106 only shrink during iteration.
14107 (compute_partial_antic_aux): Adjust for the phi_translate changes.
14108 (do_pre_regular_insertion): Likewise.
14109 (do_pre_partial_partial_insertion): Likewise.
14111 2018-03-02 Richard Biener <rguenther@suse.de>
14114 * config/darwin.c (saved_debug_info_level): New static global.
14115 (darwin_asm_lto_start): Disable debug info generation for LTO out.
14116 (darwin_asm_lto_end): Restore debug info generation settings.
14118 2018-03-01 Martin Liska <mliska@suse.cz>
14121 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
14122 volatile arguments.
14124 2018-03-01 Richard Biener <rguenther@suse.de>
14127 * dwarf2out.c (gen_variable_die): Properly handle late VLA
14128 type annotation with LTO when debug was disabled at compile-time.
14130 2018-03-01 Matthew Fortune <mfortune@gmail.com>
14132 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
14134 (mips_final_postscan_insn): Likewise.
14136 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
14138 PR rtl-optimization/84528
14139 * alias.c (init_alias_target): Add commentary.
14140 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
14141 a unique base value if the frame pointer is not eliminated
14142 to the stack pointer.
14144 2018-03-01 Tom de Vries <tom@codesourcery.com>
14146 PR rtl-optimization/83327
14147 * lra-int.h (hard_regs_spilled_into): Declare.
14148 * lra.c (hard_regs_spilled_into): Define.
14149 (init_reg_info): Init hard_regs_spilled_into.
14150 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
14151 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
14152 (process_bb_lives): Handle hard_regs_spilled_into.
14153 (lra_create_live_ranges_1): Before doing liveness propagation, clear
14154 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
14156 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
14158 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
14159 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
14160 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
14161 * config/rs6000/aix72.h: New file.
14163 2018-02-28 Jakub Jelinek <jakub@redhat.com>
14165 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
14166 instead of warning_at with conditional singular and plural messages
14170 * stor-layout.c (update_alignment_for_field): For
14171 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
14172 && !DECL_PACKED (field), do the alignment update, just use
14173 only desired_align instead of MAX (type_align, desired_align)
14175 (place_field): Don't do known_align < desired_align handling
14176 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
14177 is non-NULL, instead do it after rli->prev_field handling and
14178 only if not within a bitfield word. For DECL_PACKED (field)
14179 use type_align of BITS_PER_UNIT.
14181 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
14183 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
14184 superfluous parentheses and trailing spaces.
14186 2018-02-28 Richard Biener <rguenther@suse.de>
14188 PR tree-optimization/84584
14189 * graphite-scop-detection.c (scop_detection::add_scop): Discard
14190 SCoPs with fake exit edge.
14192 2018-02-28 Martin Liska <mliska@suse.cz>
14195 * timevar.c (timer::print): Fix format to properly print 100%
14198 2018-02-28 Richard Biener <rguenther@suse.de>
14200 PR middle-end/84607
14201 * genmatch.c (capture_info::walk_match): Do not mark
14202 captured expressions without operands as expr_p given
14203 they act more like predicates and should be subject to
14204 "lost tail" side-effect preserving.
14206 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
14208 PR rtl-optimization/81611
14209 * auto-inc-dec.c (attempt_change): Move dead note from
14210 mem_insn if it's the next use of regno
14211 (find_address): Take address use of reg holding
14212 non-incremented value. Add parm to limit search to the named
14214 (merge_in_block): Attempt to use a mem insn that is the next
14215 use of the original regno.
14217 2018-02-27 Martin Sebor <msebor@redhat.com>
14220 * doc/invoke.texi (-Wmissing-attributes): New option.
14221 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
14223 2018-02-27 Martin Sebor <msebor@redhat.com>
14225 PR translation/84207
14226 * diagnostic-core.h (warning_n, error_n, inform_n): Change
14227 n argument to unsigned HOST_WIDE_INT.
14228 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
14229 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
14230 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
14231 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
14233 2018-02-27 Richard Biener <rguenther@suse.de>
14235 PR tree-optimization/84512
14236 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
14237 Do not use the estimate returned from record_stmt_cost for
14238 the scalar iteration cost but sum properly using add_stmt_cost.
14240 2018-02-27 Richard Biener <rguenther@suse.de>
14242 PR tree-optimization/84466
14243 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
14244 Adjust last change to less strictly validate use operands.
14246 2018-02-27 Martin Liska <mliska@suse.cz>
14248 PR gcov-profile/84548
14249 * gcov.c (process_file): Allow partial overlap and consider it
14250 also as group functions.
14251 (output_lines): Properly calculate range of lines for a group.
14253 2018-02-27 Martin Liska <mliska@suse.cz>
14255 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
14256 'ggc' suffixes. Change first column width.
14257 (timer::print): Fix formatting of the column.
14259 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
14261 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
14262 preserve inline entry blocks for the sake of debug inline
14263 entry point markers alone.
14264 (remove_unused_locals): Suggest in comments a better place to
14265 force the preservation of inline entry blocks that are
14266 otherwise unused, but do not preserve them.
14268 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
14270 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
14272 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
14275 * config/i386/constraints.md (Bs): Replace
14276 ix86_indirect_branch_register with
14277 TARGET_INDIRECT_BRANCH_REGISTER.
14279 * config/i386/i386.md (indirect_jump): Likewise.
14280 (tablejump): Likewise.
14281 (*sibcall_memory): Likewise.
14282 (*sibcall_value_memory): Likewise.
14283 Peepholes of indirect call and jump via memory: Likewise.
14284 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
14285 (*sibcall_value_GOT_32): Likewise.
14286 * config/i386/predicates.md (indirect_branch_operand): Likewise.
14287 (GOT_memory_operand): Likewise.
14288 (call_insn_operand): Likewise.
14289 (sibcall_insn_operand): Likewise.
14290 (GOT32_symbol_operand): Likewise.
14291 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
14293 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
14295 PR rtl-optimization/83496
14296 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
14297 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
14298 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
14299 redundant insn, if any.
14300 (relax_delay_slots): Likewise.
14301 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
14303 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
14305 PR tree-optimization/83965
14306 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
14307 that grouped statements are part of a reduction chain. Return
14308 true if the statement is not marked as a reduction itself but
14309 is part of a group.
14310 (vect_recog_dot_prod_pattern): Don't check whether the statement
14311 is part of a group here.
14312 (vect_recog_sad_pattern): Likewise.
14313 (vect_recog_widen_sum_pattern): Likewise.
14315 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
14318 * final.c (rest_of_clean_state): Also look for calls inside sequences.
14320 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
14323 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
14325 (ix86_output_indirect_function_return): New prototype.
14326 (ix86_split_simple_return_pop_internal): Likewise.
14327 * config/i386/i386.c (indirect_return_via_cx): New.
14328 (indirect_return_via_cx_bnd): Likewise.
14329 (indirect_thunk_name): Handle return va CX_REG.
14330 (output_indirect_thunk_function): Create alias for
14331 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
14332 (ix86_output_indirect_jmp): Remove the bool argument.
14333 (ix86_output_indirect_function_return): New function.
14334 (ix86_split_simple_return_pop_internal): Likewise.
14335 * config/i386/i386.md (*indirect_jump): Don't pass false
14336 to ix86_output_indirect_jmp.
14337 (*tablejump_1): Likewise.
14338 (simple_return_pop_internal): Change it to define_insn_and_split.
14339 Call ix86_split_simple_return_pop_internal to split it for
14340 -mfunction-return=.
14341 (simple_return_indirect_internal): Call
14342 ix86_output_indirect_function_return instead of
14343 ix86_output_indirect_jmp.
14345 2018-02-26 Jakub Jelinek <jakub@redhat.com>
14348 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
14349 memset and value initialization afterwards.
14351 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
14353 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
14355 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
14358 * common/config/aarch64/aarch64-common.c
14359 (aarch_option_optimization_table[]): Switch
14360 off fomit-frame-pointer
14362 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
14363 Chung-Ju Wu <jasonwucj@gmail.com>
14365 * config/nds32/nds32-multiple.md (load_multiple): Disallow
14367 (store_multiple): Ditto.
14369 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
14371 * config.gcc: Add --with-cpu support for nds32 target.
14372 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
14373 * config/nds32/nds32.opt: Add -mcpu= option.
14375 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
14377 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
14378 isel=yes): Warn for these deprecated options.
14380 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
14382 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
14383 ISA_2_5_MASKS_EMBEDDED.
14385 2018-02-23 Jakub Jelinek <jakub@redhat.com>
14387 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
14388 p->max as pointers rather than using iterative_hash_expr.
14390 2018-02-23 Carl Love <cel@us.ibm.com>
14392 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
14393 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
14395 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
14396 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
14397 P8V_BUILTIN_VEC_VUNSIGNED2.
14399 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
14402 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
14403 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
14405 * lra-constraints.c (curr_insn_transform): Set up
14406 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
14408 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
14409 * lra-eliminations.c (spill_pseudos): Ditto.
14410 (process_insn_for_elimination): Ditto.
14411 * lra-lives.c (reg_early_clobber_p): Use the new macros.
14412 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
14413 LRA_NON_CLOBBERED_ALT.
14415 2018-02-22 Martin Sebor <msebor@redhat.com>
14417 PR tree-optimization/84480
14418 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
14419 to maybe_diag_stxncpy_trunc. Call it.
14420 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
14421 from gimple_fold_builtin_strcpy. Print inlining stack.
14422 (handle_builtin_stxncpy): Print inlining stack.
14423 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
14425 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
14428 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
14429 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
14430 and -fcheck-pointer-bounds are used together.
14431 (indirect_thunk_prefix): New enum.
14432 (indirect_thunk_need_prefix): New function.
14433 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
14434 "_nt" instead of "_bnd" for NOTRACK prefix.
14435 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
14436 (output_indirect_thunk_function): Likewise.
14438 (ix86_code_end): Update output_indirect_thunk_function calls.
14439 (ix86_output_indirect_branch_via_reg): Replace
14440 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
14441 (ix86_output_indirect_branch_via_push): Likewise.
14442 (ix86_output_function_return): Likewise.
14443 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
14444 incompatible with -fcf-protection=branch and
14445 -fcheck-pointer-bounds.
14447 2018-02-22 Steve Ellcey <sellcey@cavium.com>
14450 * config/aarch64/aarch64.c (aarch64_print_address_internal):
14451 Change gcc_assert call to output_operand_lossage.
14453 2018-02-22 Steve Ellcey <sellcey@cavium.com>
14455 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
14457 2018-02-22 DJ Delorie <dj@redhat.com>
14458 Sebastian Perta <sebastian.perta@renesas.com>
14459 Oleg Endo <olegendo@gcc.gnu.org>
14461 * config/rx/rx.c (rx_rtx_costs): New function.
14462 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
14464 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
14466 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
14468 2018-02-22 Martin Liska <mliska@suse.cz>
14471 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
14472 Add "native" as a possible value.
14474 2018-02-22 Martin Liska <mliska@suse.cz>
14477 * config/i386/i386.c (ix86_option_override_internal):
14478 Add "native" as a possible value for -march and -mtune.
14480 2018-02-22 Jakub Jelinek <jakub@redhat.com>
14483 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
14484 to all type variants.
14486 PR tree-optimization/84503
14487 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
14488 width as info->bitpos + info->bitsize - start.
14489 (merged_store_group::merge_overlapping): Simplify width computation.
14490 (check_no_overlap): New function.
14491 (imm_store_chain_info::try_coalesce_bswap): Compute expected
14492 start + width and last_order of the group, fail if check_no_overlap
14494 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
14495 to group if check_no_overlap fails.
14497 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
14499 * config/rs6000/altivec.md: Delete contraint arguments to
14500 define_expand, define_split, and define_peephole2, and in
14501 define_insn_and_split if always unused.
14502 * config/rs6000/darwin.md: Ditto.
14503 * config/rs6000/dfp.md: Ditto.
14504 * config/rs6000/rs6000.md: Ditto.
14505 * config/rs6000/sync.md: Ditto.
14506 * config/rs6000/vector.md: Ditto.
14507 * config/rs6000/vsx.md: Ditto.
14509 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
14511 * config/rs6000/altivec.md: Write output control strings as braced
14512 blocks instead of double-quoted strings.
14513 * config/rs6000/darwin.md: Ditto.
14514 * config/rs6000/rs6000.md: Ditto.
14515 * config/rs6000/vector.md: Ditto.
14516 * config/rs6000/vsx.md: Ditto.
14518 2018-02-21 Jason Merrill <jason@redhat.com>
14520 PR c++/84314 - ICE with templates and fastcall attribute.
14521 * attribs.c (build_type_attribute_qual_variant): Remove assert.
14523 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
14525 * ipa-cp.c (determine_versionability): Fix comment typos.
14527 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
14530 * ipa-cp.c (determine_versionability): Do not version functions caling
14533 2018-02-21 Martin Liska <mliska@suse.cz>
14536 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
14537 Add "native" as a possible value.
14538 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
14539 the macro when native cpu detection is available.
14541 2018-02-21 Martin Liska <mliska@suse.cz>
14544 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
14545 Add "native" as a possible value.
14546 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
14547 when native cpu detection is available.
14549 2018-02-21 Jakub Jelinek <jakub@redhat.com>
14550 Martin Sebor <msebor@redhat.com>
14552 PR tree-optimization/84478
14553 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
14555 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
14556 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
14557 support which is conservatively correct, for 2 only stay conservative
14558 for maxlen. Formatting and comment capitalization fixes. Add STRICT
14559 argument to the 2 argument get_range_strlen, adjust 6 arg
14560 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
14562 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
14563 (gimple_fold_builtin_strlen): Pass true as last argument to
14566 2018-02-20 Martin Sebor <msebor@redhat.com>
14568 PR middle-end/84095
14569 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
14570 (builtin_memref::set_base_and_offset): Same. Handle inner references.
14571 (builtin_memref::builtin_memref): Factor out parts into
14572 set_base_and_offset and call it.
14574 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
14576 PR middle-end/84406
14577 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
14578 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
14579 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
14580 search at the associated MODE_INT.
14582 2018-02-20 Jeff Law <law@redhat.com>
14584 PR middle-end/82123
14585 PR tree-optimization/81592
14586 PR middle-end/79257
14587 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
14588 for range data rather than using global data.
14589 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
14590 range data rather than using global data.
14591 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
14592 pass it to children as needed.
14593 (struct directive::fmtresult): Similarly.
14594 (struct directive::set_width): Similarly.
14595 (struct directive::set_precision): Similarly.
14596 (format_integer, format_directive, parse_directive): Similarly.
14597 (format_none): Accept unnamed vr_values parameter.
14598 (format_percent, format_floating, format_character): Similarly.
14599 (format_string, format_plain): Similarly.
14600 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
14601 the EVRP range analyzer for range data rather than using global data.
14602 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
14603 gimple-ssa-evrp-analyze.h
14604 (class sprintf_dom_walker): Add after_dom_children member function.
14605 Add evrp_range_analyzer member.
14606 (sprintf_dom_walker::before_dom_children): Call into the EVRP
14607 range analyzer as needed.
14608 (sprintf_dom_walker::after_dom_children): New member function.
14609 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
14611 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
14612 (evrp_range_analyzer::pop_to_marker): Likewise.
14614 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
14616 PR tree-optimization/84419
14617 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
14618 with the required type if its current type is compatible but
14621 2018-02-20 Jakub Jelinek <jakub@redhat.com>
14623 PR middle-end/82004
14624 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
14625 after vectorization.
14627 2018-02-20 Martin Liska <mliska@suse.cz>
14630 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
14631 possible values if we don't have a hint.
14633 2018-02-20 Martin Liska <mliska@suse.cz>
14637 * final.c (shorten_branches): Build align_tab array with one
14639 * opts.c (finish_options): Add alignment option limit check.
14640 (MAX_CODE_ALIGN): Likewise.
14641 (MAX_CODE_ALIGN_VALUE): Likewise.
14642 * doc/invoke.texi: Document maximum allowed option value for
14643 all -falign-* options.
14645 2018-02-19 Jakub Jelinek <jakub@redhat.com>
14648 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
14649 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
14650 * var-tracking.c (emit_note_insn_var_location): Remove all references
14651 to NOTE_INSN_CALL_ARG_LOCATION.
14652 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
14653 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
14654 Use copy_rtx_if_shared.
14655 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
14656 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
14657 (dwarf2out_var_location): Remove handling of
14658 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
14660 * final.c (final_scan_insn): Remove all references to
14661 NOTE_INSN_CALL_ARG_LOCATION.
14662 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
14663 before dumping final insns.
14664 * except.c (emit_note_eh_region_end): Remove all references to
14665 NOTE_INSN_CALL_ARG_LOCATION.
14666 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
14667 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
14668 * config/arc/arc.c (hwloop_optimize): Likewise.
14669 * config/arm/arm.c (create_fix_barrier): Likewise.
14670 * config/s390/s390.c (s390_chunkify_start): Likewise.
14671 * config/sh/sh.c (find_barrier): Likewise.
14672 * config/i386/i386.c (rest_of_insert_endbranch,
14673 ix86_seh_fixup_eh_fallthru): Likewise.
14674 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
14675 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
14676 * config/frv/frv.c (frv_function_prologue): Likewise.
14677 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
14679 (note_outside_basic_block_p): Remove all references to
14680 NOTE_INSN_CALL_ARG_LOCATION.
14681 * gengtype.c (adjust_field_rtx_def): Likewise.
14682 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
14684 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
14685 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
14688 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
14691 PR tree-optimization/84452
14692 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
14693 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
14696 2018-02-19 Martin Liska <mliska@suse.cz>
14699 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
14701 2018-02-19 Martin Liska <mliska@suse.cz>
14702 Richard Sandiford <richard.sandiford@linaro.org>
14704 PR tree-optimization/82491
14705 * gimple-fold.c (get_base_constructor): Make earlier bail out
14708 2018-02-19 Carl Love <cel@us.ibm.com>
14710 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
14711 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
14713 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
14714 P8V_BUILTIN_VEC_NEG.
14716 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
14718 * config/rl78/rl78.md (movdf): New define expand.
14720 2018-02-19 Martin Liska <mliska@suse.cz>
14723 * doc/invoke.texi: Fix typo.
14724 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
14726 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
14728 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
14729 handle rs6000_single_float and rs6000_double_float specially for
14732 2018-02-16 Jeff Law <law@redhat.com>
14734 * config/rx/rx.c (add_pop_cfi_notes): New function.;
14735 (pop_regs): Use it.
14737 2018-02-16 Jakub Jelinek <jakub@redhat.com>
14740 * ipa-inline.c (inline_small_functions): Fix a typo.
14742 2018-02-16 Nathan Sidwell <nathan@acm.org>
14744 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
14746 2018-02-16 Carl Love <cel@us.ibm.com>
14748 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
14749 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
14750 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
14751 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
14752 expansion to P8V_BUILTIN_VEC_FLOAT2.
14754 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
14756 PR rtl-optimization/70023
14757 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
14758 src_regno into account.
14760 2018-02-16 Carl Love <cel@us.ibm.com>
14762 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
14763 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
14764 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
14765 * config/rs6000/rs6000.c: Remove case statements for
14766 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
14767 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
14768 and P9V_BUILTIN_VEC_VINSERT4B.
14769 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
14770 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
14771 * config/rs6000/vsx.md:
14772 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
14775 2018-02-16 Carl Love <cel@us.ibm.com>
14777 * config/rs6000/altivec.h: Add builtin names vec_extract4b
14779 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
14781 * config/rs6000/rs6000-c.c: Add the definitions for
14782 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
14783 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
14784 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
14785 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
14786 definition for insert4b and define insn *insert3b_internal.
14787 * doc/extend.texi: Add documentation for vec_extract4b.
14789 2018-02-16 Nathan Sidwell <nathan@acm.org>
14791 * doc/extend.texi (Backwards Compatibility): Mention friend
14792 injection. Note for-scope is deprecated.
14793 * doc/invoke.texi (-ffriend-injection): Deprecate.
14795 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
14797 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
14798 that moved to I2, also allow destinations that are a paradoxical
14799 subreg (instead of a normal reg).
14801 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
14804 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
14807 2018-02-16 Richard Biener <rguenther@suse.de>
14809 PR tree-optimization/84037
14810 PR tree-optimization/84016
14812 * config/i386/i386.c (ix86_builtin_vectorization_cost):
14813 Adjust vec_construct for the fact we need additional higher latency
14814 128bit inserts for AVX256 and AVX512 vector builds.
14815 (ix86_add_stmt_cost): Scale vector construction cost for
14818 2018-02-16 Richard Biener <rguenther@suse.de>
14820 PR tree-optimization/84417
14821 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
14822 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
14823 (non_rewritable_lvalue_p): Likewise, use poly-ints.
14825 2018-02-16 Martin Liska <mliska@suse.cz>
14828 * internal-fn.def (ASAN_CHECK): Set proper flags.
14829 (ASAN_MARK): Likewise.
14831 2018-02-16 Julia Koval <julia.koval@intel.com>
14833 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
14834 from PTA_CANNONLAKE.
14836 2018-02-16 Jakub Jelinek <jakub@redhat.com>
14839 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
14840 Use ++iter rather than iter++ for std::list iterators.
14841 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
14842 defer deleting them until all nodes in the forest are processed. Do
14843 free even leaf nodes. Change to_process into auto_vec.
14846 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
14847 * vec.h (vec_default_construct): Use memset instead of placement new
14848 if BROKEN_VALUE_INITIALIZATION is defined.
14849 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
14850 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
14853 PR rtl-optimization/83723
14854 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
14855 * lra.c (lra_substitute_pseudo): Likewise. If true, use
14856 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
14858 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
14860 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
14862 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
14864 PR rtl-optimization/81443
14865 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
14866 from inner REGs to paradoxical SUBREGs.
14868 2018-02-16 Richard Biener <rguenther@suse.de>
14870 PR tree-optimization/84399
14871 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
14872 For operands we can analyze at their definition make sure we can
14873 analyze them at each use as well.
14875 2018-02-16 Richard Biener <rguenther@suse.de>
14877 PR tree-optimization/84190
14878 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
14879 volatile accesses if the decl isn't volatile.
14881 2018-02-15 Jason Merrill <jason@redhat.com>
14883 PR c++/84314 - ICE with templates and fastcall attribute.
14884 * attribs.c (build_type_attribute_qual_variant): Don't clobber
14885 TYPE_CANONICAL on an existing type.
14887 2018-02-15 Jakub Jelinek <jakub@redhat.com>
14889 PR tree-optimization/84383
14890 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
14891 dstoff nor call operand_equal_p if dstbase is NULL.
14893 PR tree-optimization/84334
14894 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
14895 also a CONSTANT_CLASS_P, punt.
14897 2018-02-14 Jim Wilson <jimw@sifive.com>
14899 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
14900 first SMALL_OPERAND check. New local min_second_step. Move assert
14901 to where locals are set. Add TARGET_RVC support.
14902 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
14904 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
14906 * doc/invoke.texi: Correct -Wformat-overflow code sample.
14908 2018-02-14 Martin Sebor <msebor@redhat.com>
14910 PR tree-optimization/83698
14911 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
14912 arrays constrain the offset range to their bounds.
14913 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
14914 (builtin_access::overlap): Avoid setting the size of overlap if it's
14916 (maybe_diag_overlap): Also consider arrays when deciding what values
14917 of offsets to include in diagnostics.
14919 2018-02-14 Martin Sebor <msebor@redhat.com>
14922 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
14923 that correspond to the kind of a declaration.
14925 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
14928 * config/pa/pa.md: Load address of PIC label using the linkage table
14929 if the label is nonlocal.
14931 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
14933 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
14934 warning message if user requests -maltivec=be.
14935 * doc/invoke.texi: Document deprecation of -maltivec=be.
14937 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
14940 * config/rs6000/rs6000-c.c: Update definitions for
14941 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
14942 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
14944 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
14947 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
14948 add _get_ssp intrinsics. Remove argument from
14949 __builtin_ia32_rdssp[d|q].
14950 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
14951 * config/i386/i386-builtin.def: Remove argument from
14952 __builtin_ia32_rdssp[d|q].
14953 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
14954 ix86_expand_special_args_builtin for _rdssp[d|q].
14955 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
14956 Clear register before usage.
14957 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
14958 Add documentation for new _get_ssp and _inc_ssp intrinsics.
14960 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
14962 PR tree-optimization/84357
14963 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
14964 operand 1 of an ARRAY_REF too.
14966 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
14969 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
14970 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
14972 (set_of_reg): New struct.
14973 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
14974 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
14975 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
14977 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
14978 Split into bitclr, bitset, bitinvert patterns if appropriate.
14979 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
14980 use rx_fuse_in_memory_bitop.
14981 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
14982 to named insn, correct maximum insn length.
14984 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
14987 * machmode.def: Define a complex mode for PARTIAL_INT.
14988 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
14990 * doc/rtl.texi: Document CSPImode.
14991 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
14993 (msp430_hard_regno_nregs_with_padding): Likewise.
14995 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
14998 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
15000 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
15002 PR rtl-optimization/84169
15003 * combine.c (try_combine): New variable split_i2i3. Set it to true if
15004 we generated a parallel as new i3 and we split that to new i2 and i3
15005 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
15006 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
15007 those to i2, not i1. Partially rewrite this scan code.
15009 2018-02-13 Jakub Jelinek <jakub@redhat.com>
15012 * stor-layout.c (place_field): For variable length fields, adjust
15013 offset_align afterwards not just based on the field's alignment,
15014 but also on the size.
15016 PR middle-end/84309
15017 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
15018 of exps and logs in the use_exp2 case.
15020 2018-02-13 Jeff Law <law@redhat.com>
15022 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
15023 entry for "vector".
15025 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
15028 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
15032 * common.opt (gas-loc-support, gas-locview-support): New.
15033 (ginline-points, ginternal-reset-location-views): New.
15034 * doc/invoke.texi: Document them. Use @itemx where intended.
15035 (gvariable-location-views): Adjust.
15036 * target.def (reset_location_view): New.
15037 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
15038 (TARGET_RESET_LOCATION_VIEW): New.
15039 * doc/tm.texi: Rebuilt.
15040 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
15041 (dwarf2out_default_as_locview_support): New.
15042 (output_asm_line_debug_info): Use option variables.
15043 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
15044 (output_loc_list): Likewise.
15045 (add_high_low_attributes): Check option variables.
15046 Don't output entry view attribute in strict mode.
15047 (gen_inlined_subroutine_die): Check option variables.
15048 (dwarf2out_inline_entry): Likewise.
15049 (init_sections_and_labels): Likewise.
15050 (dwarf2out_early_finish): Likewise.
15051 (maybe_reset_location_view): New, from...
15052 (dwarf2out_var_location): ... here. Call it.
15053 * debug.h (dwarf2out_default_as_loc_support): Declare.
15054 (dwarf2out_default_as_locview_support): Declare.
15055 * hooks.c (hook_int_rtx_insn_0): New.
15056 * hooks.h (hook_int_rtx_insn_0): Declare.
15057 * toplev.c (process_options): Take -gas-loc-support and
15058 -gas-locview-support from dwarf2out. Enable
15059 -gvariable-location-views by default only with locview
15060 assembler support. Enable -ginternal-reset-location-views by
15061 default only if the target defines the corresponding hook.
15062 Enable -ginline-points by default if location views are
15063 enabled; force it disabled if statement frontiers are
15065 * tree-inline.c (expand_call_inline): Check option variables.
15066 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
15068 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
15070 PR tree-optimization/84321
15071 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
15072 handling. Also check whether the anti-range contains any values
15073 that satisfy the mask; switch to a VR_RANGE if not.
15075 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
15078 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
15080 2018-02-13 Martin Jambor <mjambor@suse.cz>
15083 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
15084 of call statements, also set location of a load to a temporary.
15086 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
15088 * config/rl78/rl78.c (add_vector_labels): New function.
15089 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
15090 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
15091 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
15092 which checks that no arguments are passed.
15093 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
15094 * doc/extend.texi: Documentation for the new attribute.
15096 2018-02-13 Andreas Schwab <schwab@suse.de>
15098 * config/riscv/linux.h (CPP_SPEC): Define.
15100 2018-02-13 Jakub Jelinek <jakub@redhat.com>
15103 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
15104 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
15105 OPTION_MASK_ISA_AES as first argument to def_builtin_const
15106 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
15107 instead of OPTION_MASK_ISA_PCLMUL as first argument to
15108 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
15109 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
15110 temporarily for AES and PCLMUL builtins.
15112 PR tree-optimization/84339
15113 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
15114 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
15117 PR middle-end/84309
15118 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
15119 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
15120 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
15122 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
15124 * omp-simd-clone.h: New file.
15125 * omp-simd-clone.c: Include omp-simd-clone.h.
15126 (expand_simd_clones): No longer static.
15127 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
15128 cgraph.h and omp-simd-clone.h.
15129 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
15130 (vect_recog_widen_shift_pattern): Formatting fix.
15131 (vect_pattern_recog_1): Don't check optab for calls.
15134 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
15135 operands[2] into a REG before using gen_lowpart on it.
15137 2018-02-12 Jeff Law <law@redhat.com>
15140 * config/sh/sh.c (find_barrier): Consider a sibling call
15143 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
15144 successfully back substituting a reg.
15146 2018-02-12 Richard Biener <rguenther@suse.de>
15148 PR tree-optimization/84037
15149 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
15150 parameter, move visited init to caller.
15151 (vect_slp_analyze_operations): Separate cost from validity
15152 check, initialize visited once for all instances.
15153 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
15155 * tree-vect-stmts.c (vect_model_simple_cost): Make early
15157 (vect_model_promotion_demotion_cost): Likewise.
15158 (vectorizable_bswap): Guard cost modeling with !slp_node
15159 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
15161 (vectorizable_call): Likewise.
15162 (vectorizable_conversion): Likewise.
15163 (vectorizable_assignment): Likewise.
15164 (vectorizable_shift): Likewise.
15165 (vectorizable_operation): Likewise.
15166 (vectorizable_store): Likewise.
15167 (vectorizable_load): Likewise.
15168 (vectorizable_condition): Likewise.
15169 (vectorizable_comparison): Likewise.
15171 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
15174 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
15175 (ASAN_MARK): Fix fnspec to account for return value, change pointer
15176 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
15178 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
15180 PR middle-end/83665
15181 * params.def (inline-min-speedup): Increase from 8 to 15.
15182 (max-inline-insns-auto): Decrease from 40 to 30.
15183 * ipa-split.c (consider_split): Add some buffer for function to
15184 be considered inlining candidate.
