1 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3 * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
4 * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
6 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
8 * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
9 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
10 (vec_info::free_stmt_vec_info): New private member functions.
11 (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
12 (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
13 * tree-parloops.c (gather_scalar_reductions): Remove calls to
14 set_stmt_vec_info_vec and free_stmt_vec_infos.
15 * tree-vect-loop.c (_loop_vec_info): Remove call to
16 set_stmt_vec_info_vec.
17 * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
18 (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
19 * tree-vectorizer.c (vec_info::new_stmt_vec_info)
20 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
21 (vec_info::free_stmt_vec_info): ...these new functions. Remove
22 assignments in {vec_info::,}new_stmt_vec_info that are redundant
23 with the clearing in the xcalloc.
24 (stmt_vec_info_vec): Delete.
25 (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
26 (vectorize_loops): Likewise.
27 (vec_info::~vec_info): Remove argument from call to
29 (vec_info::add_stmt): Remove vinfo argument from call to
32 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
34 * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
35 rather than a gimple stmt.
36 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. Don't free
37 information for pattern statements when passed the original
38 statement; instead wait to be passed the pattern statement itself.
39 Don't call set_vinfo_for_stmt here.
40 (free_stmt_vec_infos): Update call to free_stmt_vec_info.
41 * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
43 * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
44 * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
47 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
49 * tree-vectorizer.h (vec_info::replace_stmt): Declare.
50 * tree-vectorizer.c (vec_info::replace_stmt): New function.
51 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
52 * tree-vect-stmts.c (vectorizable_call): Likewise.
53 (vectorizable_simd_clone_call): Likewise.
55 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
57 * tree-vectorizer.h (vec_info::remove_stmt): Declare.
58 * tree-vectorizer.c (vec_info::remove_stmt): New function.
59 * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
60 * tree-vect-loop.c (vect_transform_loop): Likewise.
61 * tree-vect-slp.c (vect_schedule_slp): Likewise.
62 * tree-vect-stmts.c (vect_remove_stores): Likewise.
64 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
66 * tree-vectorizer.h (vec_info::lookup_dr): New member function.
67 (vect_dr_stmt): Delete.
68 * tree-vectorizer.c (vec_info::lookup_dr): New function.
69 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
71 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
72 (vect_analyze_data_ref_dependence, vect_record_base_alignments)
73 (vect_verify_datarefs_alignment, vect_peeling_supportable)
74 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
75 (vect_analyze_data_refs): Likewise.
76 (vect_slp_analyze_data_ref_dependence): Likewise. Take a vec_info
78 (vect_find_same_alignment_drs): Likewise.
79 (vect_slp_analyze_node_dependences): Update calls accordingly.
80 (vect_analyze_data_refs_alignment): Likewise. Use vec_info::lookup_dr
81 instead of DR_VECT_AUX.
82 (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
83 of a vector data references. Use vec_info::lookup_dr instead of
85 (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
86 (vect_enhance_data_refs_alignment): Likewise. Use vec_info::lookup_dr
87 instead of DR_VECT_AUX.
89 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
91 * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
93 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
95 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
96 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
97 (vect_gen_prolog_loop_niters): Likewise.
99 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
101 * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
102 (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
103 (vect_known_alignment_in_bytes, vect_dr_behavior)
104 (vect_get_scalar_dr_size): Take references as dr_vec_infos
105 instead of data_references. Update calls to other routines for
106 which the same change has been made.
107 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
108 dr_vec_infos instead of stmt_vec_infos.
109 (vect_analyze_data_ref_dependence): Update call accordingly.
110 (vect_slp_analyze_data_ref_dependence)
111 (vect_record_base_alignments): Use DR_VECT_AUX.
112 (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
113 (vect_update_misalignment_for_peel, verify_data_ref_alignment)
114 (vector_alignment_reachable_p, vect_get_data_access_cost)
115 (vect_peeling_supportable, vect_analyze_group_access_1)
116 (vect_analyze_group_access, vect_analyze_data_ref_access)
117 (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
118 (vect_compile_time_alias, vect_small_gap_p)
119 (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
120 (vect_supportable_dr_alignment): Take references as dr_vec_infos
121 instead of data_references. Update calls to other routines for
122 which the same change has been made.
123 (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
124 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
125 (vect_slp_analyze_and_verify_node_alignment)
126 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
127 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
128 (vect_setup_realignment): Use dr_vec_infos. Update calls after
130 (_vect_peel_info::dr): Replace with...
131 (_vect_peel_info::dr_info): ...this new field.
132 (vect_peeling_hash_get_most_frequent)
133 (vect_peeling_hash_choose_best_peeling): Update accordingly.
134 (vect_peeling_hash_get_lowest_cost):
135 (vect_enhance_data_refs_alignment): Likewise. Update calls to other
136 routines for which the same change has been made.
137 (vect_peeling_hash_insert): Likewise. Take a dr_vec_info instead of a
139 * tree-vect-loop-manip.c (get_misalign_in_elems)
140 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after
142 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
143 * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
144 (vect_truncate_gather_scatter_offset, compare_step_with_zero)
145 (get_group_load_store_type, get_negative_load_store_type)
146 (vect_get_data_ptr_increment, vectorizable_store)
147 (vectorizable_load): Likewise.
148 (ensure_base_align): Take a dr_vec_info instead of a data_reference.
149 Update calls to other routines for which the same change has been made.
151 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
153 * tree-vectorizer.h (vec_info::move_dr): New member function.
154 (dataref_aux): Rename to...
155 (dr_vec_info): ...this and add "dr" and "stmt" fields.
156 (_stmt_vec_info::dr_aux): Update accordingly.
157 (_stmt_vec_info::data_ref_info): Delete.
158 (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
159 (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
160 (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
161 (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
163 (STMT_VINFO_DATA_REF): Likewise. Turn into an lvalue.
164 (STMT_VINFO_DR_INFO): New macro.
165 (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
166 (set_dr_misalignment): Update after rename of dataref_aux.
167 (vect_dr_stmt): Move earlier in file. Return dr_aux.stmt.
168 * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
169 initialization of STMT_VINFO_DATA_REF.
170 * tree-vectorizer.c (vec_info::move_dr): New function.
171 * tree-vect-patterns.c (vect_recog_bool_pattern)
172 (vect_recog_mask_conversion_pattern)
173 (vect_recog_gather_scatter_pattern): Use it.
174 * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
175 the "dr" and "stmt" fields of dr_vec_info instead of
178 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
180 * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
181 (is_pattern_stmt_p): Use it.
182 * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
183 on pattern statements.
185 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
187 * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
188 original stmt as a stmt_vec_info rather than a gimple stmt.
189 (vect_pattern_recog_1): Take the statement directly as a
190 stmt_vec_info, rather than via a gimple_stmt_iterator.
191 Update call to vect_mark_pattern_stmts.
192 (vect_pattern_recog): Update calls accordingly.
194 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
196 * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
197 (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
198 a vect_def_type for the first argument.
199 * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
200 (vect_get_vec_def_for_stmt_copy): Likewise. Return the original
201 operand if it isn't defined by a vectorized statement.
202 (vect_build_gather_load_calls): Remove the mask_dt argument and
203 update calls to vect_get_vec_def_for_stmt_copy.
204 (vectorizable_bswap): Likewise the dt argument.
205 (vectorizable_call): Update calls to vectorizable_bswap and
206 vect_get_vec_def_for_stmt_copy.
207 (vectorizable_simd_clone_call, vectorizable_assignment)
208 (vectorizable_shift, vectorizable_operation, vectorizable_condition)
209 (vectorizable_comparison): Update calls to
210 vect_get_vec_def_for_stmt_copy.
211 (vectorizable_store): Likewise. Remove now-unnecessary calls to
213 (vect_get_loop_based_defs): Remove dt argument and update call
214 to vect_get_vec_def_for_stmt_copy.
215 (vectorizable_conversion): Update calls to vect_get_loop_based_defs
216 and vect_get_vec_def_for_stmt_copy.
217 (vectorizable_load): Update calls to vect_build_gather_load_calls
218 and vect_get_vec_def_for_stmt_copy.
219 * tree-vect-loop.c (vect_create_epilog_for_reduction)
220 (vectorizable_reduction, vectorizable_live_operation): Update calls
221 to vect_get_vec_def_for_stmt_copy.
223 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
225 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
226 and gimple stmt arguments with a stmt_vec_info.
227 (vect_record_base_alignments): Update calls accordingly.
228 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
229 and gimple stmt arguments with a stmt_vec_info.
230 (vect_build_slp_tree_1): Remove vinfo argument and update call
231 to vect_record_max_nunits.
232 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
233 and vect_record_max_nunits.
235 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
237 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
238 file and take a stmt_vec_info instead of a gimple stmt.
239 (supportable_widening_operation, vect_finish_replace_stmt)
240 (vect_finish_stmt_generation, vect_get_store_rhs)
241 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
242 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
243 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
244 (vect_get_smallest_scalar_type, vect_check_gather_scatter)
245 (vect_create_data_ref_ptr, bump_vector_ptr)
246 (vect_permute_store_chain, vect_setup_realignment)
247 (vect_transform_grouped_load, vect_record_grouped_load_vectors)
248 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
249 (vectorizable_reduction, vectorizable_induction)
250 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
251 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
252 than gimple stmts as arguments.
253 * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
254 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
255 (can_group_stmts_p, vect_check_gather_scatter)
256 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
257 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
258 (vect_permute_load_chain, vect_shift_permute_load_chain)
259 (vect_transform_grouped_load)
260 (vect_record_grouped_load_vectors): Likewise.
261 * tree-vect-loop.c (vect_fixup_reduc_chain)
262 (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
263 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
264 (vectorizable_reduction, vectorizable_induction)
265 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
266 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
267 (vect_get_load_store_mask): Likewise.
268 * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
269 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
270 * tree-vect-stmts.c (vect_mark_relevant)
271 (is_simple_and_all_uses_invariant)
272 (exist_non_indexing_operands_for_use_p, process_use)
273 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
274 (vect_get_vec_def_for_operand, vect_get_vec_defs)
275 (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
276 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
277 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
278 (get_negative_load_store_type, get_load_store_type)
279 (vect_check_load_store_mask, vect_check_store_rhs)
280 (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
281 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
282 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
283 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
284 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
285 (vectorizable_load, vectorizable_condition, vectorizable_comparison)
286 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
287 (supportable_widening_operation): Likewise.
289 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
291 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
292 a stmt_vec_info instead of a gcall.
293 (vect_check_gather_scatter): Update call accordingly.
294 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
296 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
297 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
298 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
299 instead of a gimple stmt.
300 (vect_transform_loop): Update calls accordingly.
301 * tree-vect-slp.c (vect_split_slp_store_group): Take and return
302 stmt_vec_infos instead of gimple stmts.
303 (vect_analyze_slp_instance): Update use accordingly.
304 * tree-vect-stmts.c (read_vector_array, write_vector_array)
305 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
306 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
307 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
308 (vect_gen_widened_results_half, vect_get_loop_based_defs)
309 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
310 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
313 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
315 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
316 of the worklist from a vector of gimple stmts to a vector of
318 * tree-vect-stmts.c (vect_mark_relevant, process_use)
319 (vect_mark_stmts_to_be_vectorized): Likewise
321 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
323 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
324 statement before passing it to vect_analyze_stmt.
325 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
326 the chain of phi vector definitions. Track the exit phi via its
328 (vectorizable_reduction): Set cond_stmt_vinfo directly from the
329 STMT_VINFO_REDUC_DEF.
330 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
331 stmt_vec_infos to handle the statement chains.
332 (vect_get_slp_defs): Record the first statement in the node
333 using a stmt_vec_info.
334 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
335 statements here and pass their stmt_vec_info down to subroutines.
336 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
337 down to vect_finish_stmt_generation.
338 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
339 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
340 stmt_vec_infos to subroutines.
341 (vect_remove_stores): Use stmt_vec_infos to handle the statement
344 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
346 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
347 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
348 (vect_permute_store_chain, vect_setup_realignment)
349 (vect_permute_load_chain, vect_shift_permute_load_chain)
350 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
351 stmts internally, and when passing values to other vectorizer routines.
352 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
353 * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
354 (vect_analyze_loop_operations, get_initial_def_for_reduction)
355 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
356 (vectorizable_reduction, vectorizable_induction)
357 (vectorizable_live_operation, vect_transform_loop_stmt)
358 (vect_transform_loop): Likewise.
359 * tree-vect-patterns.c (vect_reassociating_reduction_p)
360 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
361 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
362 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
363 (vect_slp_analyze_node_operations_1): Likewise.
364 * tree-vect-stmts.c (vect_mark_relevant, process_use)
365 (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
366 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
367 (vect_finish_stmt_generation_1, get_group_load_store_type)
368 (get_load_store_type, vect_build_gather_load_calls)
369 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
370 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
371 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
372 (vectorizable_store, vectorizable_load, vectorizable_condition)
373 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
374 (supportable_widening_operation): Likewise.
375 (vect_get_vector_types_for_stmt): Likewise.
376 * tree-vectorizer.h (vect_dr_behavior): Likewise.
378 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
380 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
381 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
382 (vect_permute_store_chain, vect_permute_load_chain)
383 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
384 repeated stmt_vec_info lookups.
385 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
386 (vect_update_ivs_after_vectorizer): Likewise.
387 * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
388 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
389 * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
390 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
391 (vect_bb_slp_scalar_cost): Likewise.
392 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
394 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
396 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
397 gcall rather than the generic gimple stmt to gimple_call_internal_fn.
398 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
399 to get gassigns and gcalls, rather than operating on generc gimple
401 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
402 (vect_mark_stmts_to_be_vectorized, vectorizable_store)
403 (vectorizable_load, vect_analyze_stmt): Likewise.
404 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
406 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
408 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
409 return stmt_vec_infos rather than gimple stmts. Do not accept
411 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
413 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
414 Update use of get_later_stmt.
415 (vect_get_constant_vectors): Update call accordingly.
416 (vect_schedule_slp_instance): Likewise
417 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
418 (vect_slp_analyze_instance_dependence): Likewise.
419 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
421 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
423 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
424 (stmt_info_for_cost::stmt_info): ...this new field.
425 (add_stmt_costs): Update accordingly.
426 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
427 (vect_get_known_peeling_cost): Likewise.
428 (vect_estimate_min_profitable_iters): Likewise.
429 * tree-vect-stmts.c (record_stmt_cost): Likewise.
431 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
433 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
434 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
435 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
437 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
439 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
441 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
442 a gimple stmt to a stmt_vec_info.
443 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
445 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
447 * tree-vectorizer.h (vec_info::grouped_stores): Change from
448 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
449 (_loop_vec_info::reduction_chains): Likewise.
450 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
452 * tree-vect-slp.c (vect_analyze_slp): Likewise.
454 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
456 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
457 a gimple stmt to a stmt_vec_info.
458 (_stmt_vec_info::next_element): Likewise.
459 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
460 (vect_slp_analyze_and_verify_node_alignment)
461 (vect_analyze_group_access_1, vect_analyze_group_access)
462 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
463 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
464 (vect_supportable_dr_alignment): Update accordingly.
465 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
466 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
467 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
468 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
469 * tree-vect-slp.c (vect_build_slp_tree_1)
470 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
471 (vect_split_slp_store_group, vect_analyze_slp_instance)
472 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
473 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
474 (get_group_load_store_type, get_load_store_type)
475 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
476 (vect_transform_stmt, vect_remove_stores): Likewise.
478 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
480 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
482 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
483 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
484 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
485 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
486 (vector_alignment_reachable_p, vect_get_data_access_cost)
487 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
488 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
489 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
490 (vect_analyze_group_access_1, vect_analyze_group_access)
491 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
492 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
493 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
494 result of vect_dr_stmt and use the stmt_vec_info instead of
495 the associated gimple stmt.
496 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
497 (vect_gen_prolog_loop_niters): Likewise.
498 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
500 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
502 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
503 to a vec<stmt_vec_info>.
504 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
505 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
507 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
508 to a vec<stmt_vec_info>.
509 (bst_traits::value_type, bst_traits::value_type): Likewise.
510 (bst_traits::hash): Update accordingly.
511 (vect_get_and_check_slp_defs): Change the stmts parameter from
512 a vec<gimple *> to a vec<stmt_vec_info>.
513 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
514 (vect_build_slp_tree): Likewise.
515 (vect_build_slp_tree_2): Likewise. Update uses of
516 SLP_TREE_SCALAR_STMTS.
517 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
518 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
519 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
520 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
521 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
522 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
523 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
524 (vect_get_constant_vectors, vect_get_slp_defs)
525 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
526 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
527 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
528 instead of gimple stmts.
529 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
530 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
531 (vect_slp_analyze_instance_dependence): Update uses of
532 SLP_TREE_SCALAR_STMTS.
533 (vect_slp_analyze_and_verify_node_alignment): Likewise.
534 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
535 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
536 (get_initial_defs_for_reduction): Likewise.
537 (vect_create_epilog_for_reduction): Likewise.
538 (vectorize_fold_left_reduction): Likewise.
539 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
540 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
541 (can_vectorize_live_stmts): Likewise.
543 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
545 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
546 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
547 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
549 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
551 (gather_scalar_reductions): Update after above interface changes.
552 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
553 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
555 (vect_force_simple_reduction): Likewise.
556 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
557 LOOP_VINFO_REDUCTIONS.
558 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
560 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
562 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
563 a gimple stmt to a stmt_vec_info.
564 * tree-vect-loop.c (vect_active_double_reduction_p)
565 (vect_force_simple_reduction, vectorizable_reduction): Update
568 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
570 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
571 vec<gimple *> to a vec<stmt_vec_info>.
572 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
573 the reduction_phis argument from a vec<gimple *> to a
575 (vectorizable_reduction): Likewise the phis local variable that
576 is passed to vect_create_epilog_for_reduction. Update for new type
577 of SLP_TREE_VEC_STMTS.
578 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
579 (vectorizable_live_operation): Likewise.
580 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
581 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
583 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
585 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
586 a gimple stmt to a stmt_vec_info.
587 (vectorizable_condition, vectorizable_live_operation)
588 (vectorizable_reduction, vectorizable_induction): Pass back the
589 vectorized statement as a stmt_vec_info.
590 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
591 use of STMT_VINFO_VEC_STMT.
592 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
593 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
594 as stmt_vec_infos rather than gimple stmts.
595 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
597 (vectorizable_live_operation): Likewise.
598 (vectorizable_reduction, vectorizable_induction): Likewise,
599 updating use of STMT_VINFO_VEC_STMT.
600 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
601 of STMT_VINFO_VEC_STMT.
602 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
603 (vectorizable_simd_clone_call, vectorizable_conversion)
604 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
605 (vectorizable_store, vectorizable_load, vectorizable_condition)
606 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
607 from a gimple stmt to a stmt_vec_info.
608 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
609 pointer to a stmt_vec_info to the vectorizable_* routines.
611 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
613 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
614 a gimple stmt to a stmt_vec_info.
615 (is_pattern_stmt_p): Update accordingly.
616 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
617 (vect_record_grouped_load_vectors): Likewise.
618 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
619 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
620 (vect_model_reduction_cost): Likewise.
621 (vect_create_epilog_for_reduction): Likewise.
622 (vectorizable_reduction, vectorizable_induction): Likewise.
623 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
624 Return the stmt_vec_info for the pattern statement.
625 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
626 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
627 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
628 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
629 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
630 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
631 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
632 (free_stmt_vec_info, vect_is_simple_use): Likewise.
634 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
636 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
637 (vect_finish_stmt_generation): Likewise.
638 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
639 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
640 (vect_build_gather_load_calls): Use the return value of the above
641 functions instead of a separate call to vinfo_for_stmt. Use narrow
642 scopes for the input gimple stmt and wider scopes for the associated
643 stmt_vec_info. Use vec_info::lookup_def when setting these
644 stmt_vec_infos from an SSA_NAME definition.
645 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
646 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
647 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
648 (vectorizable_store, vectorizable_load, vectorizable_condition)
649 (vectorizable_comparison): Likewise.
650 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
651 (vectorizable_reduction): Likewise.
653 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
655 * tree-vectorizer.h (vect_is_simple_use): Add an optional
656 stmt_vec_info * parameter before the optional gimple **.
657 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
658 (process_use, vect_get_vec_def_for_operand_1): Update callers.
659 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
660 * tree-vect-loop.c (vectorizable_reduction): Likewise.
661 (vectorizable_live_operation): Likewise.
662 * tree-vect-patterns.c (type_conversion_p): Likewise.
663 (vect_look_through_possible_promotion): Likewise.
664 (vect_recog_rotate_pattern): Likewise.
665 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
667 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
669 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
670 a typedef to a wrapper class.
671 (NULL_STMT_VEC_INFO): New macro.
672 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
673 (stmt_vec_info::operator*): New function.
674 (stmt_vec_info::operator gimple *): Likewise.
675 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
676 (add_stmt_costs): Likewise.