15185 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
15188 2018-02-12 Richard Biener <rguenther@suse.de>
15190 PR tree-optimization/84037
15191 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
15192 matched stmts if we cannot swap the non-matched ones.
15194 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
15196 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
15197 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
15198 _mm_maskz_scalef_round_ss): New intrinsics.
15199 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
15200 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
15201 __builtin_ia32_scalefss_round): Remove.
15202 (__builtin_ia32_scalefsd_mask_round,
15203 __builtin_ia32_scalefss_mask_round): New intrinsics.
15204 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
15205 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
15206 ((match_operand:VF_128 2 "<round_nimm_predicate>"
15207 "<round_constraint>")): Changed to ...
15208 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
15209 "<round_scalar_constraint>")): ... this.
15210 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
15211 %0, %1, %2<round_op3>}"): Changed to ...
15212 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
15213 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
15214 %2<round_scalar_mask_op3>}"): ... this.
15215 * config/i386/subst.md (round_scalar_nimm_predicate): New.
15217 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
15219 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
15220 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
15221 (_mm_maskz_sqrt_round_ss): New intrinsics.
15222 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
15223 (__builtin_ia32_sqrtsd_mask_round)
15224 (__builtin_ia32_sqrtss_mask_round): New builtins.
15225 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
15226 (__builtin_ia32_sqrtss_round): Remove.
15227 (__builtin_ia32_sqrtsd_mask_round)
15228 (__builtin_ia32_sqrtss_mask_round): New builtins.
15229 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
15230 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
15231 ((match_operand:VF_128 1 "vector_operand"
15232 "xBm,<round_constraint>")): Changed to ...
15233 ((match_operand:VF_128 1 "vector_operand"
15234 "xBm,<round_scalar_constraint>")): ... this.
15235 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
15236 %0, %2, %<iptr>1<round_op3>}): Changed to ...
15237 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
15238 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
15239 %<iptr>1<round_scalar_mask_op3>}): ... this.
15240 ((set_attr "prefix" "<round_prefix>")): Changed to ...
15241 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
15243 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
15246 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
15247 Cast vec_cmpeq result to correct type.
15248 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
15249 Cast vec_cmpgt result to correct type.
15251 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
15253 * final.c (final_scan_insn_1): Renamed from...
15254 (final_scan_insn): ... this. New wrapper, to recover
15255 seen from the outermost call in recursive ones.
15256 * config/sparc/sparc.c (output_return): Drop seen from call.
15257 (output_sibcall): Likewise.
15258 * config/visium/visium.c (output_branch): Likewise.
15260 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
15262 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
15265 2018-02-10 Alan Modra <amodra@gmail.com>
15268 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
15269 Specify LR as an input.
15271 2018-02-10 Jakub Jelinek <jakub@redhat.com>
15274 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
15275 remove_member_access_dummy_vars): New functions.
15276 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
15277 lower_omp_1, execute_lower_omp): Use them.
15279 PR rtl-optimization/84308
15280 * shrink-wrap.c (spread_components): Release todo vector.
15282 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
15284 PR rtl-optimization/57193
15285 * ira-color.c (struct allocno_color_data): Add member
15286 conflict_allocno_hard_prefs.
15287 (update_conflict_allocno_hard_prefs): New.
15288 (bucket_allocno_compare_func): Add a preference based on
15289 conflict_allocno_hard_prefs.
15290 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
15291 (color_allocnos): Remove a dead code. Initiate
15292 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
15294 2018-02-09 Jakub Jelinek <jakub@redhat.com>
15297 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
15298 constraint from =wa to wa. Avoid a subreg on the output operand,
15299 instead use a pseudo and subreg it in a move.
15300 (p9_xxbrd_<mode>): Changed to ...
15301 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
15302 (p9_xxbrd_v2df): New expander.
15303 (p9_xxbrw_<mode>): Changed to ...
15304 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
15305 (p9_xxbrw_v4sf): New expander.
15307 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
15309 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
15311 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
15314 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
15315 multiply in 32-bit mode.
15316 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
15317 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
15320 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
15322 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
15323 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
15324 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
15325 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
15327 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
15330 * dwarf2out.c (is_trivial_indirect_ref): New function.
15331 (dwarf2out_late_global_decl): Do not generate a location
15332 attribute for variables that have a non-trivial DECL_VALUE_EXPR
15333 and that are not defined in the current unit.
15335 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
15337 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
15338 instead of a libcall for UNORDERED.
15340 2018-02-09 Tamar Christina <tamar.christina@arm.com>
15343 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
15344 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
15346 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
15349 * config/s390/s390.c (s390_set_current_function): Invoke
15350 s390_indirect_branch_settings also if fndecl didn't change.
15352 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
15354 * config/rs6000/rs6000.md (blockage): Set length to zero.
15356 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
15358 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
15360 2018-02-09 Jakub Jelinek <jakub@redhat.com>
15363 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
15364 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
15368 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
15369 PARALLEL incoming that failed vt_get_decl_and_offset check.
15371 PR middle-end/84237
15372 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
15373 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
15375 (get_variable_section): For decls in named .bss* sections pass true as
15376 second argument to bss_initializer_p.
15378 2018-02-09 Marek Polacek <polacek@redhat.com>
15379 Jakub Jelinek <jakub@redhat.com>
15382 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
15383 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
15384 Sync some changes from cxx_fold_indirect_ref.
15386 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
15388 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
15390 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
15391 (BLOCK_INLINE_ENTRY_LABEL): New.
15392 (dwarf2out_var_location): Disregard inline entry markers.
15393 (inline_entry_data): New struct.
15394 (inline_entry_data_hasher): New hashtable type.
15395 (inline_entry_data_hasher::hash): New.
15396 (inline_entry_data_hasher::equal): New.
15397 (inline_entry_data_table): New variable.
15398 (add_high_low_attributes): Add DW_AT_entry_pc and
15399 DW_AT_GNU_entry_view attributes if a pending entry is found
15400 in inline_entry_data_table. Add old entry_pc attribute only
15401 if debug nonbinding markers are disabled.
15402 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
15403 markers are enabled.
15404 (block_within_block_p, dwarf2out_inline_entry): New.
15405 (dwarf2out_finish): Check that no entries remained in
15406 inline_entry_data_table.
15407 * final.c (reemit_insn_block_notes): Handle inline entry notes.
15408 (final_scan_insn, notice_source_line): Likewise.
15409 (rest_of_clean_state): Skip inline entry markers.
15410 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
15412 * gimple.c (gimple_build_debug_inline_entry): New.
15413 * gimple.h (enum gimple_debug_subcode): Add
15414 GIMPLE_DEBUG_INLINE_ENTRY.
15415 (gimple_build_debug_inline_entry): Declare.
15416 (gimple_debug_inline_entry_p): New.
15417 (gimple_debug_nonbind_marker_p): Adjust.
15418 * insn-notes.def (INLINE_ENTRY): New.
15419 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
15420 inline entry marker notes.
15421 (print_insn): Likewise.
15422 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
15423 (INSN_DEBUG_MARKER_KIND): Likewise.
15424 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
15425 * tree-inline.c (expand_call_inline): Build and insert
15426 debug_inline_entry stmt.
15427 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
15428 inline entry blocks early, if nonbind markers are enabled.
15429 (dump_scope_block): Dump fragment info.
15430 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
15431 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
15432 (gimple_build_debug_inline_entry): New.
15433 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
15434 Enable/disable inline entry points too.
15435 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
15436 (DEBUG_INSN): Describe inline entry markers.
15438 * common.opt (gvariable-location-views): New.
15439 (gvariable-location-views=incompat5): New.
15440 * config.in: Rebuilt.
15441 * configure: Rebuilt.
15442 * configure.ac: Test assembler for view support.
15443 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
15444 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
15445 * dwarf2out.c (var_loc_view): New typedef.
15446 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
15447 (dwarf2out_locviews_in_attribute): New.
15448 (dwarf2out_locviews_in_loclist): New.
15449 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
15450 (enum dw_line_info_opcode): Add LI_adv_address.
15451 (struct dw_line_info_table): Add view.
15452 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
15453 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
15454 (zero_view_p): New variable.
15455 (ZERO_VIEW_P): New macro.
15456 (output_asm_line_debug_info): New.
15457 (struct var_loc_node): Add view.
15458 (add_AT_view_list, AT_loc_list): New.
15459 (add_var_loc_to_decl): Add view param. Test it against last.
15460 (new_loc_list): Add view params. Record them.
15461 (AT_loc_list_ptr): Handle loc and view lists.
15462 (view_list_to_loc_list_val_node): New.
15463 (print_dw_val): Handle dw_val_class_view_list.
15464 (size_of_die): Likewise.
15465 (value_format): Likewise.
15466 (loc_list_has_views): New.
15467 (gen_llsym): Set vl_symbol too.
15468 (maybe_gen_llsym, skip_loc_list_entry): New.
15469 (dwarf2out_maybe_output_loclist_view_pair): New.
15470 (output_loc_list): Output view list or entries too.
15471 (output_view_list_offset): New.
15472 (output_die): Handle dw_val_class_view_list.
15473 (output_dwarf_version): New.
15474 (output_compilation_unit_header): Use it.
15475 (output_skeleton_debug_sections): Likewise.
15476 (output_rnglists, output_line_info): Likewise.
15477 (output_pubnames, output_aranges): Update version comments.
15478 (output_one_line_info_table): Output view numbers in asm comments.
15479 (dw_loc_list): Determine current endview, pass it to new_loc_list.
15480 Call maybe_gen_llsym.
15481 (loc_list_from_tree_1): Adjust.
15482 (add_AT_location_description): Create view list attribute if
15483 needed, check it's absent otherwise.
15484 (convert_cfa_to_fb_loc_list): Adjust.
15485 (maybe_emit_file): Call output_asm_line_debug_info for test.
15486 (dwarf2out_var_location): Reset views as needed. Precompute
15487 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
15488 attribute. Set view.
15489 (new_line_info_table): Reset next view.
15490 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
15491 (dwarf2out_source_line): Likewise. Output view resets and labels to
15492 the assembler, or select appropriate line info opcodes.
15493 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
15494 (optimize_string_length): Catch it. Adjust.
15495 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
15496 dw_val_class_view_list, and remove it if no longer needed.
15497 (hash_loc_list): Hash view numbers.
15498 (loc_list_hasher::equal): Compare them.
15499 (optimize_location_lists): Check whether a view list symbol is
15500 needed, and whether the locview attribute is present, and
15501 whether they match. Remove the locview attribute if no longer
15503 (index_location_lists): Call skip_loc_list_entry for test.
15504 (dwarf2out_finish): Call output_asm_line_debug_info for test.
15505 Use output_dwarf_version.
15506 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
15507 (struct dw_val_node): Add val_view_list.
15508 * final.c (SEEN_NEXT_VIEW): New.
15509 (set_next_view_needed): New.
15510 (clear_next_view_needed): New.
15511 (maybe_output_next_view): New.
15512 (final_start_function): Rename to...
15513 (final_start_function_1): ... this. Take pointer to FIRST,
15514 add SEEN parameter. Emit param bindings in the initial view.
15515 (final_start_function): Reintroduce SEEN-less interface.
15516 (final): Rename to...
15517 (final_1): ... this. Take SEEN parameter. Output final pending
15518 next view at the end.
15519 (final): Reintroduce seen-less interface.
15520 (final_scan_insn): Output pending next view before switching
15521 sections or ending a block. Mark the next view as needed when
15522 outputting variable locations. Notify debug backend of section
15523 changes, and of location view changes.
15524 (rest_of_handle_final): Adjust.
15525 * toplev.c (process_options): Autodetect value for debug variable
15526 location views option. Warn on incompat5 without -gdwarf-5.
15527 * doc/invoke.texi (gvariable-location-views): New.
15528 (gvariable-location-views=incompat5): New.
15529 (gno-variable-location-views): New.
15531 2018-02-08 David Malcolm <dmalcolm@redhat.com>
15533 PR tree-optimization/84136
15534 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
15535 that the result of find_edge is non-NULL.
15537 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
15540 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
15541 storing integer register in SImode. Fix cost of 256 and 512
15542 byte aligned SSE register store.
15544 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
15546 * config/i386/i386.c (ix86_multiplication_cost): Fix
15547 multiplication cost for TARGET_AVX512DQ.
15549 2018-02-08 Marek Polacek <polacek@redhat.com>
15551 PR tree-optimization/84238
15552 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
15555 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
15557 PR tree-optimization/84265
15558 * tree-vect-stmts.c (vectorizable_store): Don't treat
15559 VMAT_CONTIGUOUS accesses as grouped.
15560 (vectorizable_load): Likewise.
15562 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
15564 PR tree-optimization/81635
15565 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
15566 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
15567 (test_round_for_mask): New functions.
15568 (wide_int_cc_tests): Call test_round_for_mask.
15569 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
15570 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
15571 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
15572 range returned by get_range_info.
15574 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
15577 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
15578 * symtab.c: Include builtins.h
15579 (symtab_node::output_to_lto_symbol_table_p): Move here
15580 from lto-streamer-out.c:output_symbol_p.
15581 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
15582 (output_symbol_p): Move all logic to symtab.c
15583 (produce_symtab): Update.
15585 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
15587 * config/s390/s390-opts.h (enum indirect_branch): Define.
15588 * config/s390/s390-protos.h (s390_return_addr_from_memory)
15589 (s390_indirect_branch_via_thunk)
15590 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
15591 (enum s390_indirect_branch_type): Define.
15592 * config/s390/s390.c (struct s390_frame_layout, struct
15593 machine_function): Remove.
15594 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
15595 (indirect_branch_table_label_no, indirect_branch_table_name):
15597 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
15598 (enum s390_indirect_branch_option): Define.
15599 (s390_return_addr_from_memory): New function.
15600 (s390_handle_string_attribute): New function.
15601 (s390_attribute_table): Add new attribute handler.
15602 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
15603 (s390_indirect_branch_via_thunk): New function.
15604 (s390_indirect_branch_via_inline_thunk): New function.
15605 (s390_function_ok_for_sibcall): When jumping via thunk disallow
15606 sibling call optimization for non z10 compiles.
15607 (s390_emit_call): Force indirect branch target to be a single
15608 register. Add r1 clobber for non-z10 compiles.
15609 (s390_emit_epilogue): Emit return jump via return_use expander.
15610 (s390_reorg): Handle JUMP_INSNs as execute targets.
15611 (s390_option_override_internal): Perform validity checks for the
15612 new command line options.
15613 (s390_indirect_branch_attrvalue): New function.
15614 (s390_indirect_branch_settings): New function.
15615 (s390_set_current_function): Invoke s390_indirect_branch_settings.
15616 (s390_output_indirect_thunk_function): New function.
15617 (s390_code_end): Implement target hook.
15618 (s390_case_values_threshold): Implement target hook.
15619 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
15621 * config/s390/s390.h (struct s390_frame_layout)
15622 (struct machine_function): Move here from s390.c.
15623 (TARGET_INDIRECT_BRANCH_NOBP_RET)
15624 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
15625 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
15626 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
15627 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
15628 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
15629 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
15630 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
15631 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
15632 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
15633 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
15634 (mnemonic attribute): Add values which aren't recognized
15636 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
15637 pattern for branch conversion. Fix mnemonic attribute.
15638 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
15639 indirect branch via thunk if requested.
15640 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
15641 ("*indirect_jump"): Disable for branch conversion using out of
15643 ("indirect_jump_via_thunk<mode>_z10")
15644 ("indirect_jump_via_thunk<mode>")
15645 ("indirect_jump_via_inlinethunk<mode>_z10")
15646 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
15647 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
15648 ("casesi_jump_via_inlinethunk<mode>_z10")
15649 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
15650 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
15651 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
15652 ("*indirect2_jump"): Disable for branch conversion.
15653 ("casesi_jump"): Turn into expander and expand patterns for branch
15655 ("return_use"): New expander.
15656 ("*return"): Emit return via thunk and rename it to ...
15657 ("*return<mode>"): ... this one.
15658 * config/s390/s390.opt: Add new options and and enum for the
15661 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
15663 * lra-constraints.c (match_reload): Unconditionally use
15664 gen_lowpart_SUBREG, rather than selecting between that
15665 and equivalent gen_rtx_SUBREG code.
15667 2018-02-08 Richard Biener <rguenther@suse.de>
15669 PR tree-optimization/84233
15670 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
15671 changed flag instead of boguously re-using phi_inserted.
15673 2018-02-08 Martin Jambor <mjambor@suse.cz>
15675 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
15676 static local variables.
15678 2018-02-08 Richard Biener <rguenther@suse.de>
15680 PR tree-optimization/84278
15681 * tree-vect-stmts.c (vectorizable_store): When looking for
15682 smaller vector types to perform grouped strided loads/stores
15683 make sure the mode is supported by the target.
15684 (vectorizable_load): Likewise.
15686 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
15688 * config/aarch64/aarch64.c (aarch64_components_for_bb):
15689 Increase LDP/STP opportunities by adding adjacent callee-saves.
15691 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
15693 PR rtl-optimization/84068
15694 PR rtl-optimization/83459
15695 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
15697 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
15699 PR tree-optimization/84224
15700 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
15701 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
15702 non-zero arguments.
15704 2018-02-07 Iain Sandoe <iain@codesourcery.com>
15707 * config/rs6000/altivec.md (*restore_world): Remove LR use.
15708 * config/rs6000/predicates.md (restore_world_operation): Adjust op
15709 count, remove one USE.
15711 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
15713 * doc/install.texi (Configuration): Document the
15714 --with-long-double-format={ibm,ieee} PowerPC configuration
15718 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
15719 Convert from define_expand to be define_insn_and_split. Rework
15720 float/double/_Float128 conversions to QI/HI/SImode to work with
15721 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
15722 conversions to QI/HImode types did a store and then a load to
15723 truncate the value. For conversions to VSX registers, don't split
15724 the insn, instead emit the code directly. Use the code iterator
15725 any_fix to combine signed and unsigned conversions.
15726 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
15727 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
15728 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
15729 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
15730 (fix_<mode>di2_hw): Likewise.
15731 (fixuns_<mode>di2_hw): Likewise.
15732 (fix_<mode>si2_hw): Likewise.
15733 (fixuns_<mode>si2_hw): Likewise.
15734 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
15735 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
15736 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
15737 fix<uns>_trunc<SFDF:mode>si2_p8.
15738 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
15739 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
15740 (fix<uns>_<mode>_mem): Likewise.
15741 (fctiw<u>z_<mode>_mem): Likewise.
15742 (fix<uns>_<mode>_mem): Likewise.
15743 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
15744 the register allocator from doing a direct move to the GPRs to do
15745 a store, and instead use the ISA 3.0 store byte/half-word from
15746 vector register instruction. For IEEE 128-bit floating point,
15747 also optimize stores of 32-bit ints.
15748 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
15750 2018-02-07 Alan Hayward <alan.hayward@arm.com>
15752 * genextract.c (push_pathstr_operand): New function to support
15754 (walk_rtx): Call push_pathstr_operand.
15755 (print_path): Support [a-zA-Z].
15757 2018-02-07 Richard Biener <rguenther@suse.de>
15759 PR tree-optimization/84037
15760 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
15761 (cse_and_gimplify_to_preheader): Declare.
15762 (vect_get_place_in_interleaving_chain): Likewise.
15763 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15765 (_loop_vec_info::~_loop_vec_info): Delete it.
15766 (cse_and_gimplify_to_preheader): New function.
15767 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
15768 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
15769 (vectorizable_load): Likewise. For grouped stores always base
15770 the IV on the first element.
15771 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
15772 condition before gimplifying.
15774 2018-02-07 Jakub Jelinek <jakub@redhat.com>
15776 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
15777 *DIV_EXPR and *MOD_EXPR.
15779 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
15782 * config/i386/i386.c (ix86_option_override_internal): Mask out
15783 the CF_SET bit when checking -fcf-protection.
15785 2018-02-07 Tom de Vries <tom@codesourcery.com>
15788 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
15791 2018-02-07 Richard Biener <rguenther@suse.de>
15793 PR tree-optimization/84204
15794 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
15797 PR tree-optimization/84205
15798 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
15799 special-case isl_ast_op_zdiv_r.
15801 PR tree-optimization/84223
15802 * graphite-scop-detection.c (gather_bbs::before_dom_children):
15803 Only add conditions from within the region.
15804 (gather_bbs::after_dom_children): Adjust.
15806 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
15809 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
15810 * config/avr/avr.md: Only post-reload split REG-REG moves if
15811 either register is GENERAL_REG_P.
15813 2018-02-07 Jakub Jelinek <jakub@redhat.com>
15815 PR tree-optimization/84235
15816 * tree-ssa-scopedtables.c
15817 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
15818 if the subtraction is performed in floating point type where NaNs are
15819 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
15820 build 1. Formatting fix.
15822 2018-02-06 Jakub Jelinek <jakub@redhat.com>
15825 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
15826 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
15827 and skip it regardless of bb boundaries. Use CALL_P macro,
15828 don't test INSN_P (insn) together with CALL_P or JUMP_P check
15829 unnecessarily, formatting fix.
15831 2018-02-06 Michael Collison <michael.collison@arm.com>
15833 * config/arm/thumb2.md:
15834 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
15835 (*thumb_mov_notscc): Ditto.
15837 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
15840 * config/rs6000/rs6000.md (su code attribute): Use "u" for
15841 unsigned_fix, not "s".
15843 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15845 * configure.ac (gcc_fn_eh_frame_ro): New function.
15846 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
15847 correct .eh_frame permissions.
15848 * configure: Regenerate.
15850 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
15852 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
15853 irrelevant options.
15855 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15857 * config/rs6000/rs6000.c (rs6000_option_override_internal):
15858 Display warning message for -mno-speculate-indirect-jumps.
15860 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
15862 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
15864 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
15866 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
15868 PR tree-optimization/84225
15869 * tree-eh.c (find_trapping_overflow): Only call
15870 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
15872 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
15875 * config/i386/i386.c: Reimplement the check of possible options
15876 -mibt/-mshstk conbination. Change error messages.
15877 * doc/invoke.texi: Fix a typo: remove extra '='.
15879 2018-02-06 Marek Polacek <polacek@redhat.com>
15881 PR tree-optimization/84228
15882 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
15884 2018-02-06 Tamar Christina <tamar.christina@arm.com>
15887 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
15888 emitted arch directives.
15889 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
15890 __ARM_FEATURE_COPROC before changing architectures.
15892 2018-02-06 Richard Biener <rguenther@suse.de>
15894 * config/i386/i386.c (print_reg): Fix typo.
15895 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
15897 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
15899 * configure: Regenerate.
15901 2018-02-05 Martin Sebor <msebor@redhat.com>
15903 PR tree-optimization/83369
15904 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
15907 2018-02-05 Martin Liska <mliska@suse.cz>
15909 * doc/invoke.texi: Cherry-pick upstream r323995.
15911 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
15913 * ira.c (ira_init_register_move_cost): Adjust comment.
15915 2018-02-05 Martin Liska <mliska@suse.cz>
15917 PR gcov-profile/84137
15918 * doc/gcov.texi: Fix typo in documentation.
15920 2018-02-05 Martin Liska <mliska@suse.cz>
15922 PR gcov-profile/83879
15923 * doc/gcov.texi: Document necessity of --dynamic-list-data when
15924 using dlopen functionality.
15926 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
15928 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
15929 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
15930 _mm_maskz_range_ss, _mm_mask_range_round_ss,
15931 _mm_maskz_range_round_ss): New intrinsics.
15932 (__builtin_ia32_rangesd128_round)
15933 (__builtin_ia32_rangess128_round): Remove.
15934 (__builtin_ia32_rangesd128_mask_round,
15935 __builtin_ia32_rangess128_mask_round): New builtins.
15936 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
15937 __builtin_ia32_rangess128_round): Remove.
15938 (__builtin_ia32_rangesd128_mask_round,
15939 __builtin_ia32_rangess128_mask_round): New builtins.
15940 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
15941 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
15942 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
15943 "<round_saeonly_constraint>")): Changed to ...
15944 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
15945 "<round_saeonly_scalar_constraint>")): ... this.
15946 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
15947 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
15948 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
15949 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
15950 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
15952 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
15954 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
15956 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
15957 Remove all values except native, 8540 and 8548.
15959 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
15961 * config/i386/i386.c (ix86_output_function_return): Pass
15962 INVALID_REGNUM, instead of -1, as invalid register number to
15963 indirect_thunk_name and output_indirect_thunk.
15965 2018-02-02 Julia Koval <julia.koval@intel.com>
15967 * config.gcc: Add -march=icelake.
15968 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
15969 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
15970 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
15971 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
15972 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
15973 (processor_target_table): Add icelake.
15974 (ix86_option_override_internal): Handle new PTAs.
15975 (get_builtin_code_for_version): Handle icelake.
15976 (M_INTEL_COREI7_ICELAKE): New.
15977 (fold_builtin_cpu): Handle icelake.
15978 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
15979 * doc/invoke.texi: Add -march=icelake.
15981 2018-02-02 Julia Koval <julia.koval@intel.com>
15983 * config/i386/i386.c (ix86_option_override_internal): Change flags type
15984 to wide_int_bitmask.
15985 * wide-int-bitmask.h: New.
15987 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
15990 * config/i386/i386.md: Replace Pmode with word_mode in
15991 builtin_setjmp_setup and builtin_longjmp to support x32.
15993 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
15997 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
15999 (rs6000_supported_cpu_names): New static variable.
16000 (linux_cpu_translation_table): Likewise.
16001 (elf_platform) <cpu>: Define new static variable and use it.
16002 Translate kernel AT_PLATFORM name to canonical name if needed.
16003 Error if platform name is unknown.
16005 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
16008 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
16010 2018-02-01 Jeff Law <law@redhat.com>
16013 * config/i386/i386.c (release_scratch_register_on_entry): Add new
16014 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
16015 the scratch if RELEASE_VIA_POP is false.
16016 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
16017 If we have to save a temporary register, decrement SIZE appropriately.
16018 Pass new arguments to release_scratch_register_on_entry.
16019 (ix86_adjust_stack_and_probe): Likewise.
16020 (ix86_emit_probe_stack_range): Pass new arguments to
16021 release_scratch_register_on_entry.
16023 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
16025 PR rtl-optimization/84157
16026 * combine.c (change_zero_ext): Use REG_P predicate in
16027 front of HARD_REGISTER_P predicate.
16029 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
16031 * config/avr/avr.c (avr_option_override): Move disabling of
16032 -fdelete-null-pointer-checks to...
16033 * common/config/avr/avr-common.c (avr_option_optimization_table):
16036 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
16038 PR tree-optimization/81635
16039 * tree-data-ref.c (split_constant_offset_1): For types that
16040 wrap on overflow, try to use range info to prove that wrapping
16043 2018-02-01 Renlin Li <renlin.li@arm.com>
16046 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
16047 TAILCALL_ADDR_REGS.
16048 (aarch64_register_move_cost): Likewise.
16049 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
16050 TAILCALL_ADDR_REGS.
16051 (REG_CLASS_NAMES): Likewise.
16052 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
16053 TAILCALL_ADDR_REGS. Remove IP registers.
16054 * config/aarch64/aarch64.md (Ucs): Update register constraint.
16056 2018-02-01 Richard Biener <rguenther@suse.de>
16058 * domwalk.h (dom_walker::dom_walker): Add additional constructor
16059 for specifying RPO order and allow NULL for that.
16060 * domwalk.c (dom_walker::dom_walker): Likewise.
16061 (dom_walker::walk): Handle NULL RPO order.
16062 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
16064 (rewrite_update_dom_walker): Likewise.
16065 (mark_def_dom_walker): Likewise.
16067 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
16069 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
16070 (aarch64_maybe_expand_sve_subreg_move): Declare.
16071 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
16072 * config/aarch64/predicates.md (aarch64_any_register_operand): New
16074 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
16075 that are semantically a reverse operation.
16076 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
16077 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
16078 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
16080 (aarch64_can_change_mode_class): For big-endian, forbid changes
16081 between two SVE modes if they have different element sizes.
16083 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
16085 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
16086 the TImode handling for big-endian targets.
16088 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
16090 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
16091 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
16093 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
16094 Remove BSWAP handing for big-endian targets and use the form of
16095 LD1RQ appropariate for the mode.
16097 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
16099 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
16100 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
16101 duplicated element.
16103 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
16106 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
16107 check for operands that need to go through aarch64_sve_reload_be.
16109 2018-02-01 Jakub Jelinek <jakub@redhat.com>
16111 PR tree-optimization/81661
16112 PR tree-optimization/84117
16113 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
16114 * tree-eh.c: Include gimplify.h.
16115 (find_trapping_overflow, replace_trapping_overflow,
16116 rewrite_to_non_trapping_overflow): New functions.
16117 * tree-vect-loop.c: Include tree-eh.h.
16118 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
16119 * tree-data-ref.c: Include tree-eh.h.
16120 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
16122 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
16124 PR rtl-optimization/84123
16125 * combine.c (change_zero_ext): Check if hard register satisfies
16126 can_change_dest_mode before calling gen_lowpart_SUBREG.
16128 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
16131 * ira.c (ira_init_register_move_cost): Remove assert.
16133 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
16135 PR rtl-optimization/84071
16136 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
16137 * doc/tm.texi: Regenerate.
16139 2018-01-31 Richard Biener <rguenther@suse.de>
16141 PR tree-optimization/84132
16142 * tree-data-ref.c (analyze_miv_subscript): Properly
16143 check whether evolution_function_is_affine_multivariate_p
16144 before calling gcd_of_steps_may_divide_p.
16146 2018-01-31 Julia Koval <julia.koval@intel.com>
16149 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
16150 * config/i386/i386.md (rdpid_rex64) New.
16151 (rdpid): Make 32bit only.
16153 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
16156 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
16157 an IDENTIFIER_NODE for FUNCTION_TYPE's.
16159 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
16162 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
16164 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
16166 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
16168 PR rtl-optimization/84071
16169 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
16170 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
16172 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
16174 * config/arc/arc.c (arc_handle_aux_attribute): New function.
16175 (arc_attribute_table): Add 'aux' attribute.
16176 (arc_in_small_data_p): Consider aux like variables.
16177 (arc_is_aux_reg_p): New function.
16178 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
16179 (arc_get_aux_arg): New function.
16180 (prepare_move_operands): Handle aux-register access.
16181 (arc_handle_aux_attribute): New function.
16182 * doc/extend.texi (ARC Variable attributes): Add subsection.
16184 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
16186 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
16187 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
16188 (arc_attribute_table): Add 'uncached' attribute.
16189 (arc_print_operand): Print '.di' flag for uncached memory
16191 (arc_in_small_data_p): Do not consider for small data the uncached
16193 (arc_is_uncached_mem_p): New function.