677 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
678 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
679 (vect_get_known_peeling_cost): Likewise.
680 (vect_estimate_min_profitable_iters): Likewise.
681 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
682 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
683 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
684 (vectorizable_store, free_stmt_vec_infos): Likewise.
685 (new_stmt_vec_info): Change return type of xcalloc to
688 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
690 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
691 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
692 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
693 a single_imm_use-based sequence.
694 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
696 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
698 * tree-vectorizer.h (vec_info::lookup_def): Declare.
699 * tree-vectorizer.c (vec_info::lookup_def): New function.
700 * tree-vect-patterns.c (vect_get_internal_def): Use it.
701 (vect_widened_op_tree): Likewise.
702 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
703 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
704 (vectorizable_reduction): Likewise.
705 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
707 (vect_is_slp_reduction): Update calls accordingly. Use
708 vec_info::lookup_def.
709 (vect_is_simple_reduction): Likewise
710 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
712 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
714 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
715 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
716 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
718 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
719 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
720 (vect_update_vf_for_slp, vect_analyze_loop_operations)
721 (vect_is_slp_reduction, vectorizable_induction)
722 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
723 * tree-vect-patterns.c (vect_init_pattern_stmt):
724 (vect_determine_min_output_precision_1, vect_determine_precisions)
725 (vect_pattern_recog): Likewise.
726 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
727 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
728 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
729 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
730 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
731 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
732 info field from a loop to a loop_vec_info.
734 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
736 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
737 (vec_info::add_stmt): Declare.
738 * tree-vectorizer.c (vec_info::add_stmt): New function.
739 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
740 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
741 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
742 (vectorizable_induction): Likewise.
743 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
744 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
745 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
746 (vectorizable_load): Likewise.
747 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
748 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
749 (vect_recog_gather_scatter_pattern): Likewise.
750 (append_pattern_def_seq): Likewise. Remove a check that is
751 performed by add_stmt itself.
753 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
755 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
756 which make_ssa_name was called with new_stmt before new_stmt
759 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
761 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
763 (vect_is_slp_reduction): ...here...
764 (vect_is_simple_reduction): ...and here. Remove repetition of tests
765 that are already known to be false.
767 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
769 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
770 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
771 STMT_VINFO_NUM_SLP_USES when it's true.
772 (vect_free_slp_instance): Add a final_p parameter and pass it to
774 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
775 (vect_analyze_slp_instance): Likewise.
776 (vect_slp_analyze_operations): Likewise.
777 (vect_slp_analyze_bb_1): Likewise.
778 * tree-vectorizer.c (vec_info): Likewise.
779 * tree-vect-loop.c (vect_transform_loop): Likewise.
781 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
783 * tree-vect-loop.c (vectorizable_reduction): Assert that the
784 function is not called for second and subsequent members of
787 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
789 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
790 cases for nested loops from here to ...
791 (vect_create_epilog_for_reduction): ...here. Only call
792 vect_is_simple_use for inner-loop reductions.
794 2018-07-31 Martin Liska <mliska@suse.cz>
796 PR gcov-profile/85338
797 PR gcov-profile/85350
798 PR gcov-profile/85372
799 * profile.c (struct location_triplet): New.
800 (struct location_triplet_hash): Likewise.
801 (output_location): Do not output a BB that
802 is already recorded for a line.
803 (branch_prob): Use streamed_locations.
805 2018-07-31 Martin Liska <mliska@suse.cz>
807 PR gcov-profile/85370
808 * coverage.c (coverage_begin_function): Do not mark target
809 clones as artificial functions.
811 2018-07-31 Martin Liska <mliska@suse.cz>
813 PR gcov-profile/83813
814 PR gcov-profile/84758
815 PR gcov-profile/85217
816 PR gcov-profile/85332
817 * profile.c (branch_prob): Do not record GOTO expressions
818 for GIMPLE statements which locations are already streamed.
820 2018-07-31 Olivier Hainque <hainque@adacore.com>
822 * gcc.c (handle_spec_function): Accept a soft_matched_part
823 argument, as do_spec_1. Pass it down to ...
824 (eval_spec_function): Accept a soft_matched_part argument,
825 and pass it down to ...
826 (do_spec_2): Accept a soft_matched_part argument, and pass
827 it down to do_spec_1.
828 (do_spec_1): Pass soft_matched_part to handle_spec_function.
829 (handle_braces): Update call to handle_spec_function.
830 (driver::set_up_specs): Update calls to do_spec_2.
831 (compare_debug_dump_opt_spec_function): Likewise.
832 (compare_debug_self_opt_spec_function): Likewise.
834 2018-07-31 Olivier Hainque <hainque@adacore.com>
836 * common.opt (nolibc): New option.
837 * doc/invoke.texi (Link Options): Document it.
838 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
839 * config/alpha/linux.h: Likewise.
840 * config/arc/elf.h: Likewise.
841 * config/arm/uclinux-elf.h: Likewise.
842 * config/arm/unknown-elf.h: Likewise.
843 * config/avr/avrlibc.h: Likewise.
844 * config/bfin/bfin.h: Likewise.
845 * config/bfin/linux.h: Likewise.
846 * config/bfin/uclinux.h: Likewise.
847 * config/darwin.h: Likewise.
848 * config/darwin10.h: Likewise.
849 * config/darwin12.h: Likewise.
850 * config/gnu-user.h: Likewise.
851 * config/lm32/uclinux-elf.h: Likewise.
852 * config/pa/pa-hpux11.h: Likewise.
853 * config/pa/pa64-hpux.h: Likewise.
854 * config/sparc/sparc.h: Likewise.
856 2018-07-31 Olivier Hainque <hainque@adacore.com>
858 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
861 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
864 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
867 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
869 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
870 terminated string literal.
872 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
874 PR rtl-optimization/85160
875 * combine.c (is_just_move): New function.
876 (try_combine): Allow combining two instructions into two if neither of
877 the original instructions was a move.
879 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
882 * doc/extend.texi (Global Register Variables): Discourage use of type
884 (Local Register Variables): Likewise.
886 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
888 PR tree-optimization/86506
889 * hwint.h (ceil_log2): Resync with hwint.c implementation.
891 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
894 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
895 hard_regno, make sure no insn between `from` and `to` clobbers it.
897 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
898 Tom de Vries <tdevries@suse.de>
900 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
901 (PTX_DEFAULT_RUNTIME_DIM): ... this.
902 (nvptx_goacc_validate_dims): Set default worker and gang dims to
903 PTX_DEFAULT_RUNTIME_DIM.
904 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
906 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
908 * config/pa/pa.c (pa_output_addr_vec): Align address table.
909 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
910 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
912 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
914 * config/rs6000/constraints.md (wG constraint): Delete, no longer
916 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
917 predicate to reflect toc fusion has been deleted.
918 (toc_fusion_mem_raw): Delete, no longer used.
919 (toc_fusion_mem_wrapped): Likewise.
920 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
922 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
923 Delete, no longer used.
924 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
925 meant to be used for toc fusion.
926 (rs6000_debug_print_mode): Delete toc fusion debugging.
927 (rs6000_debug_reg_global): Likewise.
928 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
929 fusion and secondary reload support that were never used.
930 (rs6000_option_override_internal): Delete TOC fusion, that was only
931 partially defined, and it did not work unless you also used the
933 (rs6000_legitimate_address_p): Delete TOC fusion support.
934 (rs6000_opt_masks): Likewise.
935 (fusion_wrap_memory_address): Delete function, no longer used.
936 (fusion_split_address); Delete TOC fusion support.
937 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
938 longer used with toc fusion being deleted.
939 (TARGET_TOC_FUSION_FP): Likewise.
940 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
942 (toc fusion spliter): Delete TOC fusion support.
943 (toc_fusionload_<mode>): Likewise.
944 (toc_fusionload_di): Likewise.
945 (fusion_gpr_load_<mode>): Delete generator function, this insn no
946 longer needs to be named. Rename predicate to delete TOC fusion.
947 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
948 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
949 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
950 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
951 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
953 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
955 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
956 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
957 __int128 in built-in function prototypes.
958 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
959 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
961 2018-07-27 Martin Sebor <msebor@redhat.com>
963 PR tree-optimization/86696
964 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
965 types, including enums.
966 (handle_char_store): Be prepared for the above function to fail.
968 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
970 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
971 when optimization level is lower than 2 or optimize for size.
973 2018-07-26 Martin Sebor <msebor@redhat.com>
975 PR tree-optimization/86043
976 PR tree-optimization/86042
977 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
978 (get_string_cst_length): Rename...
979 (get_min_string_length): ...to this. Add argument.
980 (handle_char_store): Extend to handle multi-character stores by
982 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
983 * tree.h (initializer_zerop): Add argument.
985 2018-07-26 Jakub Jelinek <jakub@redhat.com>
988 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
989 declare target to variables if they have always,{to,from,tofrom} map
992 2018-07-26 Martin Liska <mliska@suse.cz>
995 * lto-wrapper.c: Add linker_output as prefix
996 for ltrans_output_file.
998 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
1000 PR rtl-optimization/85805
1001 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
1002 value for hard registers if that was written in the same mode.
1004 2018-07-26 Martin Liska <mliska@suse.cz>
1006 PR gcov-profile/86536
1007 * gcov.c (format_gcov): Use printf format %.*f directly
1008 and do not handle special values.
1010 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
1012 * common/config/arc/arc-common.c (arc_option_optimization_table):
1013 Update default optimizations for size.
1015 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
1017 * config/arc/arc.md (movsf_insn): Add short instruction selection.
1018 * config/arc/constraints.md (CfZ): New constraint.
1019 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
1020 (subsf3_fpu): Likewise.
1021 (cmpsf_fpu): Likewise.
1022 (cmpsf_fpu_uneq): Likewise.
1024 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
1026 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
1028 (arc_is_uncached_mem_p): uncached applies to both the variable and
1031 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
1033 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
1036 2018-07-25 David Malcolm <dmalcolm@redhat.com>
1038 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
1039 field "m_scopes" from vec to auto_vec.
1041 2018-07-25 Martin Liska <mliska@suse.cz>
1043 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
1046 2018-07-25 Richard Biener <rguenther@suse.de>
1049 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
1050 special wrt context_die late.
1051 (gen_subprogram_die): Re-use DIEs in local scope.
1053 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
1055 PR tree-optimization/86644
1056 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
1058 2018-07-25 Martin Liska <mliska@suse.cz>
1061 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
1062 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
1064 2018-07-25 Martin Liska <mliska@suse.cz>
1067 * params.def: Explain ASan abbreviation and provide
1068 a documentation link.
1070 2018-07-24 Martin Sebor <msebor@redhat.com>
1072 PR tree-optimization/86622
1073 PR tree-optimization/86532
1074 * builtins.h (string_length): Declare.
1075 * builtins.c (c_strlen): Correct handling of non-constant offsets.
1076 (check_access): Be prepared for non-constant length ranges.
1077 (string_length): Make extern.
1078 * expr.c (string_constant): Only handle the minor non-constant
1079 array index. Use string_constant to compute the length of
1080 a generic string constant.
1082 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
1084 PR tree-optimization/86618
1085 * tree-vect-stmts.c (vectorizable_call): Don't take the address
1086 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
1088 2018-07-24 David Malcolm <dmalcolm@redhat.com>
1090 PR tree-optimization/86636
1091 * json.cc (json::object::set): Fix comment. Add assertions.
1092 (json::array::append): Move here from json.h. Add comment and an
1094 (json::string::string): Likewise.
1095 * json.h (json::array::append): Move to json.cc.
1096 (json::string::string): Likewise.
1097 * optinfo-emit-json.cc
1098 (optrecord_json_writer::impl_location_to_json): Assert that we
1099 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
1100 wrapper around it. Expand the location once, rather than three
1102 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
1103 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
1105 (optrecord_json_writer::optinfo_to_json): Likewise, in four
1106 places. Fix some overlong lines.
1108 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
1110 * config/aarch64/aarch64-simd.md
1111 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
1112 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
1113 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
1114 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
1115 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
1116 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
1117 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
1118 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
1119 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
1121 2018-07-24 Jakub Jelinek <jakub@redhat.com>
1124 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
1125 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
1126 and abs_d == d, do the power of two handling if profitable.
1128 2018-07-24 Richard Biener <rguenther@suse.de>
1130 * match.pd: Add BIT_FIELD_REF canonicalizations.
1132 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
1135 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
1137 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
1139 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
1140 terminated STRING_CST object.
1142 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
1144 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
1146 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
1148 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
1149 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
1150 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
1151 the elements into a register.
1152 (rs6000_split_v4si_init_di_reg): Delete.
1153 (rs6000_split_v4si_init): Delete.
1154 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
1155 (vsx_init_v4si): Rewrite as a define_expand.
1157 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
1159 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
1160 zero_extend argument from memory): New.
1162 2018-07-22 Martin Sebor <msebor@redhat.com>
1165 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
1166 diagnosing calls with unknown arguments unless -Walloca-larger-than
1167 is restricted to less than PTRDIFF_MAX bytes.
1169 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
1171 * doc/gcov.texi (Invoking Gcov): Editorial changes.
1173 2018-07-20 David Malcolm <dmalcolm@redhat.com>
1175 * pretty-print.c (text_info::set_location): Remove redundant
1176 "line_table" parameter from call to rich_location::set_range.
1178 2018-07-20 Martin Sebor <msebor@redhat.com>
1181 * builtins.c (expand_builtin_alloca): Adjust.
1182 * calls.c (alloc_max_size): Simplify.
1183 * cgraphunit.c (cgraph_node::expand): Adjust.
1184 * common.opt (larger_than_size, warn_frame_larger_than): Remove
1186 (frame_larger_than_size): Same.
1187 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
1188 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
1190 * doc/invoke.texi (GCC Command Options): Document option arguments.
1191 Explain byte-size arguments and suffixes.
1192 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
1193 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
1194 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
1195 * doc/options.texi (UInteger): Expand.
1196 (Host_Wide_Int, ByteSize): Document new properties.
1197 * final.c (final_start_function_1): Include sizes in an error message.
1198 * function.c (frame_offset_overflow): Same.
1199 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
1200 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
1201 Diagnose unbounded alloca calls only for limits of less than
1203 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
1204 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
1206 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
1207 only for limits of less than PTRDIFF_MAX.
1208 * langhooks-def.h (lhd_handle_option): Change function argument
1210 * langhooks.c (lhd_handle_option): Same.
1211 * langhooks.h (handle_option): Same.
1212 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
1214 (var_type, var_type_struct): Same.
1215 (var_set): Handle ByteSize flag.
1216 * optc-gen.awk: Add comments to output to ease debugging. Make
1217 use of HOST_WIDE_INT where appropriate.
1218 * opts-gen-save.awk: Use %lx to format unsigned long.
1219 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
1220 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
1221 arguments. Parse bytes-size suffixes.
1222 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
1223 (enum_value_to_arg): Same.
1224 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
1225 (handle_option): Adjust.
1226 (generate_option): Change function argument to HOST_WIDE_INT.
1227 (cmdline_handle_error): Adjust.
1228 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
1229 (set_option): Change function argument to HOST_WIDE_INT.
1230 (option_enabled): Handle cl_host_wide_int.
1231 (get_option_state): Handle CLVC_SIZE.
1232 (control_warning_option): Same.
1233 * opts.c (common_handle_option): Change function argument to
1234 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
1235 OPT_Wvla_larger_than_.
1236 * opts.h (enum cl_var_type): Add an enumerator.
1237 * stor-layout.c (layout_decl): Print a more meaningful warning.
1238 * toplev.c (output_stack_usage): Adjust.
1240 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
1242 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
1243 call to inline_expand_builtin_string_cmp.
1244 (expand_builtin_strcmp): Likewise.
1245 (expand_builtin_strncmp): Likewise.
1246 (inline_string_cmp): Delete the last parameter, change char_type_node
1247 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
1249 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
1250 the inlining expansion on target where the type of the call has same or
1251 narrower precision than unsigned char.
1253 2018-07-20 David Malcolm <dmalcolm@redhat.com>
1255 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
1256 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
1257 * common.opt (fsave-optimization-record): New option.
1258 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
1259 * doc/invoke.texi (-fsave-optimization-record): New option.
1260 * dumpfile.c: Include "optinfo-emit-json.h".
1261 (struct kv_pair): Move to coretypes.h.
1262 (optgroup_options): Make non-static.
1263 (dump_context::end_scope): Call
1264 optimization_records_maybe_pop_dump_scope.
1265 * dumpfile.h (optgroup_options): New decl.
1266 * json.cc: New file.
1268 * optinfo-emit-json.cc: New file.
1269 * optinfo-emit-json.h: New file.
1270 * optinfo.cc: Include "optinfo-emit-json.h".
1271 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
1272 (optinfo_enabled_p): Check optimization_records_enabled_p.
1273 (optinfo_wants_inlining_info_p): Likewise.
1274 * optinfo.h: Update comment.
1275 * profile-count.c (profile_quality_as_string): New function.
1276 * profile-count.h (profile_quality_as_string): New decl.
1277 (profile_count::quality): New accessor.
1278 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
1279 and optinfo_emit_json_cc_tests.
1280 * selftest.h (selftest::json_cc_tests): New decl.
1281 (selftest::optinfo_emit_json_cc_tests): New decl.
1282 * toplev.c: Include "optinfo-emit-json.h".
1283 (compile_file): Call optimization_records_finish.
1284 (do_compile): Call optimization_records_start.
1285 * tree-ssa-live.c: Include optinfo.h.
1286 (remove_unused_scope_block_p): Retain inlining information if
1287 optinfo_wants_inlining_info_p returns true.
1289 2018-07-20 Richard Biener <rguenther@suse.de>
1292 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
1293 to cover -flto-partition=none.
1295 2018-07-20 Martin Liska <mliska@suse.cz>
1297 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
1298 (get_decl_source_range): Remove unused function.
1300 2018-07-20 Richard Biener <rguenther@suse.de>
1302 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
1303 (struct vn_phi_s): Likewise.
1304 (struct vn_reference_s): Likewise.
1305 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
1306 for searching the slot of an entry known to be in the hash itself.
1307 (vn_phi_hasher::equal): Likewise.
1308 (vn_reference_hasher::equal): Likewise.
1309 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
1311 (optimistic_info, current_info): Remove, keeping only valid_info.
1312 (vn_reference_lookup_1): Remove fallback lookup.
1313 (vn_reference_lookup_2): Likewise.
1314 (vn_nary_op_lookup_1): Likewise.
1315 (vn_phi_lookup): Likewise.
1316 (vn_nary_build_or_lookup_1): Make sure to not chain the built
1318 (vn_reference_insert): Adjust, chain the inserted hash element
1319 at last_inserted_ref.
1320 (vn_reference_insert_pieces): Likewise.
1321 (visit_reference_op_call): Likewise.
1322 (vn_nary_op_insert_into): Chain the inserted hash element at
1324 (vn_nary_op_insert_pieces): Adjust.
1325 (vn_nary_op_insert): Likewise.
1326 (vn_nary_op_insert_stmt): Likewise.
1327 (vn_phi_insert): Adjust, chain the inserted hash element at
1329 (process_scc): Remove clearing and copying the optimistic
1330 table. Instead remove elements inserted during an optimistic
1331 iteration from the single table we maintain.
1332 (init_scc_vn): Adjust.
1333 (free_scc_vn): Likewise.
1334 (sccvn_dom_walker::record_cond): Likewise.
1335 (sccvn_dom_walker::after_dom_children): Likewise.
1337 2018-07-19 Martin Sebor <msebor@redhat.com>
1339 PR tree-optimization/84047
1340 PR tree-optimization/83776
1341 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
1342 (check_array_bounds): Call it.
1344 2018-07-19 Martin Sebor <msebor@redhat.com>
1346 * align.h (align_flags): Use member initialization.
1348 2018-07-19 David Malcolm <dmalcolm@redhat.com>
1350 * Makefile.in (OBJS): Add optinfo.o.
1351 * coretypes.h (class symtab_node): New forward decl.
1352 (struct cgraph_node): New forward decl.
1353 (class varpool_node): New forward decl.
1354 * dump-context.h: New file.
1355 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
1357 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
1358 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
1359 (set_alt_dump_file): Likewise.
1360 (dump_context::~dump_context): New dtor.
1361 (dump_gimple_stmt): Move implementation to...
1362 (dump_context::dump_gimple_stmt): ...this new member function.
1363 Add the stmt to any pending optinfo, creating one if need be.
1364 (dump_gimple_stmt_loc): Move implementation to...
1365 (dump_context::dump_gimple_stmt_loc): ...this new member function.
1366 Start a new optinfo and add the stmt to it.
1367 (dump_gimple_expr): Move implementation to...
1368 (dump_context::dump_gimple_expr): ...this new member function.
1369 Add the stmt to any pending optinfo, creating one if need be.
1370 (dump_gimple_expr_loc): Move implementation to...
1371 (dump_context::dump_gimple_expr_loc): ...this new member function.
1372 Start a new optinfo and add the stmt to it.
1373 (dump_generic_expr): Move implementation to...
1374 (dump_context::dump_generic_expr): ...this new member function.
1375 Add the tree to any pending optinfo, creating one if need be.
1376 (dump_generic_expr_loc): Move implementation to...
1377 (dump_context::dump_generic_expr_loc): ...this new member
1378 function. Add the tree to any pending optinfo, creating one if
1380 (dump_printf): Move implementation to...
1381 (dump_context::dump_printf_va): ...this new member function. Add
1382 the text to any pending optinfo, creating one if need be.
1383 (dump_printf_loc): Move implementation to...
1384 (dump_context::dump_printf_loc_va): ...this new member function.
1385 Start a new optinfo and add the stmt to it.
1386 (dump_dec): Move implementation to...
1387 (dump_context::dump_dec): ...this new member function. Add the
1388 value to any pending optinfo, creating one if need be.
1389 (dump_context::dump_symtab_node): New member function.
1390 (dump_context::get_scope_depth): New member function.
1391 (dump_context::begin_scope): New member function.
1392 (dump_context::end_scope): New member function.
1393 (dump_context::ensure_pending_optinfo): New member function.
1394 (dump_context::begin_next_optinfo): New member function.
1395 (dump_context::end_any_optinfo): New member function.
1396 (dump_context::s_current): New global.
1397 (dump_context::s_default): New global.
1398 (dump_scope_depth): Delete global.
1399 (dumpfile_ensure_any_optinfo_are_flushed): New function.
1400 (dump_symtab_node): New function.
1401 (get_dump_scope_depth): Reimplement in terms of dump_context.
1402 (dump_begin_scope): Likewise.
1403 (dump_end_scope): Likewise.
1404 (selftest::temp_dump_context::temp_dump_context): New ctor.
1405 (selftest::temp_dump_context::~temp_dump_context): New dtor.
1406 (selftest::verify_item): New function.
1407 (ASSERT_IS_TEXT): New macro.
1408 (ASSERT_IS_TREE): New macro.
1409 (ASSERT_IS_GIMPLE): New macro.
1410 (selftest::test_capture_of_dump_calls): New test.
1411 (selftest::dumpfile_c_tests): Call it.
1412 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
1413 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
1414 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
1415 descriptive comment.
1416 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
1417 (dump_node, dump_bb): Move these unrelated decls.
1418 (class dump_manager): Add leading comment.
1419 * optinfo.cc: New file.
1420 * optinfo.h: New file.
1422 2018-07-19 Michael Collison <michael.collison@arm.com>
1423 Richard Henderson <rth@redhat.com>
1425 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
1426 (subti): Handle op1 zero.
1427 (subvti4, usub4ti4): New.
1428 (*sub<GPI>3_compare1_imm): New.
1429 (sub<GPI>3_carryinCV): New.
1430 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
1431 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
1433 2018-07-19 Michael Collison <michael.collison@arm.com>
1434 Richard Henderson <rth@redhat.com>
1436 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
1437 (addti3): Create simpler code if low part is already known to be 0.
1438 (addvti4, uaddvti4): New.
1439 (*add<GPI>3_compareC_cconly_imm): New.
1440 (*add<GPI>3_compareC_cconly): New.
1441 (*add<GPI>3_compareC_imm): New.
1442 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
1443 handle constants within this pattern..
1444 (*add<GPI>3_compareV_cconly_imm): New.
1445 (*add<GPI>3_compareV_cconly): New.
1446 (*add<GPI>3_compareV_imm): New.
1447 (add<GPI>3_compareV): New.
1448 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
1449 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
1450 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
1451 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
1452 with 'comparison' operator.
1453 (*add<GPI>3_compareV_cconly_imm): Ditto.
1454 (*add<GPI>3_compareV_cconly): Ditto.
1455 (*add<GPI>3_compareV_imm): Ditto.
1456 (add<GPI>3_compareV): Ditto.
1457 (add<mode>3_carryinC): Ditto.
1458 (*add<mode>3_carryinC_zero): Ditto.
1459 (*add<mode>3_carryinC): Ditto.
1460 (add<mode>3_carryinV): Ditto.
1461 (*add<mode>3_carryinV_zero): Ditto.
1462 (*add<mode>3_carryinV): Ditto.
1464 2018-07-19 Michael Collison <michael.collison@arm.com>
1465 Richard Henderson <rth@redhat.com>
1467 * config/aarch64/aarch64-modes.def (CC_V): New.
1468 * config/aarch64/aarch64-protos.h
1469 (aarch64_addti_scratch_regs): Declare
1470 (aarch64_subvti_scratch_regs): Declare.
1471 (aarch64_expand_subvti): Declare.
1472 (aarch64_gen_unlikely_cbranch): Declare
1473 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
1474 for signed overflow using CC_Vmode.
1475 (aarch64_get_condition_code_1): Handle CC_Vmode.
1476 (aarch64_gen_unlikely_cbranch): New function.
1477 (aarch64_addti_scratch_regs): New function.
1478 (aarch64_subvti_scratch_regs): New function.
1479 (aarch64_expand_subvti): New function.
1481 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
1483 * config/aarch64/aarch64-option-extensions.def: New entry for profile
1485 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
1486 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
1489 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
1492 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
1493 address check not strict.
1495 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
1497 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
1499 (store_pair_lanes<mode>): Likewise.
1500 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
1501 enum value 'ADDR_QUERY_LDP_STP_N'.
1502 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
1503 (aarch64_print_address_internal): Add declaration.
1504 (aarch64_print_ldpstp_address): Remove.
1505 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
1506 (aarch64_print_operand): Change printing of 'y'.
1507 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
1508 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
1509 'true' rather than '1'.
1510 * config/aarch64/constraints.md (Uml): Likewise.
1511 (Uml): Rename to Umn.
1514 2018-07-19 Richard Biener <rguenther@suse.de>
1516 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
1518 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
1519 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
1520 (vn_reference_hasher): Likewise.
1521 (struct vn_tables_s): Remove obstack and alloc-pool members.
1522 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
1523 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
1524 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
1525 (vn_reference_insert_pieces): Likewise.
1526 (alloc_vn_nary_op_noinit): Adjust.
1527 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
1528 (vn_phi_eq): Adjust.
1529 (shared_lookup_phiargs): Remove.
1530 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
1531 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
1532 (visit_reference_op_call): Likewise.
1533 (copy_nary, copy_phi, copy_reference): Remove.
1534 (process_scc): Rewind the obstack when iterating. Do not
1535 copy the elements to valid_info but just move them from one
1536 hashtable to the other.
1537 (allocate_vn_table): Adjust.
1538 (free_vn_table): Likewise.
1539 (init_scc_vn): Likewise.
1540 (free_scc_vn): Likewise.
1542 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
1545 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
1546 indirect_return as function type attribute.
1547 (ix86_attribute_table): Change indirect_return to function
1549 * doc/extend.texi: Update indirect_return attribute.
1551 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
1553 * wide-int.h (widest2_int): New.
1554 * gimple-fold.c (arith_overflowed_p): Use it.
1555 * tree.h (widest2_int_cst): New.
1556 * tree-vrp.c (wide_int_binop_overflow): Rename from
1557 vrp_int_const_binop.
1558 Rewrite to work on trees.
1559 (extract_range_from_multiplicative_op_1): Abstract code to...
1560 (wide_int_range_min_max): ...here.
1561 (wide_int_range_cross_product): ...and here.
1562 (extract_range_from_binary_expr_1): Abstract overflow code to...
1563 (wide_int_range_mult_wrapping): ...here.
1564 * tree-vrp.h (wide_int_range_cross_product): New.
1565 (wide_int_range_mult_wrapping): New.
1567 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
1568 Julia Koval <julia.koval@intel.com>
1570 * config/i386/x86-tune-costs.h (skylake_memcpy,
1571 skylake_memset): Replace rep_prefix with unrolling for size 512.
1573 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
1576 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle comparision with EQ_EXPR
1579 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
1581 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
1582 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
1583 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
1584 previously known as "PowerPC AltiVec Built-in Functions". Move
1585 some material to new subsubsections "PowerPC AltiVec Built-in
1586 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
1588 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
1589 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
1590 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
1591 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
1593 2018-07-18 Richard Biener <rguenther@suse.de>
1595 PR tree-optimization/86557
1596 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
1599 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
1601 * config/s390/s390.c (s390_function_profiler): Generate CFI.
1603 2018-07-17 Jeff Law <law@redhat.com>
1605 * config/arm/arm.c (get_label_padding): Update for recent
1606 changes to label_to_alignment.
1608 PR tree-optimization/86010
1609 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
1611 * config/mips/mips.c (vr4130_align_insns): Update for recent
1612 changes to label_to_alignment.
1614 * config/frv/frv.c (frv_label_align): Update for recent changes
1615 to label_to_alignment.
1617 * config/nios2/nios2.c (nios2_label_align): Update for recent
1618 changes which dropped ALIGN_LABELS_LOG.
1620 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
1622 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
1623 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
1625 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
1627 * config/arc/arc.c (arc_label_align): Use align_labels instead of
1628 deprecated align_labels_log.
1630 2018-07-17 Richard Biener <rguenther@suse.de>
1633 * dwarf2out.c (init_sections_and_labels): Always generate
1634 a debug_line_str_section for early LTO debug.
1635 (dwarf2out_finish): Reset debug_line_str_hash output early.
1636 Bump counter for extra dwarf5 .debug_loc labels to not conflict
1638 (dwarf2out_early_finish): Output debug_line_str.
1640 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
1642 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
1643 index register on z196 or later.
1645 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
1647 * config/s390/s390.c (s390_default_align): Set default function
1649 (s390_override_options_after_change): Call s390_default align.
1650 (s390_option_override_internal): Call s390_default_align.
1651 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
1653 2018-07-17 Jakub Jelinek <jakub@redhat.com>
1656 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
1657 to _looptemp_ clauses, other than the first two.
1659 2018-07-17 Martin Liska <mliska@suse.cz>
1661 * opts.c: Do not enable OPT_falign_* for -Os.
1663 2018-07-17 Martin Liska <mliska@suse.cz>
1665 * align.h (MAX_CODE_ALIGN): New.
1666 (MAX_CODE_ALIGN_VALUE): New.
1667 * common/config/i386/i386-common.c (ix86_handle_option):
1668 (MAX_CODE_ALIGN): Moved to align.h.
1669 * final.c (MAX_CODE_ALIGN): Likewise.
1670 * opts.c (parse_and_check_align_values):
1671 (MAX_CODE_ALIGN): Likewise.
1672 (MAX_CODE_ALIGN_VALUE): Likewise.
1674 2018-07-17 Martin Liska <mliska@suse.cz>
1676 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
1677 in order to fulfil coding style.
1678 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
1679 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1680 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
1681 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
1682 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
1683 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
1684 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
1685 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
1686 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1688 2018-07-17 Martin Liska <mliska@suse.cz>
1690 * align.h: New file.
1691 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions directly.
1692 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
1693 align_flags of label_to_alignment.
1694 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into align_flags
1696 * config/m68k/m68k.c: Do not use removed align_labels_value and
1698 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
1699 (LOOP_ALIGN): Likewise.
1700 (LABEL_ALIGN): Likewise.
1701 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
1702 Remove not used macro.
1703 (rs6000_loop_align): Change return type to align_flags.
1704 (rs6000_loop_align_max_skip): Remove.
1705 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
1706 Change return type to align_flags.
1707 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
1708 Remove not used macro.
1709 (rs6000_loop_align): Change return type to align_flags.
1710 (rs6000_loop_align_max_skip): Remove.
1711 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
1712 * config/rx/rx-protos.h (rx_align_for_label): Make it
1714 * config/rx/rx.c (rx_align_for_label): Change return type
1716 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
1718 into align_flags class.
1719 (LABEL_ALIGN): Likewise.
1720 (LOOP_ALIGN): Likewise.
1721 * config/s390/s390.c (s390_label_align): Use align_flags
1723 (s390_asm_output_function_label): Likewise.
1724 * config/sh/sh.c (sh_override_options_after_change):
1725 Use align_flags class directly without macros.
1726 (find_barrier): Likewise.
1727 (barrier_align): Likewise.
1728 (sh_loop_align): Likewise.
1729 * config/spu/spu.c (spu_option_override):
1730 Use align_flags_tuple::get_value instead of removed macros.
1731 (spu_sched_init): Likewise.
1732 * config/spu/spu.h (GTY): Likewise.
1733 * config/visium/visium.c (visium_option_override):
1734 Set "8" as default secondary alignment.
1735 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
1736 in order to guarantee secondary alignment of 8.
1737 * coretypes.h: Include align.h header file.
1738 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
1739 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
1740 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
1741 * doc/tm.texi.in: Likewise.
1742 * final.c (struct label_alignment): Remove not used structure.
1743 (LABEL_ALIGN): Change type to align_flags.
1744 (LOOP_ALIGN): Likewise.
1745 (JUMP_ALIGN): Likewise.
1746 (default_loop_align_max_skip): Remove.
1747 (default_label_align_max_skip): Likewise.
1748 (default_jump_align_max_skip): Likewise.
1749 (default_label_align_after_barrier_max_skip):
1750 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
1751 (LABEL_TO_MAX_SKIP): Remove.
1752 (label_to_alignment): Return align_flags type instead of integer.
1753 (label_to_max_skip): Remove.
1754 (align_fuzz): Use align_flags type.
1755 (compute_alignments): Use align_flags type and use align_flags::max
1756 to combine multiple alignments.
1757 (grow_label_align): Grow vec instead of C array.
1758 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
1759 (shorten_branches): Use align_flags type and use align_flags::max
1760 to combine multiple alignments.
1761 (final_scan_insn_1): Remove usage of secondary alignment that comes
1762 from label alignment, but instead use proper secondary alignment
1763 which is computed in grow_label_align.
1764 * flags.h (struct align_flags_tuple): Move to align.h.
1765 (struct align_flags): Likewise.
1766 (state_align_loops): Rename to align_loops.
1767 (state_align_jumps): Rename to align_jumps.
1768 (state_align_labels): Rename to align_labels.
1769 (state_align_functions): Rename to align_functions.
1770 (align_loops_log): Remove.
1771 (align_jumps_log): Remove.
1772 (align_labels_log): Remove.
1773 (align_functions_log): Remove.
1774 (align_loops_max_skip): Remove.
1775 (align_jumps_max_skip): Remove.
1776 (align_labels_max_skip): Remove.
1777 (align_functions_max_skip): Remove.
1778 (align_loops_value): Remove.
1779 (align_jumps_value): Remove.
1780 (align_labels_value): Remove.
1781 (align_functions_value): Remove.
1782 * output.h (label_to_alignment): Change return type to align_flags.
1783 (label_to_max_skip): Remove.
1784 * target.def: Remove loop_align_max_skip, label_align_max_skip,
1785 jump_align_max_skip macros.
1786 * targhooks.h (default_loop_align_max_skip): Remove.
1787 (default_label_align_max_skip): Likewise.
1788 (default_jump_align_max_skip): Likewise.
1789 (default_label_align_after_barrier_max_skip): Remove.
1790 * toplev.c (read_log_maxskip): Use ::normalize function.
1791 (parse_N_M): Remove not used argument and also call ::normalize.
1792 (parse_alignment_opts): Do not pass unused arguments.
1793 * varasm.c (assemble_start_function): Use directly align_functions
1794 instead of removed macros.
1795 * system.h: Do not poison removed macros.
1797 2018-07-17 Jakub Jelinek <jakub@redhat.com>
1800 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
1801 and cond temporaries don't have reference type if iterator has
1802 pointer type. For init use &for_pre_body instead of pre_p if
1803 for_pre_body is non-empty.
1805 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
1807 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
1808 double-double modes to SFmode directly directly.
1809 (trunc<mode>sf2_fprs): Delete.
1811 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
1813 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
1814 for conversions between IFmode and the decimal floating point modes.
1815 (init_float128_ieee): Use the correct names for conversions between
1816 KFmode and the decimal floating point modes.
1818 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
1820 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
1821 for the conversions between TDmode and IFmode.
1822 (init_float128_ieee): Use more correct names for the conversions
1823 between TDmode and KFmode.
1825 2018-07-16 Jakub Jelinek <jakub@redhat.com>
1827 PR tree-optimization/86526
1828 * builtins.c (expand_builtin_memcmp): Formatting fixes.
1829 (inline_expand_builtin_string_cmp): Likewise.
1830 (inline_string_cmp): Likewise. Use c_readstr instead of
1831 builtin_memcpy_read_str. Add unit_mode temporary.
1833 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
1836 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
1837 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
1839 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
1841 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
1842 Alphabetize prototypes of built-in functions, separating out
1843 built-in functions that are listed in this section but should be
1844 described elsewhere.
1846 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
1849 * expmed.c (emit_store_flag): Do not emit setcc followed by a
1850 conditional move when trapping comparison was split to a
1851 non-trapping one (and vice versa).
1853 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
1855 * config/s390/s390.c (s390_function_profiler): Generate nops
1856 instead of profiler call sequences.
1857 * config/s390/s390.opt: Add the new option.
1859 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
1861 * config/s390/s390.c (s390_function_profiler): Generate
1862 __mcount_loc section.
1863 * config/s390/s390.opt: Add the new option.
1865 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
1867 * common.opt: Add the new warning.
1868 * config/s390/s390.c (s390_function_profiler): Emit "brasl
1869 %r0,__fentry__" when -mfentry is specified.
1870 (s390_option_override_internal): Disallow -mfentry for 31-bit
1872 * config/s390/s390.opt: Add the new option.
1874 2018-07-16 Richard Biener <rguenther@suse.de>
1877 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
1878 for function-local FUNCTION_DECL and RESULT_DECL immediately.
1880 2018-07-16 Martin Liska <mliska@suse.cz>
1883 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
1886 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
1888 * config/arc/arcHS.md: Update ARCHS scheduling rules.
1890 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
1892 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
1894 * config/arc/arc-cpus.def (hs4x): New cpu.
1896 * config/arc/arc-tables.opt: Regenerate.
1897 * config/arc/arc.c (arc_sched_issue_rate): New function.
1898 (TARGET_SCHED_ISSUE_RATE): Define.
1899 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
1900 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
1902 (attr tune): Add ARCHS4x tune values.
1903 (attr tune_dspmpy): Define.
1904 (*tst): Correct instruction type.
1905 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
1906 * config/arc/arcHS4x.md: New file.
1907 * config/arc/fpu.md: Update instruction type attributes.
1908 * config/arc/t-multilib: Regenerate.
1910 2018-07-16 Tom de Vries <tdevries@suse.de>
1913 * var-tracking.c (vt_initialize): Fix pre_dec handling.
1915 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1917 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
1920 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
1922 PR tree-optimization/86514
1923 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
1924 conversion to a boolean type from a type with greater precision.
1926 2018-07-16 Tom de Vries <tdevries@suse.de>
1928 * var-tracking.c (vt_initialize): Print adjusted insn slim if
1929 dump_flags request TDF_SLIM.
1931 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
1933 * fold-const.c (int_const_binop_1): Abstract...
1934 (wide_int_binop): ...wide int code here.
1935 (poly_int_binop): ...poly int code here.
1936 Abstract the rest of int_const_binop_1 into int_const_binop.
1937 * fold-const.h (wide_int_binop): New.
1938 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
1939 Remove useless PLUS/MINUS_EXPR case.
1940 (zero_nonzero_bits_from_vr): Move wide int code...
1941 (zero_nonzero_bits_from_bounds): ...here.
1942 (extract_range_from_binary_expr_1): Move mask optimization code...
1943 (range_easy_mask_min_max): ...here.
1944 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
1945 (range_easy_mask_min_max): New.
1947 2018-07-15 Jeff Law <law@redhat.com>
1950 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
1953 2018-07-14 Jim Wilson <jimw@sifive.com>
1955 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
1957 2018-07-14 Paul Koning <ni1d@arrl.net>
1959 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
1961 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
1963 * lto-streamer-out.c (copy_function_or_variable): Dump info about
1966 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
1967 Steve Munroe <munroesj52@gmail.com>
1969 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
1970 (_mm_andnot_si128): Likewise.
1971 (_mm_or_si128): Likewise.
1972 (_mm_xor_si128): Likewise.
1974 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
1977 * builtins.c (expand_builtin_memcmp): Inline the calls first
1978 when result_eq is false.
1979 (expand_builtin_strcmp): Inline the calls first.
1980 (expand_builtin_strncmp): Likewise.
1981 (inline_string_cmp): New routine. Expand a string compare
1982 call by using a sequence of char comparison.
1983 (inline_expand_builtin_string_cmp): New routine. Inline expansion
1984 a call to str(n)cmp/memcmp.
1985 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
1987 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
1989 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
1991 * config/arm/driver-arm.c: Include arm-native.h.