16194 * config/arc/predicates.md (compact_store_memory_operand): Check
16195 for uncached memory accesses.
16196 (nonvol_nonimm_operand): Likewise.
16197 * doc/extend.texi (ARC Type Attribute): New subsection.
16199 2018-01-31 Jakub Jelinek <jakub@redhat.com>
16202 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
16203 falign-loops=): Add Optimization flag.
16205 2018-01-30 Jeff Law <law@redhat.com>
16208 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
16209 INT_REGISTERS_SAVED. Check it prior to calling
16210 get_scratch_register_on_entry.
16211 (ix86_adjust_stack_and_probe): Similarly.
16212 (ix86_emit_probe_stack_range): Similarly.
16213 (ix86_expand_prologue): Corresponding changes.
16215 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16218 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
16219 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
16221 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
16224 * lra-constraints.c (curr_insn_transform): Process AND in the
16227 2018-01-30 Jakub Jelinek <jakub@redhat.com>
16229 PR rtl-optimization/83986
16230 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
16231 dependence against last_pending_memory_flush in addition to
16232 pending_jump_insns.
16234 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
16236 PR tree-optimization/81611
16237 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
16240 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16243 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
16246 2018-01-30 Richard Biener <rguenther@suse.de>
16247 Jakub Jelinek <jakub@redhat.com>
16249 PR tree-optimization/84111
16250 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
16251 inner loops added during recursion, as they don't have up-to-date
16254 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
16257 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
16258 (can_inline_edge_by_limits_p): ... here.
16259 (can_early_inline_edge_p, check_callers,
16260 update_caller_keys, update_callee_keys, recursive_inlining,
16261 add_new_edges_to_heap, speculation_useful_p,
16262 inline_small_functions,
16263 inline_small_functions, flatten_function,
16264 inline_to_all_callers_1): Update.
16266 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
16268 * profile-count.c (profile_count::combine_with_ipa_count): Handle
16271 2018-01-30 Richard Biener <rguenther@suse.de>
16273 PR tree-optimization/83008
16274 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
16275 invariant and constant vector uses in stmts when they need
16276 more than one stmt.
16278 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16281 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
16282 * configure: Regenerate.
16284 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
16286 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
16288 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
16289 Use gen_rtx_REG rather than gen_lowpart.
16291 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
16293 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
16294 rather than 0 when creating partial subregs.
16296 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
16298 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
16301 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
16304 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
16305 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
16306 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
16307 flags. This restores the settings used before the 2017-07-24.
16308 Turning off pre increment/decrement/modify allows IVOPTS to
16309 optimize DF/SF loops where the index is an int.
16311 2018-01-29 Richard Biener <rguenther@suse.de>
16312 Kelvin Nilsen <kelvin@gcc.gnu.org>
16315 * tree-vect-stmts.c (vectorizable_call): Don't call
16316 targetm.vectorize_builtin_md_vectorized_function if callee is
16319 2018-01-22 Carl Love <cel@us.ibm.com>
16321 * doc/extend.tex: Fix typo in second arg in
16322 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
16324 2018-01-29 Richard Biener <rguenther@suse.de>
16326 PR tree-optimization/84086
16327 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
16328 (flush_ssaname_freelist): When SSA names were released reset
16329 the SCEV hash table.
16331 2018-01-29 Richard Biener <rguenther@suse.de>
16333 PR tree-optimization/84057
16334 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
16335 removed paths when removing edges.
16337 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
16339 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
16340 -mfunction-return=@var{choice}.
16342 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
16344 PR diagnostic/84034
16345 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
16346 Handle CR like TAB.
16347 (layout::print_source_line): Likewise.
16348 (test_get_line_width_without_trailing_whitespace): Add test cases.
16350 2018-01-27 Jakub Jelinek <jakub@redhat.com>
16352 PR middle-end/84040
16353 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
16356 2018-01-26 Jim Wilson <jimw@sifive.com>
16358 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
16360 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
16363 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16365 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
16366 and CMP + SUB-immediate -> SUBS.
16368 2018-01-26 Martin Sebor <msebor@redhat.com>
16370 PR tree-optimization/83896
16371 * tree-ssa-strlen.c (get_string_len): Rename...
16372 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
16373 Avoid assuming length is constant.
16374 (handle_char_store): Use HOST_WIDE_INT for string length.
16376 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
16379 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
16380 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
16382 2018-01-26 Richard Biener <rguenther@suse.de>
16384 PR rtl-optimization/84003
16385 * dse.c (record_store): Only record redundant stores when
16386 the earlier store aliases at least all accesses the later one does.
16388 2018-01-26 Jakub Jelinek <jakub@redhat.com>
16390 PR rtl-optimization/83985
16391 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
16392 REG_CFA_RESTORE insns.
16393 (delete_unmarked_insns): Don't ignore separate shrink wrapping
16394 REG_CFA_RESTORE insns here.
16397 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
16398 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
16400 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
16402 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
16403 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
16404 (arc_init): Likewise.
16405 (arc_override_options): Likewise.
16406 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
16408 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
16410 * config/arc/arc.h (TARGET_DBNZ): Define.
16411 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
16412 properly set the tune attribute.
16413 (dbnz): Use TARGET_DBNZ guard.
16414 * config/arc/arc.opt (mtune): Add core3 option.
16416 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
16418 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
16419 recognize new pic like addresses.
16420 (arc_delegitimize_address): Clean up.
16422 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
16424 * config/arc/arc-arches.def: Option mrf16 valid for all
16426 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
16427 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
16428 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
16429 * config/arc/arc-tables.opt: Regenerate.
16430 * config/arc/arc.c (arc_conditional_register_usage): Handle
16431 reduced register file case.
16432 (arc_file_start): Set must have build attributes.
16433 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
16434 mrf16 option value.
16435 * config/arc/arc.opt (mrf16): Add new option.
16436 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
16437 * config/arc/genmultilib.awk: Handle new mrf16 option.
16438 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
16439 * config/arc/t-multilib: Regenerate.
16440 * doc/invoke.texi (ARC Options): Document mrf16 option.
16442 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
16444 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
16445 * config/arc/arc.c (arc_handle_secure_attribute): New function.
16446 (arc_attribute_table): Add 'secure_call' attribute.
16447 (arc_print_operand): Print secure call operand.
16448 (arc_function_ok_for_sibcall): Don't optimize tail calls when
16450 (arc_is_secure_call_p): New function. * config/arc/arc.md
16451 (call_i): Add support for sjli instruction.
16452 (call_value_i): Likewise.
16453 * config/arc/constraints.md (Csc): New constraint.
16455 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
16456 John Eric Martin <John.Martin@emmicro-us.com>
16458 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
16459 * config/arc/arc.c (_arc_jli_section): New struct.
16460 (arc_jli_section): New type.
16461 (rc_jli_sections): New static variable.
16462 (arc_handle_jli_attribute): New function.
16463 (arc_attribute_table): Add jli_always and jli_fixed attribute.
16464 (arc_file_end): New function.
16465 (TARGET_ASM_FILE_END): Define.
16466 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
16467 (arc_add_jli_section): New function.
16468 (jli_call_scan): Likewise.
16469 (arc_reorg): Call jli_call_scan.
16470 (arc_output_addsi): Remove 'S' from printing asm operand.
16471 (arc_is_jli_call_p): New function.
16472 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
16474 (movhi_insn): Likewise.
16475 (movsi_insn): Likewise.
16476 (movsi_set_cc_insn): Likewise.
16477 (loadqi_update): Likewise.
16478 (load_zeroextendqisi_update): Likewise.
16479 (load_signextendqisi_update): Likewise.
16480 (loadhi_update): Likewise.
16481 (load_zeroextendhisi_update): Likewise.
16482 (load_signextendhisi_update): Likewise.
16483 (loadsi_update): Likewise.
16484 (loadsf_update): Likewise.
16485 (movsicc_insn): Likewise.
16486 (bset_insn): Likewise.
16487 (bxor_insn): Likewise.
16488 (bclr_insn): Likewise.
16489 (bmsk_insn): Likewise.
16490 (bicsi3_insn): Likewise.
16491 (cmpsi_cc_c_insn): Likewise.
16492 (movsi_ne): Likewise.
16493 (movsi_cond_exec): Likewise.
16494 (clrsbsi2): Likewise.
16495 (norm_f): Likewise.
16505 (call_i): Remove 'S' asm letter, add jli instruction.
16506 (call_value_i): Likewise.
16507 * config/arc/arc.op (mjli-always): New option.
16508 * config/arc/constraints.md (Cji): New constraint.
16509 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
16511 (subsf3_fpx): Likewise.
16512 (mulsf3_fpx): Likewise.
16513 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
16515 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
16516 function attrbutes.
16517 * doc/invoke.texi (ARC): Document mjli-always option.
16519 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
16521 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
16522 avoid addition with 0 and use incw and decw where possible.
16524 2018-01-26 Richard Biener <rguenther@suse.de>
16526 PR tree-optimization/81082
16527 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
16528 association if it requires casting to unsigned.
16529 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
16530 from fold_plusminus_mult_expr to catch important cases late when
16531 range info is available.
16533 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16535 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
16536 * configure.ac (hidden_linkonce): New test.
16537 * configure: Regenerate.
16538 * config.in: Regenerate.
16540 2018-01-26 Julia Koval <julia.koval@intel.com>
16542 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
16543 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
16544 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
16545 _mm_mask_bitshuffle_epi64_mask): Fix type.
16546 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
16547 USI_FTYPE_V4DI_V4DI_USI): Remove.
16548 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
16549 __builtin_ia32_vpshufbitqmb256_mask,
16550 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
16551 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
16552 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
16554 2018-01-26 Alan Modra <amodra@gmail.com>
16557 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
16558 UNSPEC_VBPERMQ. Sort other unspecs.
16560 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
16562 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
16564 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
16566 PR middle-end/83055
16567 * predict.c (drop_profile): Do not push/pop cfun; update also
16569 (handle_missing_profiles): Fix logic looking for zero profiles.
16571 2018-01-25 Jakub Jelinek <jakub@redhat.com>
16573 PR middle-end/83977
16574 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
16575 on functions with #pragma omp declare simd or functions with simd
16577 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
16578 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
16579 Remove trailing \n from warning_at calls.
16581 2018-01-25 Tom de Vries <tom@codesourcery.com>
16584 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
16585 for neutered workers.
16587 2018-01-24 Joseph Myers <joseph@codesourcery.com>
16590 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
16591 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
16593 2018-01-24 Jeff Law <law@redhat.com>
16596 * i386.c (get_probe_interval): Move to earlier point.
16597 (ix86_compute_frame_layout): If -fstack-clash-protection and
16598 the frame is larger than the probe interval, then use pushes
16599 to save registers rather than reg->mem moves.
16600 (ix86_expand_prologue): Remove conditional for int_registers_saved
16603 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
16606 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
16607 min/max for never referenced object.
16609 2018-01-24 Jakub Jelinek <jakub@redhat.com>
16611 PR middle-end/83977
16612 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
16614 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
16615 attributes from DECL_ATTRIBUTES (decl) without affecting
16616 DECL_ATTRIBUTES (current_function_decl).
16617 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
16618 functions with non-NULL DECL_ABSTRACT_ORIGIN.
16620 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
16622 PR tree-optimization/83979
16623 * fold-const.c (fold_comparison): Use constant_boolean_node
16624 instead of boolean_{true,false}_node.
16626 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
16628 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
16631 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16633 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
16634 Simplify the clause that sets the length attribute.
16635 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
16636 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
16637 clause that sets the length attribute.
16638 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
16640 2018-01-24 Tom de Vries <tom@codesourcery.com>
16643 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
16644 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
16645 Add strict parameter.
16646 (prevent_branch_around_nothing): Insert dummy insn between branch to
16647 label and label with no ptx insn inbetween.
16648 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
16650 2018-01-24 Tom de Vries <tom@codesourcery.com>
16653 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
16654 for neutered threads in warp.
16655 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
16657 2018-01-24 Richard Biener <rguenther@suse.de>
16659 PR tree-optimization/83176
16660 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
16663 2018-01-24 Richard Biener <rguenther@suse.de>
16665 PR tree-optimization/82819
16666 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
16667 code generating pluses that are no-ops in the target precision.
16669 2018-01-24 Richard Biener <rguenther@suse.de>
16671 PR middle-end/84000
16672 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
16674 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
16676 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
16677 to merge probabilities.
16678 * predict.c (probably_never_executed): Also mark as cold functions
16679 with global 0 profile and guessed local profile.
16680 * profile-count.c (profile_probability::combine_with_count): New
16682 * profile-count.h (profile_probability::operator*,
16683 profile_probability::operator*=, profile_probability::operator/,
16684 profile_probability::operator/=): Reduce precision to adjusted
16685 and set value to guessed on contradictory divisions.
16686 (profile_probability::combine_with_freq): Remove.
16687 (profile_probability::combine_wiht_count): Declare.
16688 (profile_count::force_nonzero):: Set to adjusted.
16689 (profile_count::probability_in):: Set quality to adjusted.
16690 * tree-ssa-tail-merge.c (replace_block_by): Use
16691 combine_with_count.
16693 2018-01-23 Andrew Waterman <andrew@sifive.com>
16694 Jim Wilson <jimw@sifive.com>
16696 * config/riscv/riscv.c (riscv_stack_boundary): New.
16697 (riscv_option_override): Set riscv_stack_boundary. Handle
16698 riscv_preferred_stack_boundary_arg.
16699 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
16700 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
16701 (STACK_BOUNDARY): Set to riscv_stack_boundary.
16702 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
16703 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
16704 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
16706 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
16709 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
16710 of struct ix86_frame.
16711 (ix86_expand_epilogue): Likewise. Add a local variable for
16712 the reg_save_offset field in struct ix86_frame.
16714 2018-01-23 Bin Cheng <bin.cheng@arm.com>
16716 PR tree-optimization/82604
16717 * tree-loop-distribution.c (enum partition_kind): New enum item
16718 PKIND_PARTIAL_MEMSET.
16719 (partition_builtin_p): Support above new enum item.
16720 (generate_code_for_partition): Ditto.
16721 (compute_access_range): Differentiate cases that equality can be
16722 proven at all loops, the innermost loops or no loops.
16723 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
16724 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
16725 (finalize_partitions, distribute_loop): Don't fuse partition of
16726 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
16727 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
16728 parloop is enabled.
16730 2018-01-23 Martin Liska <mliska@suse.cz>
16732 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
16733 order to ignore the predictor.
16734 (PRED_POLYMORPHIC_CALL): Likewise.
16735 (PRED_RECURSIVE_CALL): Likewise.
16737 2018-01-23 Martin Liska <mliska@suse.cz>
16739 * tree-profile.c (tree_profiling): Print function header to
16740 aware reader which function we are working on.
16741 * value-prof.c (gimple_find_values_to_profile): Do not print
16742 not interesting value histograms.
16744 2018-01-23 Martin Liska <mliska@suse.cz>
16746 * profile-count.h (enum profile_quality): Add
16747 profile_uninitialized as the first value. Do not number values
16748 as they are zero based.
16749 (profile_count::verify): Update sanity check.
16750 (profile_probability::verify): Likewise.
16752 2018-01-23 Nathan Sidwell <nathan@acm.org>
16754 * doc/invoke.texi (ffor-scope): Deprecate.
16756 2018-01-23 David Malcolm <dmalcolm@redhat.com>
16758 PR tree-optimization/83510
16759 * domwalk.c (set_all_edges_as_executable): New function.
16760 (dom_walker::dom_walker): Convert bool param
16761 "skip_unreachable_blocks" to enum reachability. Move setup of
16762 edge flags to set_all_edges_as_executable and only do it when
16763 reachability is REACHABLE_BLOCKS.
16764 * domwalk.h (enum dom_walker::reachability): New enum.
16765 (dom_walker::dom_walker): Convert bool param
16766 "skip_unreachable_blocks" to enum reachability.
16767 (set_all_edges_as_executable): New decl.
16768 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
16769 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
16771 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
16772 but converting true to REACHABLE_BLOCKS.
16773 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
16775 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
16776 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
16777 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
16779 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
16780 if check_all_array_refs will be called.
16782 2018-01-23 David Malcolm <dmalcolm@redhat.com>
16784 * tree.c (selftest::test_location_wrappers): Add more test
16787 2018-01-23 David Malcolm <dmalcolm@redhat.com>
16789 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
16790 (selftest::test_bit_in_range): Likewise.
16792 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
16795 * doc/sourcebuild.texi (vect_float): Say that the selector
16796 only describes the situation when -funsafe-math-optimizations is on.
16797 (vect_float_strict): Document.
16799 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
16801 PR tree-optimization/83965
16802 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
16803 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
16804 instead of checking only for a reduction.
16805 (vect_recog_widen_sum_pattern): Likewise.
16807 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
16809 * predict.c (probably_never_executed): Only use precise profile info.
16810 (compute_function_frequency): Skip after inlining hack since we now
16811 have quality checking.
16813 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
16815 * profile-count.h (profile_probability::very_unlikely,
16816 profile_probability::unlikely, profile_probability::even): Set
16817 precision to guessed.
16819 2018-01-23 Richard Biener <rguenther@suse.de>
16821 PR tree-optimization/83963
16822 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
16823 Properly terminate dominator walk when crossing the exit edge not
16824 when visiting its source block.
16826 2018-01-23 Jakub Jelinek <jakub@redhat.com>
16829 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
16830 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
16832 2018-01-22 Jakub Jelinek <jakub@redhat.com>
16834 PR tree-optimization/83957
16835 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
16836 semicolon after for body surrounded by braces.
16838 PR tree-optimization/83081
16839 * profile-count.h (profile_probability::split): New method.
16840 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
16841 Use profile_probability::split.
16842 (do_compare_rtx_and_jump): Fix adjustment of probabilities
16843 when splitting a single conditional jump into 2.
16845 2018-01-22 David Malcolm <dmalcolm@redhat.com>
16847 PR tree-optimization/69452
16848 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
16851 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16853 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
16854 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
16855 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
16857 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16859 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
16861 * config/rl78/rl78.md (movdi): New define_expand.
16862 * config/rl78/rl78.c (rl78_split_movdi): New function.
16864 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
16867 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
16869 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
16870 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
16871 128-bit to produce an UNSPEC move to get the double word with the
16872 signbit and then a shift directly to do signbit.
16873 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
16874 implementation with a new version that just does either a direct
16875 move or a regular move. Move memory interface to separate insns.
16876 Move insns so they are next to the expander.
16877 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
16878 with signbit move. Split big and little endian case.
16879 (signbit<mode>2_dm_mem_le): Likewise.
16880 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
16881 (signbit<mode>2_dm2): Likewise.
16883 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16885 * config/rl78/rl78.md (anddi3): New define_expand.
16887 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16889 * config/rl78/rl78.md (umindi3): New define_expand.
16891 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16893 * config/rl78/rl78.md (smindi3): New define_expand.
16895 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16897 * config/rl78/rl78.md (smaxdi3): New define_expand.
16899 2018-01-22 Carl Love <cel@us.ibm.com>
16901 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
16902 LVX_V1TI): Add macro expansion.
16903 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
16904 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
16905 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
16906 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
16907 Change check to determine if the instruction is a byte reversing
16908 entry. Fix typo in comment.
16909 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
16910 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
16911 Add def_builtin calls for new builtins.
16912 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
16913 Add define_insn expansion.
16915 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16917 * config/rl78/rl78.md (umaxdi3): New define_expand.
16919 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16921 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
16922 for non-QImode registers.
16924 2018-01-22 Richard Biener <rguenther@suse.de>
16926 PR tree-optimization/83963
16927 * graphite-scop-detection.c (scop_detection::get_sese): Delay
16928 including the loop exit block.
16929 (scop_detection::merge_sese): Likewise.
16930 (scop_detection::add_scop): Do it here instead.
16932 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16934 * doc/sourcebuild.texi (arm_softfloat): Document.
16936 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
16939 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
16940 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
16941 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
16943 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16944 David Edelsohn <dje.gcc@gmail.com>
16947 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
16948 Change "crset eq" to "crset 2".
16949 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
16950 (*call_indirect_aix<mode>_nospec): Likewise.
16951 (*call_value_indirect_aix<mode>_nospec): Likewise.
16952 (*call_indirect_elfv2<mode>_nospec): Likewise.
16953 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
16954 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
16955 change assembly output from . to $.
16956 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
16957 (indirect_jump<mode>_nospec): Change assembly output from . to $.
16958 (*tablejump<mode>_internal1_nospec): Likewise.
16960 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
16963 * config/sh/sh_optimize_sett_clrt.cc:
16964 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
16966 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
16968 PR tree-optimization/83940
16969 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
16970 offset_dt to vect_constant_def rather than vect_unknown_def_type.
16971 (vect_check_load_store_mask): Add a mask_dt_out parameter and
16972 use it to pass back the definition type.
16973 (vect_check_store_rhs): Likewise rhs_dt_out.
16974 (vect_build_gather_load_calls): Add a mask_dt argument and use
16975 it instead of a call to vect_is_simple_use.
16976 (vectorizable_store): Update calls to vect_check_load_store_mask
16977 and vect_check_store_rhs. Use the dt returned by the latter instead
16978 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
16979 instead of calls to vect_is_simple_use. Pass the scalar rather
16980 than the vector operand to vect_is_simple_use when handling
16981 second and subsequent copies of an rhs value.
16982 (vectorizable_load): Update calls to vect_check_load_store_mask
16983 and vect_build_gather_load_calls. Use the cached mask_dt and
16984 gs_info.offset_dt instead of calls to vect_is_simple_use.
16986 2018-01-20 Jakub Jelinek <jakub@redhat.com>
16988 PR middle-end/83945
16989 * tree-emutls.c: Include gimplify.h.
16990 (lower_emutls_2): New function.
16991 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
16992 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
16993 it before further processing.
16996 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
16997 UINTVAL (trueop1) instead of INTVAL (op1).
16999 2018-01-19 Jakub Jelinek <jakub@redhat.com>
17003 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
17004 INCOMING_FRAME_SP_OFFSET if not defined.
17005 (scan_trace): Add ENTRY argument. If true and
17006 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
17007 emit a note to adjust the CFA offset.
17008 (create_cfi_notes): Adjust scan_trace callers.
17009 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
17010 INCOMING_FRAME_SP_OFFSET in the CIE.
17011 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
17012 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
17014 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
17015 * doc/tm.texi: Regenerated.
17017 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
17019 PR rtl-optimization/83147
17020 * lra-constraints.c (remove_inheritance_pseudos): Use
17021 lra_substitute_pseudo_within_insn.
17023 2018-01-19 Tom de Vries <tom@codesourcery.com>
17024 Cesar Philippidis <cesar@codesourcery.com>
17027 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
17029 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
17032 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
17033 spaces for function labels.
17035 2018-01-19 Martin Liska <mliska@suse.cz>
17037 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
17038 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
17039 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
17040 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
17041 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
17042 (PRED_CONST_RETURN): Change from 69 to 65.
17043 (PRED_NULL_RETURN): Change from 91 to 71.
17044 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
17045 (PRED_LOOP_GUARD): Change from 66 to 73.
17047 2018-01-19 Martin Liska <mliska@suse.cz>
17049 * predict.c (predict_insn_def): Add new assert.
17050 (struct branch_predictor): Change type to signed integer.
17051 (test_prediction_value_range): Amend test to cover
17052 PROB_UNINITIALIZED.
17053 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
17054 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
17055 (PRED_LOOP_ITERATIONS_MAX): Likewise.
17056 (PRED_LOOP_IV_COMPARE): Likewise.
17057 * predict.h (PROB_UNINITIALIZED): Define new constant.
17059 2018-01-19 Martin Liska <mliska@suse.cz>
17061 * predict.c (dump_prediction): Add new format for
17062 analyze_brprob.py script which is enabled with -details
17064 * profile-count.h (precise_p): New function.
17066 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
17068 PR tree-optimization/83922
17069 * tree-vect-loop.c (vect_verify_full_masking): Return false if
17070 there are no statements that need masking.
17071 (vect_active_double_reduction_p): New function.
17072 (vect_analyze_loop_operations): Use it when handling phis that
17073 are not in the loop header.
17075 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
17077 PR tree-optimization/83914
17078 * tree-vect-loop.c (vectorizable_induction): Don't convert
17079 init_expr or apply the peeling adjustment for inductions
17080 that are nested within the vectorized loop.
17082 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17084 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
17087 2018-01-18 Jakub Jelinek <jakub@redhat.com>
17091 * function.h (gimplify_parameters): Add gimple_seq * argument.
17092 * function.c: Include gimple.h and options.h.
17093 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
17094 for the added local temporaries if needed.
17095 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
17096 if there are any parameter cleanups, wrap whole body into a
17097 try/finally with the cleanups.
17099 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
17102 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
17103 Use GET_MODE_CLASS for scalar floating point.
17105 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
17109 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
17110 Fix call of call_cgraph_insertion_hooks.
17112 2018-01-18 Martin Sebor <msebor@redhat.com>
17114 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
17116 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
17119 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
17122 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
17125 * common.opt: (-ffile-prefix-map): New option.
17126 * opts.c (common_handle_option): Defer it.
17127 * opts-global.c (handle_common_deferred_options): Handle it.
17128 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
17129 * file-prefix-map.h: New file.
17130 (remap_debug_filename, add_debug_prefix_map): ...here.
17131 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
17132 * final.c (debug_prefix_map, add_debug_prefix_map
17133 remap_debug_filename): Move to...
17134 * file-prefix-map.c: New file.
17135 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
17136 generalize, get rid of alloca(), use strrchr() instead of strchr().
17137 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
17138 Implement in terms of add_prefix_map().
17139 (remap_macro_filename, remap_debug_filename): Implement in term of
17141 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
17142 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
17143 * dbxout.c: Include file-prefix-map.h.
17144 * varasm.c: Likewise.
17145 * vmsdbgout.c: Likewise.
17146 * xcoffout.c: Likewise.
17147 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
17148 * doc/cppopts.texi (-fmacro-prefix-map): Document.
17149 * doc/invoke.texi (-ffile-prefix-map): Document.
17150 (-fdebug-prefix-map): Update description.
17152 2018-01-18 Martin Liska <mliska@suse.cz>
17154 * config/i386/i386.c (indirect_thunk_name): Document that also
17156 (output_indirect_thunk): Document why both instructions
17157 (pause and lfence) are generated.
17159 2018-01-18 Richard Biener <rguenther@suse.de>
17161 PR tree-optimization/83887
17162 * graphite-scop-detection.c
17163 (scop_detection::get_nearest_dom_with_single_entry): Remove.
17164 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
17165 (scop_detection::merge_sese): Re-implement with a flood-fill
17166 algorithm that properly finds a SESE region if it exists.
17168 2018-01-18 Jakub Jelinek <jakub@redhat.com>
17171 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
17172 pointer_diff optimizations use view_convert instead of convert.
17174 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17176 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
17177 Generate different code for -mno-speculate-indirect-jumps.
17178 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
17179 (*call_indirect_aix<mode>): Disable for
17180 -mno-speculate-indirect-jumps.
17181 (*call_indirect_aix<mode>_nospec): New define_insn.
17182 (*call_value_indirect_aix<mode>): Disable for
17183 -mno-speculate-indirect-jumps.
17184 (*call_value_indirect_aix<mode>_nospec): New define_insn.
17185 (*sibcall_nonlocal_sysv<mode>): Generate different code for
17186 -mno-speculate-indirect-jumps.
17187 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
17189 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
17191 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
17192 long double type, set the flags for noting the default long double
17193 type, even if we don't pass or return a long double type.
17195 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
17198 * ipa-inline.c (flatten_function): Do not overwrite final inlining
17201 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
17203 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
17204 support for merge[hl].
17205 (fold_mergehl_helper): New helper function.
17206 (tree-vector-builder.h): New #include for tree_vector_builder usage.
17207 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
17208 (altivec_vmrglw_direct): Add xxmrglw insn.
17210 2018-01-17 Andrew Waterman <andrew@sifive.com>
17212 * config/riscv/riscv.c (riscv_conditional_register_usage): If
17213 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
17215 2018-01-17 David Malcolm <dmalcolm@redhat.com>
17218 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
17219 call the lto_location_cache before reading the
17220 DECL_SOURCE_LOCATION of the types.
17222 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
17223 Richard Sandiford <richard.sandiford@linaro.org>
17225 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
17226 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
17227 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
17228 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
17229 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
17231 * config/aarch64/constraints.md (aarch64_movti_operand):
17233 * config/aarch64/predicates.md (Uti): Add new constraint.
17235 2018-01-17 Carl Love <cel@us.ibm.com>
17237 * config/rs6000/vsx.md (define_expand xl_len_r,
17238 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
17239 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
17241 (define_expand, define_insn): Move the shift left from the
17242 define_insn to the define_expand for lxvl and stxvl instructions.
17243 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
17244 and XL_LEN_R definitions to PURE.
17246 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
17248 * config/i386/i386.c (indirect_thunk_name): Declare regno
17249 as unsigned int. Compare regno with INVALID_REGNUM.
17250 (output_indirect_thunk): Ditto.
17251 (output_indirect_thunk_function): Ditto.
17252 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
17253 in the call to output_indirect_thunk_function.
17255 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
17257 PR middle-end/83884
17258 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
17259 rather than the size of inner_type to determine the stack slot size
17260 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
17262 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
17265 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
17268 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
17270 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
17271 endian Linux systems to optionally enable multilibs for selecting
17272 the long double type if the user configured an explicit type.
17273 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
17274 have no long double multilibs if not defined.
17275 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
17276 warn if the user used -mabi={ieee,ibm}longdouble and we built
17277 multilibs for long double.
17278 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
17279 appropriate multilib option.
17280 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
17282 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
17283 for building long double multilibs.
17284 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
17286 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
17288 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
17291 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
17293 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
17296 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
17299 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
17302 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
17304 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
17305 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
17307 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
17309 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
17310 different rtl trees depending on TARGET_64BIT.
17311 (rs6000_gen_lvx): Likewise.
17313 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
17315 * config/visium/visium.md (nop): Tweak comment.
17316 (hazard_nop): Likewise.
17318 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17320 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
17321 -mspeculate-indirect-jumps.
17322 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
17323 for -mno-speculate-indirect-jumps.
17324 (*call_indirect_elfv2<mode>_nospec): New define_insn.
17325 (*call_value_indirect_elfv2<mode>): Disable for
17326 -mno-speculate-indirect-jumps.
17327 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
17328 (indirect_jump): Emit different RTL for
17329 -mno-speculate-indirect-jumps.