1992 (host_detect_local_cpu): Use auto-generated data tables.
1993 (vendors, arm_cpu_table): Delete. Move part information to ...
1994 * config/arm/arm-cpus.in: ... here.
1995 * config/arm/parsecpu.awk (gen_native): New function.
1996 (vendor, part): New CPU fields.
1997 (END): Add support for building the native CPU detection tables.
1998 * config/arm/t-arm (arm-native.h): Add build rule.
1999 (driver-arm.o): Add dependency on arm-native.h.
2001 2018-07-13 Richard Biener <rguenther@suse.de>
2004 * match.pd (addr1 - addr2): Allow either of the operand to
2007 2018-07-13 Tom de Vries <tdevries@suse.de>
2009 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
2012 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
2014 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
2015 arrays instead of numbered variables.
2017 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
2019 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
2020 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
2022 2018-07-13 Richard Biener <rguenther@suse.de>
2025 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
2026 instead of get_context_die.
2028 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
2029 Richard Biener <rguenther@suse.de>
2032 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
2033 that the loop latch destination where phi is defined.
2035 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
2037 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
2038 (riscv_expand_epilogue): Add assertion to check interrupt mode.
2039 (riscv_set_current_function): Extract getting interrupt type to new
2041 (riscv_get_interrupt_type): New function.
2042 (riscv_merge_decl_attributes): New function, checking interrupt type is
2044 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
2046 2018-07-12 Paul Koning <ni1d@arrl.net>
2048 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
2051 2018-07-12 Paul Koning <ni1d@arrl.net>
2053 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
2054 zero reference, add doloop_end instead.
2055 * doc/md.texi (decrement_and_branch_until_zero): Remove.
2056 (Looping patterns): Remove decrement_and_branch_until_zero. Add
2057 detail for doloop_end.
2059 2018-07-12 Martin Sebor <msebor@redhat.com>
2062 * attribs.c (decl_attributes): Reject conflicting attributes before
2063 calling attribute handlers.
2065 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
2067 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
2069 (gcc::dump_manager::get_dump_file_name): likewise.
2070 (dump_begin): Likewise.
2071 * dumpfile.h (dump_begin): Update prototype.
2072 (gcc::dump_manager::get_dump_file_name,
2073 gcc::dump_manager::get_dump_file_name): Update prototype.
2075 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
2077 * internal-fn.h (vectorizable_internal_fn_p): New function.
2078 * tree-vect-slp.c (compatible_calls_p): Likewise.
2079 (vect_build_slp_tree_1): Remove nops argument. Handle calls
2080 to internal functions.
2081 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
2083 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
2085 * fold-const.h (inverse_conditions_p): Declare.
2086 * fold-const.c (inverse_conditions_p): New function.
2087 * match.pd: Use inverse_conditions_p. Add folds of view_converts
2088 that test the inverse condition of a conditional internal function.
2089 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
2090 * internal-fn.c (internal_fn_mask_index): Handle conditional
2092 (vectorized_internal_fn_supported_p): New function.
2093 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
2094 (any_pred_load_store): Replace with...
2095 (need_to_predicate): ...this new variable.
2096 (redundant_ssa_names): New variable.
2097 (ifcvt_can_use_mask_load_store): Move initial checks to...
2098 (ifcvt_can_predicate): ...this new function. Handle tree codes
2099 for which a conditional internal function exists.
2100 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
2101 instead of ifcvt_can_use_mask_load_store. Update after variable
2103 (predicate_load_or_store): New function, split out from
2104 predicate_mem_writes.
2105 (check_redundant_cond_expr): New function.
2106 (value_available_p): Likewise.
2107 (predicate_rhs_code): Likewise.
2108 (predicate_mem_writes): Rename to...
2109 (predicate_statements): ...this. Use predicate_load_or_store
2110 and predicate_rhs_code.
2111 (combine_blocks, tree_if_conversion): Update after above name changes.
2112 (ifcvt_local_dce): Handle redundant_ssa_names.
2113 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
2114 general conditional functions.
2115 * tree-vect-stmts.c (vectorizable_call): Likewise.
2117 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
2118 Alan Hayward <alan.hayward@arm.com>
2119 David Sherwood <david.sherwood@arm.com>
2121 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
2122 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
2123 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
2124 plus and minus and convert them into IFN_COND_FMA-based sequences.
2125 (convert_mult_to_fma): Handle conditional plus and minus.
2127 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
2129 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
2130 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
2131 (cond_fnms_optab): New optabs.
2132 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
2134 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
2135 * internal-fn.h (get_conditional_internal_fn): Declare.
2136 (get_unconditional_internal_fn): Likewise.
2137 * internal-fn.c (cond_ternary_direct): New macro.
2138 (expand_cond_ternary_optab_fn): Likewise.
2139 (direct_cond_ternary_optab_supported_p): Likewise.
2140 (FOR_EACH_COND_FN_PAIR): Likewise.
2141 (get_conditional_internal_fn): New function.
2142 (get_unconditional_internal_fn): Likewise.
2143 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
2144 (gimple_match_op::gimple_match_op): Add a new overload for 5
2146 (gimple_match_op::set_op): Likewise.
2147 (gimple_resimplify5): Declare.
2148 * genmatch.c (decision_tree::gen): Generate simplifications for
2150 * gimple-match-head.c (gimple_simplify): Define an overload for
2151 5 operands. Handle calls with 5 arguments in the top-level overload.
2152 (convert_conditional_op): Handle conversions from unconditional
2153 internal functions to conditional ones.
2154 (gimple_resimplify5): New function.
2155 (build_call_internal): Pass a fifth operand.
2156 (maybe_push_res_to_seq): Likewise.
2157 (try_conditional_simplification): Try converting conditional
2158 internal functions to unconditional internal functions.
2159 Handle 3-operand unconditional forms.
2160 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
2161 Define ternary equivalents of the current rules for binary conditional
2163 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
2165 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
2166 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
2167 (optab): Handle them.
2168 (SVE_COND_FP_TERNARY): New int iterator.
2169 (sve_fmla_op, sve_fmad_op): New int attributes.
2170 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
2171 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
2172 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
2174 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
2176 * target.def (preferred_else_value): New target hook.
2177 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
2178 * doc/tm.texi: Regenerate.
2179 * targhooks.h (default_preferred_else_value): Declare.
2180 * targhooks.c (default_preferred_else_value): New function.
2181 * internal-fn.h (conditional_internal_fn_code): Declare.
2182 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
2183 (get_conditional_internal_fn): Use it.
2184 (conditional_internal_fn_code): New function.
2185 * gimple-match.h (gimple_match_cond): New struct.
2186 (gimple_match_op): Add a cond member function.
2187 (gimple_match_op::gimple_match_op): Update all forms to take a
2189 * genmatch.c (expr::gen_transform): Use the same condition as res_op
2190 for the suboperation, but don't specify a particular else_value.
2191 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
2192 (visit_nary_op, visit_reference_op_load): Pass
2193 gimple_match_cond::UNCOND to the gimple_match_op constructor.
2194 * gimple-match-head.c: Include tree-eh.h
2195 (convert_conditional_op): New function.
2196 (maybe_resimplify_conditional_op): Likewise.
2197 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
2198 (gimple_resimplify2): Likewise.
2199 (gimple_resimplify3): Likewise.
2200 (gimple_resimplify4): Likewise.
2201 (maybe_push_res_to_seq): Return null for conditional operations.
2202 (try_conditional_simplification): New function.
2203 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
2205 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
2207 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
2209 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
2211 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
2213 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
2215 (hash_tree): Do not hash DECL_FCONTEXT
2216 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
2217 Do not stream DECL_FCONTEXT.
2218 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
2219 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
2221 2018-07-12 Richard Biener <rguenther@suse.de>
2224 * dwarf2out.c (gen_block_die): Only output blocks when they have
2225 at least one !DECL_IGNORED_P variable.
2227 2018-07-12 Richard Biener <rguenther@suse.de>
2230 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
2231 Remove -mieee-fp handling.
2233 2018-07-12 Richard Biener <rguenther@suse.de>
2235 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
2236 left-over from last patch.
2238 2018-07-12 Jakub Jelinek <jakub@redhat.com>
2240 PR tree-optimization/86492
2241 * gimple-ssa-store-merging.c
2242 (imm_store_chain_info::coalesce_immediate_stores): Call
2243 check_no_overlap even for the merge_overlapping case. Formatting fix.
2245 2018-07-12 Richard Biener <rguenther@suse.de>
2248 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
2249 move possibly trapping operations into the conditional.
2251 2018-07-12 Richard Biener <rguenther@suse.de>
2253 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
2254 (vn_lookup_simplify_result): Remove recursion limit applied
2256 (vn_nary_build_or_lookup_1): Adjust.
2257 (try_to_simplify): Likewise.
2258 * gimple-match-head.c (gimple_resimplify1): Instead apply one
2260 (gimple_resimplify2): Likewise.
2261 (gimple_resimplify3): Likewise.
2262 (gimple_resimplify4): Likewise.
2264 2018-07-11 Jakub Jelinek <jakub@redhat.com>
2266 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
2267 Use __mmask64 type instead of __mmask8 for __M argument.
2268 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
2269 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
2271 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
2272 __mmask16 for __M argument.
2273 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
2274 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
2275 to __mmask16 instead of __mmask8.
2276 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
2277 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
2278 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
2279 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
2280 instead of __mmask16 for __U argument.
2281 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
2282 __mmask16 instead of __mmask8 for __U argument.
2283 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
2285 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
2287 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
2289 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
2291 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
2293 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
2295 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
2297 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
2299 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
2301 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
2303 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
2304 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
2305 return type as well as __M argument type and all casts from __mmask8
2307 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
2308 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
2309 return type as well as __M argument type and all casts from __mmask8
2311 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
2312 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
2313 return type as well as __M argument type and all casts from __mmask8
2315 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
2316 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
2317 return type as well as __M argument type and all casts from __mmask8
2319 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
2320 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
2323 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
2325 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
2326 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
2329 2018-07-11 Paul Koning <ni1d@arrl.net>
2331 * doc/md.texi (define_subst): Document how multiple occurrences of
2332 the same argument in the replacement pattern are handled.
2334 2018-07-11 Paul Koning <ni1d@arrl.net>
2336 * doc/extend.texi (Common Variable Attributes): Move "mode" into
2338 (Common Type Attributes): Add "mode" attribute.
2340 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
2342 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
2343 stream DECL_ORIGINAL_TYPE.
2344 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
2345 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
2346 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
2347 Do not walk original type.
2348 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
2350 (write_ts_decl_non_common_tree_pointers): Do not stream
2352 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
2353 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
2355 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
2357 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
2358 thread twice from the same starting edge.
2360 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
2362 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
2363 * gimple.c (gimple_call_nonnull_result_p): ...here...
2364 (gimple_call_nonnull_arg): ...and here.
2365 * gimple.h (gimple_call_nonnull_result_p): New.
2366 (gimple_call_nonnull_arg): New.
2368 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
2370 * config/arm/arm-cpus.in: Move information from fpu field of each
2371 cpu definition to the isa field.
2372 * config/arm/parsecpu.awk (fpu): Delete match rule.
2373 (gen_comm_data): Don't add bits from the CPU's FPU entry.
2375 2018-07-11 Richard Biener <rguenther@suse.de>
2378 * dwarf2out.c (init_sections_and_labels): Use
2379 output_asm_line_debug_info consistently.
2380 (dwarf2out_early_finish): Likewise.
2381 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
2384 2018-07-11 Richard Biener <rguenther@suse.de>
2386 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
2387 Rework father_bb setting in a way to avoid propagating constants
2388 multiple times on a loop body.
2390 2018-07-10 Mark Wielaard <mark@klomp.org>
2393 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
2395 2018-07-10 Richard Biener <rguenther@suse.de>
2397 * hash-map.h (hash_map::iterator::operator*): Return
2398 references to key and value.
2400 2018-07-10 Jakub Jelinek <jakub@redhat.com>
2403 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
2404 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
2405 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
2406 (gimplify_omp_for): For composite loops, move outer
2407 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
2408 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
2409 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
2410 TREE_LIST for both the original class iterator and the "last" helper
2411 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
2412 loop, remember has_decl_expr from outer composite loops for the
2413 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
2415 2018-07-09 Martin Sebor <msebor@redhat.com>
2419 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
2420 accessing implicitly initialized array elements.
2421 * expr.c (string_constant): Handle string initializers of
2422 character arrays within aggregates.
2423 * gimple-fold.c (fold_array_ctor_reference): Add argument.
2424 Store element offset. As a special case, handle zero size.
2425 (fold_nonarray_ctor_reference): Same.
2426 (fold_ctor_reference): Add argument. Store subobject offset.
2427 * gimple-fold.h (fold_ctor_reference): Add argument.
2429 2018-07-09 Paul Koning <ni1d@arrl.net>
2431 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
2432 (pdp11_insn_cost): New function.
2433 (pdp11_md_asm_adjust): New function.
2434 (TARGET_INVALID_WITHIN_DOLOOP): Define.
2435 (pdp11_rtx_costs): Update to match machine better.
2436 (output_addr_const_pdp11): Correct format mismatch warnings.
2437 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
2438 * config/pdp11/pdp11.md: General change to add base_cost and/or
2439 length attributes for use by new pdp11_insn_cost function.
2440 (MIN_BRANCH): Correct definition.
2442 (doloop_end): Use standard pattern name for looping pattern.
2443 (doloop_end_nocc): New.
2444 (movsf): Add another constraint alternative.
2445 (zero_extendqihi2): Add constraint alternatives for not in place
2447 (zero_extendhisi2): Remove.
2448 (shift patterns): Add CC handling variants.
2452 (define_peephole2): New peephole to recognize mov that sets CC for
2455 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2457 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
2458 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
2461 2018-07-09 Jakub Jelinek <jakub@redhat.com>
2464 * real.c (real_nextafter): Return true if result is denormal.
2466 2018-07-09 Martin Liska <mliska@suse.cz>
2468 * common.opt: Add back wrongly removed attribute.
2470 2018-07-09 Richard Biener <rguenther@suse.de>
2473 * dwarf2out.c (gen_block_die): For an early generated DIE
2474 always output high/low PC attributes.
2476 2018-07-09 Tom de Vries <tdevries@suse.de>
2478 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
2479 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
2480 onto VAR_DECL with abstract origin.
2482 2018-07-07 Jim Wilson <jimw@sifive.com>
2484 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
2486 2018-07-07 Tom de Vries <tdevries@suse.de>
2488 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
2490 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
2492 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
2494 (combine_bound): Use wide-int overflow calculation instead of
2496 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
2498 * fold-const.c (int_const_binop_2): Same.
2499 (extract_muldiv_1): Same.
2500 (fold_div_compare): Same.
2501 (fold_abs_const): Same.
2503 * poly-int.h (add): Same.
2507 * predict.c (predict_iv_comparison): Same.
2508 * profile-count.c (slow_safe_scale_64bit): Same.
2509 * simplify-rtx.c (simplify_const_binary_operation): Same.
2510 * tree-chrec.c (tree_fold_binomial): Same.
2511 * tree-data-ref.c (split_constant_offset_1): Same.
2512 * tree-if-conv.c (idx_within_array_bound): Same.
2513 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
2514 * tree-ssa-phiopt.c (minmax_replacement): Same.
2515 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
2516 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
2517 * vr-values.c (vr_values::adjust_range_with_scev): Same.
2518 * wide-int.cc (wi::add_large): Same.
2519 (wi::mul_internal): Same.
2520 (wi::sub_large): Same.
2521 (wi::divmod_internal): Same.
2522 * wide-int.h: Change overflow type to overflow_type for neg, add,
2523 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
2524 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
2525 mul_internal, divmod_internal.
2526 (overflow_type): New enum.
2527 (accumulate_overflow): New.
2529 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
2531 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
2532 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
2534 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
2536 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
2537 argument is checked for zero before entering loop, avoid checking again.
2539 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
2541 * gimplify.h (generic_expr_could_trap_p): Set as global function.
2542 * gimplify.h (generic_expr_could_trap_p): Likwise.
2543 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
2545 2018-07-06 Jakub Jelinek <jakub@redhat.com>
2547 PR tree-optimization/86401
2548 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
2549 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
2550 (fold_bit_and_mask): ... here. New helper function for match.pd.
2551 * fold-const.h (fold_bit_and_mask): Declare.
2552 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
2554 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
2557 * target.def (translate_mode_attribute): New hook.
2558 * targhooks.h (default_translate_mode_attribute): Declare.
2559 * targhooks.c (default_translate_mode_attribute): New function.
2560 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
2561 * doc/tm.texi: Regenerate.
2562 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
2563 (rs6000_translate_mode_attribute): New function.
2565 2018-07-06 Paul Koning <ni1d@arrl.net>
2567 * doc/md.texi (define_split): Document DONE and FAIL.
2568 (define_peephole2): Ditto.
2570 2018-07-05 Jeff Law <law@redhat.com>
2572 PR tree-optimization/86010
2573 * tree-ssa-dse.c (compute_trims): More aggressively trim at
2574 both the head and tail of mem* and str* calls.
2576 2018-07-05 Jim Wilson <jimw@sifive.com>
2578 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
2580 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
2582 * config/aarch64/aarch64-simd.md: correct flags text for
2583 MIN_EXPR replacement.
2585 2018-07-05 James Clarke <jrtc27@jrtc27.com>
2587 * configure: Regenerated.
2589 2018-07-05 Carl Love <cel@us.ibm.com>
2591 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
2592 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
2593 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
2594 VSX_BUILTIN_DOUBLEL_V4SF.
2596 2018-07-05 Martin Sebor <msebor@redhat.com>
2599 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
2600 than its domain to compute its the upper bound of a char array.
2602 2018-07-05 Nathan Sidwell <nathan@acm.org>
2604 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
2605 * doc/cpp.texi: Update comment.
2606 * doc/tm.texi: Rebuilt.
2607 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
2608 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
2609 * doc/extend.texi (Backwards Compatibility): Clarify it is system
2610 headers affected by extern "C".
2611 * system.h: Poison NO_IMPLICIT_EXTERN_C.
2612 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
2613 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
2614 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
2615 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
2616 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
2617 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
2618 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
2619 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
2620 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
2621 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
2622 config/sparc/sp64-elf.h, config/spu/spu.h,
2623 config/stormy16/stormy16.h, config/v850/v850.h,
2624 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
2625 define NO_IMPLICIT_EXTERN_C.
2626 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
2628 2018-07-05 Tamar Christina <tamar.christina@arm.com>
2631 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
2632 instead of GET_MODE_SIZE when comparing Units.
2634 2018-07-05 Tamar Christina <tamar.christina@arm.com>
2637 * rtlanal.c (set_noop_p): Constrain on mode change,
2638 include hard-reg-set.h
2640 2018-07-05 Tamar Christina <tamar.christina@arm.com>
2642 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
2644 2018-07-05 Jakub Jelinek <jakub@redhat.com>
2647 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
2650 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
2652 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
2654 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
2657 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
2659 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
2661 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
2664 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
2666 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
2668 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
2669 neg_*_op* variables.
2671 2018-07-04 Martin Liska <mliska@suse.cz>
2673 * tree-switch-conversion.c: Define
2674 max_ratio_for_speed and max_ratio_for_size constants.
2676 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
2677 Martin Liska <mliska@suse.cz>
2682 * common.opt: Rename align options with 'str_' prefix.
2683 * common/config/i386/i386-common.c (set_malign_value): New
2685 (ix86_handle_option): Use it to set -falign-* options/
2686 * config/aarch64/aarch64-protos.h (struct tune_params): Change
2687 type from int to string.
2688 * config/aarch64/aarch64.c: Update default values from int
2690 * config/alpha/alpha.c (alpha_override_options_after_change):
2692 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
2693 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
2694 max skip conditionally.
2695 * config/i386/freebsd.h (SUBALIGN_LOG): New.
2696 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
2697 max skip conditionally.
2698 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
2699 max skip conditionally.
2700 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
2701 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
2702 max skip conditionally.
2703 * config/i386/i386.c (struct ptt): Change type from int to
2705 (ix86_default_align): Set default values.
2706 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
2707 max skip conditionally.
2708 * config/i386/iamcu.h (SUBALIGN_LOG): New.
2709 (ASM_OUTPUT_MAX_SKIP_ALIGN):
2710 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
2711 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
2712 max skip conditionally.
2713 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
2714 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
2715 * config/i386/x86-64.h (SUBALIGN_LOG): New.
2716 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
2717 max skip conditionally.
2718 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
2719 * config/ia64/ia64.c (ia64_option_override): Set default values
2720 for alignment options.
2721 * config/m68k/m68k.c: Handle new str_align_* options.
2722 * config/mips/mips.c (mips_set_compression_mode): Change
2724 (mips_option_override): Set default values for options.
2725 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
2727 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2729 * config/rx/rx.c (rx_option_override): Likewise.
2730 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
2731 (LABEL_ALIGN): Use align_labels_log.