17330 (*indirect_jump<mode>): Disable for
17331 -mno-speculate-indirect-jumps.
17332 (*indirect_jump<mode>_nospec): New define_insn.
17333 (tablejump): Emit different RTL for
17334 -mno-speculate-indirect-jumps.
17335 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
17336 (tablejumpsi_nospec): New define_expand.
17337 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
17338 (tablejumpdi_nospec): New define_expand.
17339 (*tablejump<mode>_internal1): Disable for
17340 -mno-speculate-indirect-jumps.
17341 (*tablejump<mode>_internal1_nospec): New define_insn.
17342 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
17345 2018-01-16 Artyom Skrobov tyomitch@gmail.com
17347 * caller-save.c (insert_save): Drop unnecessary parameter. All
17350 2018-01-16 Jakub Jelinek <jakub@redhat.com>
17351 Richard Biener <rguenth@suse.de>
17354 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
17355 return early, inline manually is_gimple_sizepos. Make sure if we
17356 call gimplify_expr we don't end up with a gimple constant.
17357 * tree.c (variably_modified_type_p): Don't return true for
17358 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
17359 * gimplify.h (is_gimple_sizepos): Remove.
17361 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
17363 PR tree-optimization/83857
17364 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
17365 vectorizable_live_operation for pure SLP statements.
17366 (vectorizable_live_operation): Handle PHIs.
17368 2018-01-16 Richard Biener <rguenther@suse.de>
17370 PR tree-optimization/83867
17371 * tree-vect-stmts.c (vect_transform_stmt): Precompute
17372 nested_in_vect_loop_p since the scalar stmt may get invalidated.
17374 2018-01-16 Jakub Jelinek <jakub@redhat.com>
17377 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
17378 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
17379 If off is not INTEGER_CST, issue a may not be aligned warning
17380 rather than isn't aligned. Use isn%'t rather than isn't.
17381 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
17383 <case MULT_EXPR>: Improve the case when bottom and one of the
17384 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
17385 operand, in that case check if the other operand is multiple of
17386 bottom divided by the INTEGER_CST operand.
17388 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
17391 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
17392 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
17393 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
17394 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
17395 * config/pa/pa.c (pa_function_arg_advance): Likewise.
17396 (pa_function_arg, pa_arg_partial_bytes): Likewise.
17397 (pa_function_arg_size): New function.
17399 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
17401 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
17402 in a separate statement.
17404 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
17406 PR tree-optimization/83847
17407 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
17408 group gathers and scatters.
17410 2018-01-16 Jakub Jelinek <jakub@redhat.com>
17412 PR rtl-optimization/86620
17413 * params.def (max-sched-ready-insns): Bump minimum value to 1.
17415 PR rtl-optimization/83213
17416 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
17417 to last if both are JUMP_INSNs.
17419 PR tree-optimization/83843
17420 * gimple-ssa-store-merging.c
17421 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
17422 store_immediate_info for bswap/nop orig_stores.
17424 2018-01-15 Andrew Waterman <andrew@sifive.com>
17426 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
17428 <UDIV>: Increase cost if !TARGET_DIV.
17430 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
17432 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
17433 (define_attr "cr_logical_3op"): New.
17434 (cceq_ior_compare): Adjust.
17435 (cceq_ior_compare_complement): Adjust.
17436 (*cceq_rev_compare): Adjust.
17437 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
17438 (is_cracked_insn): Adjust.
17439 (insn_must_be_first_in_group): Adjust.
17440 * config/rs6000/40x.md: Adjust.
17441 * config/rs6000/440.md: Adjust.
17442 * config/rs6000/476.md: Adjust.
17443 * config/rs6000/601.md: Adjust.
17444 * config/rs6000/603.md: Adjust.
17445 * config/rs6000/6xx.md: Adjust.
17446 * config/rs6000/7450.md: Adjust.
17447 * config/rs6000/7xx.md: Adjust.
17448 * config/rs6000/8540.md: Adjust.
17449 * config/rs6000/cell.md: Adjust.
17450 * config/rs6000/e300c2c3.md: Adjust.
17451 * config/rs6000/e500mc.md: Adjust.
17452 * config/rs6000/e500mc64.md: Adjust.
17453 * config/rs6000/e5500.md: Adjust.
17454 * config/rs6000/e6500.md: Adjust.
17455 * config/rs6000/mpc.md: Adjust.
17456 * config/rs6000/power4.md: Adjust.
17457 * config/rs6000/power5.md: Adjust.
17458 * config/rs6000/power6.md: Adjust.
17459 * config/rs6000/power7.md: Adjust.
17460 * config/rs6000/power8.md: Adjust.
17461 * config/rs6000/power9.md: Adjust.
17462 * config/rs6000/rs64.md: Adjust.
17463 * config/rs6000/titan.md: Adjust.
17465 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
17467 * config/i386/predicates.md (indirect_branch_operand): Rewrite
17468 ix86_indirect_branch_register logic.
17470 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
17472 * config/i386/constraints.md (Bs): Update
17473 ix86_indirect_branch_register check. Don't check
17474 ix86_indirect_branch_register with GOT_memory_operand.
17476 * config/i386/predicates.md (GOT_memory_operand): Don't check
17477 ix86_indirect_branch_register here.
17478 (GOT32_symbol_operand): Likewise.
17480 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
17482 * config/i386/predicates.md (constant_call_address_operand):
17483 Rewrite ix86_indirect_branch_register logic.
17484 (sibcall_insn_operand): Likewise.
17486 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
17488 * config/i386/constraints.md (Bs): Replace
17489 ix86_indirect_branch_thunk_register with
17490 ix86_indirect_branch_register.
17492 * config/i386/i386.md (indirect_jump): Likewise.
17493 (tablejump): Likewise.
17494 (*sibcall_memory): Likewise.
17495 (*sibcall_value_memory): Likewise.
17496 Peepholes of indirect call and jump via memory: Likewise.
17497 * config/i386/i386.opt: Likewise.
17498 * config/i386/predicates.md (indirect_branch_operand): Likewise.
17499 (GOT_memory_operand): Likewise.
17500 (call_insn_operand): Likewise.
17501 (sibcall_insn_operand): Likewise.
17502 (GOT32_symbol_operand): Likewise.
17504 2018-01-15 Jakub Jelinek <jakub@redhat.com>
17506 PR middle-end/83837
17507 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
17508 type rather than type addr's type points to.
17509 (expand_omp_atomic_mutex): Likewise.
17510 (expand_omp_atomic): Likewise.
17512 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
17515 * config/i386/i386.c (output_indirect_thunk_function): Use
17516 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
17517 for __x86_return_thunk.
17519 2018-01-15 Richard Biener <rguenther@suse.de>
17521 PR middle-end/83850
17522 * expmed.c (extract_bit_field_1): Fix typo.
17524 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17527 * config/arm/iterators.md (VF): New mode iterator.
17528 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
17529 Remove integer-related logic from pattern.
17530 (neon_vabd<mode>_3): Likewise.
17532 2018-01-15 Jakub Jelinek <jakub@redhat.com>
17534 PR middle-end/82694
17535 * common.opt (fstrict-overflow): No longer an alias.
17536 (fwrapv-pointer): New option.
17537 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
17538 also for pointer types based on flag_wrapv_pointer.
17539 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
17540 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
17541 opts->x_flag_wrapv got set.
17542 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
17543 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
17544 POINTER_TYPE_OVERFLOW_UNDEFINED.
17545 * match.pd: Likewise in address comparison pattern.
17546 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
17548 2018-01-15 Richard Biener <rguenther@suse.de>
17551 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
17552 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
17553 Reset type names to their identifier if their TYPE_DECL doesn't
17554 have linkage (and thus is used for ODR and devirt).
17555 (save_debug_info_for_decl): Remove.
17556 (save_debug_info_for_type): Likewise.
17557 (add_tree_to_fld_list): Adjust.
17558 * tree-pretty-print.c (dump_generic_node): Make dumping of
17559 type names more robust.
17561 2018-01-15 Richard Biener <rguenther@suse.de>
17563 * BASE-VER: Bump to 8.0.1.
17565 2018-01-14 Martin Sebor <msebor@redhat.com>
17568 * builtins.c (check_access): Avoid warning when the no-warning bit
17571 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
17573 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
17574 * ira-color (allocno_hard_regs_compare): Likewise.
17576 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
17579 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
17580 Use .pushsection/.popsection.
17582 2018-01-14 Martin Sebor <msebor@redhat.com>
17585 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
17587 2018-01-14 Jakub Jelinek <jakub@redhat.com>
17589 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
17590 entry from extra_headers.
17591 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
17592 extra_headers, make the list bitwise identical to the i?86-*-* one.
17594 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
17596 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
17597 -mcmodel=large with -mindirect-branch=thunk,
17598 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
17599 -mfunction-return=thunk-extern.
17600 * doc/invoke.texi: Document -mcmodel=large is incompatible with
17601 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
17602 -mfunction-return=thunk and -mfunction-return=thunk-extern.
17604 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
17606 * config/i386/i386.c (print_reg): Print the name of the full
17607 integer register without '%'.
17608 (ix86_print_operand): Handle 'V'.
17609 * doc/extend.texi: Document 'V' modifier.
17611 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
17613 * config/i386/constraints.md (Bs): Disallow memory operand for
17614 -mindirect-branch-register.
17616 * config/i386/predicates.md (indirect_branch_operand): Likewise.
17617 (GOT_memory_operand): Likewise.
17618 (call_insn_operand): Likewise.
17619 (sibcall_insn_operand): Likewise.
17620 (GOT32_symbol_operand): Likewise.
17621 * config/i386/i386.md (indirect_jump): Call convert_memory_address
17622 for -mindirect-branch-register.
17623 (tablejump): Likewise.
17624 (*sibcall_memory): Likewise.
17625 (*sibcall_value_memory): Likewise.
17626 Disallow peepholes of indirect call and jump via memory for
17627 -mindirect-branch-register.
17628 (*call_pop): Replace m with Bw.
17629 (*call_value_pop): Likewise.
17630 (*sibcall_pop_memory): Replace m with Bs.
17631 * config/i386/i386.opt (mindirect-branch-register): New option.
17632 * doc/invoke.texi: Document -mindirect-branch-register option.
17634 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
17636 * config/i386/i386-protos.h (ix86_output_function_return): New.
17637 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
17638 set function_return_type.
17639 (indirect_thunk_name): Add ret_p to indicate thunk for function
17641 (output_indirect_thunk_function): Pass false to
17642 indirect_thunk_name.
17643 (ix86_output_indirect_branch_via_reg): Likewise.
17644 (ix86_output_indirect_branch_via_push): Likewise.
17645 (output_indirect_thunk_function): Create alias for function
17646 return thunk if regno < 0.
17647 (ix86_output_function_return): New function.
17648 (ix86_handle_fndecl_attribute): Handle function_return.
17649 (ix86_attribute_table): Add function_return.
17650 * config/i386/i386.h (machine_function): Add
17651 function_return_type.
17652 * config/i386/i386.md (simple_return_internal): Use
17653 ix86_output_function_return.
17654 (simple_return_internal_long): Likewise.
17655 * config/i386/i386.opt (mfunction-return=): New option.
17656 (indirect_branch): Mention -mfunction-return=.
17657 * doc/extend.texi: Document function_return function attribute.
17658 * doc/invoke.texi: Document -mfunction-return= option.
17660 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
17662 * config/i386/i386-opts.h (indirect_branch): New.
17663 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
17664 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
17665 with local indirect jump when converting indirect call and jump.
17666 (ix86_set_indirect_branch_type): New.
17667 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
17668 (indirectlabelno): New.
17669 (indirect_thunk_needed): Likewise.
17670 (indirect_thunk_bnd_needed): Likewise.
17671 (indirect_thunks_used): Likewise.
17672 (indirect_thunks_bnd_used): Likewise.
17673 (INDIRECT_LABEL): Likewise.
17674 (indirect_thunk_name): Likewise.
17675 (output_indirect_thunk): Likewise.
17676 (output_indirect_thunk_function): Likewise.
17677 (ix86_output_indirect_branch_via_reg): Likewise.
17678 (ix86_output_indirect_branch_via_push): Likewise.
17679 (ix86_output_indirect_branch): Likewise.
17680 (ix86_output_indirect_jmp): Likewise.
17681 (ix86_code_end): Call output_indirect_thunk_function if needed.
17682 (ix86_output_call_insn): Call ix86_output_indirect_branch if
17684 (ix86_handle_fndecl_attribute): Handle indirect_branch.
17685 (ix86_attribute_table): Add indirect_branch.
17686 * config/i386/i386.h (machine_function): Add indirect_branch_type
17687 and has_local_indirect_jump.
17688 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
17690 (tablejump): Likewise.
17691 (*indirect_jump): Use ix86_output_indirect_jmp.
17692 (*tablejump_1): Likewise.
17693 (simple_return_indirect_internal): Likewise.
17694 * config/i386/i386.opt (mindirect-branch=): New option.
17695 (indirect_branch): New.
17698 (thunk-inline): Likewise.
17699 (thunk-extern): Likewise.
17700 * doc/extend.texi: Document indirect_branch function attribute.
17701 * doc/invoke.texi: Document -mindirect-branch= option.
17703 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
17706 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
17708 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
17710 * ipa-inline.c (want_inline_small_function_p): Return false if
17711 inlining has already failed with CIF_FINAL_ERROR.
17712 (update_caller_keys): Call want_inline_small_function_p before
17714 (update_callee_keys): Likewise.
17716 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
17718 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
17720 (rs6000_quadword_masked_address_p): Likewise.
17721 (quad_aligned_load_p): Likewise.
17722 (quad_aligned_store_p): Likewise.
17723 (const_load_sequence_p): Add comment to describe the outer-most loop.
17724 (mimic_memory_attributes_and_flags): New function.
17725 (rs6000_gen_stvx): Likewise.
17726 (replace_swapped_aligned_store): Likewise.
17727 (rs6000_gen_lvx): Likewise.
17728 (replace_swapped_aligned_load): Likewise.
17729 (replace_swapped_load_constant): Capitalize argument name in
17730 comment describing this function.
17731 (rs6000_analyze_swaps): Add a third pass to search for vector loads
17732 and stores that access quad-word aligned addresses and replace
17733 with stvx or lvx instructions when appropriate.
17734 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
17735 New function prototype.
17736 (rs6000_quadword_masked_address_p): Likewise.
17737 (rs6000_gen_lvx): Likewise.
17738 (rs6000_gen_stvx): Likewise.
17739 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
17740 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
17741 when memory address is aligned.
17742 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
17743 this split to select lvx instruction when memory address is aligned.
17744 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
17745 instruction when memory address is aligned.
17746 (*vsx_le_perm_load_v16qi): Likewise.
17747 (four unnamed splitters): Modify to select the stvx instruction
17748 when memory is aligned.
17750 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
17752 * predict.c (determine_unlikely_bbs): Handle correctly BBs
17753 which appears in the queue multiple times.
17755 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17756 Alan Hayward <alan.hayward@arm.com>
17757 David Sherwood <david.sherwood@arm.com>
17759 * tree-vectorizer.h (vec_lower_bound): New structure.
17760 (_loop_vec_info): Add check_nonzero and lower_bounds.
17761 (LOOP_VINFO_CHECK_NONZERO): New macro.
17762 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
17763 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
17764 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
17765 fields. Make seg_len the distance travelled, not including the
17767 (dr_direction_indicator): Declare.
17768 (dr_zero_step_indicator): Likewise.
17769 (dr_known_forward_stride_p): Likewise.
17770 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
17772 (runtime_alias_check_p): Allow runtime alias checks with
17774 (operator ==): Compare access_size and align.
17775 (prune_runtime_alias_test_list): Rework for new distinction between
17776 the access_size and seg_len.
17777 (create_intersect_range_checks_index): Likewise. Cope with polynomial
17779 (get_segment_min_max): New function.
17780 (create_intersect_range_checks): Use it.
17781 (dr_step_indicator): New function.
17782 (dr_direction_indicator): Likewise.
17783 (dr_zero_step_indicator): Likewise.
17784 (dr_known_forward_stride_p): Likewise.
17785 * tree-loop-distribution.c (data_ref_segment_size): Return
17786 DR_STEP * (niters - 1).
17787 (compute_alias_check_pairs): Update call to the dr_with_seg_len
17789 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
17790 (vect_preserves_scalar_order_p): New function, split out from...
17791 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
17792 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
17793 (vect_vfa_access_size): New function.
17794 (vect_vfa_align): Likewise.
17795 (vect_compile_time_alias): Take access_size_a and access_b arguments.
17796 (dump_lower_bound): New function.
17797 (vect_check_lower_bound): Likewise.
17798 (vect_small_gap_p): Likewise.
17799 (vectorizable_with_step_bound_p): Likewise.
17800 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
17801 depencies if the vectorization factor is 1. Convert the checks
17802 for nonzero steps into checks on the bounds of DR_STEP. Try using
17803 a bunds check for variable steps if the minimum required step is
17804 relatively small. Update calls to the dr_with_seg_len
17805 constructor and to vect_compile_time_alias.
17806 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
17808 (vect_loop_versioning): Call it.
17809 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
17811 (vect_estimate_min_profitable_iters): Account for any bounds checks.
17813 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17814 Alan Hayward <alan.hayward@arm.com>
17815 David Sherwood <david.sherwood@arm.com>
17817 * doc/sourcebuild.texi (vect_scatter_store): Document.
17818 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
17820 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
17822 * genopinit.c (main): Add supports_vec_scatter_store and
17823 supports_vec_scatter_store_cached to target_optabs.
17824 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
17825 IFN_MASK_SCATTER_STORE.
17826 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
17828 * internal-fn.h (internal_store_fn_p): Declare.
17829 (internal_fn_stored_value_index): Likewise.
17830 * internal-fn.c (scatter_store_direct): New macro.
17831 (expand_scatter_store_optab_fn): New function.
17832 (direct_scatter_store_optab_supported_p): New macro.
17833 (internal_store_fn_p): New function.
17834 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
17835 IFN_MASK_SCATTER_STORE.
17836 (internal_fn_mask_index): Likewise.
17837 (internal_fn_stored_value_index): New function.
17838 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
17839 for scatter stores.
17840 * optabs-query.h (supports_vec_scatter_store_p): Declare.
17841 * optabs-query.c (supports_vec_scatter_store_p): New function.
17842 * tree-vectorizer.h (vect_get_store_rhs): Declare.
17843 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
17844 true for scatter stores.
17845 (vect_gather_scatter_fn_p): Handle scatter stores too.
17846 (vect_check_gather_scatter): Consider using scatter stores if
17847 supports_vec_scatter_store_p.
17848 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
17849 scatter stores too.
17850 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
17851 internal_fn_stored_value_index.
17852 (check_load_store_masking): Handle scatter stores too.
17853 (vect_get_store_rhs): Make public.
17854 (vectorizable_call): Use internal_store_fn_p.
17855 (vectorizable_store): Handle scatter store internal functions.
17856 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
17857 when deciding whether the end of the group has been reached.
17858 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
17859 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
17860 (mask_scatter_store<mode>): New insns.
17862 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17863 Alan Hayward <alan.hayward@arm.com>
17864 David Sherwood <david.sherwood@arm.com>
17866 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
17867 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
17868 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
17870 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
17871 Use vect_truncate_gather_scatter_offset if we can't treat the
17872 operation as a normal gather load or scatter store.
17873 (get_group_load_store_type): Take the gather_scatter_info
17874 as argument. Try using a gather load or scatter store for
17875 single-element groups.
17876 (get_load_store_type): Update calls to get_group_load_store_type
17877 and vect_use_strided_gather_scatters_p.
17879 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17880 Alan Hayward <alan.hayward@arm.com>
17881 David Sherwood <david.sherwood@arm.com>
17883 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
17884 optional tree argument.
17885 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
17887 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
17888 but continue to use the current value as a fallback.
17889 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
17890 to compare the updates.
17891 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
17892 (get_load_store_type): Use it when handling a strided access.
17893 (vect_get_strided_load_store_ops): New function.
17894 (vect_get_data_ptr_increment): Likewise.
17895 (vectorizable_load): Handle strided gather loads. Always pass
17896 a step to vect_create_data_ref_ptr and bump_vector_ptr.
17898 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17899 Alan Hayward <alan.hayward@arm.com>
17900 David Sherwood <david.sherwood@arm.com>
17902 * doc/md.texi (gather_load@var{m}): Document.
17903 (mask_gather_load@var{m}): Likewise.
17904 * genopinit.c (main): Add supports_vec_gather_load and
17905 supports_vec_gather_load_cached to target_optabs.
17906 * optabs-tree.c (init_tree_optimization_optabs): Use
17907 ggc_cleared_alloc to allocate target_optabs.
17908 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
17909 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
17911 * internal-fn.h (internal_load_fn_p): Declare.
17912 (internal_gather_scatter_fn_p): Likewise.
17913 (internal_fn_mask_index): Likewise.
17914 (internal_gather_scatter_fn_supported_p): Likewise.
17915 * internal-fn.c (gather_load_direct): New macro.
17916 (expand_gather_load_optab_fn): New function.
17917 (direct_gather_load_optab_supported_p): New macro.
17918 (direct_internal_fn_optab): New function.
17919 (internal_load_fn_p): Likewise.
17920 (internal_gather_scatter_fn_p): Likewise.
17921 (internal_fn_mask_index): Likewise.
17922 (internal_gather_scatter_fn_supported_p): Likewise.
17923 * optabs-query.c (supports_at_least_one_mode_p): New function.
17924 (supports_vec_gather_load_p): Likewise.
17925 * optabs-query.h (supports_vec_gather_load_p): Declare.
17926 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
17927 and memory_type field.
17928 (NUM_PATTERNS): Bump to 15.
17929 * tree-vect-data-refs.c: Include internal-fn.h.
17930 (vect_gather_scatter_fn_p): New function.
17931 (vect_describe_gather_scatter_call): Likewise.
17932 (vect_check_gather_scatter): Try using internal functions for
17933 gather loads. Recognize existing calls to a gather load function.
17934 (vect_analyze_data_refs): Consider using gather loads if
17935 supports_vec_gather_load_p.
17936 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
17937 (vect_get_gather_scatter_offset_type): Likewise.
17938 (vect_convert_mask_for_vectype): Likewise.
17939 (vect_add_conversion_to_patterm): Likewise.
17940 (vect_try_gather_scatter_pattern): Likewise.
17941 (vect_recog_gather_scatter_pattern): New pattern recognizer.
17942 (vect_vect_recog_func_ptrs): Add it.
17943 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
17944 internal_fn_mask_index and internal_gather_scatter_fn_p.
17945 (check_load_store_masking): Take the gather_scatter_info as an
17946 argument and handle gather loads.
17947 (vect_get_gather_scatter_ops): New function.
17948 (vectorizable_call): Check internal_load_fn_p.
17949 (vectorizable_load): Likewise. Handle gather load internal
17951 (vectorizable_store): Update call to check_load_store_masking.
17952 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
17953 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
17954 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
17955 (aarch64_gather_scale_operand_d): New predicates.
17956 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
17957 (mask_gather_load<mode>): New insns.
17959 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17960 Alan Hayward <alan.hayward@arm.com>
17961 David Sherwood <david.sherwood@arm.com>
17963 * optabs.def (fold_left_plus_optab): New optab.
17964 * doc/md.texi (fold_left_plus_@var{m}): Document.
17965 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
17966 * internal-fn.c (fold_left_direct): Define.
17967 (expand_fold_left_optab_fn): Likewise.
17968 (direct_fold_left_optab_supported_p): Likewise.
17969 * fold-const-call.c (fold_const_fold_left): New function.
17970 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
17971 * tree-parloops.c (valid_reduction_p): New function.
17972 (gather_scalar_reductions): Use it.
17973 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
17974 (vect_finish_replace_stmt): Declare.
17975 * tree-vect-loop.c (fold_left_reduction_fn): New function.
17976 (needs_fold_left_reduction_p): New function, split out from...
17977 (vect_is_simple_reduction): ...here. Accept reductions that
17978 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
17979 (vect_force_simple_reduction): Also store the reduction type in
17980 the assignment's STMT_VINFO_REDUC_TYPE.
17981 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
17982 (merge_with_identity): New function.
17983 (vect_expand_fold_left): Likewise.
17984 (vectorize_fold_left_reduction): Likewise.
17985 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
17986 scalar phi in place for it. Check for target support and reject
17987 cases that would reassociate the operation. Defer the transform
17988 phase to vectorize_fold_left_reduction.
17989 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
17990 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
17991 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
17993 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17995 * tree-if-conv.c (predicate_mem_writes): Remove redundant
17996 call to ifc_temp_var.
17998 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17999 Alan Hayward <alan.hayward@arm.com>
18000 David Sherwood <david.sherwood@arm.com>
18002 * target.def (legitimize_address_displacement): Take the original
18003 offset as a poly_int.
18004 * targhooks.h (default_legitimize_address_displacement): Update
18006 * targhooks.c (default_legitimize_address_displacement): Likewise.
18007 * doc/tm.texi: Regenerate.
18008 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
18009 as an argument, moving assert of ad->disp == ad->disp_term to...
18010 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
18011 Try calling targetm.legitimize_address_displacement before expanding
18012 the address rather than afterwards, and adjust for the new interface.
18013 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
18014 Match the new hook interface. Handle SVE addresses.
18015 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
18016 new hook interface.
18018 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18020 * Makefile.in (OBJS): Add early-remat.o.
18021 * target.def (select_early_remat_modes): New hook.
18022 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
18023 * doc/tm.texi: Regenerate.
18024 * targhooks.h (default_select_early_remat_modes): Declare.
18025 * targhooks.c (default_select_early_remat_modes): New function.
18026 * timevar.def (TV_EARLY_REMAT): New timevar.
18027 * passes.def (pass_early_remat): New pass.
18028 * tree-pass.h (make_pass_early_remat): Declare.
18029 * early-remat.c: New file.
18030 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
18032 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
18034 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18035 Alan Hayward <alan.hayward@arm.com>
18036 David Sherwood <david.sherwood@arm.com>
18038 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
18039 vfm1 with a bound_epilog parameter.
18040 (vect_do_peeling): Update calls accordingly, and move the prologue
18041 call earlier in the function. Treat the base bound_epilog as 0 for
18042 fully-masked loops and retain vf - 1 for other loops. Add 1 to
18043 this base when peeling for gaps.
18044 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
18045 with fully-masked loops.
18046 (vect_estimate_min_profitable_iters): Handle the single peeled
18047 iteration in that case.
18049 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18050 Alan Hayward <alan.hayward@arm.com>
18051 David Sherwood <david.sherwood@arm.com>
18053 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
18054 single-element interleaving even if the size is not a power of 2.
18055 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
18056 accesses for single-element interleaving if the group size is
18059 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18060 Alan Hayward <alan.hayward@arm.com>
18061 David Sherwood <david.sherwood@arm.com>
18063 * doc/md.texi (fold_extract_last_@var{m}): Document.
18064 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
18065 * optabs.def (fold_extract_last_optab): New optab.
18066 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
18067 * internal-fn.c (fold_extract_direct): New macro.
18068 (expand_fold_extract_optab_fn): Likewise.
18069 (direct_fold_extract_optab_supported_p): Likewise.
18070 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
18071 * tree-vect-loop.c (vect_model_reduction_cost): Handle
18072 EXTRACT_LAST_REDUCTION.
18073 (get_initial_def_for_reduction): Do not create an initial vector
18074 for EXTRACT_LAST_REDUCTION reductions.
18075 (vectorizable_reduction): Leave the scalar phi in place for
18076 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
18077 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
18078 epilogue code for EXTRACT_LAST_REDUCTION and defer the
18079 transform phase to vectorizable_condition.
18080 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
18082 (vect_finish_stmt_generation): ...here.
18083 (vect_finish_replace_stmt): New function.
18084 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
18085 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
18087 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
18089 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18090 Alan Hayward <alan.hayward@arm.com>
18091 David Sherwood <david.sherwood@arm.com>
18093 * doc/md.texi (extract_last_@var{m}): Document.
18094 * optabs.def (extract_last_optab): New optab.
18095 * internal-fn.def (EXTRACT_LAST): New internal function.
18096 * internal-fn.c (cond_unary_direct): New macro.
18097 (expand_cond_unary_optab_fn): Likewise.
18098 (direct_cond_unary_optab_supported_p): Likewise.
18099 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
18100 loops using EXTRACT_LAST.
18101 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
18102 (extract_last_<mode>): ...this optab.
18103 (vec_extract<mode><Vel>): Update accordingly.
18105 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18106 Alan Hayward <alan.hayward@arm.com>
18107 David Sherwood <david.sherwood@arm.com>
18109 * target.def (empty_mask_is_expensive): New hook.
18110 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
18111 * doc/tm.texi: Regenerate.
18112 * targhooks.h (default_empty_mask_is_expensive): Declare.
18113 * targhooks.c (default_empty_mask_is_expensive): New function.
18114 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
18115 if the target says that empty masks are expensive.
18116 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
18118 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
18120 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18121 Alan Hayward <alan.hayward@arm.com>
18122 David Sherwood <david.sherwood@arm.com>
18124 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
18125 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
18126 (vect_use_loop_mask_for_alignment_p): New function.
18127 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
18128 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
18129 niters_skip argument. Make sure that the first niters_skip elements
18130 of the first iteration are inactive.
18131 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
18132 Update call to vect_set_loop_masks_directly.
18133 (get_misalign_in_elems): New function, split out from...
18134 (vect_gen_prolog_loop_niters): ...here.
18135 (vect_update_init_of_dr): Take a code argument that specifies whether
18136 the adjustment should be added or subtracted.
18137 (vect_update_init_of_drs): Likewise.
18138 (vect_prepare_for_masked_peels): New function.
18139 (vect_do_peeling): Skip prologue peeling if we're using a mask
18140 instead. Update call to vect_update_inits_of_drs.
18141 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
18143 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
18144 alignment. Do not include the number of peeled iterations in
18145 the minimum threshold in that case.
18146 (vectorizable_induction): Adjust the start value down by
18147 LOOP_VINFO_MASK_SKIP_NITERS iterations.
18148 (vect_transform_loop): Call vect_prepare_for_masked_peels.
18149 Take the number of skipped iterations into account when calculating
18151 * tree-vect-stmts.c (vect_gen_while_not): New function.
18153 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18154 Alan Hayward <alan.hayward@arm.com>
18155 David Sherwood <david.sherwood@arm.com>
18157 * doc/sourcebuild.texi (vect_fully_masked): Document.
18158 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
18159 default value to 0.
18160 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
18162 (vect_analyze_loop_2): ...here. Don't check the vectorization
18163 factor against the number of loop iterations if the loop is
18166 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18167 Alan Hayward <alan.hayward@arm.com>
18168 David Sherwood <david.sherwood@arm.com>
18170 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
18171 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
18172 (dump_groups): Update accordingly.
18173 (iv_use::mem_type): New member variable.
18174 (address_p): New function.
18175 (record_use): Add a mem_type argument and initialize the new
18177 (record_group_use): Add a mem_type argument. Use address_p.
18178 Remove obsolete null checks of base_object. Update call to record_use.
18179 (find_interesting_uses_op): Update call to record_group_use.
18180 (find_interesting_uses_cond): Likewise.
18181 (find_interesting_uses_address): Likewise.
18182 (get_mem_type_for_internal_fn): New function.
18183 (find_address_like_use): Likewise.
18184 (find_interesting_uses_stmt): Try find_address_like_use before
18185 calling find_interesting_uses_op.
18186 (addr_offset_valid_p): Use the iv mem_type field as the type
18187 of the addressed memory.
18188 (add_autoinc_candidates): Likewise.
18189 (get_address_cost): Likewise.
18190 (split_small_address_groups_p): Use address_p.
18191 (split_address_groups): Likewise.
18192 (add_iv_candidate_for_use): Likewise.
18193 (autoinc_possible_for_pair): Likewise.
18194 (rewrite_groups): Likewise.
18195 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
18196 (determine_group_iv_cost): Update after split of USE_ADDRESS.
18197 (get_alias_ptr_type_for_ptr_address): New function.
18198 (rewrite_use_address): Rewrite address uses in calls that were
18199 identified by find_address_like_use.
18201 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18202 Alan Hayward <alan.hayward@arm.com>
18203 David Sherwood <david.sherwood@arm.com>
18205 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
18207 * gimple-expr.h (is_gimple_addressable: Likewise.
18208 * gimple-expr.c (is_gimple_address): Likewise.
18209 * internal-fn.c (expand_call_mem_ref): New function.
18210 (expand_mask_load_optab_fn): Use it.
18211 (expand_mask_store_optab_fn): Likewise.
18213 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18214 Alan Hayward <alan.hayward@arm.com>
18215 David Sherwood <david.sherwood@arm.com>
18217 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
18218 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
18219 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
18220 (cond_umax@var{mode}): Document.
18221 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
18222 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
18223 (cond_umin_optab, cond_umax_optab): New optabs.
18224 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
18225 (COND_IOR, COND_XOR): New internal functions.
18226 * internal-fn.h (get_conditional_internal_fn): Declare.
18227 * internal-fn.c (cond_binary_direct): New macro.
18228 (expand_cond_binary_optab_fn): Likewise.
18229 (direct_cond_binary_optab_supported_p): Likewise.
18230 (get_conditional_internal_fn): New function.
18231 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
18232 Cope with reduction statements that are vectorized as calls rather
18234 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
18235 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
18236 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
18237 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
18238 (UNSPEC_COND_EOR): New unspecs.
18239 (optab): Add mappings for them.
18240 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
18241 (sve_int_op, sve_fp_op): New int attributes.
18243 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18244 Alan Hayward <alan.hayward@arm.com>
18245 David Sherwood <david.sherwood@arm.com>
18247 * optabs.def (while_ult_optab): New optab.
18248 * doc/md.texi (while_ult@var{m}@var{n}): Document.
18249 * internal-fn.def (WHILE_ULT): New internal function.
18250 * internal-fn.h (direct_internal_fn_supported_p): New override
18251 that takes two types as argument.
18252 * internal-fn.c (while_direct): New macro.
18253 (expand_while_optab_fn): New function.
18254 (convert_optab_supported_p): Likewise.
18255 (direct_while_optab_supported_p): New macro.
18256 * wide-int.h (wi::udiv_ceil): New function.
18257 * tree-vectorizer.h (rgroup_masks): New structure.
18258 (vec_loop_masks): New typedef.
18259 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
18260 and fully_masked_p.
18261 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
18262 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
18263 (vect_max_vf): New function.
18264 (slpeel_make_loop_iterate_ntimes): Delete.
18265 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
18266 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
18267 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
18268 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
18269 internal-fn.h, stor-layout.h and optabs-query.h.
18270 (vect_set_loop_mask): New function.
18271 (add_preheader_seq): Likewise.
18272 (add_header_seq): Likewise.
18273 (interleave_supported_p): Likewise.
18274 (vect_maybe_permute_loop_masks): Likewise.
18275 (vect_set_loop_masks_directly): Likewise.
18276 (vect_set_loop_condition_masked): Likewise.
18277 (vect_set_loop_condition_unmasked): New function, split out from
18278 slpeel_make_loop_iterate_ntimes.
18279 (slpeel_make_loop_iterate_ntimes): Rename to..
18280 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
18281 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
18282 (vect_do_peeling): Update call accordingly.
18283 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
18285 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
18286 mask_compare_type, can_fully_mask_p and fully_masked_p.
18287 (release_vec_loop_masks): New function.
18288 (_loop_vec_info): Use it to free the loop masks.
18289 (can_produce_all_loop_masks_p): New function.
18290 (vect_get_max_nscalars_per_iter): Likewise.
18291 (vect_verify_full_masking): Likewise.
18292 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
18293 retries, and free the mask rgroups before retrying. Check loop-wide
18294 reasons for disallowing fully-masked loops. Make the final decision
18295 about whether use a fully-masked loop or not.
18296 (vect_estimate_min_profitable_iters): Do not assume that peeling
18297 for the number of iterations will be needed for fully-masked loops.
18298 (vectorizable_reduction): Disable fully-masked loops.
18299 (vectorizable_live_operation): Likewise.
18300 (vect_halve_mask_nunits): New function.
18301 (vect_double_mask_nunits): Likewise.
18302 (vect_record_loop_mask): Likewise.
18303 (vect_get_loop_mask): Likewise.
18304 (vect_transform_loop): Handle the case in which the final loop
18305 iteration might handle a partial vector. Call vect_set_loop_condition
18306 instead of slpeel_make_loop_iterate_ntimes.
18307 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
18308 (check_load_store_masking): New function.
18309 (prepare_load_store_mask): Likewise.
18310 (vectorizable_store): Handle fully-masked loops.
18311 (vectorizable_load): Likewise.
18312 (supportable_widening_operation): Use vect_halve_mask_nunits for
18314 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
18315 (vect_gen_while): New function.
18316 * config/aarch64/aarch64.md (umax<mode>3): New expander.
18317 (aarch64_uqdec<mode>): New insn.
18319 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18320 Alan Hayward <alan.hayward@arm.com>
18321 David Sherwood <david.sherwood@arm.com>
18323 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
18324 (reduc_xor_scal_optab): New optabs.
18325 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
18326 (reduc_xor_scal_@var{m}): Document.
18327 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
18328 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
18329 internal functions.
18330 * fold-const-call.c (fold_const_call): Handle them.
18331 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
18332 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
18333 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
18334 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
18335 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
18336 (UNSPEC_XORV): New unspecs.
18337 (optab): Add entries for them.
18338 (BITWISEV): New int iterator.
18339 (bit_reduc_op): New int attributes.
18341 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18342 Alan Hayward <alan.hayward@arm.com>
18343 David Sherwood <david.sherwood@arm.com>
18345 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
18346 * internal-fn.def (VEC_SHL_INSERT): New internal function.
18347 * optabs.def (vec_shl_insert_optab): New optab.
18348 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
18349 (duplicate_and_interleave): Likewise.
18350 * tree-vect-loop.c: Include internal-fn.h.
18351 (neutral_op_for_slp_reduction): New function, split out from
18352 get_initial_defs_for_reduction.
18353 (get_initial_def_for_reduction): Handle option 2 for variable-length
18354 vectors by loading the neutral value into a vector and then shifting
18355 the initial value into element 0.
18356 (get_initial_defs_for_reduction): Replace the code argument with
18357 the neutral value calculated by neutral_op_for_slp_reduction.
18358 Use gimple_build_vector for constant-length vectors.
18359 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
18360 but the first group_size elements have a neutral value.
18361 Use duplicate_and_interleave otherwise.
18362 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
18363 Update call to get_initial_defs_for_reduction. Handle SLP
18364 reductions for variable-length vectors by creating one vector
18365 result for each scalar result, with the elements associated
18366 with other scalar results stubbed out with the neutral value.
18367 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
18368 Require IFN_VEC_SHL_INSERT for double reductions on
18369 variable-length vectors, or SLP reductions that have
18370 a neutral value. Require can_duplicate_and_interleave_p
18371 support for variable-length unchained SLP reductions if there
18372 is no neutral value, such as for MIN/MAX reductions. Also require
18373 the number of vector elements to be a multiple of the number of
18374 SLP statements when doing variable-length unchained SLP reductions.
18375 Update call to vect_create_epilog_for_reduction.
18376 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
18377 and remove initial values.
18378 (duplicate_and_interleave): Make public.
18379 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
18380 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
18382 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18383 Alan Hayward <alan.hayward@arm.com>
18384 David Sherwood <david.sherwood@arm.com>
18386 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
18387 (can_duplicate_and_interleave_p): New function.
18388 (vect_get_and_check_slp_defs): Take the vector of statements
18389 rather than just the current one. Remove excess parentheses.
18390 Restriction rejectinon of vect_constant_def and vect_external_def
18391 for variable-length vectors to boolean types, or types for which
18392 can_duplicate_and_interleave_p is false.
18393 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
18394 (duplicate_and_interleave): New function.
18395 (vect_get_constant_vectors): Use gimple_build_vector for
18396 constant-length vectors and suitable variable-length constant
18397 vectors. Use duplicate_and_interleave for other variable-length
18398 vectors. Don't defer the update when inserting new statements.
18400 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18401 Alan Hayward <alan.hayward@arm.com>
18402 David Sherwood <david.sherwood@arm.com>
18404 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
18405 min_profitable_iters doesn't go negative.
18407 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18408 Alan Hayward <alan.hayward@arm.com>
18409 David Sherwood <david.sherwood@arm.com>
18411 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
18412 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
18413 * optabs.def (vec_mask_load_lanes_optab): New optab.
18414 (vec_mask_store_lanes_optab): Likewise.
18415 * internal-fn.def (MASK_LOAD_LANES): New internal function.
18416 (MASK_STORE_LANES): Likewise.
18417 * internal-fn.c (mask_load_lanes_direct): New macro.
18418 (mask_store_lanes_direct): Likewise.
18419 (expand_mask_load_optab_fn): Handle masked operations.
18420 (expand_mask_load_lanes_optab_fn): New macro.
18421 (expand_mask_store_optab_fn): Handle masked operations.
18422 (expand_mask_store_lanes_optab_fn): New macro.
18423 (direct_mask_load_lanes_optab_supported_p): Likewise.
18424 (direct_mask_store_lanes_optab_supported_p): Likewise.
18425 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
18427 (vect_load_lanes_supported): Likewise.
18428 * tree-vect-data-refs.c (strip_conversion): New function.
18429 (can_group_stmts_p): Likewise.
18430 (vect_analyze_data_ref_accesses): Use it instead of checking
18431 for a pair of assignments.
18432 (vect_store_lanes_supported): Take a masked_p parameter.
18433 (vect_load_lanes_supported): Likewise.
18434 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
18435 vect_store_lanes_supported and vect_load_lanes_supported.
18436 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
18437 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
18438 parameter. Don't allow gaps for masked accesses.
18439 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
18440 and vect_load_lanes_supported.
18441 (get_load_store_type): Take a masked_p parameter and update
18442 call to get_group_load_store_type.
18443 (vectorizable_store): Update call to get_load_store_type.
18444 Handle IFN_MASK_STORE_LANES.
18445 (vectorizable_load): Update call to get_load_store_type.
18446 Handle IFN_MASK_LOAD_LANES.
18448 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18449 Alan Hayward <alan.hayward@arm.com>
18450 David Sherwood <david.sherwood@arm.com>
18452 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
18454 * config/aarch64/aarch64-protos.h
18455 (aarch64_sve_struct_memory_operand_p): Declare.
18456 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
18457 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
18458 (VPRED, vpred): Handle SVE structure modes.
18459 * config/aarch64/constraints.md (Utx): New constraint.
18460 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
18461 (aarch64_sve_struct_nonimmediate_operand): New predicates.
18462 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
18463 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
18464 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
18465 structure modes. Split into pieces after RA.
18466 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
18467 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
18469 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
18470 SVE structure modes.
18471 (aarch64_classify_address): Likewise.
18472 (sizetochar): Move earlier in file.
18473 (aarch64_print_operand): Handle SVE register lists.
18474 (aarch64_array_mode): New function.
18475 (aarch64_sve_struct_memory_operand_p): Likewise.
18476 (TARGET_ARRAY_MODE): Redefine.
18478 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18479 Alan Hayward <alan.hayward@arm.com>
18480 David Sherwood <david.sherwood@arm.com>
18482 * target.def (array_mode): New target hook.
18483 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
18484 * doc/tm.texi: Regenerate.
18485 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
18486 * hooks.c (hook_optmode_mode_uhwi_none): New function.
18487 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
18488 targetm.array_mode.
18489 * stor-layout.c (mode_for_array): Likewise. Support polynomial
18492 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18493 Alan Hayward <alan.hayward@arm.com>
18494 David Sherwood <david.sherwood@arm.com>
18496 * fold-const.c (fold_binary_loc): Check the argument types
18497 rather than the result type when testing for a vector operation.
18499 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18501 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
18502 * doc/tm.texi: Regenerate.
18504 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18505 Alan Hayward <alan.hayward@arm.com>
18506 David Sherwood <david.sherwood@arm.com>
18508 * doc/invoke.texi (-msve-vector-bits=): Document new option.
18509 (sve): Document new AArch64 extension.
18510 * doc/md.texi (w): Extend the description of the AArch64
18511 constraint to include SVE vectors.
18512 (Upl, Upa): Document new AArch64 predicate constraints.
18513 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
18515 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
18516 (msve-vector-bits=): New option.
18517 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
18518 SVE when these are disabled.
18519 (sve): New extension.
18520 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
18521 modes. Adjust their number of units based on aarch64_sve_vg.
18522 (MAX_BITSIZE_MODE_ANY_MODE): Define.
18523 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
18524 aarch64_addr_query_type.
18525 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
18526 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
18527 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
18528 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
18529 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
18530 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
18531 (aarch64_simd_imm_zero_p): Delete.
18532 (aarch64_check_zero_based_sve_index_immediate): Declare.
18533 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
18534 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
18535 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
18536 (aarch64_sve_float_mul_immediate_p): Likewise.
18537 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
18538 rather than an rtx.
18539 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
18540 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
18541 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
18542 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
18543 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
18544 (aarch64_regmode_natural_size): Likewise.
18545 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
18546 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
18548 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
18549 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
18550 for VG and the SVE predicate registers.
18551 (V_ALIASES): Add a "z"-prefixed alias.
18552 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
18553 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
18554 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
18555 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
18556 (REG_CLASS_NAMES): Add entries for them.
18557 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
18558 and the predicate registers.
18559 (aarch64_sve_vg): Declare.
18560 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
18561 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
18562 (REGMODE_NATURAL_SIZE): Define.
18563 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
18565 * config/aarch64/aarch64.c: Include cfgrtl.h.
18566 (simd_immediate_info): Add a constructor for series vectors,
18567 and an associated step field.
18568 (aarch64_sve_vg): New variable.
18569 (aarch64_dbx_register_number): Handle VG and the predicate registers.
18570 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
18571 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
18572 (VEC_ANY_DATA, VEC_STRUCT): New constants.
18573 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
18574 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
18575 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
18576 (aarch64_get_mask_mode): New functions.
18577 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
18578 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
18579 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
18580 predicate modes and predicate registers. Explicitly restrict
18581 GPRs to modes of 16 bytes or smaller. Only allow FP registers
18582 to store a vector mode if it is recognized by
18583 aarch64_classify_vector_mode.
18584 (aarch64_regmode_natural_size): New function.
18585 (aarch64_hard_regno_caller_save_mode): Return the original mode
18587 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
18588 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
18589 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
18590 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
18592 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
18593 does not overlap dest if the function is frame-related. Handle
18595 (aarch64_split_add_offset): New function.
18596 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
18597 them aarch64_add_offset.
18598 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
18599 and update call to aarch64_sub_sp.
18600 (aarch64_add_cfa_expression): New function.
18601 (aarch64_expand_prologue): Pass extra temporary registers to the
18602 functions above. Handle the case in which we need to emit new
18603 DW_CFA_expressions for registers that were originally saved
18604 relative to the stack pointer, but now have to be expressed
18605 relative to the frame pointer.
18606 (aarch64_output_mi_thunk): Pass extra temporary registers to the
18608 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
18609 IP0 and IP1 values for SVE frames.
18610 (aarch64_expand_vec_series): New function.
18611 (aarch64_expand_sve_widened_duplicate): Likewise.
18612 (aarch64_expand_sve_const_vector): Likewise.
18613 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
18614 Handle SVE constants. Use emit_move_insn to move a force_const_mem
18615 into the register, rather than emitting a SET directly.
18616 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
18617 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
18618 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
18619 (offset_9bit_signed_scaled_p): New functions.
18620 (aarch64_replicate_bitmask_imm): New function.
18621 (aarch64_bitmask_imm): Use it.
18622 (aarch64_cannot_force_const_mem): Reject expressions involving
18623 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
18624 (aarch64_classify_index): Handle SVE indices, by requiring
18625 a plain register index with a scale that matches the element size.
18626 (aarch64_classify_address): Handle SVE addresses. Assert that
18627 the mode of the address is VOIDmode or an integer mode.
18628 Update call to aarch64_classify_symbol.
18629 (aarch64_classify_symbolic_expression): Update call to
18630 aarch64_classify_symbol.
18631 (aarch64_const_vec_all_in_range_p): New function.
18632 (aarch64_print_vector_float_operand): Likewise.
18633 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
18634 "vN" for FP registers with SVE modes. Handle (const ...) vectors
18635 and the FP immediates 1.0 and 0.5.
18636 (aarch64_print_address_internal): Handle SVE addresses.
18637 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
18638 (aarch64_regno_regclass): Handle predicate registers.
18639 (aarch64_secondary_reload): Handle big-endian reloads of SVE
18641 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
18642 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
18643 (aarch64_convert_sve_vector_bits): New function.
18644 (aarch64_override_options): Use it to handle -msve-vector-bits=.
18645 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
18646 rather than an rtx.
18647 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
18648 Handle SVE vector and predicate modes. Accept VL-based constants
18649 that need only one temporary register, and VL offsets that require
18650 no temporary registers.
18651 (aarch64_conditional_register_usage): Mark the predicate registers
18652 as fixed if SVE isn't available.
18653 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
18654 Return true for SVE vector and predicate modes.
18655 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
18656 rather than an unsigned int. Handle SVE modes.
18657 (aarch64_preferred_simd_mode): Update call accordingly. Handle
18659 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
18661 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
18662 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
18663 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
18664 (aarch64_sve_float_mul_immediate_p): New functions.
18665 (aarch64_sve_valid_immediate): New function.
18666 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
18667 Explicitly reject structure modes. Check for INDEX constants.
18668 Handle PTRUE and PFALSE constants.
18669 (aarch64_check_zero_based_sve_index_immediate): New function.
18670 (aarch64_simd_imm_zero_p): Delete.
18671 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
18672 vector modes. Accept constants in the range of CNT[BHWD].
18673 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
18674 ask for an Advanced SIMD mode.
18675 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
18676 (aarch64_simd_vector_alignment): Handle SVE predicates.
18677 (aarch64_vectorize_preferred_vector_alignment): New function.
18678 (aarch64_simd_vector_alignment_reachable): Use it instead of
18680 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
18681 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
18683 (MAX_VECT_LEN): Delete.
18684 (expand_vec_perm_d): Add a vec_flags field.
18685 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
18686 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
18687 (aarch64_evpc_ext): Don't apply a big-endian lane correction
18689 (aarch64_evpc_rev): Rename to...
18690 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
18691 (aarch64_evpc_rev_global): New function.
18692 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
18693 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
18695 (aarch64_evpc_sve_tbl): New function.
18696 (aarch64_expand_vec_perm_const_1): Update after rename of
18697 aarch64_evpc_rev. Handle SVE permutes too, trying
18698 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
18699 than aarch64_evpc_tbl.
18700 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
18701 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
18702 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
18703 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
18704 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
18705 (aarch64_expand_sve_vcond): New functions.
18706 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
18707 of aarch64_vector_mode_p.
18708 (aarch64_dwarf_poly_indeterminate_value): New function.
18709 (aarch64_compute_pressure_classes): Likewise.
18710 (aarch64_can_change_mode_class): Likewise.
18711 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
18712 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
18713 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
18714 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
18715 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
18716 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
18717 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
18718 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
18720 (Dn, Dl, Dr): Accept const as well as const_vector.
18721 (Dz): Likewise. Compare against CONST0_RTX.
18722 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
18723 of "vector" where appropriate.
18724 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
18725 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
18726 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
18727 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
18728 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
18729 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
18730 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
18731 (v_int_equiv): Extend to SVE modes.
18732 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
18734 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
18735 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
18736 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
18737 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
18738 (SVE_COND_FP_CMP): New int iterators.
18739 (perm_hilo): Handle the new unpack unspecs.
18740 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
18742 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
18743 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
18744 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
18745 (aarch64_equality_operator, aarch64_constant_vector_operand)
18746 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
18747 (aarch64_sve_nonimmediate_operand): Likewise.
18748 (aarch64_sve_general_operand): Likewise.
18749 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
18750 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
18751 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
18752 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
18753 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
18754 (aarch64_sve_float_arith_immediate): Likewise.
18755 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
18756 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
18757 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
18758 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
18759 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
18760 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
18761 (aarch64_sve_float_arith_operand): Likewise.
18762 (aarch64_sve_float_arith_with_sub_operand): Likewise.
18763 (aarch64_sve_float_mul_operand): Likewise.
18764 (aarch64_sve_vec_perm_operand): Likewise.
18765 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
18766 (aarch64_mov_operand): Accept const_poly_int and const_vector.
18767 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
18768 as well as const_vector.
18769 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
18770 in file. Use CONST0_RTX and CONSTM1_RTX.
18771 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
18772 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
18773 Use aarch64_simd_imm_zero.
18774 * config/aarch64/aarch64-sve.md: New file.
18775 * config/aarch64/aarch64.md: Include it.
18776 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
18777 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
18778 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
18779 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
18780 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
18781 (sve): New attribute.
18782 (enabled): Disable instructions with the sve attribute unless
18784 (movqi, movhi): Pass CONST_POLY_INT operaneds through
18785 aarch64_expand_mov_immediate.
18786 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
18787 CNT[BHSD] immediates.
18788 (movti): Split CONST_POLY_INT moves into two halves.
18789 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
18790 Split additions that need a temporary here if the destination
18791 is the stack pointer.
18792 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
18793 (*add<mode>3_poly_1): New instruction.
18794 (set_clobber_cc): New expander.
18796 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18798 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
18799 parameter and use it instead of GET_MODE_SIZE (innermode). Use
18800 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
18801 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
18802 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
18803 Change innermode from fixed_mode_size to machine_mode.
18804 (simplify_subreg): Update call accordingly. Handle a constant-sized
18805 subreg of a variable-length CONST_VECTOR.
18807 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18808 Alan Hayward <alan.hayward@arm.com>
18809 David Sherwood <david.sherwood@arm.com>
18811 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
18812 (add_offset_to_base): New function, split out from...
18813 (create_mem_ref): ...here. When handling a scale other than 1,
18814 check first whether the address is valid without the offset.
18815 Add it into the base if so, leaving the index and scale as-is.
18817 2018-01-12 Jakub Jelinek <jakub@redhat.com>
18820 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
18821 fold_for_warn before checking if arg2 is INTEGER_CST.
18823 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
18825 * config/rs6000/predicates.md (load_multiple_operation): Delete.
18826 (store_multiple_operation): Delete.
18827 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
18828 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
18829 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
18830 guarded by TARGET_STRING.
18831 (rs6000_output_load_multiple): Delete.
18832 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
18833 OPTION_MASK_STRING / TARGET_STRING handling.
18834 (print_operand) <'N', 'O'>: Add comment that these are unused now.
18835 (const rs6000_opt_masks) <"string">: Change mask to 0.
18836 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
18837 (MASK_STRING): Delete.
18838 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
18840 (load_multiple): Delete.
18847 (store_multiple): Delete.
18854 (movmemsi_8reg): Delete.
18855 (corresponding unnamed define_insn): Delete.
18856 (movmemsi_6reg): Delete.
18857 (corresponding unnamed define_insn): Delete.
18858 (movmemsi_4reg): Delete.
18859 (corresponding unnamed define_insn): Delete.
18860 (movmemsi_2reg): Delete.
18861 (corresponding unnamed define_insn): Delete.
18862 (movmemsi_1reg): Delete.
18863 (corresponding unnamed define_insn): Delete.
18864 * config/rs6000/rs6000.opt (mno-string): New.
18865 (mstring): Replace by deprecation warning stub.
18866 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
18868 2018-01-12 Jakub Jelinek <jakub@redhat.com>
18870 * regrename.c (regrename_do_replace): If replacing the same
18871 reg multiple times, try to reuse last created gen_raw_REG.
18874 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
18875 main to workaround a bug in GDB.
18877 2018-01-12 Tom de Vries <tom@codesourcery.com>
18880 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
18882 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
18884 PR rtl-optimization/80481
18885 * ira-color.c (get_cap_member): New function.
18886 (allocnos_conflict_by_live_ranges_p): Use it.
18887 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
18888 (setup_slot_coalesced_allocno_live_ranges): Ditto.
18890 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
18893 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
18894 (*saddl_se_1): Ditto.
18895 (*ssubsi_1): Ditto.
18896 (*ssubl_se_1): Ditto.
18898 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
18900 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
18901 rather than wi::to_widest for DR_INITs.
18902 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
18903 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
18904 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
18906 (vect_analyze_group_access_1): Note that here.
18908 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
18910 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
18911 polynomial type sizes.
18913 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
18915 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
18916 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
18917 (gimple_add_tmp_var): Likewise.
18919 2018-01-12 Martin Liska <mliska@suse.cz>
18921 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
18922 (gimple_alloc_sizes): Likewise.
18923 (dump_gimple_statistics): Use PRIu64 in printf format.
18924 * gimple.h: Change uint64_t to int.
18926 2018-01-12 Martin Liska <mliska@suse.cz>
18928 * tree-core.h: Use uint64_t instead of int.
18929 * tree.c (tree_node_counts): Likewise.
18930 (tree_node_sizes): Likewise.
18931 (dump_tree_statistics): Use PRIu64 in printf format.
18933 2018-01-12 Martin Liska <mliska@suse.cz>
18935 * Makefile.in: As qsort_chk is implemented in vec.c, add
18936 vec.o to linkage of gencfn-macros.
18937 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
18938 passing the info to record_node_allocation_statistics.
18939 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
18941 * ggc-common.c (struct ggc_usage): Add operator== and use
18942 it in operator< and compare function.
18943 * mem-stats.h (struct mem_usage): Likewise.
18944 * vec.c (struct vec_usage): Remove operator< and compare
18945 function. Can be simply inherited.
18947 2018-01-12 Martin Jambor <mjambor@suse.cz>
18950 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
18951 * tree-ssa-math-opts.c: Include domwalk.h.
18952 (convert_mult_to_fma_1): New function.
18953 (fma_transformation_info): New type.
18954 (fma_deferring_state): Likewise.
18955 (cancel_fma_deferring): New function.
18956 (result_of_phi): Likewise.
18957 (last_fma_candidate_feeds_initial_phi): Likewise.
18958 (convert_mult_to_fma): Added deferring logic, split actual
18959 transformation to convert_mult_to_fma_1.
18960 (math_opts_dom_walker): New type.
18961 (math_opts_dom_walker::after_dom_children): New method, body moved
18962 here from pass_optimize_widening_mul::execute, added deferring logic
18964 (pass_optimize_widening_mul::execute): Moved most of code to
18965 math_opts_dom_walker::after_dom_children.
18966 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
18967 * config/i386/i386.c (ix86_option_override_internal): Added
18968 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
18970 2018-01-12 Richard Biener <rguenther@suse.de>
18973 * dwarf2out.c (gen_variable_die): Do not reset old_die for
18974 inline instance vars.
18976 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
18979 * config/rx/rx.c (rx_is_restricted_memory_address):
18980 Handle SUBREG case.
18982 2018-01-12 Richard Biener <rguenther@suse.de>
18984 PR tree-optimization/80846
18985 * target.def (split_reduction): New target hook.
18986 * targhooks.c (default_split_reduction): New function.
18987 * targhooks.h (default_split_reduction): Declare.
18988 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
18989 target requests first reduce vectors by combining low and high
18991 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
18992 (get_vectype_for_scalar_type_and_size): Export.
18993 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
18994 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
18995 * doc/tm.texi: Regenerate.
18996 * config/i386/i386.c (ix86_split_reduction): Implement
18997 TARGET_VECTORIZE_SPLIT_REDUCTION.
18999 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
19002 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
19003 in PIC mode except for TARGET_VXWORKS_RTP.
19004 * config/sparc/sparc.c: Include cfgrtl.h.
19005 (TARGET_INIT_PIC_REG): Define.
19006 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
19007 (sparc_pic_register_p): New predicate.
19008 (sparc_legitimate_address_p): Use it.
19009 (sparc_legitimize_pic_address): Likewise.
19010 (sparc_delegitimize_address): Likewise.
19011 (sparc_mode_dependent_address_p): Likewise.
19012 (gen_load_pcrel_sym): Remove 4th parameter.
19013 (load_got_register): Adjust call to above. Remove obsolete stuff.
19014 (sparc_expand_prologue): Do not call load_got_register here.
19015 (sparc_flat_expand_prologue): Likewise.
19016 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
19017 (sparc_use_pseudo_pic_reg): New function.
19018 (sparc_init_pic_reg): Likewise.
19019 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
19020 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
19022 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
19024 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
19025 Add item for branch_cost.
19027 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
19029 PR rtl-optimization/83565
19030 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
19031 not extend the result to a larger mode for rotate operations.
19032 (num_sign_bit_copies1): Likewise.
19034 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
19037 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
19039 Use values-Xc.o for -pedantic.
19040 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
19042 2018-01-12 Martin Liska <mliska@suse.cz>
19045 * ipa-devirt.c (final_warning_record::grow_type_warnings):
19047 (possible_polymorphic_call_targets): Use it.
19048 (ipa_devirt): Likewise.