2732 (LOOP_ALIGN): Use align_loops_align.
2733 * config/s390/s390.c (s390_asm_output_function_label): Use new
2735 * config/sh/sh.c (sh_override_options_after_change):
2736 Change type of constants.
2737 * config/spu/spu.c (spu_sched_init): Likewise.
2738 * config/sparc/sparc.c (sparc_option_override): Set default
2740 * config/visium/visium.c (visium_option_override): Likewise.
2741 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
2742 emit p2align format with last argument if it's not needed.
2743 * doc/invoke.texi: Document extended format of -falign-*.
2744 * final.c: Use align_labels alignment.
2745 * flags.h (struct target_flag_state): Change type to use
2747 (struct align_flags_tuple): New.
2748 (struct align_flags): Likewise.
2749 (align_loops_log): Redefine macro to use new types.
2750 (align_loops_max_skip): Redefine macro to use new types.
2751 (align_jumps_log): Redefine macro to use new types.
2752 (align_jumps_max_skip): Redefine macro to use new types.
2753 (align_labels_log): Redefine macro to use new types.
2754 (align_labels_max_skip): Redefine macro to use new types.
2755 (align_functions_log): Redefine macro to use new types.
2756 (align_loops): Redefine macro to use new types.
2757 (align_jumps): Redefine macro to use new types.
2758 (align_labels): Redefine macro to use new types.
2759 (align_functions): Redefine macro to use new types.
2760 (align_functions_max_skip): Redefine macro to use new types.
2761 (align_loops_value): New macro.
2762 (align_jumps_value): New macro.
2763 (align_labels_value): New macro.
2764 (align_functions_value): New macro.
2765 * function.c (invoke_set_current_function_hook): Propagate
2766 alignment values from flags to global variables default in
2768 * ipa-icf.c (sem_function::equals_wpa): Use
2769 cl_optimization_option_eq instead of memcmp.
2770 * lto-streamer.h (cl_optimization_stream_out): Support streaming
2772 (cl_optimization_stream_in): Likewise.
2773 * optc-save-gen.awk: Support strings in cl_optimization.
2774 * opth-gen.awk: Likewise.
2775 * opts.c (finish_options): Remove error checking of invalid
2777 (MAX_CODE_ALIGN): Remove.
2778 (MAX_CODE_ALIGN_VALUE): Likewise.
2779 (parse_and_check_align_values): New function.
2780 (check_alignment_argument): Likewise.
2781 (common_handle_option): Use check_alignment_argument.
2782 * opts.h (parse_and_check_align_values): Declare.
2783 * toplev.c (init_alignments): Remove.
2784 (read_log_maxskip): New.
2785 (parse_N_M): Likewise.
2786 (parse_alignment_opts): Likewise.
2787 (backend_init_target): Remove usage of init_alignments.
2788 * toplev.h (parse_alignment_opts): Declare.
2789 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
2791 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
2792 * tree.c (cl_option_hasher::equal): New.
2793 * varasm.c: Use new global macros.
2795 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
2797 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
2798 Use a simpler align directive also if MAXSKIP = ALIGN-1.
2799 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2800 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2801 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2802 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
2803 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
2804 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
2805 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
2806 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2807 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2808 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2809 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2811 2018-07-04 Martin Liska <mliska@suse.cz>
2812 Jonathan Wakely <jwakely@redhat.com>
2814 * coverage.c: Use correct type.
2815 * doc/invoke.texi: Language correction.
2817 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
2820 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
2821 ENDBRANCH for non-tail call which may return via indirect branch.
2822 * doc/extend.texi: Document indirect_return attribute.
2824 2018-07-03 Martin Sebor <msebor@redhat.com>
2826 PR tree-optimization/86274
2827 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
2829 (format_floating): Correct handling of infinities and NaNs.
2831 2018-07-03 Martin Sebor <msebor@redhat.com>
2833 * print-tree.c (print_real_cst): New function.
2834 (print_node_brief): Call it.
2835 (print_node): Ditto.
2837 2018-07-03 Jeff Law <law@redhat.com>
2839 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
2840 into a single pattern.
2842 * config/h8300/h8300.md (ors code_iterator): New.
2843 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
2844 a single pattern and single splitter.
2845 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
2846 (iorqi3_1, xorqi3_1): Likewise.
2847 (iorqi3, xorqi3 expanders): Similarly.
2849 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
2850 (movmd_internal) into a single pattern using the P mode iterator.
2851 (movmd splitters): Similarly.
2852 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
2853 (movsd splitters): Similarly.
2855 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
2856 ADDB, ADDW and ADDL into a single ADD attribute which selects the
2857 right table based on the size of the operand.
2858 * config/h8300/h8300.md (length_table): Corresponding changes. All
2859 references to "addb", "addw" and "addl" changed to "add".
2860 (btst patterns): Merge two variants into a single pattern.
2861 (tstqi, tsthi): Likewise.
2862 (addhi3_incdec, addsi3_incdec): Likewise.
2863 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
2864 (mulhi3, mulsi3): Likewise.
2865 (udivhi3, udivsi3): Likewise.
2866 (divhi3, divsi3): Likewise.
2867 (andorqi3, andorhi3, andorsi3): Likewise.
2869 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
2872 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
2873 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
2875 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2877 PR tree-optimization/85694
2878 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
2879 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
2881 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
2882 (<u>avg<mode>3_ceil): New patterns.
2884 2018-07-03 David Malcolm <dmalcolm@redhat.com>
2886 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
2887 scan-tree-dump directive.
2888 * gcc.dg/vect/slp-perm-2.c: Likewise.
2889 * gcc.dg/vect/slp-perm-3.c: Likewise.
2890 * gcc.dg/vect/slp-perm-5.c: Likewise.
2891 * gcc.dg/vect/slp-perm-6.c: Likewise.
2892 * gcc.dg/vect/slp-perm-7.c: Likewise.
2893 * gcc.dg/vect/slp-perm-8.c: Likewise.
2895 2018-07-03 Marek Polacek <polacek@redhat.com>
2898 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
2900 2018-07-03 Richard Biener <rguenther@suse.de>
2903 * tree-ssa-structalias.c (find_func_clobbers): Properly
2904 handle indirect calls.
2906 2018-07-03 Jeff Law <law@redhat.com>
2908 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
2909 (shifts): New code iterator.
2910 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
2911 expander. Fix HImode handling on H8/SX.
2912 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
2913 (subqi3, subhi3, subsi3 expanders): Likewise.
2914 (andqi3, andhi3, andsi3 expanders): Likewise.
2915 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
2916 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
2917 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
2918 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
2919 (zero_extendqihi2, zero_extendqisi2): Likewise.
2920 (extendqihi2, extendqisi2): Likewise.
2921 (rotlqi3, rotlhi3, rotlsi3): Likewise.
2922 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
2923 (rotlqi3_1, rotlhi3_1): Likewise.
2924 (logicalhi3_sn, logicalsi3_sn): Likewise.
2925 (logicalhi3, logicalsi3): Likewise.
2927 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2929 * tree-vect-patterns.c (vect_recog_rotate_pattern)
2930 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
2931 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
2932 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
2933 type to append_pattern_def_seq instead of creating a stmt_vec_info
2935 (build_mask_conversion): Likewise. Remove vinfo argument.
2936 (vect_add_conversion_to_patterm): Likewise, renaming to...
2937 (vect_add_conversion_to_pattern): ...this.
2938 (vect_recog_mask_conversion_pattern): Update call to
2939 build_mask_conversion. Pass the vector type to
2940 append_pattern_def_seq here too.
2941 (vect_recog_gather_scatter_pattern): Update call to
2942 vect_add_conversion_to_pattern.
2944 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2946 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
2947 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
2948 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
2949 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
2950 STMT_VINFO_PATTERN_DEF_SEQ to null here.
2951 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
2952 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
2953 append_pattern_def_seq instead of new_pattern_def_seq.
2954 (vect_recog_divmod_pattern): Do both of the above.
2955 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
2958 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2960 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
2961 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
2962 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
2963 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
2964 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
2965 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
2966 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
2967 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
2968 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
2969 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
2970 parameter with a single stmt_vec_info.
2971 (vect_recog_func_ptr): Likewise.
2972 (vect_recog_gather_scatter_pattern): Likewise, folding in...
2973 (vect_try_gather_scatter_pattern): ...this.
2974 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
2975 the stmt_vec_info of the statement to be matched. Don't clear
2976 STMT_VINFO_RELATED_STMT.
2977 (vect_pattern_recog): Update call accordingly.
2979 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2981 PR tree-optimization/85694
2982 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
2983 (uavgM3_ceil): Document new optabs.
2984 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
2985 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
2987 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
2988 (savg_ceil_optab): New optabs.
2989 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
2990 (vect_vect_recog_func_ptrs): Add it.
2991 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
2992 constant directly from the associated lhs.
2994 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2996 * tree-vect-patterns.c (vect_split_statement): New function.
2997 (vect_convert_input): Use it to try to split an existing cast.
2999 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
3001 * poly-int.h (print_hex): New function.
3002 * dumpfile.h (dump_dec, dump_hex): Declare.
3003 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
3004 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
3005 min_input_precision, operation_precision and operation_sign.
3006 * tree-vect-patterns.c (vect_get_range_info): New function.
3007 (vect_same_loop_or_bb_p, vect_single_imm_use)
3008 (vect_operation_fits_smaller_type): Delete.
3009 (vect_look_through_possible_promotion): Add an optional
3010 single_use_p parameter.
3011 (vect_recog_over_widening_pattern): Rewrite to use new
3012 stmt_vec_info infomration. Handle one operation at a time.
3013 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
3014 (vect_truncatable_operation_p, vect_set_operation_type)
3015 (vect_set_min_input_precision): New functions.
3016 (vect_determine_min_output_precision_1): Likewise.
3017 (vect_determine_min_output_precision): Likewise.
3018 (vect_determine_precisions_from_range): Likewise.
3019 (vect_determine_precisions_from_users): Likewise.
3020 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
3021 (vect_vect_recog_func_ptrs): Put over_widening first.
3023 (vect_pattern_recog): Call vect_determine_precisions.
3025 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
3027 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
3028 statements that have been replaced by further pattern statements.
3029 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
3031 2018-07-03 Richard Biener <rguenther@suse.de>
3033 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
3034 always set *dt. Dump vectype in vectype overload.
3035 * dumpfile.h (dump_gimple_expr): New function.
3036 (dump_gimple_expr_loc): Likewise.
3037 * dumpfile.c (dump_gimple_expr): New function.
3038 (dump_gimple_expr_loc): Likewise.
3040 2018-07-02 Jeff Law <law@redhat.com>
3042 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
3043 the H8/300, H8/300H and H8/S variants into a single pattern.
3044 (movhi_h8300, movqi_h8300hs): Similarly.
3045 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
3046 (QHI mode iterator): New.
3048 * config/h8300/h8300.md: Remove trailing whitespace.
3050 2018-07-02 Jim Wilson <jimw@sifive.com>
3052 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
3053 instead of emit_insn for interrupt returns.
3054 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
3055 (riscv_sret, riscv_uret): Likewise.
3057 2018-07-02 David Malcolm <dmalcolm@redhat.com>
3059 * pretty-print.c (selftest::test_pp_format): Move save and restore
3060 of quotes to class auto_fix_quotes, and add an instance.
3061 * selftest.c: Include "intl.h".
3062 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
3063 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
3064 * selftest.h (selftest::auto_fix_quotes): New class.
3066 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
3068 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
3069 (aarch64_sve_prepare_conditional_op): Remove.
3070 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
3071 Allow aarch64_simd_reg_or_zero as select operand; remove
3072 the aarch64_sve_prepare_conditional_op call.
3073 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
3074 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
3075 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
3076 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
3077 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
3078 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
3079 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
3080 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
3081 and a splitters to match all of the *_any patterns.
3082 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
3084 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
3085 (SVE_COND_FP_BINARY_REV): Remove.
3086 (sve_int_op_rev, sve_fp_op_rev): New.
3087 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
3088 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
3089 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
3090 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
3091 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
3092 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
3093 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
3094 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
3095 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
3097 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
3098 Remove match_dup 1 from the inner unspec.
3099 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
3101 * config/aarch64/aarch64.md (movprfx): New attr.
3102 (length): Default movprfx to 8.
3103 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
3104 (*madd<SVE_I>, *msub<SVE_I): Likewise.
3105 (*<su>mul<SVE_I>3_highpart): Likewise.
3106 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
3107 (*v<ASHIFT><SVE_I>3): Likewise.
3108 (*<su><MAXMIN><SVE_I>3): Likewise.
3109 (*<su><MAXMIN><SVE_F>3): Likewise.
3110 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
3111 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
3112 (*div<SVE_F>4): Likewise.
3114 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
3116 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
3119 2018-07-02 Richard Biener <rguenther@suse.de>
3121 PR tree-optimization/86363
3122 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
3123 memset argument refers to a non-variable address.
3125 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
3127 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
3128 {PLUS,MINUS}_EXPR code to...
3129 (adjust_symbolic_bound): ...here,
3130 (combine_bound): ...here,
3131 (set_value_range_with_overflow): ...and here.
3133 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
3135 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
3137 (extract_range_from_abs_expr): ...here.
3139 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
3141 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
3142 -fno-omit-frame-pointer when not optimizing.
3144 2018-07-02 Martin Liska <mliska@suse.cz>
3147 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
3148 (propagate_nothrow): Likewise.
3150 2018-07-02 Martin Liska <mliska@suse.cz>
3153 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
3155 2018-07-02 David Malcolm <dmalcolm@redhat.com>
3157 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
3158 function in r262149, changing "loc" param from source_location to
3159 const dump_location_t &.
3160 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
3161 declaration, as above.
3163 2018-07-01 Paul Koning <ni1d@arrl.net>
3165 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
3166 -munit-asm, -mgnu-asm, -mdec-asm.
3167 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
3168 (pdp11_output_labelref): New.
3169 (pdp11_output_def): New.
3170 (pdp11_output_addr_vec_elt): New.
3171 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
3172 %# and %@ format codes.
3173 (pdp11_option_override): New.
3174 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
3175 (pdp11_output_ident): New.
3176 (pdp11_asm_named_section): New.
3177 (pdp11_asm_init_sections): New.
3178 (pdp11_file_start): New.
3179 (pdp11_file_end): New.
3180 (output_ascii): Use .ascii/.asciz for -mdec-asm.
3181 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
3182 %o, like %c but octal.
3183 (pdp11_option_override): New.
3184 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
3186 (DATA_SECTION_ASM_OP): Ditto.
3187 (READONLY_DATA_SECTION_ASM_OP): New.
3188 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
3189 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
3190 (ASM_OUTPUT_LABELREF): Ditto.
3191 (ASM_OUTPUT_DEF): Ditto.
3192 (ASM_OUTPUT_EXTERNAL): New.
3193 (ASM_OUTPUT_SOURCE_FILENAME): New.
3194 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
3195 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
3196 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
3197 %# and %@ format codes.
3198 * config/pdp11/pdp11.opt (mgnu-asm): New.
3199 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
3200 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
3201 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
3203 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
3205 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
3206 dereferencing path[] beyond its length.
3208 (debug_all_paths): New.
3209 (rewire_first_differing_edge): New.
3210 (adjust_paths_after_duplication): New.
3211 (duplicate_thread_path): Call adjust_paths_after_duplication.
3213 (thread_through_all_blocks): Add new argument to
3214 duplicate_thread_path.
3216 2018-06-30 Jim Wilson <jimw@sifive.com>
3218 * config/riscv/predicates.md (p2m1_shift_operand): New.
3219 (high_mask_shift_operand): New.
3220 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
3221 pattern using p2m1_shift_operand.
3222 (lshsi3_zero_extend_3+2): New combiner pattern using
3223 high_mask_shift_operand.
3225 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
3227 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
3229 (vect_recog_rotate_pattern): ...here.
3230 (vect_convert_input): Try to insert casts of invariants in the
3232 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
3233 preheader to be empty.
3235 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
3237 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
3238 vector type. If given, install it in the new statement's
3240 (vect_element_precision): New function.
3241 (vect_unpromoted_value): New struct.
3242 (vect_unpromoted_value::vect_unpromoted_value): New function.
3243 (vect_unpromoted_value::set_op): Likewise.
3244 (vect_look_through_possible_promotion): Likewise.
3245 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
3246 (vect_widened_op_tree, vect_convert_input): Likewise.
3247 (vect_convert_inputs, vect_convert_output): Likewise.
3248 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
3249 to handle the optional cast of the multiplication result and
3250 vect_widened_op_tree to detect the widened multiplication itself.
3251 Do not require the input and output of promotion casts to have
3252 the same sign, but base the signedness of the operation on the
3253 input rather than the result. If the pattern includes two
3254 promotions, check that those promotions have the same sign.
3255 Do not restrict the MULT_EXPR handling to a double-width result;
3256 handle quadruple-width results and wider. Use vect_convert_inputs
3257 to convert the inputs to the common type.
3258 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
3259 to handle the optional cast of the ABS result. Also allow a sign
3260 change or a sign extension between the ABS and MINUS.
3261 Use vect_widened_op_tree to detect the widened subtraction and use
3262 vect_convert_inputs to convert the inputs to the common type.
3263 (vect_handle_widen_op_by_const): Delete.
3264 (vect_recog_widen_op_pattern): New function.
3265 (vect_recog_widen_mult_pattern): Use it.
3266 (vect_recog_widen_shift_pattern): Likewise.
3267 (vect_recog_widen_sum_pattern): Use
3268 vect_look_through_possible_promotion to handle the promoted
3271 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
3273 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
3274 the containing gimple_seq *.
3275 * gimple-iterator.h (gsi_for_stmt): Declare it.
3276 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
3277 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
3278 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
3279 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
3280 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
3282 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
3284 (vect_mark_pattern_stmts): ...here. Handle cases in which the
3285 statement being replaced is part of an existing pattern
3286 definition sequence, inserting the new pattern statements before
3288 (vect_pattern_recog_1): Don't return a bool. If the statement
3289 is already part of a pattern, instead apply pattern matching
3290 to the pattern definition statements. Don't clear the
3291 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
3292 (vect_pattern_recog): Don't break after the first match;
3293 continue processing the pattern definition statements instead.
3294 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
3296 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
3298 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
3299 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
3300 (vect_recog_widen_sum_pattern): Use it.
3302 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
3304 * tree-vect-loop.c (vectorizable_reduction): Assert that the
3305 phi is not a pattern statement and has not been replaced by
3306 a pattern statement.
3307 * tree-vect-patterns.c (type_conversion_p): Don't check
3308 STMT_VINFO_IN_PATTERN_P.
3309 (vect_recog_vector_vector_shift_pattern): Likewise.
3310 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
3311 the pattern statement rather than the original statement; check
3312 directly for a WIDEN_MULT_EXPR here.
3313 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
3314 vect_is_simple_use to return the pattern statement rather
3315 than the original statement; use is_pattern_stmt_p to check
3316 for such a pattern statement.
3317 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
3318 to return the pattern statement rather than the original statement;
3319 don't do the same transformation here.
3320 (vect_is_simple_use): If the defining statement has been replaced
3321 by a pattern statement, return the pattern statement instead.
3322 Remove the corresponding (local) transformation from the vectype
3325 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
3327 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
3328 end and default to null.
3329 * tree-vect-loop.c (vect_create_epilog_for_reduction)
3330 (vectorizable_reduction): Update calls accordingly, dropping the
3331 gimple ** argument if the passed-back statement isn't needed.
3332 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
3333 (vect_recog_rotate_pattern): Likewise.
3334 (vect_recog_mask_conversion_pattern): Likewise.
3335 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
3336 (vect_mask_constant_operand_p): Likewise.
3337 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
3338 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
3339 (get_group_load_store_type, get_load_store_type): Likewise.
3340 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
3341 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
3342 (vectorizable_conversion, vectorizable_assignment): Likewise.
3343 (vectorizable_shift, vectorizable_operation): Likewise.
3344 (vectorizable_store, vect_is_simple_cond): Likewise.
3345 (vectorizable_condition, vectorizable_comparison): Likewise.
3346 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
3347 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
3348 and move it to the end. Cope with null def_stmt_outs.
3350 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
3352 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
3354 2018-06-29 Jeff Law <law@redhat.com>
3356 * config/v850/v850.c (v850_legitimate_address_p): Handle large
3357 displacements for TARGET_V850E2V3 and newer.
3358 (TARGET_LRA_P): Remove. Defaults to LRA now.
3359 * config/v850/v850.md (sign23byte_load): Remove.
3360 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
3361 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
3363 2018-06-29 Martin Liska <mliska@suse.cz>
3366 * coverage.c (coverage_init): Mangle full path name.
3367 * doc/invoke.texi: Document the change.
3368 * gcov-io.c (mangle_path): New.
3369 * gcov-io.h (mangle_path): Likewise.