19050 2018-01-12 Martin Liska <mliska@suse.cz>
19052 * profile-count.h (enum profile_quality): Use 0 as invalid
19053 enum value of profile_quality.
19055 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
19057 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
19058 -mext-string options.
19060 2018-01-12 Richard Biener <rguenther@suse.de>
19062 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
19063 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
19064 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
19066 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
19068 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
19070 * configure.ac (--with-long-double-format): Add support for the
19071 configuration option to change the default long double format on
19073 * config.gcc (powerpc*-linux*-*): Likewise.
19074 * configure: Regenerate.
19075 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
19076 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
19077 used without modification.
19079 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19081 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
19082 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
19083 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
19084 MISC_BUILTIN_SPEC_BARRIER.
19085 (rs6000_init_builtins): Likewise.
19086 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
19088 (speculation_barrier): New define_insn.
19089 * doc/extend.texi: Document __builtin_speculation_barrier.
19091 2018-01-11 Jakub Jelinek <jakub@redhat.com>
19094 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
19095 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
19096 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
19098 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
19099 integral modes instead of "ss" and "sd".
19100 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
19101 vectors with 32-bit and 64-bit elements.
19102 (vecdupssescalarmodesuffix): New mode attribute.
19103 (vec_dup<mode>): Use it.
19105 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
19108 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
19109 frame if argument is passed on stack.
19111 2018-01-11 Jakub Jelinek <jakub@redhat.com>
19114 * ree.c (combine_reaching_defs): Optimize also
19115 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
19116 reg2=any_extend(exp); reg1=reg2;, formatting fix.
19118 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
19120 PR middle-end/83189
19121 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
19123 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
19125 PR middle-end/83718
19126 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
19127 after they are computed.
19129 2018-01-11 Bin Cheng <bin.cheng@arm.com>
19131 PR tree-optimization/83695
19132 * gimple-loop-linterchange.cc
19133 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
19134 reset cached scev information after interchange.
19135 (pass_linterchange::execute): Remove call to scev_reset_htab.
19137 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19139 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
19140 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
19141 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
19142 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
19143 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
19144 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
19145 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
19146 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
19147 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
19148 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
19149 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
19150 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
19151 (V_lane_reg): Likewise.
19152 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
19154 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
19155 (vfmal_lane_low<mode>_intrinsic,
19156 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
19157 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
19158 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
19159 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
19160 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
19161 vfmsl_lane_high<mode>_intrinsic): New define_insns.
19163 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19165 * config/arm/arm-cpus.in (fp16fml): New feature.
19166 (ALL_SIMD): Add fp16fml.
19167 (armv8.2-a): Add fp16fml as an option.
19168 (armv8.3-a): Likewise.
19169 (armv8.4-a): Add fp16fml as part of fp16.
19170 * config/arm/arm.h (TARGET_FP16FML): Define.
19171 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
19173 * config/arm/arm-modes.def (V2HF): Define.
19174 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
19175 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
19176 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
19177 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
19178 vfmsl_low, vfmsl_high): New set of builtins.
19179 * config/arm/iterators.md (PLUSMINUS): New code iterator.
19180 (vfml_op): New code attribute.
19181 (VFMLHALVES): New int iterator.
19182 (VFML, VFMLSEL): New mode attributes.
19183 (V_reg): Define mapping for V2HF.
19184 (V_hi, V_lo): New mode attributes.
19185 (VF_constraint): Likewise.
19186 (vfml_half, vfml_half_selector): New int attributes.
19187 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
19189 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
19190 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
19192 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
19193 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
19194 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
19195 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
19197 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
19198 Document new effective target and option set.
19200 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19202 * config/arm/arm-cpus.in (armv8_4): New feature.
19203 (ARMv8_4a): New fgroup.
19204 (armv8.4-a): New arch.
19205 * config/arm/arm-tables.opt: Regenerate.
19206 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
19207 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
19208 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
19209 Add matching rules for -march=armv8.4-a and extensions.
19210 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
19212 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
19215 * config/rx/rx.md (BW): New mode attribute.
19216 (sync_lock_test_and_setsi): Add mode suffix to insn output.
19218 2018-01-11 Richard Biener <rguenther@suse.de>
19220 PR tree-optimization/83435
19221 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
19222 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
19223 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
19225 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
19226 Alan Hayward <alan.hayward@arm.com>
19227 David Sherwood <david.sherwood@arm.com>
19229 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
19231 (aarch64_classify_address): Initialize it. Track polynomial offsets.
19232 (aarch64_print_address_internal): Use it to check for a zero offset.
19234 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
19235 Alan Hayward <alan.hayward@arm.com>
19236 David Sherwood <david.sherwood@arm.com>
19238 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
19239 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
19240 Return a poly_int64 rather than a HOST_WIDE_INT.
19241 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
19242 rather than a HOST_WIDE_INT.
19243 * config/aarch64/aarch64.h (aarch64_frame): Protect with
19244 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
19245 hard_fp_offset, frame_size, initial_adjust, callee_offset and
19246 final_offset from HOST_WIDE_INT to poly_int64.
19247 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
19248 to_constant when getting the number of units in an Advanced SIMD
19250 (aarch64_builtin_vectorized_function): Check for a constant number
19252 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
19254 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
19255 attribute instead of GET_MODE_NUNITS.
19256 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
19257 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
19258 GET_MODE_SIZE for fixed-size registers.
19259 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
19260 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
19261 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
19262 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
19263 (aarch64_print_operand, aarch64_print_address_internal)
19264 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
19265 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
19266 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
19267 Handle polynomial GET_MODE_SIZE.
19268 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
19269 wider than SImode without modification.
19270 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
19271 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
19272 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
19273 passing and returning SVE modes.
19274 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
19275 rather than GEN_INT.
19276 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
19277 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
19278 (aarch64_allocate_and_probe_stack_space): Likewise.
19279 (aarch64_layout_frame): Cope with polynomial offsets.
19280 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
19281 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
19282 polynomial offsets.
19283 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
19284 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
19285 poly_int64 rather than a HOST_WIDE_INT.
19286 (aarch64_get_separate_components, aarch64_process_components)
19287 (aarch64_expand_prologue, aarch64_expand_epilogue)
19288 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
19289 (aarch64_anchor_offset): New function, split out from...
19290 (aarch64_legitimize_address): ...here.
19291 (aarch64_builtin_vectorization_cost): Handle polynomial
19292 TYPE_VECTOR_SUBPARTS.
19293 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
19295 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
19296 number of elements from the PARALLEL rather than the mode.
19297 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
19298 rather than GET_MODE_BITSIZE.
19299 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
19300 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
19301 (aarch64_expand_vec_perm_const_1): Handle polynomial
19302 d->perm.length () and d->perm elements.
19303 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
19304 Apply to_constant to d->perm elements.
19305 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
19306 polynomial CONST_VECTOR_NUNITS.
19307 (aarch64_move_pointer): Take amount as a poly_int64 rather
19309 (aarch64_progress_pointer): Avoid temporary variable.
19310 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
19311 the mode attribute instead of GET_MODE.
19313 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
19314 Alan Hayward <alan.hayward@arm.com>
19315 David Sherwood <david.sherwood@arm.com>
19317 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
19318 x exists before using it.
19319 (aarch64_add_constant_internal): Rename to...
19320 (aarch64_add_offset_1): ...this. Replace regnum with separate
19321 src and dest rtxes. Handle the case in which they're different,
19322 including when the offset is zero. Replace scratchreg with an rtx.
19323 Use 2 additions if there is no spare register into which we can
19324 move a 16-bit constant.
19325 (aarch64_add_constant): Delete.
19326 (aarch64_add_offset): Replace reg with separate src and dest
19327 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
19328 Use aarch64_add_offset_1.
19329 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
19330 an rtx rather than an int. Take the delta as a poly_int64
19331 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
19332 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
19333 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
19334 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
19335 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
19336 and aarch64_add_sp.
19337 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
19338 aarch64_add_constant.
19340 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
19342 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
19343 Use scalar_float_mode.
19345 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
19347 * config/aarch64/aarch64-simd.md
19348 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
19349 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
19350 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
19351 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
19352 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
19353 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
19354 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
19355 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
19356 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
19357 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
19359 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
19362 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
19363 targ_options->x_arm_arch_string is non NULL.
19365 2018-01-11 Tamar Christina <tamar.christina@arm.com>
19367 * config/aarch64/aarch64.h
19368 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
19370 2018-01-11 Sudakshina Das <sudi.das@arm.com>
19373 * expmed.c (emit_store_flag_force): Swap if const op0
19374 and change VOIDmode to mode of op0.
19376 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
19378 PR rtl-optimization/83761
19379 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
19380 than bytes to mode_for_size.
19382 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
19384 PR middle-end/83189
19385 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
19386 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
19389 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
19391 PR middle-end/83575
19392 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
19393 when in layout mode.
19394 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
19395 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
19398 2018-01-10 Michael Collison <michael.collison@arm.com>
19400 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
19401 * config/aarch64/aarch64-option-extension.def: Add
19402 AARCH64_OPT_EXTENSION of 'fp16fml'.
19403 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
19404 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
19405 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
19406 * config/aarch64/constraints.md (Ui7): New constraint.
19407 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
19408 (VFMLA_SEL_W): Ditto.
19411 (VFMLA16_LOW): New int iterator.
19412 (VFMLA16_HIGH): Ditto.
19413 (UNSPEC_FMLAL): New unspec.
19414 (UNSPEC_FMLSL): Ditto.
19415 (UNSPEC_FMLAL2): Ditto.
19416 (UNSPEC_FMLSL2): Ditto.
19417 (f16mac): New code attribute.
19418 * config/aarch64/aarch64-simd-builtins.def
19419 (aarch64_fmlal_lowv2sf): Ditto.
19420 (aarch64_fmlsl_lowv2sf): Ditto.
19421 (aarch64_fmlalq_lowv4sf): Ditto.
19422 (aarch64_fmlslq_lowv4sf): Ditto.
19423 (aarch64_fmlal_highv2sf): Ditto.
19424 (aarch64_fmlsl_highv2sf): Ditto.
19425 (aarch64_fmlalq_highv4sf): Ditto.
19426 (aarch64_fmlslq_highv4sf): Ditto.
19427 (aarch64_fmlal_lane_lowv2sf): Ditto.
19428 (aarch64_fmlsl_lane_lowv2sf): Ditto.
19429 (aarch64_fmlal_laneq_lowv2sf): Ditto.
19430 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
19431 (aarch64_fmlalq_lane_lowv4sf): Ditto.
19432 (aarch64_fmlsl_lane_lowv4sf): Ditto.
19433 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
19434 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
19435 (aarch64_fmlal_lane_highv2sf): Ditto.
19436 (aarch64_fmlsl_lane_highv2sf): Ditto.
19437 (aarch64_fmlal_laneq_highv2sf): Ditto.
19438 (aarch64_fmlsl_laneq_highv2sf): Ditto.
19439 (aarch64_fmlalq_lane_highv4sf): Ditto.
19440 (aarch64_fmlsl_lane_highv4sf): Ditto.
19441 (aarch64_fmlalq_laneq_highv4sf): Ditto.
19442 (aarch64_fmlsl_laneq_highv4sf): Ditto.
19443 * config/aarch64/aarch64-simd.md:
19444 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
19445 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
19446 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
19447 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
19448 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
19449 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
19450 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
19451 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
19452 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
19453 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
19454 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
19455 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
19456 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
19457 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
19458 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
19459 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
19460 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
19461 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
19462 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
19463 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
19464 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
19465 (vfmlsl_low_u32): Ditto.
19466 (vfmlalq_low_u32): Ditto.
19467 (vfmlslq_low_u32): Ditto.
19468 (vfmlal_high_u32): Ditto.
19469 (vfmlsl_high_u32): Ditto.
19470 (vfmlalq_high_u32): Ditto.
19471 (vfmlslq_high_u32): Ditto.
19472 (vfmlal_lane_low_u32): Ditto.
19473 (vfmlsl_lane_low_u32): Ditto.
19474 (vfmlal_laneq_low_u32): Ditto.
19475 (vfmlsl_laneq_low_u32): Ditto.
19476 (vfmlalq_lane_low_u32): Ditto.
19477 (vfmlslq_lane_low_u32): Ditto.
19478 (vfmlalq_laneq_low_u32): Ditto.
19479 (vfmlslq_laneq_low_u32): Ditto.
19480 (vfmlal_lane_high_u32): Ditto.
19481 (vfmlsl_lane_high_u32): Ditto.
19482 (vfmlal_laneq_high_u32): Ditto.
19483 (vfmlsl_laneq_high_u32): Ditto.
19484 (vfmlalq_lane_high_u32): Ditto.
19485 (vfmlslq_lane_high_u32): Ditto.
19486 (vfmlalq_laneq_high_u32): Ditto.
19487 (vfmlslq_laneq_high_u32): Ditto.
19488 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
19489 (AARCH64_FL_FOR_ARCH8_4): New.
19490 (AARCH64_ISA_F16FML): New ISA flag.
19491 (TARGET_F16FML): New feature flag for fp16fml.
19492 (doc/invoke.texi): Document new fp16fml option.
19494 2018-01-10 Michael Collison <michael.collison@arm.com>
19496 * config/aarch64/aarch64-builtins.c:
19497 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
19498 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
19499 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
19500 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
19501 (AARCH64_ISA_SHA3): New ISA flag.
19502 (TARGET_SHA3): New feature flag for sha3.
19503 * config/aarch64/iterators.md (sha512_op): New int attribute.
19504 (CRYPTO_SHA512): New int iterator.
19505 (UNSPEC_SHA512H): New unspec.
19506 (UNSPEC_SHA512H2): Ditto.
19507 (UNSPEC_SHA512SU0): Ditto.
19508 (UNSPEC_SHA512SU1): Ditto.
19509 * config/aarch64/aarch64-simd-builtins.def
19510 (aarch64_crypto_sha512hqv2di): New builtin.
19511 (aarch64_crypto_sha512h2qv2di): Ditto.
19512 (aarch64_crypto_sha512su0qv2di): Ditto.
19513 (aarch64_crypto_sha512su1qv2di): Ditto.
19514 (aarch64_eor3qv8hi): Ditto.
19515 (aarch64_rax1qv2di): Ditto.
19516 (aarch64_xarqv2di): Ditto.
19517 (aarch64_bcaxqv8hi): Ditto.
19518 * config/aarch64/aarch64-simd.md:
19519 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
19520 (aarch64_crypto_sha512su0qv2di): Ditto.
19521 (aarch64_crypto_sha512su1qv2di): Ditto.
19522 (aarch64_eor3qv8hi): Ditto.
19523 (aarch64_rax1qv2di): Ditto.
19524 (aarch64_xarqv2di): Ditto.
19525 (aarch64_bcaxqv8hi): Ditto.
19526 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
19527 (vsha512h2q_u64): Ditto.
19528 (vsha512su0q_u64): Ditto.
19529 (vsha512su1q_u64): Ditto.
19530 (veor3q_u16): Ditto.
19531 (vrax1q_u64): Ditto.
19532 (vxarq_u64): Ditto.
19533 (vbcaxq_u16): Ditto.
19534 * config/arm/types.md (crypto_sha512): New type attribute.
19535 (crypto_sha3): Ditto.
19536 (doc/invoke.texi): Document new sha3 option.
19538 2018-01-10 Michael Collison <michael.collison@arm.com>
19540 * config/aarch64/aarch64-builtins.c:
19541 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
19542 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
19543 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
19544 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
19545 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
19546 (AARCH64_ISA_SM4): New ISA flag.
19547 (TARGET_SM4): New feature flag for sm4.
19548 * config/aarch64/aarch64-simd-builtins.def
19549 (aarch64_sm3ss1qv4si): Ditto.
19550 (aarch64_sm3tt1aq4si): Ditto.
19551 (aarch64_sm3tt1bq4si): Ditto.
19552 (aarch64_sm3tt2aq4si): Ditto.
19553 (aarch64_sm3tt2bq4si): Ditto.
19554 (aarch64_sm3partw1qv4si): Ditto.
19555 (aarch64_sm3partw2qv4si): Ditto.
19556 (aarch64_sm4eqv4si): Ditto.
19557 (aarch64_sm4ekeyqv4si): Ditto.
19558 * config/aarch64/aarch64-simd.md:
19559 (aarch64_sm3ss1qv4si): Ditto.
19560 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
19561 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
19562 (aarch64_sm4eqv4si): Ditto.
19563 (aarch64_sm4ekeyqv4si): Ditto.
19564 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
19565 (sm3part_op): Ditto.
19566 (CRYPTO_SM3TT): Ditto.
19567 (CRYPTO_SM3PART): Ditto.
19568 (UNSPEC_SM3SS1): New unspec.
19569 (UNSPEC_SM3TT1A): Ditto.
19570 (UNSPEC_SM3TT1B): Ditto.
19571 (UNSPEC_SM3TT2A): Ditto.
19572 (UNSPEC_SM3TT2B): Ditto.
19573 (UNSPEC_SM3PARTW1): Ditto.
19574 (UNSPEC_SM3PARTW2): Ditto.
19575 (UNSPEC_SM4E): Ditto.
19576 (UNSPEC_SM4EKEY): Ditto.
19577 * config/aarch64/constraints.md (Ui2): New constraint.
19578 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
19579 * config/arm/types.md (crypto_sm3): New type attribute.
19580 (crypto_sm4): Ditto.
19581 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
19582 (vsm3tt1aq_u32): Ditto.
19583 (vsm3tt1bq_u32): Ditto.
19584 (vsm3tt2aq_u32): Ditto.
19585 (vsm3tt2bq_u32): Ditto.
19586 (vsm3partw1q_u32): Ditto.
19587 (vsm3partw2q_u32): Ditto.
19588 (vsm4eq_u32): Ditto.
19589 (vsm4ekeyq_u32): Ditto.
19590 (doc/invoke.texi): Document new sm4 option.
19592 2018-01-10 Michael Collison <michael.collison@arm.com>
19594 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
19595 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
19596 (AARCH64_FL_FOR_ARCH8_4): New.
19597 (AARCH64_FL_V8_4): New flag.
19598 (doc/invoke.texi): Document new armv8.4-a option.
19600 2018-01-10 Michael Collison <michael.collison@arm.com>
19602 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
19603 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
19604 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
19605 * config/aarch64/aarch64-option-extension.def: Add
19606 AARCH64_OPT_EXTENSION of 'sha2'.
19607 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
19608 (crypto): Disable sha2 and aes if crypto disabled.
19609 (crypto): Enable aes and sha2 if enabled.
19610 (simd): Disable sha2 and aes if simd disabled.
19611 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
19613 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
19614 (TARGET_SHA2): New feature flag for sha2.
19615 (TARGET_AES): New feature flag for aes.
19616 * config/aarch64/aarch64-simd.md:
19617 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
19618 conditional on TARGET_AES.
19619 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
19620 (aarch64_crypto_sha1hsi): Make pattern conditional
19622 (aarch64_crypto_sha1hv4si): Ditto.
19623 (aarch64_be_crypto_sha1hv4si): Ditto.
19624 (aarch64_crypto_sha1su1v4si): Ditto.
19625 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
19626 (aarch64_crypto_sha1su0v4si): Ditto.
19627 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
19628 (aarch64_crypto_sha256su0v4si): Ditto.
19629 (aarch64_crypto_sha256su1v4si): Ditto.
19630 (doc/invoke.texi): Document new aes and sha2 options.
19632 2018-01-10 Martin Sebor <msebor@redhat.com>
19634 PR tree-optimization/83781
19635 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
19638 2018-01-11 Martin Sebor <msebor@gmail.com>
19639 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
19641 PR tree-optimization/83501
19642 PR tree-optimization/81703
19644 * tree-ssa-strlen.c (get_string_cst): Rename...
19645 (get_string_len): ...to this. Handle global constants.
19646 (handle_char_store): Adjust.
19648 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
19649 Jim Wilson <jimw@sifive.com>
19651 * config/riscv/riscv-protos.h (riscv_output_return): New.
19652 * config/riscv/riscv.c (struct machine_function): New naked_p field.
19653 (riscv_attribute_table, riscv_output_return),
19654 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
19655 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
19656 (riscv_compute_frame_info): Only compute frame->mask if not a naked
19658 (riscv_expand_prologue): Add early return for naked function.
19659 (riscv_expand_epilogue): Likewise.
19660 (riscv_function_ok_for_sibcall): Return false for naked function.
19661 (riscv_set_current_function): New.
19662 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
19663 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
19664 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
19665 * doc/extend.texi (RISC-V Function Attributes): New.
19667 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
19669 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
19670 check for 128-bit long double before checking TCmode.
19671 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
19672 128-bit long doubles before checking TFmode or TCmode.
19673 (FLOAT128_IBM_P): Likewise.
19675 2018-01-10 Martin Sebor <msebor@redhat.com>
19677 PR tree-optimization/83671
19678 * builtins.c (c_strlen): Unconditionally return zero for the empty
19680 Use -Warray-bounds for warnings.
19681 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
19682 for non-constant array indices with COMPONENT_REF, arrays of
19683 arrays, and pointers to arrays.
19684 (gimple_fold_builtin_strlen): Determine and set length range for
19685 non-constant character arrays.
19687 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
19689 PR middle-end/81897
19690 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
19693 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
19695 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
19697 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
19700 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
19701 VECTOR_MEM_ALTIVEC_OR_VSX_P.
19702 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
19703 indexed_or_indirect_operand predicate.
19704 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
19705 (*vsx_le_perm_load_v8hi): Likewise.
19706 (*vsx_le_perm_load_v16qi): Likewise.
19707 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
19708 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
19709 (*vsx_le_perm_store_v8hi): Likewise.
19710 (*vsx_le_perm_store_v16qi): Likewise.
19711 (eight unnamed splitters): Likewise.
19713 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
19715 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
19716 * config/rs6000/emmintrin.h: Likewise.
19717 * config/rs6000/mmintrin.h: Likewise.
19718 * config/rs6000/xmmintrin.h: Likewise.
19720 2018-01-10 David Malcolm <dmalcolm@redhat.com>
19723 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
19725 * tree.c (tree_nop_conversion): Return true for location wrapper
19727 (maybe_wrap_with_location): New function.
19728 (selftest::check_strip_nops): New function.
19729 (selftest::test_location_wrappers): New function.
19730 (selftest::tree_c_tests): Call it.
19731 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
19732 (maybe_wrap_with_location): New decl.
19733 (EXPR_LOCATION_WRAPPER_P): New macro.
19734 (location_wrapper_p): New inline function.
19735 (tree_strip_any_location_wrapper): New inline function.
19737 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
19740 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
19741 stack_realign_offset for the largest alignment of stack slot
19743 (ix86_find_max_used_stack_alignment): New function.
19744 (ix86_finalize_stack_frame_flags): Use it. Set
19745 max_used_stack_alignment if we don't realign stack.
19746 * config/i386/i386.h (machine_function): Add
19747 max_used_stack_alignment.
19749 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
19751 * config/arm/arm.opt (-mbranch-cost): New option.
19752 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
19755 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
19758 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
19759 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
19761 2018-01-10 Richard Biener <rguenther@suse.de>
19764 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
19765 early out so it also covers the case where we have a non-NULL
19768 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
19770 PR tree-optimization/83753
19771 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
19772 for non-strided grouped accesses if the number of elements is 1.
19774 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
19777 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
19778 * i386.h (TARGET_USE_GATHER): Define.
19779 * x86-tune.def (X86_TUNE_USE_GATHER): New.
19781 2018-01-10 Martin Liska <mliska@suse.cz>
19784 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
19785 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
19787 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
19788 CLEANUP_NO_PARTITIONING is not set.
19790 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
19792 * doc/rtl.texi: Remove documentation of (const ...) wrappers
19793 for vectors, as a partial revert of r254296.
19794 * rtl.h (const_vec_p): Delete.
19795 (const_vec_duplicate_p): Don't test for vector CONSTs.
19796 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
19797 * expmed.c (make_tree): Likewise.
19800 * common.md (E, F): Use CONSTANT_P instead of checking for
19802 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
19803 checking for CONST_VECTOR.
19805 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
19807 PR middle-end/83575
19808 * predict.c (force_edge_cold): Handle in more sane way edges
19809 with no prediction.
19811 2018-01-09 Carl Love <cel@us.ibm.com>
19813 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
19815 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
19816 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
19817 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
19818 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
19819 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
19820 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
19821 * config/rs6000/rs6000-protos.h: Add extern defition for
19822 rs6000_generate_float2_double_code.
19823 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
19825 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
19826 (float2_v2df): Add define_expand.
19828 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
19831 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
19832 op_mode in the force_to_mode call.
19834 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
19836 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
19837 instead of checking each element individually.
19838 (aarch64_evpc_uzp): Likewise.
19839 (aarch64_evpc_zip): Likewise.
19840 (aarch64_evpc_ext): Likewise.
19841 (aarch64_evpc_rev): Likewise.
19842 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
19843 instead of checking each element individually. Return true without
19845 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
19846 whether all selected elements come from the same input, instead of
19847 checking each element individually. Remove calls to gen_rtx_REG,
19848 start_sequence and end_sequence and instead assert that no rtl is
19851 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
19853 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
19854 order of HIGH and CONST checks.
19856 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
19858 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
19859 if the destination isn't an SSA_NAME.
19861 2018-01-09 Richard Biener <rguenther@suse.de>
19863 PR tree-optimization/83668
19864 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
19866 (canonicalize_loop_form): ... here, renamed from ...
19867 (canonicalize_loop_closed_ssa_form): ... this and amended to
19868 swap successor edges for loop exit blocks to make us use
19869 the RPO order we need for initial schedule generation.
19871 2018-01-09 Joseph Myers <joseph@codesourcery.com>
19873 PR tree-optimization/64811
19874 * match.pd: When optimizing comparisons with Inf, avoid
19875 introducing or losing exceptions from comparisons with NaN.
19877 2018-01-09 Martin Liska <mliska@suse.cz>
19880 * asan.c (shadow_mem_size): Add gcc_assert.
19882 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
19884 Don't save registers in main().
19887 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
19888 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
19889 * config/avr/avr.c (avr_set_current_function): Don't error if
19890 naked, OS_task or OS_main are specified at the same time.
19891 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
19893 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
19895 * common/config/avr/avr-common.c (avr_option_optimization_table):
19896 Switch on -mmain-is-OS_task for optimizing compilations.
19898 2018-01-09 Richard Biener <rguenther@suse.de>
19900 PR tree-optimization/83572
19901 * graphite.c: Include cfganal.h.
19902 (graphite_transform_loops): Connect infinite loops to exit
19903 and remove fake edges at the end.
19905 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
19907 * ipa-inline.c (edge_badness): Revert accidental checkin.
19909 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
19912 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
19913 symbols; not inline clones.
19915 2018-01-09 Jakub Jelinek <jakub@redhat.com>
19918 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
19919 hard registers. Formatting fixes.
19921 PR preprocessor/83722
19922 * gcc.c (try_generate_repro): Pass
19923 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
19924 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
19927 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
19928 Kito Cheng <kito.cheng@gmail.com>
19930 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
19931 (riscv_leaf_function_p): Delete.
19932 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
19934 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19936 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
19938 (do_ifelse): New function.
19939 (do_isel): New function.
19940 (do_sub3): New function.
19941 (do_add3): New function.
19942 (do_load_mask_compare): New function.
19943 (do_overlap_load_compare): New function.
19944 (expand_compare_loop): New function.
19945 (expand_block_compare): Call expand_compare_loop() when appropriate.
19946 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
19947 option description.
19948 (-mblock-compare-inline-loop-limit): New option.
19950 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19953 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
19954 Reverse order of second and third operands in first alternative.
19955 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
19956 of first and second elements in UNSPEC_VPERMR vector.
19957 (altivec_expand_vec_perm_le): Likewise.
19959 2018-01-08 Jeff Law <law@redhat.com>
19961 PR rtl-optimizatin/81308
19962 * tree-switch-conversion.c (cfg_altered): New file scoped static.
19963 (process_switch): If group_case_labels makes a change, then set
19965 (pass_convert_switch::execute): If a switch is converted, then
19966 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
19968 PR rtl-optimization/81308
19969 * recog.c (split_all_insns): Conditionally cleanup the CFG after
19972 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
19974 PR target/83663 - Revert r255946
19975 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
19976 generation for cases where splatting a value is not useful.
19977 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
19978 across a vec_duplicate and a paradoxical subreg forming a vector
19979 mode to a vec_concat.
19981 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19983 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
19984 -march=armv8.3-a variants.
19985 * config/arm/t-multilib: Likewise.
19986 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
19988 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19990 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
19992 (cceq_ior_compare_complement): Give it a name so I can use it, and
19993 change boolean_or_operator predicate to boolean_operator so it can
19994 be used to generate a crand.
19995 (eqne): New code iterator.
19996 (bd/bd_neg): New code_attrs.
19997 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
19998 a single define_insn.
19999 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
20000 decrement (bdnzt/bdnzf/bdzt/bdzf).
20001 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
20002 with the new names of the branch decrement patterns, and added the
20003 names of the branch decrement conditional patterns.
20005 2018-01-08 Richard Biener <rguenther@suse.de>
20007 PR tree-optimization/83563
20008 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
20011 2018-01-08 Richard Biener <rguenther@suse.de>
20013 PR middle-end/83713
20014 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
20016 2018-01-08 Richard Biener <rguenther@suse.de>
20018 PR tree-optimization/83685
20019 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
20020 references to abnormals.
20022 2018-01-08 Richard Biener <rguenther@suse.de>
20025 * dwarf2out.c (output_indirect_strings): Handle empty
20026 skeleton_debug_str_hash.
20027 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
20029 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
20031 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
20032 (emit_store_direct): Likewise.
20033 (arc_trampoline_adjust_address): Likewise.
20034 (arc_asm_trampoline_template): New function.
20035 (arc_initialize_trampoline): Use asm_trampoline_template.
20036 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
20037 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
20038 * config/arc/arc.md (flush_icache): Delete pattern.
20040 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
20042 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
20043 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
20046 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
20049 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
20050 by not USED_FOR_TARGET.
20051 (make_pass_resolve_sw_modes): Likewise.
20053 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
20055 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
20058 2018-01-08 Richard Biener <rguenther@suse.de>
20060 PR middle-end/83580
20061 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
20063 2018-01-08 Richard Biener <rguenther@suse.de>
20065 PR middle-end/83517
20066 * match.pd ((t * 2) / 2) -> t): Add missing :c.
20068 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
20070 PR middle-end/81897
20071 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
20072 basic blocks with a small number of successors.