3370 * gcov.c (mangle_name): Use mangle_path for path mangling.
3372 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3374 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
3375 if starting source register is not even.
3377 2018-06-29 Martin Liska <mliska@suse.cz>
3379 PR tree-optimization/86263
3380 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
3381 Make edge redirection.
3383 2018-06-29 David Malcolm <dmalcolm@redhat.com>
3385 * dumpfile.c (dump_loc): Add indentation based on scope depth.
3386 (dump_scope_depth): New variable.
3387 (get_dump_scope_depth): New function.
3388 (dump_begin_scope): New function.
3389 (dump_end_scope): New function.
3390 * dumpfile.h (get_dump_scope_depth): New declaration.
3391 (dump_begin_scope): New declaration.
3392 (dump_end_scope): New declaration.
3393 (class auto_dump_scope): New class.
3394 (AUTO_DUMP_SCOPE): New macro.
3395 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
3398 2018-06-29 Richard Biener <rguenther@suse.de>
3400 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
3401 compute_all_dependences succeeds.
3402 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
3403 exceed --param loop-max-datarefs-for-datadeps.
3405 2018-06-29 Jakub Jelinek <jakub@redhat.com>
3407 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
3409 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
3412 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
3413 alternative 0 in preferred_for_speed attribute.
3415 2018-06-28 Paul Koning <ni1d@arrl.net>
3417 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
3418 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
3419 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
3420 * config/pdp11/pdp11.md: Correct "length" attribute calculation
3421 for shift insn patterns.
3423 2018-06-28 David Malcolm <dmalcolm@redhat.com>
3425 * cgraph.c (cgraph_node::get_body): Replace assignments to
3426 "dump_file" with calls to set_dump_file.
3427 * dumpfile.c (alt_dump_file): Make static, and group with...
3428 (alt_flags): ...this definition.
3429 (dumps_are_enabled): New variable.
3430 (refresh_dumps_are_enabled): New function.
3431 (set_dump_file): New function.
3432 (set_alt_dump_file): New function.
3433 (gcc::dump_manager::dump_start): Replace assignments to
3434 "dump_file" and "alt_dump_file" with calls to set_dump_file and
3436 (gcc::dump_manager::dump_finish): Likewise.
3437 * dumpfile.h (alt_dump_file): Delete decl.
3438 (dumps_are_enabled): New variable decl.
3439 (set_dump_file): New function decl.
3440 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
3442 * tree-nested.c (lower_nested_functions): Replace assignments to
3443 "dump_file" with calls to set_dump_file.
3445 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
3447 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
3448 goto_locus of each outgoing edge of each basic block.
3450 2018-06-28 Richard Biener <rguenther@suse.de>
3452 * dwarf2out.c (decl_scope_table): Remove.
3453 (push_decl_scope): Likewise.
3454 (pop_decl_scope): Likewise.
3455 (gen_type_die_for_member): Do not call push/pop_decl_scope.
3456 (gen_struct_or_union_type_die): Likewise.
3457 (gen_tagged_type_die): Likewise.
3458 (dwarf2out_init): Do not initialize decl_scope_table.
3459 (dwarf2out_c_finalize): Do not free it.
3461 2018-06-28 Richard Biener <rguenther@suse.de>
3463 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
3464 deciding whether to not re-use a DIE.
3466 2018-06-28 Richard Biener <rguenther@suse.de>
3468 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
3469 DW_AT_abstract_origin attribute.
3471 2018-06-28 Martin Liska <mliska@suse.cz>
3473 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
3474 Use newly introduced constants.
3475 * tree-switch-conversion.h (struct jump_table_cluster):
3476 Define max_ratio_for_size and max_ratio_for_speed.
3478 2018-06-28 Martin Liska <mliska@suse.cz>
3480 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
3481 Add new checking assert to catch invalid state.
3482 (jump_table_cluster::can_be_handled): Handle single case
3484 (jump_table_cluster::is_beneficial): Bail out for such case.
3485 (bit_test_cluster::find_bit_tests):
3486 Add new checking assert to catch invalid state.
3487 (bit_test_cluster::can_be_handled): Handle single case
3489 (bit_test_cluster::is_beneficial): Bail out for such case.
3490 (switch_decision_tree::analyze_switch_statement):
3493 2018-06-28 Martin Liska <mliska@suse.cz>
3495 * common.opt: Introduce -completion option.
3496 * gcc.c (driver_handle_option): Handle it.
3497 (driver::main): Print completions if completion
3499 * opt-suggestions.c (option_proposer::get_completions):
3501 (option_proposer::suggest_completion): Likewise.
3502 (option_proposer::find_param_completions): Likewise.
3503 (verify_autocompletions): Likewise.
3504 (test_completion_valid_options): Likewise.
3505 (test_completion_valid_params): Likewise.
3506 (in_completion_p): Likewise.
3507 (empty_completion_p): Likewise.
3508 (test_completion_partial_match): Likewise.
3509 (test_completion_garbage): Likewise.
3510 (opt_proposer_c_tests): Likewise.
3511 * opt-suggestions.h: Declare new functions.
3512 * opts.c (common_handle_option): Handle OPT__completion_.
3513 * selftest-run-tests.c (selftest::run_tests): Add
3514 opt_proposer_c_tests.
3515 * selftest.c (assert_str_startswith): New.
3516 * selftest.h (assert_str_startswith): Likewise.
3517 (opt_proposer_c_tests): New.
3518 (ASSERT_STR_STARTSWITH): Likewise.
3520 2018-06-28 Martin Liska <mliska@suse.cz>
3522 * Makefile.in: Add opt-suggestions.o.
3523 * gcc-main.c: Include opt-suggestions.h.
3524 * gcc.c (driver::driver): Likewise.
3525 (driver::~driver): Remove m_option_suggestions.
3526 (driver::build_option_suggestions): Moved to option_proposer.
3527 (driver::suggest_option): Likewise.
3528 (driver::handle_unrecognized_options): Use option_proposer.
3529 * gcc.h (class driver): Add new memver m_option_proposer.
3530 * opt-suggestions.c: New file.
3531 * opt-suggestions.h: New file.
3533 2018-06-28 Martin Liska <mliska@suse.cz>
3535 * vec.h (class auto_string_vec): New (moved from auto_argvec).
3536 (auto_string_vec::~auto_string_vec): Likewise.
3538 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
3540 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
3541 prevent_decl_creation_for_types fields up and add reset_location field.
3542 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
3543 statement if id->reset_location is true.
3544 (copy_edges_for_bb): Do not set goto_locus on the new edges if
3545 id->reset_location is true.
3546 (copy_phis_for_bb): Force input_location on the arguments if
3547 id->reset_location is true.
3548 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
3549 is set on the function to be inlined.
3551 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
3553 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
3555 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
3557 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
3558 registers for Pmode.
3559 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
3560 hard registers for the clobbered pseudo.
3562 2018-06-27 Paul Koning <ni1d@arrl.net>
3564 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
3565 mutually exclusive options.
3566 * config/pdp11/constraints.md (h): New constraint.
3567 (O): Update definition to match shift code generation.
3568 (D): New constraint.
3569 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
3571 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
3573 (output_jump): Change arguments.
3574 (pdp11_fixed_cc_regs): New function.
3575 (pdp11_cc_mode): Ditto.
3576 (pdp11_expand_shift): Ditto.
3577 (pdp11_assemble_shift): Ditto.
3578 (pdp11_small_shift): Ditto.
3579 (pdp11_branch_cost): Remove.
3580 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
3582 (pdp11_register_move_cost): Update for CC registers.
3583 (pdp11_rtx_costs): Add case for LSHIFTRT.
3584 (pdp11_output_jump): Add CCNZ mode conditional branches.
3585 (notice_update_cc_on_set): Remove.
3586 (pdp11_cc_mode): New function.
3587 (simple_memory_operand): Correct pre/post decrement case.
3588 (no_side_effect_operand): New function.
3589 (pdp11_regno_reg_class): Add CC_REGS class.
3590 (pdp11_fixed_cc_regs): New function.
3591 (pdp11_small_shift): New function.
3592 (pdp11_expand_shift): New function to expand shift insns.
3593 (pdp11_assemble_shift): New function to output shifts.
3594 (pdp11_branch_cost): Remove.
3595 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
3596 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
3597 (WCHAR_TYPE): Ditto.
3598 (PTRDIFF_TYPE): Ditto.
3599 (ADJUST_INSN_LENGTH): New macro.
3600 (FIXED_REGISTERS): Add CC registers.
3601 (CALL_USED_REGISTERS): Ditto.
3603 (REG_CLASS_NAMES): Ditto.
3604 (REG_CLASS_CONTENTS): Ditto.
3605 (SELECT_CC_MODE): Use new function.
3606 (TARGET_FLAGS_REGNUM): New macro.
3607 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
3608 (cc0_reg_rtx): Remove.
3609 (CC_STATUS_MDEP): Remove.
3610 (CC_STATUS_MDEFP_INIT): Remove.
3611 (CC_IN_FPU): Remove.
3612 (NOTICE_UPDATE_CC): Remove.
3613 (REGISTER_NAMES): Add CC registers.
3614 (BRANCH_COST): Change to constant 1.
3615 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
3617 * config/pdp11/pdp11.opt (mbcopy): Remove.
3618 (mbcopy-builtin): Remove.
3619 (mbranch-cheap): Remove.
3620 (mbranch-expensive): Remove.
3621 * config/pdp11/predicates.md (expand_shift_operand): Update to
3622 match shift code generation.
3623 (ccnz_operator): New predicate.
3624 * doc/invoke.texi (PDP-11 Options): Remove deleted options
3625 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
3626 Remove non-existent option -mabshi, -mno-abshi. Document mutually
3628 * doc/md.texi (PDP-11): Document new D and h constraints. Update
3629 description of O constraint.
3631 2018-06-27 Jeff Law <law@redhat.com>
3632 Austin Law <austinklaw@gmail.com>
3634 * config/v850/v850.md (addsi3_set_flags): New pattern.
3635 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
3636 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
3637 (zero_extendhisi2_v850_set_flags): Likewise.
3638 (zero_extendqisi2_v850_set_flags): Likewise.
3639 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
3640 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
3641 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
3643 * config/v850/v850-protos.h (notice_update_cc): Remove.
3644 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
3645 (v850_print_operand): Handle 'D' and "d".
3646 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
3647 Add handling of arithmetic/logical operations compared against zero.
3648 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
3649 Do not look at v850_compare_op, instead get mode from last argument.
3650 (v850_gen_compare): Remove
3651 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
3652 after reload for prologue insns.
3653 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
3655 (construct_save_jarl): Likewise.
3656 (TARGET_FLAGS_REGNUM): Define.
3657 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
3658 (NOTICE_UPDATE_CC): Remove.
3659 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
3660 than cc0. Conditionalize on reload_completed.
3661 (cmpsi_insn, setfcc_insn): Likewise.
3662 (tst1 splitter): Turn into define_and_split which sets the flags
3664 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
3665 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
3666 (cstoresf4, cstoredf4): Clobber the flags.
3667 (cmpsi, cmpsf, cmpdf): Remove expanders.
3668 (setf_insn): Remove pattern.
3669 (addsi3): Turn into define_and_split which clobbers the flags after
3670 reload and a suitable pattern (addsi3_clobber_flags) for use after
3672 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
3673 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
3674 (ashrsi3, ashrsi3_v850e2): Likewise.
3675 (bins): Clobber the flags.
3676 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
3677 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
3678 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
3679 (call_value_internal_short, call_value_internal_long): Likewise.
3680 (callt_save_interrupt, callt_return_interrupt): Likewise.
3681 (save_interrupt, return_interrupt): Likewise.
3682 (callt_save_all_interrupt, save_all_interrupt): Likewise.
3683 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
3684 (restore_all_interrupt, _restore_all_interrupt): Likewise.
3685 (All FP comparisons): Only allow after reload has completed.
3687 (divh, divhu): Tweak output template.
3688 (branch_z_normal, branch_z_invert): Remove
3689 (branch_nz_normal, branch_nz_invert): Likewise.
3690 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
3692 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
3693 * config/v850/v850.c (notice_update_cc): Remove.
3694 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
3695 (CC_NO_CARRY): Likewise.
3696 (NOTICE_UPDATE_CC): Define to nothing.
3697 * config/v850/v850.md: Remove block comment on cc0 handling
3698 Remove "cc" attribute from all patterns. Remove cc_status handling
3699 from all patterns. Minor formatting fixes.
3701 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3703 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
3704 (cortex-a76.cortex-a55): Likewise.
3705 * config/aarch64/aarch64-tune.md: Regenerate.
3706 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
3707 cortex-a76.cortex-a55.
3709 2018-06-27 Jeff Law <law@redhat.com>
3711 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
3712 (MULTILIB_DIRNAMES): Similarly.
3714 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
3716 * gimple.h (gimple_return_retbnd): Delete.
3717 (gimple_return_set_retbnd): Likewise.
3718 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
3719 gimple_return_set_retbnd.
3720 * gimple-pretty-print.c (dump_gimple_return): Remove call to
3721 gimple_return_retbnd and adjust.
3722 * tree-inline.h (struct copy_body_data): Remove retbnd field.
3723 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
3724 Explicitly return NULL in a couple more cases. Move assertion
3725 on debug statements and remove unreachable code.
3726 (reset_debug_binding): Do not test id->retbnd.
3727 (expand_call_inline): Do not set it.
3729 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
3731 * configure.ac: Add --disable-gcov option.
3732 * configure: Regenerate.
3733 * Makefile.in: Honour @enable_gcov@.
3734 * doc/install.texi: Document --disable-gcov.
3736 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3738 * config/arm/arm-cpus.in (cortex-a76): New entry.
3739 (cortex-a76.cortex-a55): Likewise.
3740 * config/arm/arm-tables.opt: Regenerate.
3741 * config/arm/arm-tune.md: Likewise.
3742 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
3743 * doc/invoke.texi (ARM Options): Document cortex-a76 and
3744 cortex-a76.cortex-a55.
3746 2018-06-27 Tamar Christina <tamar.christina@arm.com>
3749 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
3751 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
3753 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
3755 (EPILOGUE_USES): Likewise.
3757 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
3759 * tree-inline.c (remap_location): New function extracted from...
3760 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
3761 (copy_phis_for_bb): ...here. Call remap_location.
3762 (copy_cfg_body): Adjust call to copy_edges_for_bb.
3764 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
3766 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
3767 unaligned vsx for 16B memset.
3769 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
3772 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
3773 ieee128_float_type_node to long_double_type_node unless
3774 TARGET_LONG_DOUBLE_128 is set.
3776 2018-06-26 David Malcolm <dmalcolm@redhat.com>
3778 * cfgloop.c (get_loop_location): Convert return type from
3779 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
3780 by implicit construction from rtx_insn *, and using
3781 dump_user_location_t::from_function_decl for the fallback case.
3782 * cfgloop.h (get_loop_location): Convert return type from
3783 location_t to dump_user_location_t.
3784 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
3785 dump_printf_loc to pass in a dump_location_t rather than a
3786 location_t, via the gimple stmt.
3787 * coverage.c (get_coverage_counts): Update calls to
3788 dump_printf_loc to pass in dump_location_t rather than a
3790 * doc/optinfo.texi (Dump types): Convert example of
3791 dump_printf_loc from taking "locus" to taking "insn". Update
3792 description of the "_loc" calls to cover dump_location_t.
3793 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
3795 (dump_user_location_t::dump_user_location_t): New constructors,
3796 from gimple *stmt and rtx_insn *.
3797 (dump_user_location_t::from_function_decl): New function.
3798 (dump_loc): Make static.
3799 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
3800 const dump_location_t &.
3801 (dump_generic_expr_loc): Delete.
3802 (dump_printf_loc): Convert param "loc" from location_t to
3803 const dump_location_t &.
3804 (selftest::test_impl_location): New function.
3805 (selftest::dumpfile_c_tests): New function.
3806 * dumpfile.h: Include "profile-count.h".
3807 (class dump_user_location_t): New class.
3808 (struct dump_impl_location_t): New struct.
3809 (class dump_location_t): New class.
3810 (dump_printf_loc): Convert 2nd param from source_location to
3811 const dump_location_t &.
3812 (dump_generic_expr_loc): Delete.
3813 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
3814 const dump_location_t &.
3815 * gimple-fold.c (fold_gimple_assign): Update call to
3816 dump_printf_loc to pass in a dump_location_t rather than a
3817 location_t, via the gimple stmt.
3818 (gimple_fold_call): Likewise.
3819 * gimple-loop-interchange.cc
3820 (loop_cand::analyze_iloop_reduction_var): Update for change to
3821 check_reduction_path.
3822 (tree_loop_interchange::interchange): Update for change to
3824 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
3825 change in return-type of find_loop_location.
3826 (graphite_regenerate_ast_isl): Likewise.
3827 * graphite-optimize-isl.c (optimize_isl): Likewise.
3828 * graphite.c (graphite_transform_loops): Likewise.
3829 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
3830 pass in a dump_location_t rather than a location_t, via the
3832 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3833 * ipa.c (walk_polymorphic_call_targets): Likewise.
3834 * loop-unroll.c (report_unroll): Convert "locus" param from
3835 location_t to dump_location_t.
3836 (decide_unrolling): Update for change to get_loop_location's
3838 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
3839 location_t to dump_user_location_t.
3840 (grid_find_single_omp_among_assignments_1): Updates calls to
3841 dump_printf_loc to pass in a dump_location_t rather than a
3842 location_t, via the gimple stmt.
3843 (grid_parallel_clauses_gridifiable): Convert "tloc" from
3844 location_t to dump_location_t. Updates calls to dump_printf_loc
3845 to pass in a dump_location_t rather than a location_t, via the
3847 (grid_inner_loop_gridifiable_p): Likewise.
3848 (grid_dist_follows_simple_pattern): Likewise.
3849 (grid_gfor_follows_tiling_pattern): Likewise.
3850 (grid_target_follows_gridifiable_pattern): Likewise.
3851 (grid_attempt_target_gridification): Convert initialization
3852 of local "grid" from memset to zero-initialization; FIXME: does
3853 this require C++11? Update call to dump_printf_loc to pass in a
3854 optinfo_location rather than a location_t, via the gimple stmt.
3855 * profile.c (read_profile_edge_counts): Updates call to
3856 dump_printf_loc to pass in a dump_location_t rather than a
3858 (compute_branch_probabilities): Likewise.
3859 * selftest-run-tests.c (selftest::run_tests): Call
3861 * selftest.h (dumpfile_c_tests): New decl.
3862 * tree-loop-distribution.c (pass_loop_distribution::execute):
3863 Update for change in return type of find_loop_location.
3864 * tree-parloops.c (parallelize_loops): Likewise.
3865 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
3866 "locus" from location_t to dump_user_location_t.
3867 (canonicalize_loop_induction_variables): Likewise.
3868 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
3869 for change in return type of find_loop_location.
3870 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
3871 to dump_printf_loc to pass in a dump_location_t rather than a
3872 location_t, via the stmt.
3873 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
3875 * tree-vect-loop-manip.c (find_loop_location): Convert return
3876 type from source_location to dump_user_location_t.
3877 (vect_do_peeling): Update for above change.
3878 (vect_loop_versioning): Update for change in type of
3880 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
3881 from location_t to dump_user_location_t.
3882 (vect_estimate_min_profitable_iters): Update for change in type
3884 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
3885 location_t to dump_location_t.
3886 (vect_slp_bb): Update for change in type of vect_location.
3887 * tree-vectorizer.c (vect_location): Convert from source_location
3888 to dump_user_location_t.
3889 (try_vectorize_loop_1): Update for change in vect_location's type.
3890 (vectorize_loops): Likewise.
3891 (increase_alignment): Likewise.
3892 * tree-vectorizer.h (vect_location): Convert from source_location
3893 to dump_user_location_t.
3894 (find_loop_location): Convert return type from source_location to
3895 dump_user_location_t.
3896 (check_reduction_path): Convert 1st param from location_t to
3897 dump_user_location_t.
3898 * value-prof.c (check_counter): Update call to dump_printf_loc to
3899 pass in a dump_user_location_t rather than a location_t; update
3900 call to error_at for change in type of "locus".
3901 (check_ic_target): Update call to dump_printf_loc to
3902 pass in a dump_user_location_t rather than a location_t, via the
3905 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
3907 * config/s390/s390.h (enum processor_flags): Do not use
3908 default tune parameter when -march was specified.
3910 2018-06-26 Jakub Jelinek <jakub@redhat.com>
3913 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
3914 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
3917 2018-06-26 Richard Biener <rguenther@suse.de>
3919 PR tree-optimization/86287
3921 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
3922 (vect_analyze_loop): Initialize n_stmts.
3924 2018-06-26 Richard Biener <rguenther@suse.de>
3927 * fold-const.c (fold_convertible_p): Pointer extension
3930 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
3933 * dwarf2out.c (loc_list_has_views): Adjust comments.