20073 (convert_control_dep_chain_into_preds): Improve handling of
20075 (dump_predicates): Split apart into...
20076 (dump_pred_chain): ...here...
20077 (dump_pred_info): ...and here.
20078 (can_one_predicate_be_invalidated_p): Add debugging printfs.
20079 (can_chain_union_be_invalidated_p): Improve check for invalidation
20081 (uninit_uses_cannot_happen): Avoid unnecessary if
20082 convert_control_dep_chain_into_preds yielded nothing.
20084 2018-01-06 Martin Sebor <msebor@redhat.com>
20086 PR tree-optimization/83640
20087 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
20088 subtracting negative offset from size.
20089 (builtin_access::overlap): Adjust offset bounds of the access to fall
20090 within the size of the object if possible.
20092 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
20094 PR rtl-optimization/83699
20095 * expmed.c (extract_bit_field_1): Restrict the vector usage of
20096 extract_bit_field_as_subreg to cases in which the extracted
20097 value is also a vector.
20099 * lra-constraints.c (process_alt_operands): Test for the equivalence
20100 substitutions when detecting a possible reload cycle.
20102 2018-01-06 Jakub Jelinek <jakub@redhat.com>
20105 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
20106 by default if flag_selective_schedling{,2}. Formatting fixes.
20108 PR rtl-optimization/83682
20109 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
20110 if it has non-VECTOR_MODE element mode.
20111 (vec_duplicate_p): Likewise.
20113 PR middle-end/83694
20114 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
20115 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
20117 2018-01-05 Jakub Jelinek <jakub@redhat.com>
20120 * config/i386/i386-builtin.def
20121 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
20122 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
20123 Require also OPTION_MASK_ISA_AVX512F in addition to
20124 OPTION_MASK_ISA_GFNI.
20125 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
20126 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
20127 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
20128 to OPTION_MASK_ISA_GFNI.
20129 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
20130 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
20131 OPTION_MASK_ISA_AVX512BW.
20132 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
20133 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
20134 addition to OPTION_MASK_ISA_GFNI.
20135 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
20136 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
20137 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
20138 to OPTION_MASK_ISA_GFNI.
20139 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
20140 a requirement for all ISAs rather than any of them with a few
20142 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
20144 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
20145 bitmasks to be enabled with 3 exceptions, instead of requiring any
20146 enabled ISA with lots of exceptions.
20147 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
20148 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
20149 Change avx512bw in isa attribute to avx512f.
20150 * config/i386/sgxintrin.h: Add license boilerplate.
20151 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
20152 to __AVX512F__ and __AVX512VL to __AVX512VL__.
20153 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
20154 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
20156 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
20157 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
20158 temporarily sse2 rather than sse if not enabled already.
20161 * config/i386/sse.md (VI248_VLBW): Rename to ...
20162 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
20163 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
20164 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
20165 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
20166 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
20167 mode iterator instead of VI248_VLBW.
20169 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
20171 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
20172 (record_modified): Skip clobbers; add debug output.
20173 (param_change_prob): Use sreal frequencies.
20175 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
20177 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
20178 punt for user-aligned variables.
20180 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
20182 * tree-chrec.c (chrec_contains_symbols): Return true for
20185 2018-01-05 Sudakshina Das <sudi.das@arm.com>
20188 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
20189 of (x|y) == x for BICS pattern.
20191 2018-01-05 Jakub Jelinek <jakub@redhat.com>
20193 PR tree-optimization/83605
20194 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
20195 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
20198 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
20200 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
20201 * config/epiphany/rtems.h: New file.
20203 2018-01-04 Jakub Jelinek <jakub@redhat.com>
20204 Uros Bizjak <ubizjak@gmail.com>
20207 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
20208 QIreg_operand instead of register_operand predicate.
20209 * config/i386/i386.c (ix86_rop_should_change_byte_p,
20210 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
20211 comments instead of -fmitigate[-_]rop.
20213 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
20216 * cgraphunit.c (symbol_table::compile): Switch to text_section
20217 before calling assembly_start debug hook.
20218 * run-rtl-passes.c (run_rtl_passes): Likewise.
20221 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
20223 * tree-vrp.c (extract_range_from_binary_expr_1): Check
20224 range_int_cst_p rather than !symbolic_range_p before calling
20225 extract_range_from_multiplicative_op_1.
20227 2018-01-04 Jeff Law <law@redhat.com>
20229 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
20230 redundant test in assertion.
20232 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
20234 * doc/rtl.texi: Document machine_mode wrapper classes.
20236 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
20238 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
20239 using tree_to_uhwi.
20241 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
20243 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
20244 the VEC_PERM_EXPR fold to fail.
20246 2018-01-04 Jakub Jelinek <jakub@redhat.com>
20249 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
20250 to switched_sections.
20252 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
20255 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
20256 test for d.testing.
20258 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
20261 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
20262 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
20264 2018-01-04 Jakub Jelinek <jakub@redhat.com>
20267 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
20268 is BLKmode and bitpos not zero or mode change is needed.
20270 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
20273 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
20276 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
20279 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
20280 instead of MULT rtx. Update all corresponding splitters.
20281 (*saddl_se): Ditto.
20282 (*ssub<modesuffix>): Ditto.
20283 (*ssubl_se): Ditto.
20284 (*cmp_sadd_di): Update split patterns.
20285 (*cmp_sadd_si): Ditto.
20286 (*cmp_sadd_sidi): Ditto.
20287 (*cmp_ssub_di): Ditto.
20288 (*cmp_ssub_si): Ditto.
20289 (*cmp_ssub_sidi): Ditto.
20290 * config/alpha/predicates.md (const23_operand): New predicate.
20291 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
20292 Look for ASHIFT, not MULT inner operand.
20293 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
20295 2018-01-04 Martin Liska <mliska@suse.cz>
20297 PR gcov-profile/83669
20298 * gcov.c (output_intermediate_file): Add version to intermediate
20300 * doc/gcov.texi: Document new field 'version' in intermediate
20301 file format. Fix location of '-k' option of gcov command.
20303 2018-01-04 Martin Liska <mliska@suse.cz>
20306 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
20308 2018-01-04 Jakub Jelinek <jakub@redhat.com>
20310 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
20312 2018-01-03 Martin Sebor <msebor@redhat.com>
20314 PR tree-optimization/83655
20315 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
20316 checking calls with invalid arguments.
20318 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20320 * tree-vect-stmts.c (vect_get_store_rhs): New function.
20321 (vectorizable_mask_load_store): Delete.
20322 (vectorizable_call): Return false for masked loads and stores.
20323 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
20324 instead of gimple_assign_rhs1.
20325 (vectorizable_load): Handle IFN_MASK_LOAD.
20326 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
20328 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20330 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
20332 (vectorizable_mask_load_store): ...here.
20333 (vectorizable_load): ...and here.
20335 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20337 * tree-vect-stmts.c (vect_build_all_ones_mask)
20338 (vect_build_zero_merge_argument): New functions, split out from...
20339 (vectorizable_load): ...here.
20341 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20343 * tree-vect-stmts.c (vect_check_store_rhs): New function,
20345 (vectorizable_mask_load_store): ...here.
20346 (vectorizable_store): ...and here.
20348 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20350 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
20352 (vectorizable_mask_load_store): ...here.
20354 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20356 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
20357 (vect_model_store_cost): Take a vec_load_store_type instead of a
20359 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
20360 (vect_model_store_cost): Take a vec_load_store_type instead of a
20362 (vectorizable_mask_load_store): Update accordingly.
20363 (vectorizable_store): Likewise.
20364 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
20366 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20368 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
20369 IFN_MASK_LOAD calls here rather than...
20370 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
20372 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20373 Alan Hayward <alan.hayward@arm.com>
20374 David Sherwood <david.sherwood@arm.com>
20376 * expmed.c (extract_bit_field_1): For vector extracts,
20377 fall back to extract_bit_field_as_subreg if vec_extract
20380 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20381 Alan Hayward <alan.hayward@arm.com>
20382 David Sherwood <david.sherwood@arm.com>
20384 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
20385 they are variable or constant sized.
20386 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
20387 slots for constant-sized data.
20389 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20390 Alan Hayward <alan.hayward@arm.com>
20391 David Sherwood <david.sherwood@arm.com>
20393 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
20394 handling COND_EXPRs with boolean comparisons, try to find a better
20395 basis for the mask type than the boolean itself.
20397 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20399 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
20400 is calculated and how it can be overridden.
20401 * genmodes.c (max_bitsize_mode_any_mode): New variable.
20402 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
20404 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
20407 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20408 Alan Hayward <alan.hayward@arm.com>
20409 David Sherwood <david.sherwood@arm.com>
20411 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
20412 Remove the mode argument.
20413 (aarch64_simd_valid_immediate): Remove the mode and inverse
20415 * config/aarch64/iterators.md (bitsize): New iterator.
20416 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
20417 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
20418 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
20419 aarch64_simd_valid_immediate.
20420 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
20421 (aarch64_reg_or_bic_imm): Likewise.
20422 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
20423 with an insn_type enum and msl with a modifier_type enum.
20424 Replace element_width with a scalar_mode. Change the shift
20425 to unsigned int. Add constructors for scalar_float_mode and
20426 scalar_int_mode elements.
20427 (aarch64_vect_float_const_representable_p): Delete.
20428 (aarch64_can_const_movi_rtx_p)
20429 (aarch64_simd_scalar_immediate_valid_for_move)
20430 (aarch64_simd_make_constant): Update call to
20431 aarch64_simd_valid_immediate.
20432 (aarch64_advsimd_valid_immediate_hs): New function.
20433 (aarch64_advsimd_valid_immediate): Likewise.
20434 (aarch64_simd_valid_immediate): Remove mode and inverse
20435 arguments. Rewrite to use the above. Use const_vec_duplicate_p
20436 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
20437 and aarch64_float_const_representable_p on the result.
20438 (aarch64_output_simd_mov_immediate): Remove mode argument.
20439 Update call to aarch64_simd_valid_immediate and use of
20440 simd_immediate_info.
20441 (aarch64_output_scalar_simd_mov_immediate): Update call
20444 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20445 Alan Hayward <alan.hayward@arm.com>
20446 David Sherwood <david.sherwood@arm.com>
20448 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
20449 (mode_nunits): Likewise CONST_MODE_NUNITS.
20450 * machmode.def (ADJUST_NUNITS): Document.
20451 * genmodes.c (mode_data::need_nunits_adj): New field.
20452 (blank_mode): Update accordingly.
20453 (adj_nunits): New variable.
20454 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
20456 (emit_mode_size_inline): Set need_bytesize_adj for all modes
20457 listed in adj_nunits.
20458 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
20459 listed in adj_nunits. Don't emit case statements for such modes.
20460 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
20461 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
20462 nothing if adj_nunits is nonnull.
20463 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
20464 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
20465 (emit_mode_fbit): Update use of print_maybe_const_decl.
20466 (emit_move_size): Likewise. Treat the array as non-const
20468 (emit_mode_adjustments): Handle adj_nunits.
20470 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20472 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
20473 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
20474 (VECTOR_MODES): Use it.
20475 (make_vector_modes): Take the prefix as an argument.
20477 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20478 Alan Hayward <alan.hayward@arm.com>
20479 David Sherwood <david.sherwood@arm.com>
20481 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
20482 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
20483 for MODE_VECTOR_BOOL.
20484 * machmode.def (VECTOR_BOOL_MODE): Document.
20485 * genmodes.c (VECTOR_BOOL_MODE): New macro.
20486 (make_vector_bool_mode): New function.
20487 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
20489 * lto-streamer-in.c (lto_input_mode_table): Likewise.
20490 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
20492 * stor-layout.c (int_mode_for_mode): Likewise.
20493 * tree.c (build_vector_type_for_mode): Likewise.
20494 * varasm.c (output_constant_pool_2): Likewise.
20495 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
20496 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
20497 for MODE_VECTOR_BOOL.
20498 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
20499 of mode class checks.
20500 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
20501 instead of a list of mode class checks.
20502 (expand_vector_scalar_condition): Likewise.
20503 (type_for_widest_vector_mode): Handle BImode as an inner mode.
20505 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20506 Alan Hayward <alan.hayward@arm.com>
20507 David Sherwood <david.sherwood@arm.com>
20509 * machmode.h (mode_size): Change from unsigned short to
20511 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
20512 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
20513 or if measurement_type is not polynomial.
20514 (fixed_size_mode::includes_p): Check for constant-sized modes.
20515 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
20516 return a poly_uint16 rather than an unsigned short.
20517 (emit_mode_size): Change the type of mode_size from unsigned short
20518 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
20519 (emit_mode_adjustments): Cope with polynomial vector sizes.
20520 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
20522 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
20524 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
20525 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
20526 * caller-save.c (setup_save_areas): Likewise.
20527 (replace_reg_with_saved_mem): Likewise.
20528 * calls.c (emit_library_call_value_1): Likewise.
20529 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
20530 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
20531 (gen_lowpart_for_combine): Likewise.
20532 * convert.c (convert_to_integer_1): Likewise.
20533 * cse.c (equiv_constant, cse_insn): Likewise.
20534 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
20535 (cselib_subst_to_values): Likewise.
20536 * dce.c (word_dce_process_block): Likewise.
20537 * df-problems.c (df_word_lr_mark_ref): Likewise.
20538 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
20539 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
20540 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
20541 (rtl_for_decl_location): Likewise.
20542 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
20543 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
20544 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
20545 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
20546 (expand_expr_real_1): Likewise.
20547 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
20548 (pad_below): Likewise.
20549 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
20550 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
20551 * ira.c (get_subreg_tracking_sizes): Likewise.
20552 * ira-build.c (ira_create_allocno_objects): Likewise.
20553 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
20554 (ira_sort_regnos_for_alter_reg): Likewise.
20555 * ira-costs.c (record_operand_costs): Likewise.
20556 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
20557 (resolve_simple_move): Likewise.
20558 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
20559 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
20560 (lra_constraints): Likewise.
20561 (CONST_POOL_OK_P): Reject variable-sized modes.
20562 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
20563 (add_pseudo_to_slot, lra_spill): Likewise.
20564 * omp-low.c (omp_clause_aligned_alignment): Likewise.
20565 * optabs-query.c (get_best_extraction_insn): Likewise.
20566 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
20567 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
20568 (expand_mult_highpart, valid_multiword_target_p): Likewise.
20569 * recog.c (offsettable_address_addr_space_p): Likewise.
20570 * regcprop.c (maybe_mode_change): Likewise.
20571 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
20572 * regrename.c (build_def_use): Likewise.
20573 * regstat.c (dump_reg_info): Likewise.
20574 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
20575 (find_reloads, find_reloads_subreg_address): Likewise.
20576 * reload1.c (eliminate_regs_1): Likewise.
20577 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
20578 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
20579 (simplify_binary_operation_1, simplify_subreg): Likewise.
20580 * targhooks.c (default_function_arg_padding): Likewise.
20581 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
20582 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
20583 (verify_gimple_assign_ternary): Likewise.
20584 * tree-inline.c (estimate_move_cost): Likewise.
20585 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
20586 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
20587 (get_address_cost_ainc): Likewise.
20588 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
20589 (vect_supportable_dr_alignment): Likewise.
20590 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
20591 (vectorizable_reduction): Likewise.
20592 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
20593 (vectorizable_operation, vectorizable_load): Likewise.
20594 * tree.c (build_same_sized_truth_vector_type): Likewise.
20595 * valtrack.c (cleanup_auto_inc_dec): Likewise.
20596 * var-tracking.c (emit_note_insn_var_location): Likewise.
20597 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
20598 (ADDR_VEC_ALIGN): Likewise.
20600 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20601 Alan Hayward <alan.hayward@arm.com>
20602 David Sherwood <david.sherwood@arm.com>
20604 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
20606 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
20607 or if measurement_type is polynomial.
20608 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
20609 * combine.c (make_extraction): Likewise.
20610 * dse.c (find_shift_sequence): Likewise.
20611 * dwarf2out.c (mem_loc_descriptor): Likewise.
20612 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
20613 (extract_bit_field, extract_low_bits): Likewise.
20614 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
20615 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
20616 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
20617 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
20618 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
20619 * reload.c (find_reloads): Likewise.
20620 * reload1.c (alter_reg): Likewise.
20621 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
20622 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
20623 * tree-if-conv.c (predicate_mem_writes): Likewise.
20624 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
20625 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
20626 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
20627 * valtrack.c (dead_debug_insert_temp): Likewise.
20628 * varasm.c (mergeable_constant_section): Likewise.
20629 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
20631 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20632 Alan Hayward <alan.hayward@arm.com>
20633 David Sherwood <david.sherwood@arm.com>
20635 * expr.c (expand_assignment): Cope with polynomial mode sizes
20636 when assigning to a CONCAT.
20638 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20639 Alan Hayward <alan.hayward@arm.com>
20640 David Sherwood <david.sherwood@arm.com>
20642 * machmode.h (mode_precision): Change from unsigned short to
20644 (mode_to_precision): Return a poly_uint16 rather than an unsigned
20646 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
20647 or if measurement_type is not polynomial.
20648 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
20649 in which the mode is already known to be a scalar_int_mode.
20650 * genmodes.c (emit_mode_precision): Change the type of mode_precision
20651 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
20653 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
20654 for GET_MODE_PRECISION.
20655 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
20656 for GET_MODE_PRECISION.
20657 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
20659 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
20660 (expand_field_assignment, make_extraction): Likewise.
20661 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
20662 (get_last_value): Likewise.
20663 * convert.c (convert_to_integer_1): Likewise.
20664 * cse.c (cse_insn): Likewise.
20665 * expr.c (expand_expr_real_1): Likewise.
20666 * lra-constraints.c (simplify_operand_subreg): Likewise.
20667 * optabs-query.c (can_atomic_load_p): Likewise.
20668 * optabs.c (expand_atomic_load): Likewise.
20669 (expand_atomic_store): Likewise.
20670 * ree.c (combine_reaching_defs): Likewise.
20671 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
20672 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
20673 * tree.h (type_has_mode_precision_p): Likewise.
20674 * ubsan.c (instrument_si_overflow): Likewise.
20676 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20677 Alan Hayward <alan.hayward@arm.com>
20678 David Sherwood <david.sherwood@arm.com>
20680 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
20681 polynomial numbers of units.
20682 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
20683 (valid_vector_subparts_p): New function.
20684 (build_vector_type): Remove temporary shim and take the number
20685 of units as a poly_uint64 rather than an int.
20686 (build_opaque_vector_type): Take the number of units as a
20687 poly_uint64 rather than an int.
20688 * tree.c (build_vector_from_ctor): Handle polynomial
20689 TYPE_VECTOR_SUBPARTS.
20690 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
20691 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
20692 (build_vector_from_val): If the number of units is variable,
20693 use build_vec_duplicate_cst for constant operands and
20694 VEC_DUPLICATE_EXPR otherwise.
20695 (make_vector_type): Remove temporary is_constant ().
20696 (build_vector_type, build_opaque_vector_type): Take the number of
20697 units as a poly_uint64 rather than an int.
20698 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
20700 * cfgexpand.c (expand_debug_expr): Likewise.
20701 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
20702 (store_constructor, expand_expr_real_1): Likewise.
20703 (const_scalar_mask_from_tree): Likewise.
20704 * fold-const-call.c (fold_const_reduction): Likewise.
20705 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
20706 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
20707 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
20708 (fold_relational_const): Likewise.
20709 (native_interpret_vector): Likewise. Change the size from an
20710 int to an unsigned int.
20711 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
20712 TYPE_VECTOR_SUBPARTS.
20713 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
20714 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
20715 duplicating a non-constant operand into a variable-length vector.
20716 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
20717 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
20718 * ipa-icf.c (sem_variable::equals): Likewise.
20719 * match.pd: Likewise.
20720 * omp-simd-clone.c (simd_clone_subparts): Likewise.
20721 * print-tree.c (print_node): Likewise.
20722 * stor-layout.c (layout_type): Likewise.
20723 * targhooks.c (default_builtin_vectorization_cost): Likewise.
20724 * tree-cfg.c (verify_gimple_comparison): Likewise.
20725 (verify_gimple_assign_binary): Likewise.
20726 (verify_gimple_assign_ternary): Likewise.
20727 (verify_gimple_assign_single): Likewise.
20728 * tree-pretty-print.c (dump_generic_node): Likewise.
20729 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
20730 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
20731 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
20732 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
20733 (vect_shift_permute_load_chain): Likewise.
20734 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
20735 (expand_vector_condition, optimize_vector_constructor): Likewise.
20736 (lower_vec_perm, get_compute_type): Likewise.
20737 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
20738 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
20739 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
20740 (vect_recog_mask_conversion_pattern): Likewise.
20741 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
20742 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
20743 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
20744 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
20745 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
20746 (vectorizable_shift, vectorizable_operation, vectorizable_store)
20747 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
20748 (supportable_widening_operation): Likewise.
20749 (supportable_narrowing_operation): Likewise.
20750 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
20752 * varasm.c (output_constant): Likewise.
20754 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20755 Alan Hayward <alan.hayward@arm.com>
20756 David Sherwood <david.sherwood@arm.com>
20758 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
20759 so that both the length == 3 and length != 3 cases set up their
20760 own permute vectors. Add comments explaining why we know the
20761 number of elements is constant.
20762 (vect_permute_load_chain): Likewise.
20764 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20765 Alan Hayward <alan.hayward@arm.com>
20766 David Sherwood <david.sherwood@arm.com>
20768 * machmode.h (mode_nunits): Change from unsigned char to
20770 (ONLY_FIXED_SIZE_MODES): New macro.
20771 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
20772 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
20773 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
20775 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
20776 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
20777 or if measurement_type is not polynomial.
20778 * genmodes.c (ZERO_COEFFS): New macro.
20779 (emit_mode_nunits_inline): Make mode_nunits_inline return a
20781 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
20782 Use ZERO_COEFFS when emitting initializers.
20783 * data-streamer.h (bp_pack_poly_value): New function.
20784 (bp_unpack_poly_value): Likewise.
20785 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
20786 for GET_MODE_NUNITS.
20787 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
20788 for GET_MODE_NUNITS.
20789 * tree.c (make_vector_type): Remove temporary shim and make
20790 the real function take the number of units as a poly_uint64
20791 rather than an int.
20792 (build_vector_type_for_mode): Handle polynomial nunits.
20793 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
20794 * emit-rtl.c (const_vec_series_p_1): Likewise.
20795 (gen_rtx_CONST_VECTOR): Likewise.
20796 * fold-const.c (test_vec_duplicate_folding): Likewise.
20797 * genrecog.c (validate_pattern): Likewise.
20798 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
20799 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
20800 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
20801 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
20802 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
20803 * rtlanal.c (subreg_get_info): Likewise.
20804 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
20805 (vect_grouped_load_supported): Likewise.
20806 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
20807 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
20808 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
20809 (simplify_const_unary_operation, simplify_binary_operation_1)
20810 (simplify_const_binary_operation, simplify_ternary_operation)
20811 (test_vector_ops_duplicate, test_vector_ops): Likewise.
20812 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
20813 instead of CONST_VECTOR_NUNITS.
20814 * varasm.c (output_constant_pool_2): Likewise.
20815 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
20816 explicit-encoded elements in the XVEC for variable-length vectors.
20818 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20820 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
20822 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20823 Alan Hayward <alan.hayward@arm.com>
20824 David Sherwood <david.sherwood@arm.com>
20826 * coretypes.h (fixed_size_mode): Declare.
20827 (fixed_size_mode_pod): New typedef.
20828 * builtins.h (target_builtins::x_apply_args_mode)
20829 (target_builtins::x_apply_result_mode): Change type to
20830 fixed_size_mode_pod.
20831 * builtins.c (apply_args_size, apply_result_size, result_vector)
20832 (expand_builtin_apply_args_1, expand_builtin_apply)
20833 (expand_builtin_return): Update accordingly.
20835 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20837 * cse.c (hash_rtx_cb): Hash only the encoded elements.
20838 * cselib.c (cselib_hash_rtx): Likewise.
20839 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
20840 CONST_VECTOR encoding.
20842 2018-01-03 Jakub Jelinek <jakub@redhat.com>
20843 Jeff Law <law@redhat.com>
20846 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
20847 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
20848 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
20849 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
20852 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
20853 explicitly probe *sp in a noreturn function if there were any callee
20854 register saves or frame pointer is needed.
20856 2018-01-03 Jakub Jelinek <jakub@redhat.com>
20859 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
20860 BLKmode for ternary, binary or unary expressions.
20863 * var-tracking.c (delete_vta_debug_insn): New inline function.
20864 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
20865 insns from get_insns () to NULL instead of each bb separately.
20866 Use delete_vta_debug_insn. No longer static.
20867 (vt_debug_insns_local, variable_tracking_main_1): Adjust
20868 delete_vta_debug_insns callers.
20869 * rtl.h (delete_vta_debug_insns): Declare.
20870 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
20871 instead of variable_tracking_main.
20873 2018-01-03 Martin Sebor <msebor@redhat.com>
20875 PR tree-optimization/83603
20876 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
20877 arguments past the endof the argument list in functions declared
20878 without a prototype.
20879 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
20880 Avoid checking when arguments are null.
20882 2018-01-03 Martin Sebor <msebor@redhat.com>
20885 * doc/extend.texi (attribute const): Fix a typo.
20886 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
20887 issuing -Wsuggest-attribute for void functions.
20889 2018-01-03 Martin Sebor <msebor@redhat.com>
20891 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
20892 offset_int::from instead of wide_int::to_shwi.
20893 (maybe_diag_overlap): Remove assertion.
20894 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
20895 * gimple-ssa-sprintf.c (format_directive): Same.
20896 (parse_directive): Same.
20897 (sprintf_dom_walker::compute_format_length): Same.
20898 (try_substitute_return_value): Same.
20900 2018-01-03 Jeff Law <law@redhat.com>
20902 PR middle-end/83654
20903 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
20904 non-constant residual for zero at runtime and avoid probing in
20905 that case. Reorganize code for trailing problem to mirror handling
20908 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
20910 PR tree-optimization/83501
20911 * tree-ssa-strlen.c (get_string_cst): New.
20912 (handle_char_store): Call get_string_cst.
20914 2018-01-03 Martin Liska <mliska@suse.cz>
20916 PR tree-optimization/83593
20917 * tree-ssa-strlen.c: Include tree-cfg.h.
20918 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
20919 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
20920 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
20922 (strlen_dom_walker::before_dom_children): Call
20923 gimple_purge_dead_eh_edges. Dump tranformation with details
20925 (strlen_dom_walker::before_dom_children): Update call by adding
20926 new argument cleanup_eh.
20927 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
20929 2018-01-03 Martin Liska <mliska@suse.cz>
20932 * cif-code.def (VARIADIC_THUNK): New enum value.
20933 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
20936 2018-01-03 Jan Beulich <jbeulich@suse.com>
20938 * sse.md (mov<mode>_internal): Tighten condition for when to use
20939 vmovdqu<ssescalarsize> for TI and OI modes.
20941 2018-01-03 Jakub Jelinek <jakub@redhat.com>
20943 Update copyright years.
20945 2018-01-03 Martin Liska <mliska@suse.cz>
20948 * ipa-visibility.c (function_and_variable_visibility): Skip
20949 functions with noipa attribure.
20951 2018-01-03 Jakub Jelinek <jakub@redhat.com>
20953 * gcc.c (process_command): Update copyright notice dates.
20954 * gcov-dump.c (print_version): Ditto.
20955 * gcov.c (print_version): Ditto.
20956 * gcov-tool.c (print_version): Ditto.
20957 * gengtype.c (create_file): Ditto.
20958 * doc/cpp.texi: Bump @copying's copyright year.
20959 * doc/cppinternals.texi: Ditto.
20960 * doc/gcc.texi: Ditto.
20961 * doc/gccint.texi: Ditto.
20962 * doc/gcov.texi: Ditto.
20963 * doc/install.texi: Ditto.
20964 * doc/invoke.texi: Ditto.
20966 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20968 * vector-builder.h (vector_builder::m_full_nelts): Change from
20969 unsigned int to poly_uint64.
20970 (vector_builder::full_nelts): Update prototype accordingly.
20971 (vector_builder::new_vector): Likewise.
20972 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
20973 (vector_builder::operator ==): Likewise.
20974 (vector_builder::finalize): Likewise.
20975 * int-vector-builder.h (int_vector_builder::int_vector_builder):
20976 Take the number of elements as a poly_uint64 rather than an
20978 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
20979 from unsigned int to poly_uint64.
20980 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
20981 (vec_perm_indices::new_vector): Likewise.
20982 (vec_perm_indices::length): Likewise.
20983 (vec_perm_indices::nelts_per_input): Likewise.
20984 (vec_perm_indices::input_nelts): Likewise.
20985 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
20986 number of elements per input as a poly_uint64 rather than an
20987 unsigned int. Use the original encoding for variable-length
20988 vectors, rather than clamping each individual element.
20989 For the second and subsequent elements in each pattern,
20990 clamp the step and base before clamping their sum.
20991 (vec_perm_indices::series_p): Handle polynomial element counts.
20992 (vec_perm_indices::all_in_range_p): Likewise.
20993 (vec_perm_indices_to_tree): Likewise.
20994 (vec_perm_indices_to_rtx): Likewise.
20995 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
20996 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
20997 (tree_vector_builder::new_binary_operation): Handle polynomial
20998 element counts. Return false if we need to know the number
20999 of elements at compile time.
21000 * fold-const.c (fold_vec_perm): Punt if the number of elements
21001 isn't known at compile time.
21003 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21005 * vec-perm-indices.h (vec_perm_builder): Change element type
21006 from HOST_WIDE_INT to poly_int64.
21007 (vec_perm_indices::element_type): Update accordingly.
21008 (vec_perm_indices::clamp): Handle polynomial element_types.
21009 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
21010 (vec_perm_indices::all_in_range_p): Likewise.
21011 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
21013 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
21014 polynomial vec_perm_indices element types.
21015 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
21016 * fold-const.c (fold_vec_perm): Likewise.
21017 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
21018 * tree-vect-generic.c (lower_vec_perm): Likewise.
21019 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
21020 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
21021 element type to HOST_WIDE_INT.
21023 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21024 Alan Hayward <alan.hayward@arm.com>
21025 David Sherwood <david.sherwood@arm.com>
21027 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
21028 rather than an int. Use plus_constant.
21029 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
21030 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
21032 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21033 Alan Hayward <alan.hayward@arm.com>
21034 David Sherwood <david.sherwood@arm.com>
21036 * calls.c (emit_call_1, expand_call): Change struct_value_size from
21037 a HOST_WIDE_INT to a poly_int64.
21039 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21040 Alan Hayward <alan.hayward@arm.com>
21041 David Sherwood <david.sherwood@arm.com>
21043 * calls.c (load_register_parameters): Cope with polynomial
21044 mode sizes. Require a constant size for BLKmode parameters
21045 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
21046 forces a parameter to be padded at the lsb end in order to
21047 fill a complete number of words, require the parameter size
21048 to be ordered wrt UNITS_PER_WORD.
21050 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21051 Alan Hayward <alan.hayward@arm.com>
21052 David Sherwood <david.sherwood@arm.com>
21054 * reload1.c (spill_stack_slot_width): Change element type
21055 from unsigned int to poly_uint64_pod.
21056 (alter_reg): Treat mode sizes as polynomial.
21058 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21059 Alan Hayward <alan.hayward@arm.com>
21060 David Sherwood <david.sherwood@arm.com>
21062 * reload.c (complex_word_subreg_p): New function.
21063 (reload_inner_reg_of_subreg, push_reload): Use it.
21065 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21066 Alan Hayward <alan.hayward@arm.com>
21067 David Sherwood <david.sherwood@arm.com>
21069 * lra-constraints.c (process_alt_operands): Reject matched
21070 operands whose sizes aren't ordered.
21071 (match_reload): Refer to this check here.
21073 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21074 Alan Hayward <alan.hayward@arm.com>
21075 David Sherwood <david.sherwood@arm.com>
21077 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
21078 that the mode size is in the set {1, 2, 4, 8, 16}.
21080 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21081 Alan Hayward <alan.hayward@arm.com>
21082 David Sherwood <david.sherwood@arm.com>
21084 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
21085 Use plus_constant instead of gen_rtx_PLUS.
21087 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21088 Alan Hayward <alan.hayward@arm.com>
21089 David Sherwood <david.sherwood@arm.com>
21091 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
21092 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
21093 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
21094 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
21095 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
21096 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
21097 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
21098 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
21099 * config/i386/i386.c (ix86_push_rounding): ...this new function.
21100 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
21102 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
21103 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
21104 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
21105 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
21106 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
21107 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
21108 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
21109 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
21110 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
21111 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
21113 * expr.c (emit_move_resolve_push): Treat the input and result
21114 of PUSH_ROUNDING as a poly_int64.
21115 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
21116 (emit_push_insn): Likewise.
21117 * lra-eliminations.c (mark_not_eliminable): Likewise.
21118 * recog.c (push_operand): Likewise.
21119 * reload1.c (elimination_effects): Likewise.
21120 * rtlanal.c (nonzero_bits1): Likewise.
21121 * calls.c (store_one_arg): Likewise. Require the padding to be
21122 known at compile time.
21124 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21125 Alan Hayward <alan.hayward@arm.com>
21126 David Sherwood <david.sherwood@arm.com>
21128 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
21129 Use plus_constant instead of gen_rtx_PLUS.
21131 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21132 Alan Hayward <alan.hayward@arm.com>
21133 David Sherwood <david.sherwood@arm.com>
21135 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
21136 rather than an int.
21138 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21139 Alan Hayward <alan.hayward@arm.com>
21140 David Sherwood <david.sherwood@arm.com>
21142 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
21143 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
21144 via stack temporaries. Treat the mode size as polynomial too.
21146 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21147 Alan Hayward <alan.hayward@arm.com>
21148 David Sherwood <david.sherwood@arm.com>
21150 * expr.c (expand_expr_real_2): When handling conversions involving
21151 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
21152 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
21153 as a poly_uint64 too.
21155 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21156 Alan Hayward <alan.hayward@arm.com>
21157 David Sherwood <david.sherwood@arm.com>
21159 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
21161 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21162 Alan Hayward <alan.hayward@arm.com>
21163 David Sherwood <david.sherwood@arm.com>
21165 * combine.c (can_change_dest_mode): Handle polynomial
21166 REGMODE_NATURAL_SIZE.
21167 * expmed.c (store_bit_field_1): Likewise.
21168 * expr.c (store_constructor): Likewise.
21169 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
21170 and polynomial REGMODE_NATURAL_SIZE.
21171 (gen_lowpart_common): Likewise.
21172 * reginfo.c (record_subregs_of_mode): Likewise.
21173 * rtlanal.c (read_modify_subreg_p): Likewise.
21175 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21176 Alan Hayward <alan.hayward@arm.com>
21177 David Sherwood <david.sherwood@arm.com>
21179 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
21180 numbers of elements.
21182 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21183 Alan Hayward <alan.hayward@arm.com>
21184 David Sherwood <david.sherwood@arm.com>
21186 * match.pd: Cope with polynomial numbers of vector elements.
21188 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21189 Alan Hayward <alan.hayward@arm.com>
21190 David Sherwood <david.sherwood@arm.com>
21192 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
21193 in a POINTER_PLUS_EXPR.
21195 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21196 Alan Hayward <alan.hayward@arm.com>
21197 David Sherwood <david.sherwood@arm.com>
21199 * omp-simd-clone.c (simd_clone_subparts): New function.
21200 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
21201 (ipa_simd_modify_function_body): Likewise.
21203 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21204 Alan Hayward <alan.hayward@arm.com>
21205 David Sherwood <david.sherwood@arm.com>
21207 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
21208 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
21209 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
21210 (expand_vector_condition, vector_element): Likewise.
21211 (subparts_gt): New function.
21212 (get_compute_type): Use subparts_gt.
21213 (count_type_subparts): Delete.
21214 (expand_vector_operations_1): Use subparts_gt instead of
21215 count_type_subparts.
21217 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21218 Alan Hayward <alan.hayward@arm.com>
21219 David Sherwood <david.sherwood@arm.com>
21221 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
21222 (vect_compile_time_alias): ...this new function. Do the calculation
21223 on poly_ints rather than trees.
21224 (vect_prune_runtime_alias_test_list): Update call accordingly.
21226 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21227 Alan Hayward <alan.hayward@arm.com>
21228 David Sherwood <david.sherwood@arm.com>
21230 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
21232 (vect_schedule_slp_instance): Likewise.
21234 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21235 Alan Hayward <alan.hayward@arm.com>
21236 David Sherwood <david.sherwood@arm.com>
21238 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
21239 constant and extern definitions for variable-length vectors.
21240 (vect_get_constant_vectors): Note that the number of units
21241 is known to be constant.
21243 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21244 Alan Hayward <alan.hayward@arm.com>
21245 David Sherwood <david.sherwood@arm.com>
21247 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
21248 of units as polynomial. Choose between WIDE and NARROW based
21251 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21252 Alan Hayward <alan.hayward@arm.com>
21253 David Sherwood <david.sherwood@arm.com>
21255 * tree-vect-stmts.c (simd_clone_subparts): New function.
21256 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
21258 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21259 Alan Hayward <alan.hayward@arm.com>
21260 David Sherwood <david.sherwood@arm.com>
21262 * tree-vect-stmts.c (vectorizable_call): Treat the number of
21263 vectors as polynomial. Use build_index_vector for
21264 IFN_GOMP_SIMD_LANE.
21266 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21267 Alan Hayward <alan.hayward@arm.com>
21268 David Sherwood <david.sherwood@arm.com>
21270 * tree-vect-stmts.c (get_load_store_type): Treat the number of
21271 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
21272 for variable-length vectors.
21273 (vectorizable_mask_load_store): Treat the number of units as
21274 polynomial, asserting that it is constant if the condition has
21275 already been enforced.
21276 (vectorizable_store, vectorizable_load): Likewise.
21278 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21279 Alan Hayward <alan.hayward@arm.com>
21280 David Sherwood <david.sherwood@arm.com>
21282 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
21283 of units as polynomial. Punt if we can't tell at compile time
21284 which vector contains the final result.
21286 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21287 Alan Hayward <alan.hayward@arm.com>
21288 David Sherwood <david.sherwood@arm.com>
21290 * tree-vect-loop.c (vectorizable_induction): Treat the number
21291 of units as polynomial. Punt on SLP inductions. Use an integer
21292 VEC_SERIES_EXPR for variable-length integer reductions. Use a
21293 cast of such a series for variable-length floating-point
21296 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21297 Alan Hayward <alan.hayward@arm.com>
21298 David Sherwood <david.sherwood@arm.com>
21300 * tree.h (build_index_vector): Declare.
21301 * tree.c (build_index_vector): New function.
21302 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
21303 of units as polynomial, forcibly converting it to a constant if
21304 vectorizable_reduction has already enforced the condition.
21305 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
21306 to create a {1,2,3,...} vector.
21307 (vectorizable_reduction): Treat the number of units as polynomial.
21308 Choose vectype_in based on the largest scalar element size rather
21309 than the smallest number of units. Enforce the restrictions
21312 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21313 Alan Hayward <alan.hayward@arm.com>
21314 David Sherwood <david.sherwood@arm.com>
21316 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
21317 number of units as polynomial.
21319 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21320 Alan Hayward <alan.hayward@arm.com>
21321 David Sherwood <david.sherwood@arm.com>
21323 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
21324 * target.def (autovectorize_vector_sizes): Return the vector sizes
21325 by pointer, using vector_sizes rather than a bitmask.
21326 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
21327 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
21328 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
21330 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
21331 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
21332 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
21333 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
21334 * omp-general.c (omp_max_vf): Likewise.
21335 * omp-low.c (omp_clause_aligned_alignment): Likewise.
21336 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
21337 * tree-vect-loop.c (vect_analyze_loop): Likewise.
21338 * tree-vect-slp.c (vect_slp_bb): Likewise.
21339 * doc/tm.texi: Regenerate.
21340 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
21342 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
21343 the vector size as a poly_uint64 rather than an unsigned int.
21344 (current_vector_size): Change from an unsigned int to a poly_uint64.
21345 (get_vectype_for_scalar_type): Update accordingly.
21346 * tree.h (build_truth_vector_type): Take the size and number of
21347 units as a poly_uint64 rather than an unsigned int.
21348 (build_vector_type): Add a temporary overload that takes
21349 the number of units as a poly_uint64 rather than an unsigned int.
21350 * tree.c (make_vector_type): Likewise.
21351 (build_truth_vector_type): Take the number of units as a poly_uint64
21352 rather than an unsigned int.
21354 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21355 Alan Hayward <alan.hayward@arm.com>
21356 David Sherwood <david.sherwood@arm.com>
21358 * target.def (get_mask_mode): Take the number of units and length
21359 as poly_uint64s rather than unsigned ints.
21360 * targhooks.h (default_get_mask_mode): Update accordingly.
21361 * targhooks.c (default_get_mask_mode): Likewise.
21362 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
21363 * doc/tm.texi: Regenerate.
21365 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21366 Alan Hayward <alan.hayward@arm.com>
21367 David Sherwood <david.sherwood@arm.com>
21369 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
21370 * omp-general.c (omp_max_vf): Likewise.
21371 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
21372 (expand_omp_simd): Handle polynomial safelen.
21373 * omp-low.c (omplow_simd_context): Add a default constructor.
21374 (omplow_simd_context::max_vf): Change from int to poly_uint64.
21375 (lower_rec_simd_input_clauses): Update accordingly.
21376 (lower_rec_input_clauses): Likewise.
21378 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21379 Alan Hayward <alan.hayward@arm.com>
21380 David Sherwood <david.sherwood@arm.com>
21382 * tree-vectorizer.h (vect_nunits_for_cost): New function.
21383 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
21384 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
21385 (vect_analyze_slp_cost): Likewise.
21386 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
21387 (vect_model_load_cost): Likewise.
21389 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21390 Alan Hayward <alan.hayward@arm.com>
21391 David Sherwood <david.sherwood@arm.com>
21393 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
21394 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
21395 from an unsigned int * to a poly_uint64_pod *.
21396 (calculate_unrolling_factor): New function.
21397 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
21399 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21400 Alan Hayward <alan.hayward@arm.com>
21401 David Sherwood <david.sherwood@arm.com>
21403 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
21404 from an unsigned int to a poly_uint64.
21405 (_loop_vec_info::slp_unrolling_factor): Likewise.
21406 (_loop_vec_info::vectorization_factor): Change from an int
21408 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
21409 (vect_get_num_vectors): New function.
21410 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
21411 (vect_get_num_copies): Use vect_get_num_vectors.
21412 (vect_analyze_data_ref_dependences): Change max_vf from an int *
21413 to an unsigned int *.
21414 (vect_analyze_data_refs): Change min_vf from an int * to a
21416 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
21417 than an unsigned HOST_WIDE_INT.
21418 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
21419 (vect_analyze_data_ref_dependence): Change max_vf from an int *
21420 to an unsigned int *.
21421 (vect_analyze_data_ref_dependences): Likewise.
21422 (vect_compute_data_ref_alignment): Handle polynomial vf.
21423 (vect_enhance_data_refs_alignment): Likewise.
21424 (vect_prune_runtime_alias_test_list): Likewise.
21425 (vect_shift_permute_load_chain): Likewise.
21426 (vect_supportable_dr_alignment): Likewise.
21427 (dependence_distance_ge_vf): Take the vectorization factor as a
21428 poly_uint64 rather than an unsigned HOST_WIDE_INT.
21429 (vect_analyze_data_refs): Change min_vf from an int * to a
21431 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
21432 vfm1 as a poly_uint64 rather than an int. Make the same change
21433 for the returned bound_scalar.
21434 (vect_gen_vector_loop_niters): Handle polynomial vf.
21435 (vect_do_peeling): Likewise. Update call to
21436 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
21437 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
21439 * tree-vect-loop.c (vect_determine_vectorization_factor)
21440 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
21441 (vect_get_known_peeling_cost): Likewise.
21442 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
21443 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
21444 (vect_transform_loop): Likewise. Use the lowest possible VF when
21445 updating the upper bounds of the loop.
21446 (vect_min_worthwhile_factor): Make static. Return an unsigned int
21447 rather than an int.
21448 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
21449 polynomial unroll factors.
21450 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
21451 (vect_make_slp_decision): Likewise.
21452 (vect_supported_load_permutation_p): Likewise, and polynomial
21454 (vect_analyze_slp_cost): Handle polynomial vf.
21455 (vect_slp_analyze_node_operations): Likewise.
21456 (vect_slp_analyze_bb_1): Likewise.
21457 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
21458 than an unsigned HOST_WIDE_INT.
21459 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
21460 (vectorizable_load): Handle polynomial vf.
21461 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
21463 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
21465 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21466 Alan Hayward <alan.hayward@arm.com>
21467 David Sherwood <david.sherwood@arm.com>
21469 * match.pd: Handle bit operations involving three constants
21470 and try to fold one pair.
21472 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21474 * tree-vect-loop-manip.c: Include gimple-fold.h.
21475 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
21476 niters_maybe_zero parameters. Handle other cases besides a step of 1.
21477 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
21478 Add a path that uses a step of VF instead of 1, but disable it
21480 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
21481 and niters_no_overflow parameters. Update calls to
21482 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
21483 Create a new SSA name if the latter choses to use a ste other
21484 than zero, and return it via niters_vector_mult_vf_var.
21485 * tree-vect-loop.c (vect_transform_loop): Update calls to
21486 vect_do_peeling, vect_gen_vector_loop_niters and
21487 slpeel_make_loop_iterate_ntimes.
21488 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
21489 (vect_gen_vector_loop_niters): Update declarations after above changes.
21491 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
21493 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
21494 128-bit round to integer instructions.
21495 (ceil<mode>2): Likewise.
21496 (btrunc<mode>2): Likewise.
21497 (round<mode>2): Likewise.
21499 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
21501 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
21502 unaligned VSX load/store on P8/P9.
21503 (expand_block_clear): Allow the use of unaligned VSX
21504 load/store on P8/P9.
21506 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21508 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
21510 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
21511 swap associated with both a load and a store.
21513 2018-01-02 Andrew Waterman <andrew@sifive.com>
21515 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
21516 * config/riscv/riscv.md (clear_cache): Use it.
21518 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
21520 * web.c: Remove out-of-date comment.
21522 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21524 * expr.c (fixup_args_size_notes): Check that any existing
21525 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
21526 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
21527 (emit_single_push_insn): ...here.
21529 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21531 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
21532 (const_vector_encoded_nelts): New function.
21533 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
21534 (const_vector_int_elt, const_vector_elt): Declare.
21535 * emit-rtl.c (const_vector_int_elt_1): New function.
21536 (const_vector_elt): Likewise.
21537 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
21538 of CONST_VECTOR_ELT.
21540 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21542 * expr.c: Include rtx-vector-builder.h.
21543 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
21544 directly on the tree encoding.
21545 (const_vector_from_tree): Likewise.
21546 * optabs.c: Include rtx-vector-builder.h.
21547 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
21548 sequence of "u" values.
21549 * vec-perm-indices.c: Include rtx-vector-builder.h.
21550 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
21551 directly on the vec_perm_indices encoding.
21553 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21555 * doc/rtl.texi (const_vector): Describe new encoding scheme.
21556 * Makefile.in (OBJS): Add rtx-vector-builder.o.
21557 * rtx-vector-builder.h: New file.
21558 * rtx-vector-builder.c: Likewise.
21559 * rtl.h (rtx_def::u2): Add a const_vector field.
21560 (CONST_VECTOR_NPATTERNS): New macro.
21561 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
21562 (CONST_VECTOR_DUPLICATE_P): Likewise.
21563 (CONST_VECTOR_STEPPED_P): Likewise.
21564 (CONST_VECTOR_ENCODED_ELT): Likewise.
21565 (const_vec_duplicate_p): Check for a duplicated vector encoding.
21566 (unwrap_const_vec_duplicate): Likewise.
21567 (const_vec_series_p): Check for a non-duplicated vector encoding.
21568 Say that the function only returns true for integer vectors.
21569 * emit-rtl.c: Include rtx-vector-builder.h.
21570 (gen_const_vec_duplicate_1): Delete.
21571 (gen_const_vector): Call gen_const_vec_duplicate instead of
21572 gen_const_vec_duplicate_1.
21573 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
21574 (gen_const_vec_duplicate): Use rtx_vector_builder.
21575 (gen_const_vec_series): Likewise.
21576 (gen_rtx_CONST_VECTOR): Likewise.
21577 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
21578 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
21579 Build a new vector rather than modifying a CONST_VECTOR in-place.
21580 (handle_special_swappables): Update call accordingly.
21581 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
21582 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
21583 Build a new vector rather than modifying a CONST_VECTOR in-place.
21584 (handle_special_swappables): Update call accordingly.
21586 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21588 * simplify-rtx.c (simplify_const_binary_operation): Use
21589 CONST_VECTOR_ELT instead of XVECEXP.
21591 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21593 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
21594 the selector elements to be different from the data elements
21595 if the selector is a VECTOR_CST.
21596 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
21597 ssizetype for the selector.
21599 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21601 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
21602 before testing each element individually.
21603 * tree-vect-generic.c (lower_vec_perm): Likewise.
21605 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21607 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
21608 * selftest-run-tests.c (selftest::run_tests): Call it.
21609 * vector-builder.h (vector_builder::operator ==): New function.
21610 (vector_builder::operator !=): Likewise.
21611 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
21612 (vec_perm_indices::all_from_input_p): New function.
21613 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
21614 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
21615 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
21616 instead of reading the VECTOR_CST directly. Detect whether both
21617 vector inputs are the same before constructing the vec_perm_indices,
21618 and update the number of inputs argument accordingly. Use the
21619 utility functions added above. Only construct sel2 if we need to.
21621 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21623 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
21624 the broadcast of the low byte.
21625 (expand_mult_highpart): Use an explicit encoding for the permutes.
21626 * optabs-query.c (can_mult_highpart_p): Likewise.
21627 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
21628 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
21629 (vectorizable_bswap): Likewise.
21630 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
21631 explicit encoding for the power-of-2 permutes.
21632 (vect_permute_store_chain): Likewise.
21633 (vect_grouped_load_supported): Likewise.
21634 (vect_permute_load_chain): Likewise.
21636 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21638 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
21639 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
21640 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
21641 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
21642 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
21643 (vect_gen_perm_mask_any): Likewise.
21645 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21647 * int-vector-builder.h: New file.
21648 * vec-perm-indices.h: Include int-vector-builder.h.
21649 (vec_perm_indices): Redefine as an int_vector_builder.
21650 (auto_vec_perm_indices): Delete.
21651 (vec_perm_builder): Redefine as a stand-alone class.
21652 (vec_perm_indices::vec_perm_indices): New function.
21653 (vec_perm_indices::clamp): Likewise.
21654 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
21655 (vec_perm_indices::new_vector): New function.
21656 (vec_perm_indices::new_expanded_vector): Update for new
21657 vec_perm_indices class.
21658 (vec_perm_indices::rotate_inputs): New function.
21659 (vec_perm_indices::all_in_range_p): Operate directly on the
21660 encoded form, without computing elided elements.
21661 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
21662 encoding. Update for new vec_perm_indices class.
21663 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
21664 the given vec_perm_builder.
21665 (expand_vec_perm_var): Update vec_perm_builder constructor.
21666 (expand_mult_highpart): Use vec_perm_builder instead of
21667 auto_vec_perm_indices.
21668 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
21669 vec_perm_indices instead of auto_vec_perm_indices. Use a single
21670 or double series encoding as appropriate.
21671 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
21672 vec_perm_indices instead of auto_vec_perm_indices.
21673 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
21674 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
21675 (vect_permute_store_chain): Likewise.
21676 (vect_grouped_load_supported): Likewise.
21677 (vect_permute_load_chain): Likewise.
21678 (vect_shift_permute_load_chain): Likewise.
21679 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
21680 (vect_transform_slp_perm_load): Likewise.
21681 (vect_schedule_slp_instance): Likewise.
21682 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
21683 (vectorizable_mask_load_store): Likewise.
21684 (vectorizable_bswap): Likewise.
21685 (vectorizable_store): Likewise.
21686 (vectorizable_load): Likewise.
21687 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
21688 vec_perm_indices instead of auto_vec_perm_indices. Use
21689 tree_to_vec_perm_builder to read the vector from a tree.
21690 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
21691 vec_perm_builder instead of a vec_perm_indices.
21692 (have_whole_vector_shift): Use vec_perm_builder and
21693 vec_perm_indices instead of auto_vec_perm_indices. Leave the
21694 truncation to calc_vec_perm_mask_for_shift.
21695 (vect_create_epilog_for_reduction): Likewise.
21696 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
21697 from auto_vec_perm_indices to vec_perm_indices.
21698 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
21699 instead of changing individual elements.
21700 (aarch64_vectorize_vec_perm_const): Use new_vector to install
21701 the vector in d.perm.
21702 * config/arm/arm.c (expand_vec_perm_d::perm): Change
21703 from auto_vec_perm_indices to vec_perm_indices.
21704 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
21705 instead of changing individual elements.
21706 (arm_vectorize_vec_perm_const): Use new_vector to install
21707 the vector in d.perm.
21708 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
21709 Update vec_perm_builder constructor.
21710 (rs6000_expand_interleave): Likewise.
21711 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
21712 (rs6000_expand_interleave): Likewise.
21714 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21716 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
21717 to qimode could truncate the indices.
21718 * optabs.c (expand_vec_perm_var): Likewise.
21720 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21722 * Makefile.in (OBJS): Add vec-perm-indices.o.
21723 * vec-perm-indices.h: New file.
21724 * vec-perm-indices.c: Likewise.
21725 * target.h (vec_perm_indices): Replace with a forward class
21727 (auto_vec_perm_indices): Move to vec-perm-indices.h.
21728 * optabs.h: Include vec-perm-indices.h.
21729 (expand_vec_perm): Delete.
21730 (selector_fits_mode_p, expand_vec_perm_var): Declare.
21731 (expand_vec_perm_const): Declare.
21732 * target.def (vec_perm_const_ok): Replace with...
21733 (vec_perm_const): ...this new hook.
21734 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
21735 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
21736 * doc/tm.texi: Regenerate.
21737 * optabs.def (vec_perm_const): Delete.
21738 * doc/md.texi (vec_perm_const): Likewise.
21739 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
21740 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
21741 expand_vec_perm for constant permutation vectors. Assert that
21742 the mode of variable permutation vectors is the integer equivalent
21743 of the mode that is being permuted.
21744 * optabs-query.h (selector_fits_mode_p): Declare.
21745 * optabs-query.c: Include vec-perm-indices.h.
21746 (selector_fits_mode_p): New function.
21747 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
21748 is defined, instead of checking whether the vec_perm_const_optab
21749 exists. Use targetm.vectorize.vec_perm_const instead of
21750 targetm.vectorize.vec_perm_const_ok. Check whether the indices
21751 fit in the vector mode before using a variable permute.
21752 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
21753 vec_perm_indices instead of an rtx.
21754 (expand_vec_perm): Replace with...
21755 (expand_vec_perm_const): ...this new function. Take the selector
21756 as a vec_perm_indices rather than an rtx. Also take the mode of
21757 the selector. Update call to shift_amt_for_vec_perm_mask.
21758 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
21759 Use vec_perm_indices::new_expanded_vector to expand the original
21760 selector into bytes. Check whether the indices fit in the vector
21761 mode before using a variable permute.
21762 (expand_vec_perm_var): Make global.
21763 (expand_mult_highpart): Use expand_vec_perm_const.
21764 * fold-const.c: Includes vec-perm-indices.h.
21765 * tree-ssa-forwprop.c: Likewise.
21766 * tree-vect-data-refs.c: Likewise.
21767 * tree-vect-generic.c: Likewise.
21768 * tree-vect-loop.c: Likewise.
21769 * tree-vect-slp.c: Likewise.
21770 * tree-vect-stmts.c: Likewise.
21771 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
21773 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
21774 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
21775 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
21776 (aarch64_vectorize_vec_perm_const): ...this new function.
21777 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
21778 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
21779 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
21780 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
21781 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
21782 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
21783 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
21785 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
21786 check for NEON modes.
21787 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
21788 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
21789 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
21790 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
21792 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
21793 the old VEC_PERM_CONST conditions.
21794 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
21795 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
21796 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
21797 (ia64_vectorize_vec_perm_const_ok): Merge into...
21798 (ia64_vectorize_vec_perm_const): ...this new function.
21799 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
21800 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
21801 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
21802 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
21803 * config/mips/mips.c (mips_expand_vec_perm_const)
21804 (mips_vectorize_vec_perm_const_ok): Merge into...
21805 (mips_vectorize_vec_perm_const): ...this new function.
21806 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
21807 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
21808 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
21809 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
21810 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
21811 (rs6000_expand_vec_perm_const): Delete.
21812 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
21814 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
21815 (altivec_expand_vec_perm_const_le): Take each operand individually.
21816 Operate on constant selectors rather than rtxes.
21817 (altivec_expand_vec_perm_const): Likewise. Update call to
21818 altivec_expand_vec_perm_const_le.
21819 (rs6000_expand_vec_perm_const): Delete.
21820 (rs6000_vectorize_vec_perm_const_ok): Delete.
21821 (rs6000_vectorize_vec_perm_const): New function.
21822 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
21823 an element count and rtx array.
21824 (rs6000_expand_extract_even): Update call accordingly.
21825 (rs6000_expand_interleave): Likewise.
21826 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
21827 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
21828 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
21829 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
21830 (rs6000_expand_vec_perm_const): Delete.
21831 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
21832 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
21833 (altivec_expand_vec_perm_const_le): Take each operand individually.
21834 Operate on constant selectors rather than rtxes.
21835 (altivec_expand_vec_perm_const): Likewise. Update call to
21836 altivec_expand_vec_perm_const_le.
21837 (rs6000_expand_vec_perm_const): Delete.
21838 (rs6000_vectorize_vec_perm_const_ok): Delete.
21839 (rs6000_vectorize_vec_perm_const): New function. Remove stray
21840 reference to the SPE evmerge intructions.
21841 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
21842 an element count and rtx array.
21843 (rs6000_expand_extract_even): Update call accordingly.
21844 (rs6000_expand_interleave): Likewise.
21845 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
21846 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
21848 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
21850 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21852 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
21853 vector mode and that that mode matches the mode of the data
21855 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
21856 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
21857 directly using expand_vec_perm_1 when forcing selectors into
21859 (expand_vec_perm_var): New function, split out from expand_vec_perm.
21861 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21863 * optabs-query.h (can_vec_perm_p): Delete.
21864 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
21865 * optabs-query.c (can_vec_perm_p): Split into...
21866 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
21867 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
21868 particular selector is valid.
21869 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
21870 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
21871 (vect_grouped_load_supported): Likewise.
21872 (vect_shift_permute_load_chain): Likewise.
21873 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
21874 (vect_transform_slp_perm_load): Likewise.
21875 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
21876 (vectorizable_bswap): Likewise.
21877 (vect_gen_perm_mask_checked): Likewise.
21878 * fold-const.c (fold_ternary_loc): Likewise. Don't take
21879 implementations of variable permutation vectors into account
21880 when deciding which selector to use.
21881 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
21882 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
21883 with a false third argument.
21884 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
21885 to test whether the constant selector is valid and can_vec_perm_var_p
21886 to test whether a variable selector is valid.
21888 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21890 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
21891 * optabs-query.c (can_vec_perm_p): Likewise.
21892 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
21893 instead of vec_perm_indices.
21894 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
21895 (vect_gen_perm_mask_checked): Likewise,
21896 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
21897 (vect_gen_perm_mask_checked): Likewise,
21899 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21901 * optabs-query.h (qimode_for_vec_perm): Declare.
21902 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
21903 (qimode_for_vec_perm): ...this new function.
21904 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
21906 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
21908 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
21909 does not have a conditional at the top.
21911 2018-01-02 Richard Biener <rguenther@suse.de>
21913 * ipa-inline.c (big_speedup_p): Fix expression.
21915 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
21918 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
21921 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
21925 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
21926 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
21927 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
21928 cond_taken_branch_cost 3->4.
21930 2018-01-01 Jakub Jelinek <jakub@redhat.com>
21932 PR tree-optimization/83581
21933 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
21934 TODO_cleanup_cfg if any changes have been made.
21936 PR middle-end/83608
21937 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
21938 convert_modes if target mode has the right side, but different mode
21941 PR middle-end/83609
21942 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
21943 last argument when extracting from CONCAT. If either from_real or
21944 from_imag is NULL, use expansion through memory. If result is not
21945 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
21946 the parts directly to inner mode, if even that fails, use expansion
21949 PR middle-end/83623
21950 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
21951 check for bswap in mode rather than HImode and use that in expand_unop
21954 Copyright (C) 2018 Free Software Foundation, Inc.
21956 Copying and distribution of this file, with or without modification,
21957 are permitted in any medium without royalty provided the copyright
21958 notice and this notice are preserved.