3934 (dw_loc_list): Split single cross-partition range with
3937 2018-06-25 Jeff Law <law@redhat.com>
3939 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
3940 on -mbig-switch by default.
3942 * config/v850/predicates.md (const_float_1_operand): Fix match_code
3944 (const_float_0_operand): Remove unused predicate.
3945 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
3946 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
3947 (recipsf2): New expander. Original pattern now called
3949 (recipdf2, recipdf2_insn): Similarly.
3950 (rsqrtsf2, rsqrtsf2_insn): Similarly
3951 (rsqrtdf2, rsqrtdf2_insn): Similarly
3953 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
3955 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
3956 Simplify logic for FreeBSD (twice).
3958 2018-06-25 Martin Sebor <msebor@redhat.com>
3960 PR tree-optimization/86204
3961 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
3962 a strnlen result if it's less than the length of the string.
3964 2018-06-25 Martin Sebor <msebor@redhat.com>
3966 PR tree-optimization/85700
3967 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
3968 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
3969 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
3971 2018-06-25 Martin Sebor <msebor@redhat.com>
3973 * doc/extend.texi (Zero-length arrays): Update and clarify.
3975 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
3977 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
3978 added IEEE/IBM long double multilib support on PowerPC little
3979 endian Linux systems.
3980 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
3981 (MULTILIB_DEFAULTS): Likewise.
3982 * config/rs6000/rs6000.c (rs6000_option_override_internal):
3984 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
3985 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
3986 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
3988 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
3991 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
3992 (REORDER_45): Likewise.
3994 2018-06-25 Jeff Law <law@redhat.com>
3996 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
3997 dividend to 32 bits. Adjust length.
3998 (udivmodhi4): Cleanup output template. Fix length.
4000 2018-06-25 Carl Love <cel@us.ibm.com>
4002 * config/rs6000/vsx.md: Change word selector to prefered location.
4004 2018-06-25 Richard Biener <rguenther@suse.de>
4006 PR tree-optimization/86304
4007 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
4008 epilogue-if-converted loops as well.
4010 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
4012 * lto-section-out.c (lto_begin_section): Do not print section
4013 name for noaddr and unnumbered dumps.
4015 2018-06-25 Richard Biener <rguenther@suse.de>
4017 * tree-vectorizer.h (struct vec_info_shared): New structure
4018 with parts split out from struct vec_info and loop_nest from
4019 struct _loop_vec_info.
4020 (struct vec_info): Adjust accordingly.
4021 (struct _loop_vec_info): Likewise.
4022 (LOOP_VINFO_LOOP_NEST): Adjust.
4023 (LOOP_VINFO_DATAREFS): Likewise.
4024 (LOOP_VINFO_DDRS): Likewise.
4025 (struct _bb_vec_info): Likewise.
4026 (BB_VINFO_DATAREFS): Likewise.
4027 (BB_VINFO_DDRS): Likewise.
4028 (struct _stmt_vec_info): Add dr_aux member.
4029 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
4030 (DR_MISALIGNMENT_UNINITIALIZED): New.
4031 (set_dr_misalignment): Adjust.
4032 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
4033 (vect_analyze_loop): Adjust prototype.
4034 (vect_analyze_loop_form): Likewise.
4035 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
4036 Compute dependences lazily.
4037 (vect_record_base_alignments): Use shared datarefs/ddrs.
4038 (vect_verify_datarefs_alignment): Likewise.
4039 (vect_analyze_data_refs_alignment): Likewise.
4040 (vect_analyze_data_ref_accesses): Likewise.
4041 (vect_analyze_data_refs): Likewise.
4042 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
4043 constructor parameter for shared part.
4044 (vect_analyze_loop_form): Pass in shared part and adjust.
4045 (vect_analyze_loop_2): Pass in storage for the number of
4046 stmts. Move loop nest finding to the caller. Compute
4048 (vect_analyze_loop): Pass in shared part.
4049 (vect_transform_loop): Verify shared datarefs are unchanged.
4050 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
4051 constructor parameter for shared part.
4052 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
4053 (vect_slp_bb): Verify shared datarefs are unchanged before
4055 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
4057 (new_stmt_vec_info): Initialize DR_AUX misalignment to
4058 DR_MISALIGNMENT_UNINITIALIZED.
4059 * tree-vectorizer.c (vec_info::vec_info): Add constructor
4060 parameter for shared part.
4061 (vec_info::~vec_info): Adjust.
4062 (vec_info_shared::vec_info_shared): New.
4063 (vec_info_shared::~vec_info_shared): Likewise.
4064 (vec_info_shared::save_datarefs): Likewise.
4065 (vec_info_shared::check_datarefs): Likewise.
4066 (try_vectorize_loop_1): Construct shared part live for analyses
4067 of a single loop for multiple vector sizes.
4068 * tree-parloops.c (gather_scalar_reductions): Adjust.
4070 2018-06-25 Richard Biener <rguenther@suse.de>
4072 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
4073 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
4074 (vect_analyze_data_refs): Remove similar code from here and
4075 simplify accordingly.
4077 2018-06-25 Richard Biener <rguenther@suse.de>
4079 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
4080 for reverse storage order accesses rather than asserting
4081 they cannot happen here.
4083 2018-06-25 Tom de Vries <tdevries@suse.de>
4086 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
4087 Use data16 instead of .byte for insn prefix.
4089 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
4092 * parser.c (make_char_string_pack): Pass this literal chars
4093 through cpp_interpret_string.
4094 (cp_parser_userdef_numeric_literal): Check the result of
4095 make_char_string_pack.
4097 2018-06-24 Maya Rashish <coypu@sdf.org>
4099 * ginclude/stddef.h: Simplify conditions around avoiding
4100 re-definition of __size_t.
4102 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
4104 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
4105 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
4107 2018-06-22 Maya Rashish <coypu@sdf.org>
4109 * doc/invoke.texi (mno-fancy-math-387): Update for changes
4110 made to OpenBSD and NetBSD through the years.
4112 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
4114 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
4115 behavior of vec_pack (vector double, vector double) to match
4116 behavior of vec_float2 (vector double, vector double).
4118 2018-06-22 Olivier Hainque <hainque@adacore.com>
4120 * gimplify.c (gimplify_function_tree): Prevent creation
4121 of a trampoline for the address of the current function
4122 passed to entry/exit instrumentation hooks.
4124 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
4127 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
4130 2018-06-22 Martin Liska <mliska@suse.cz>
4132 PR tree-optimization/86263
4133 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
4134 Bail out if is_enabled is false.
4135 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
4137 (jump_table_cluster::is_enabled): New function.
4139 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
4141 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
4142 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
4143 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
4144 (lto_input_ts_binfo_tree_pointers): Likewise.
4145 * tree-streamer-out.c (streamer_write_tree_bitfields,
4146 write_ts_binfo_tree_pointers): Likewise.
4147 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
4149 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
4151 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
4153 2018-06-22 Martin Liska <mliska@suse.cz>
4155 * symbol-summary.h (get): Make it pure and inline move
4156 functionality from ::get function.
4157 (get): Remove and inline into ::get and ::get_create.
4158 (get_create): Move code from ::get function.
4160 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4163 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
4164 -x assembler-with-cpp.
4166 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4168 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
4169 _FILE_OFFSET_BITS=64 for C++.
4171 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
4173 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
4174 conversion insn that shows up when pr85657-3.c is compiled using
4175 IEEE 128-bit long double.
4176 (neg<mode>2_internal): Use the correct mode to check whether the
4177 mode is IBM extended.
4178 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
4179 multiply and divide external functions from being created more
4182 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
4184 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
4186 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
4187 the edge can be forwarded.
4188 (cfg_layout_merge_blocks): Likewise.
4190 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
4192 * except.c (finish_eh_generation): Commit edge insertions only after
4193 the EH edges have been redirected from post-landing to landing pads.
4195 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
4197 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
4198 create_tmp_var_for to create the FRAME decl.
4199 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
4201 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
4203 * tree-inline.c (copy_edges_for_bb): Minor tweak.
4204 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
4205 debug statement when resetting its value.
4206 (expand_call_inline): Copy the locus of the call onto the assignment
4207 of the return value, if any. Use local variable in more cases.
4209 2018-06-21 Martin Liska <mliska@suse.cz>
4211 * ipa-pure-const.c (propagate_nothrow): Use
4212 funct_state_summaries->get.
4213 (dump_malloc_lattice): Likewise.
4214 (propagate_malloc): Likewise.
4216 2018-06-21 Richard Biener <rguenther@suse.de>
4218 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
4219 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
4220 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
4222 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
4223 BLOCK_ABSTRACT_ORIGIN unconditionally.
4225 2018-06-21 David Malcolm <dmalcolm@redhat.com>
4227 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
4229 * ipa-reference.c (ipa_reference_c_finalize): Delete
4230 ipa_ref_opt_sum_summaries and set it to NULL.
4232 2018-06-21 Tom de Vries <tdevries@suse.de>
4234 PR tree-optimization/85859
4235 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
4236 test with comment from bb_no_side_effects_p.
4238 2018-06-21 Richard Biener <rguenther@suse.de>
4240 PR tree-optimization/86232
4241 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
4242 max for constant niter.
4244 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
4246 * config/aarch64/aarch64-simd.md
4247 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
4249 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
4251 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
4252 Make opernads of the unspec commutative.
4254 2018-06-21 Richard Biener <rguenther@suse.de>
4256 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
4257 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
4258 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
4259 (vect_analyze_data_ref_dependence): Re-order checks to deal with
4261 (vect_record_base_alignments): Do not record base alignment
4262 for gathers or scatters.
4263 (vect_compute_data_ref_alignment): Drop return value that is always
4264 true. Bail out early for gathers or scatters.
4265 (vect_enhance_data_refs_alignment): Bail out early for gathers
4267 (vect_find_same_alignment_drs): Likewise.
4268 (vect_analyze_data_refs_alignment): Remove dead code.
4269 (vect_slp_analyze_and_verify_node_alignment): Likewise.
4270 (vect_analyze_data_refs): For possible gathers or scatters do
4271 not create an alternate DR, just check their possible validity
4272 and mark them. Adjust DECL_NONALIASED handling to not rely
4274 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
4275 update inits of gathers or scatters.
4276 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
4277 Also copy gather/scatter flag to pattern vinfo.
4279 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
4281 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
4282 behavior of vec_packsu (vector unsigned long long, vector unsigned
4283 long long) to match behavior of vec_packs with same signature.
4285 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
4286 Thomas Schwinge <thomas@codesourcery.com>
4287 Cesar Philippidis <cesar@codesourcery.com>
4289 * gimplify.c (gimplify_scan_omp_clauses): Add support for
4290 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
4291 (gimplify_adjust_omp_clauses): Likewise.
4292 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
4293 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
4294 (gimplify_omp_target_update): Update handling of acc update and
4296 * omp-low.c (install_var_field): Remove unused parameter
4297 base_pointers_restrict.
4298 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
4299 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
4301 (omp_target_base_pointers_restrict_p): Delete.
4302 (scan_omp_target): Update call to scan_sharing_clauses.
4303 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
4305 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
4306 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
4307 (convert_local_omp_clauses): Likewise.
4308 * tree-pretty-print.c (dump_omp_clause): Likewise.
4309 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
4311 (omp_clause_code_name): Likewise.
4313 2018-06-20 Jakub Jelinek <jakub@redhat.com>
4316 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
4319 PR tree-optimization/86231
4320 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
4321 anti-range don't overwrite *vr0min before using it to compute *vr0max.
4323 2018-06-20 Tom de Vries <tdevries@suse.de>
4325 PR tree-optimization/86097
4326 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
4327 iv type if signedness of iv type is not the same as that of *nit.
4329 2018-06-20 Jakub Jelinek <jakub@redhat.com>
4331 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
4332 EDGE_EH edges, verify they are all EDGE_EH.
4334 2018-06-20 Maya Rashish <coypu@sdf.org>
4336 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
4338 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4340 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
4341 * config/aarch64/aarch64.c (xgene1_tunings): Add
4342 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
4343 (aarch64_mode_valid_for_sched_fusion_p):
4344 Allow 16-byte modes.
4345 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
4346 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
4348 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
4350 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
4351 * config/aarch64/iterators.md (VQ2): New mode iterator.
4353 2018-06-20 Martin Liska <mliska@suse.cz>
4355 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
4356 Change default ratio from 10 to 8.
4358 2018-06-20 Martin Liska <mliska@suse.cz>
4360 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
4362 (bit_test_cluster::find_bit_tests): Likewise.
4363 (switch_decision_tree::analyze_switch_statement): Find clusters.
4364 * tree-switch-conversion.h (struct jump_table_cluster): Document
4367 2018-06-20 Martin Liska <mliska@suse.cz>
4369 * tree-switch-conversion.c (switch_conversion::collect):
4370 Record m_uniq property.
4371 (switch_conversion::expand): Bail out for special conditions.
4372 (group_cluster::~group_cluster): New.
4373 (group_cluster::group_cluster): Likewise.
4374 (group_cluster::dump): Likewise.
4375 (jump_table_cluster::emit): New.
4376 (switch_decision_tree::fix_phi_operands_for_edges): New.
4377 (struct case_node): Remove struct.
4378 (jump_table_cluster::can_be_handled): New.
4379 (case_values_threshold): Moved to header.
4380 (reset_out_edges_aux): Likewise.
4381 (jump_table_cluster::is_beneficial): New.
4382 (bit_test_cluster::can_be_handled): Likewise.
4383 (add_case_node): Remove.
4384 (bit_test_cluster::is_beneficial): New.
4385 (case_bit_test::cmp): New.
4386 (bit_test_cluster::emit): New.
4387 (expand_switch_as_decision_tree_p): Remove.
4388 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
4389 (fix_phi_operands_for_edge): Likewise.
4390 (switch_decision_tree::analyze_switch_statement): New.
4391 (compute_cases_per_edge): Move ...
4392 (switch_decision_tree::compute_cases_per_edge): ... here.
4393 (try_switch_expansion): Likewise.
4394 (switch_decision_tree::try_switch_expansion): Likewise.
4395 (record_phi_operand_mapping): Likewise.
4396 (switch_decision_tree::record_phi_operand_mapping): Likewise.
4397 (emit_case_decision_tree): Likewise.
4398 (switch_decision_tree::emit): Likewise.
4399 (balance_case_nodes): Likewise.
4400 (switch_decision_tree::balance_case_nodes): Likewise.
4401 (dump_case_nodes): Likewise.
4402 (switch_decision_tree::dump_case_nodes): Likewise.
4403 (emit_jump): Likewise.
4404 (switch_decision_tree::emit_jump): Likewise.
4405 (emit_cmp_and_jump_insns): Likewise.
4406 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
4407 (emit_case_nodes): Likewise.
4408 (switch_decision_tree::emit_case_nodes): Likewise.
4409 (conditional_probability): Remove.
4410 * tree-switch-conversion.h (enum cluster_type): New.
4412 (struct cluster): Likewise.
4413 (cluster::cluster): Likewise.
4414 (struct simple_cluster): Likewise.
4415 (simple_cluster::simple_cluster): Likewise.
4416 (struct group_cluster): Likewise.
4417 (struct jump_table_cluster): Likewise.
4418 (struct bit_test_cluster): Likewise.
4419 (struct min_cluster_item): Likewise.
4420 (struct case_tree_node): Likewise.
4421 (case_tree_node::case_tree_node): Likewise.
4422 (jump_table_cluster::case_values_threshold): Likewise.
4423 (struct case_bit_test): Likewise.
4424 (struct switch_decision_tree): Likewise.
4425 (struct switch_conversion): Likewise.
4426 (switch_decision_tree::reset_out_edges_aux): Likewise.
4428 2018-06-20 Martin Liska <mliska@suse.cz>
4430 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
4431 (hoist_edge_and_branch_if_true): Likewise.
4432 (expand_switch_using_bit_tests_p): Likewise.
4433 (struct case_bit_test): Likewise.
4434 (case_bit_test_cmp): Likewise.
4435 (emit_case_bit_tests): Likewise.
4436 (switch_conversion::switch_conversion): New class.
4437 (struct switch_conv_info): Remove old struct.
4438 (collect_switch_conv_info): More to ...
4439 (switch_conversion::collect): ... this.
4440 (check_range): Likewise.
4441 (switch_conversion::check_range): Likewise.
4442 (check_all_empty_except_final): Likewise.
4443 (switch_conversion::check_all_empty_except_final): Likewise.
4444 (check_final_bb): Likewise.
4445 (switch_conversion::check_final_bb): Likewise.
4446 (create_temp_arrays): Likewise.
4447 (switch_conversion::create_temp_arrays): Likewise.
4448 (free_temp_arrays): Likewise.
4449 (gather_default_values): Likewise.
4450 (switch_conversion::gather_default_values): Likewise.
4451 (build_constructors): Likewise.
4452 (switch_conversion::build_constructors): Likewise.
4453 (constructor_contains_same_values_p): Likewise.
4454 (switch_conversion::contains_same_values_p): Likewise.
4455 (array_value_type): Likewise.
4456 (switch_conversion::array_value_type): Likewise.
4457 (build_one_array): Likewise.
4458 (switch_conversion::build_one_array): Likewise.
4459 (build_arrays): Likewise.
4460 (switch_conversion::build_arrays): Likewise.
4461 (gen_def_assigns): Likewise.
4462 (switch_conversion::gen_def_assigns): Likewise.
4463 (prune_bbs): Likewise.
4464 (switch_conversion::prune_bbs): Likewise.
4465 (fix_phi_nodes): Likewise.
4466 (switch_conversion::fix_phi_nodes): Likewise.
4467 (gen_inbound_check): Likewise.
4468 (switch_conversion::gen_inbound_check): Likewise.
4469 (process_switch): Use the newly created class.
4470 (switch_conversion::expand): New.
4471 (switch_conversion::~switch_conversion): New.
4472 * tree-switch-conversion.h: New file.
4474 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4476 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
4477 tree-vect-patterns.c.
4478 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
4479 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
4480 (vect_recog_sad_pattern): Likewise.
4481 (vect_recog_widen_sum_pattern): Likewise.
4482 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
4483 (vect_recog_widen_shift_pattern): Remove the type_in argument.
4484 (vect_recog_rotate_pattern): Likewise.
4485 (vect_recog_mult_pattern): Likewise.
4486 (vect_recog_vector_vector_shift_pattern): Likewise.
4487 (vect_recog_divmod_pattern): Likewise.
4488 (vect_recog_mixed_size_cond_pattern): Likewise.
4489 (vect_recog_bool_pattern): Likewise.
4490 (vect_recog_mask_conversion_pattern): Likewise.
4491 (vect_try_gather_scatter_pattern): Likewise.
4492 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
4493 (vect_recog_over_widening_pattern): Likewise.
4494 (vect_recog_gather_scatter_pattern): Likewise.
4495 (vect_recog_func_ptr): Move from tree-vectorizer.h
4496 (vect_vect_recog_func_ptrs): Move further down the file.
4497 (vect_recog_func): Likewise. Remove the third argument.
4498 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
4499 (vect_pattern_recog_1): Expect the pattern function to do any
4500 necessary target tests. Also expect it to provide a vector type.
4501 Remove the type_in handling.
4503 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4505 * tree-vect-patterns.c (vect_pattern_detected): New function.
4506 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
4507 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
4508 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
4509 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
4510 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
4511 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
4512 (vect_recog_mask_conversion_pattern)
4513 (vect_try_gather_scatter_pattern): Likewise.
4515 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4517 * tree-vect-patterns.c (vect_get_internal_def): New function.
4518 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
4519 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
4520 (search_type_for_mask_1): Use it.
4522 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4524 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
4525 redundant WIDEN_SUM_EXPR handling.
4526 (vect_recog_sad_pattern): Likewise.
4528 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4530 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
4531 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
4532 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
4533 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
4534 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
4536 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4538 * tree-vect-stmts.c (vectorizable_call): Make sure that we
4539 use the stmt_vec_info of the original bb statement for the
4540 new zero assignment, even if the call is part of a pattern.
4542 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4544 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
4545 that the sequence is attached to the original statement rather
4546 than the pattern statement.
4547 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
4548 PATTERN_DEF_SEQ from the original statement rather than
4549 the main pattern statement.
4550 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
4551 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
4552 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
4554 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4556 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
4557 definition statements before the early exit for statements that aren't
4559 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
4561 (vect_transform_loop): ...here. Process pattern definition
4562 statements without first checking whether the main pattern
4563 statement is live or relevant.
4565 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
4567 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
4568 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
4570 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
4572 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
4573 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
4574 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
4575 (expand_block_compare): Change select_block_compare_mode call.
4576 (expand_strncmp_align_check): Use new functions, fix comment.
4577 (emit_final_str_compare_gpr): New function.
4578 (expand_strn_compare): Refactor and clean up code.
4579 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
4581 2018-06-19 Tony Reix <tony.reix@atos.com>
4582 Damien Bergamini <damien.bergamini@atos.com>
4583 David Edelsohn <dje.gcc@gmail.com>
4585 * collect2.c (static_obj): New variable.
4586 (static_libs): New variable.
4587 (is_in_list): Uncomment declaration.
4588 (main): Track AIX libraries linked statically.
4589 (is_in_list): Uncomment definition.
4590 (scan_prog_file): Don't add AIX shared libraries initializer
4591 to constructor list if linking statically.
4593 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
4595 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
4597 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
4599 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
4601 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
4604 2018-06-19 Martin Liska <mliska@suse.cz>
4606 * config/i386/i386.c (ix86_can_inline_p): Do not use
4607 ipa_fn_summaries::get_create.
4608 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
4610 (devirtualization_time_bonus): Likewise.
4611 (ipcp_propagate_stage): Likewise.
4612 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
4613 (edge_set_predicate): Likewise.
4614 (evaluate_conditions_for_known_args): Likewise.
4615 (evaluate_properties_for_edge): Likewise.
4616 (ipa_call_summary::reset): Tranform to ...
4617 (ipa_call_summary::~ipa_call_summary): ... this.
4618 (ipa_fn_summary::reset): Transform to ...
4619 (ipa_fn_summary::~ipa_fn_summary): ... this.
4620 (ipa_fn_summary_t::remove): Rename to ...
4621 (ipa_fn_summary_t::remove_callees): ... this.
4622 (ipa_fn_summary_t::duplicate): Use placement new
4623 instead of memory copy.
4624 (ipa_call_summary_t::duplicate): Likewise.
4625 (ipa_call_summary_t::remove): Remove.
4626 (dump_ipa_call_summary): Change get_create to get.
4627 (ipa_dump_fn_summary): Dump only when summary exists.
4628 (analyze_function_body): Use symbol_summary::get instead
4630 (compute_fn_summary): Likewise.
4631 (estimate_edge_devirt_benefit): Likewise.
4632 (estimate_edge_size_and_time): Likewise.
4633 (inline_update_callee_summaries): Likewise.
4634 (remap_edge_change_prob): Likewise.
4635 (remap_edge_summaries): Likewise.
4636 (ipa_merge_fn_summary_after_inlining): Likewise.
4637 (write_ipa_call_summary): Likewise.
4638 (ipa_fn_summary_write): Likewise.
4639 (ipa_free_fn_summary): Likewise.
4640 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
4641 (struct ipa_call_summary): Likewise.
4642 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
4644 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
4645 (estimate_size_after_inlining): Likewise.
4646 (estimate_growth): Likewise.
4647 (growth_likely_positive): Likewise.
4648 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
4649 (inline_call): Likewise.
4650 * ipa-inline.c (caller_growth_limits): Likewise.
4651 (can_inline_edge_p): Likewise.
4652 (can_inline_edge_by_limits_p): Likewise.
4653 (compute_uninlined_call_time): Likewise.
4654 (compute_inlined_call_time): Likewise.
4655 (want_inline_small_function_p): Likewise.
4656 (edge_badness): Likewise.
4657 (update_caller_keys): Likewise.
4658 (update_callee_keys): Likewise.
4659 (inline_small_functions): Likewise.
4660 (inline_to_all_callers_1): Likewise.
4661 (dump_overall_stats): Likewise.
4662 (early_inline_small_functions): Likewise.
4663 (early_inliner): Likewise.
4664 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
4665 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
4666 * ipa-pure-const.c (malloc_candidate_p): Likewise.
4667 * ipa-split.c (execute_split_functions): Likewise.
4668 * symbol-summary.h: Likewise.
4669 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
4671 2018-06-19 Richard Biener <rguenther@suse.de>
4673 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
4674 (vectorize_loops): ... here. Fix dbgcnt handling.
4675 (try_vectorize_loop): Wrap try_vectorize_loop_1.
4677 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
4680 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
4681 ieee128 argument takes up only one (vector) register, not two (floating
4684 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
4686 * gimplify.c (gimplify_init_constructor): Really never clear for an
4687 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
4689 2018-06-19 Richard Biener <rguenther@suse.de>
4691 PR tree-optimization/86179
4692 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
4693 after failed recognition.
4695 2018-06-18 Martin Sebor <msebor@redhat.com>
4698 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
4699 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
4700 Handle integer subtraction.
4701 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
4702 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
4704 2018-06-18 David Malcolm <dmalcolm@redhat.com>
4706 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
4707 param from rtx to rtx_insn *.
4708 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
4710 (frv_ifcvt_modify_insn): Likwise.
4711 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
4712 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
4713 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
4714 as_a <rtx_insn *> cast to local "unprotected_region" once
4715 it's been established that it's not NULL or pc_rtx.
4716 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
4717 param "sethi" from rtx to rtx_insn *.
4718 (nds32_group_float_insns): Likewise for param "insn".
4719 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
4721 (vax_output_int_subtract): Likewise.
4722 * config/vax/vax.c (vax_output_int_add): Likewise for param
4724 (vax_output_int_subtract): Likewise.
4725 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
4726 (emit_pattern_after): Likewise for param "after".
4727 (emit_insn_after): Likewise.
4728 (emit_jump_insn_after): Likewise.
4729 (emit_call_insn_after): Likewise.
4730 (emit_debug_insn_after): Likewise.
4731 (emit_pattern_before): Likewise for param "before".
4732 (emit_insn_before): Likewise.
4733 (emit_jump_insn_before): Likewise.
4734 * final.c (get_insn_template): Likewise for param "insn", removing
4736 * output.h (get_insn_template): Likewise for 2nd param.
4737 * rtl.h (emit_insn_before): Likewise.
4738 (emit_jump_insn_before): Likewise.
4739 (emit_debug_insn_before_noloc): Likewise.
4740 (emit_insn_after): Likewise.
4741 (emit_jump_insn_after): Likewise.
4742 (emit_call_insn_after): Likewise.
4743 (emit_debug_insn_after): Likewise.
4744 (set_insn_deleted): Likewise for param.
4746 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
4749 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
4750 floating point modes, so that IFmode is numerically greater than
4751 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
4752 to declare the ordering. This prevents IFmode from being
4753 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
4754 machine. Include rs6000-modes.h to share the fractional values
4755 between genmodes* and the rest of the compiler.
4759 * config/rs6000/rs6000-modes.h: New file.
4760 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
4761 meaning of rs6000_long_double_size so that 126..128 selects an
4762 appropriate 128-bit floating point type.
4763 (rs6000_option_override_internal): Likewise.
4764 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
4765 (TARGET_LONG_DOUBLE_128): Change the meaning of
4766 rs6000_long_double_size so that 126..128 selects an appropriate
4767 128-bit floating point type.
4768 (LONG_DOUBLE_TYPE_SIZE): Update comment.
4769 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
4770 source and destination to match the standard usage.
4771 (truncifkf2): Likewise.
4772 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
4773 ISA 2.07 to use an explicit clobber, instead of passing in a
4775 (copysign<mode>3_soft): Likewise.
4777 2018-06-18 David Malcolm <dmalcolm@redhat.com>
4779 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
4780 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
4781 (vect_slp_analyze_instance_dependence): Likewise.
4782 (vect_enhance_data_refs_alignment): Likewise.
4783 (vect_analyze_data_refs_alignment): Likewise.
4784 (vect_slp_analyze_and_verify_instance_alignment
4785 (vect_analyze_data_ref_accesses): Likewise.
4786 (vect_prune_runtime_alias_test_list): Likewise.
4787 (vect_analyze_data_refs): Likewise.
4788 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
4789 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
4790 (vect_analyze_scalar_cycles_1): Likewise.
4791 (vect_get_loop_niters): Likewise.
4792 (vect_analyze_loop_form_1): Likewise.
4793 (vect_update_vf_for_slp): Likewise.
4794 (vect_analyze_loop_operations): Likewise.
4795 (vect_analyze_loop): Likewise.
4796 (vectorizable_induction): Likewise.
4797 (vect_transform_loop): Likewise.
4798 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
4799 * tree-vect-slp.c (vect_analyze_slp): Likewise.
4800 (vect_make_slp_decision): Likewise.
4801 (vect_detect_hybrid_slp): Likewise.
4802 (vect_slp_analyze_operations): Likewise.
4803 (vect_slp_bb): Likewise.
4804 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
4805 (vectorizable_bswap): Likewise.
4806 (vectorizable_call): Likewise.
4807 (vectorizable_simd_clone_call): Likewise.
4808 (vectorizable_conversion): Likewise.
4809 (vectorizable_assignment): Likewise.
4810 (vectorizable_shift): Likewise.
4811 (vectorizable_operation): Likewise.
4812 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
4814 2018-06-18 Martin Sebor <msebor@redhat.com>
4816 PR tree-optimization/81384
4817 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
4818 * builtins.c (expand_builtin_strnlen): New function.
4819 (expand_builtin): Call it.
4820 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
4821 * builtins.def (BUILT_IN_STRNLEN): New.
4822 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
4823 Warn for bounds in excess of maximum object size.
4824 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
4825 single-value ranges. Handle strnlen.
4826 (handle_builtin_strlen): Handle strnlen.
4827 (strlen_check_and_optimize_stmt): Same.
4828 * doc/extend.texi (Other Builtins): Document strnlen.
4830 2018-06-18 Maya Rashish <coypu@sdf.org>
4832 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
4833 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
4834 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
4836 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
4838 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
4840 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4842 * tree.c (escaped_string::escape): Replace cast to char * by
4843 const_cast<char *> (unescaped).
4845 2018-06-18 Nick Clifton <nickc@redhat.com>
4848 * tree.c (escaped_string): New class. Converts an unescaped
4849 string into its escaped equivalent.
4850 (warn_deprecated_use): Use the new class to convert the
4851 deprecation message, if present.
4852 (test_escaped_strings): New self test.
4853 (test_c_tests): Add test_escaped_strings.
4854 * doc/extend.texi (deprecated): Add a note that the
4855 deprecation message is affected by the -fmessage-length
4856 option, and that control characters will be escaped.
4857 (#pragma GCC error): Document this pragma.
4858 (#pragma GCC warning): Likewise.
4859 * doc/invoke.texi (-fmessage-length): Document this option's
4860 effect on the #warning and #error preprocessor directives and
4861 the deprecated attribute.
4863 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
4865 * tree.c (decl_value_expr_lookup): Revert latest change.
4866 (decl_value_expr_insert): Likewise.
4868 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
4870 * gimplify.c (nonlocal_vlas): Delete.
4871 (nonlocal_vla_vars): Likewise.
4872 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
4874 (gimplify_body): Do not create and destroy nonlocal_vlas.
4875 * tree-nested.c: Include diagnostic.h.
4876 (use_pointer_in_frame): Tweak.
4877 (lookup_field_for_decl): Add assertion and declare the transformation.
4878 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
4879 internal error when the reference is in a wrong context. Do not
4880 create a debug decl by default.
4881 (note_nonlocal_block_vlas): Delete.
4882 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
4883 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
4884 create a debug decl by default.
4885 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
4886 call is in a wrong context.
4887 (fixup_vla_decls): New function.
4888 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
4889 debug variables were created.
4890 * tree.c (decl_value_expr_lookup): Add checking assertion.
4891 (decl_value_expr_insert): Likewise.
4893 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
4896 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
4897 * tree-scalar-evolution.c (interpret_expr): Likewise.
4898 (expression_expensive_p): Likewise.
4899 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
4900 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
4901 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
4902 (ssa_defined_by_minus_one_stmt_p): New.
4904 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
4907 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
4908 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
4909 * dojump.c (do_jump): Likewise.
4910 * expr.c (expand_expr_real_2): Check operand type's sign.
4911 * fold-const.c (const_unop): Handle ABSU_EXPR.
4912 (fold_abs_const): Likewise.
4913 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
4914 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
4915 (strip_sign_op_1): Likesise.
4916 * match.pd: Add new pattern to generate ABSU_EXPR.
4917 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
4918 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
4919 * tree-eh.c (operation_could_trap_helper_p): Likewise.
4920 * tree-inline.c (estimate_operator_cost): Likewise.
4921 * tree-pretty-print.c (dump_generic_node): Likewise.
4922 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
4923 * tree.def (ABSU_EXPR): New.
4925 2018-06-16 Jakub Jelinek <jakub@redhat.com>
4928 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
4929 documented as preserved for backward compatibility only.
4930 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
4932 PR rtl-optimization/86108
4933 * bb-reorder.c (create_forwarder_block): Renamed to ...
4934 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
4935 jump from new landing pad to the second part.
4936 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
4939 2018-06-15 Jakub Jelinek <jakub@redhat.com>
4942 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
4943 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
4944 Only call store_expr for halves if the mode is the same.
4947 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
4948 Fix up comment formatting.
4950 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
4952 * typed-splay-tree.h (typed_splay_tree::remove): New function.
4953 (typed_splay_tree::closure,
4954 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
4955 (typed_splay_tree::typed_splay_tree,
4956 typed_splay_tree::operator =): Declared private.
4957 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
4958 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
4959 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
4960 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
4961 typed_splay_tree::splay_tree_splay,
4962 typed_splay_tree::splay_tree_foreach_helper,
4963 typed_splay_tree::splay_tree_insert,
4964 typed_splay_tree::splay_tree_remove,
4965 typed_splay_tree::splay_tree_lookup,
4966 typed_splay_tree::splay_tree_predecessor,
4967 typed_splay_tree::splay_tree_successor,
4968 typed_splay_tree::splay_tree_min,
4969 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
4970 (typed_splay_tree::root, typed_splay_tree::comp,
4971 typed_splay_tree::delete_key,
4972 typed_splay_tree::delete_value): New data members.
4973 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
4974 typed_splay_tree::remove.
4976 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
4978 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
4979 -mginv and -mno-ginv to the assembler.
4980 * config/mips/mips.opt (-mcrc): New option.
4982 * doc/invoke.text (-mcrc): Document.
4985 2018-06-15 Nick Clifton <nickc@redhat.com>
4988 * tree.c (escaped_string): New class. Converts an unescaped
4989 string into its escaped equivalent.
4990 (warn_deprecated_use): Use the new class to convert the
4991 deprecation message, if present.
4992 (test_escaped_strings): New self test.
4993 (test_c_tests): Add test_escaped_strings.
4994 * doc/extend.texi (deprecated): Add a note that the
4995 deprecation message is affected by the -fmessage-length
4996 option, and that control characters will be escaped.
4997 (#pragma GCC error): Document this pragma.
4998 (#pragma GCC warning): Likewise.
4999 * doc/invoke.texi (-fmessage-length): Document this option's
5000 effect on the #warning and #error preprocessor directives and
5001 the deprecated attribute.
5003 2018-06-15 Richard Biener <rguenther@suse.de>
5005 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
5006 here, also noting vector size used.
5007 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
5008 size used in MSG_OPTIMIZED_LOCATIONS dump.
5009 (pass_slp_vectorize::execute): Adjust.
5011 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
5014 * config/arc/arc.c (arc_return_address_register): Fix
5017 2018-06-15 Richard Biener <rguenther@suse.de>
5020 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
5021 leave useless conversion stripping to force_gimple_operand_gsi.
5022 (gimplify_build2): Likewise.
5023 (gimplify_build1): Likewise.
5025 2018-06-15 Richard Biener <rguenther@suse.de>
5028 * tree-cfg.c (move_stmt_op): unshare invariant addresses
5029 before adjusting their block.
5031 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
5033 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
5034 multilibs for *-*-rtems*.
5035 * config/riscv/t-rtems: New file.
5037 2018-06-14 Jakub Jelinek <jakub@redhat.com>
5040 * match.pd ((A +- CST1) +- CST2): Punt if last resort
5041 unsigned_type_for returns NULL.
5044 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
5045 subregs of multi-word pseudos unless the float mode has word size.
5047 2018-06-14 Richard Biener <rguenther@suse.de>
5050 * tree-vect-generic.c (build_word_mode_vector_type): Remove
5051 duplicate and harmful type_hash_canon.
5052 * tree.c (type_hash_canon): Assert we didn't find ourselves.
5054 2018-06-14 Richard Biener <rguenther@suse.de>
5057 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
5060 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
5062 * config/rtems.h (STDINT_LONG32): Define.
5064 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
5065 Prachi Godbole <prachi.godbole@imgtec.com>
5067 * config/mips/mips-cpus.def: Define P6600.
5068 * config/mips/mips-tables.opt: Regenerate.
5069 * config/mips/mips.c (mips_ucbranch_type): New enum.
5070 (mips_rtx_cost_data): Add support for P6600.
5071 (mips_issue_rate): Likewise.
5072 (mips_multipass_dfa_lookahead): Likewise.
5073 (mips_avoid_hazard): Likewise.
5074 (mips_reorg_process_insns): Likewise.
5075 (mips_classify_branch_p6600): New function.
5076 * config/mips/mips.h (TUNE_P6600): New define.
5077 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
5078 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
5079 * config/mips/mips.md: Include p6600.md.
5080 (processor): Add p6600.
5081 * config/mips/p6600.md: New file.
5082 * doc/invoke.texi: Add p6600 to supported architectures.
5084 2018-06-13 Martin Sebor <msebor@redhat.com>
5086 PR tree-optimization/86114
5087 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
5089 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
5091 2018-06-13 Richard Biener <rguenther@suse.de>
5093 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
5094 Properly set vector type of the intermediate stmt.
5095 * tree-vect-stmts.c (vectorizable_operation): The destination
5096 var always has vectype_out type.
5098 2018-06-13 Jeff Law <law@redhat.com>
5100 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
5101 integer 0 for argument to print_rtl_with_bb.
5102 (rl78_reorg): Likewise.
5104 2018-06-13 David Malcolm <dmalcolm@redhat.com>
5106 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
5107 from rtx to rtx_insn *.
5108 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
5110 (add_sched_insns_for_speculation): Likewise for local "target",
5111 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
5112 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
5113 from rtx_insn ** to rtx_code_label **.
5114 (reorg_emit_nops): Likewise.
5115 (c6x_reorg): Likewise for local "call_labels".
5116 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
5118 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
5119 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
5120 the loops over LABEL_REFS.
5121 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
5123 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
5124 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
5125 (split_branches): Strengthen local "olabel" from rtx to
5126 rtx_insn *, adding a safe_as_a cast.
5127 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
5129 (add_insn_after): Likewise for first two params.
5130 (add_insn_before): Likewise.
5131 (remove_insn): Likewise for param.
5132 (emit_pattern_before_noloc): Likewise for second and third params.
5133 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
5134 (emit_call_insn_before_noloc): Likewise.
5135 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
5137 (emit_barrier_before): Likewise.
5138 (emit_label_before): Strengthen "label" param from "rtx" to
5139 "rtx_code_label *". Strengthen "before" param from "rtx" to
5141 (emit_insn_after_1): Strengthen "after" param from "rtx" to
5143 (emit_pattern_after_noloc): Likewise.
5144 (emit_insn_after_noloc): Likewise.
5145 (emit_jump_insn_after_noloc): Likewise.
5146 (emit_call_insn_after_noloc): Likewise.
5147 (emit_debug_insn_after_noloc): Likewise.
5148 (emit_barrier_after): Likewise.
5149 (emit_label_after): Likewise for both params.
5150 (emit_pattern_after_setloc): Likewise for "after" param. Convert
5151 "loc" param from "int" to "location_t".
5152 (emit_insn_after_setloc): Likewise.
5153 (emit_jump_insn_after_setloc): Likewise.
5154 (emit_call_insn_after_setloc): Likewise.
5155 (emit_debug_insn_after_setloc): Likewise.
5156 (emit_pattern_before_setloc): Likewise for "before" param. Convert
5157 "loc" param from "int" to "location_t".
5158 (emit_pattern_before): Convert NULL_RTX to NULL.
5159 (emit_insn_before_setloc): Convert "loc" param from "int" to
5161 (emit_jump_insn_before_setloc): Likewise.
5162 (emit_call_insn_before_setloc): Likewise.
5163 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
5164 rtx_insn *. Convert "loc" param from "int" to "location_t".
5165 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
5166 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
5167 Convert 3rd param from "int" to "location_t".
5168 (emit_barrier_before, emit_barrier_after, next_real_insn):
5169 Strengthen param from rtx to rtx_insn *.
5170 (emit_label_before): Strengthen 1st param from "rtx" to
5171 "rtx_code_label *". Strengthen 2nd param from "rtx" to
5173 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
5174 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
5175 Strengthen 2nd param from "rtx" to "rtx_insn *".
5176 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
5177 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
5178 Likewise. Convert 3rd param from "int" to "location_t".
5179 (emit_label_after): Strengthen 1st param from "rtx" to
5181 (next_real_insn, remove_insn): Strengthen param from "rtx" to
5183 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
5184 from "rtx" to "rtx_insn *".
5186 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
5188 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
5189 bodies streamed in with -Q.
5190 * dumpfile.c (dump_files): Add lto-stream-out dump file.
5191 * dumpfile.h (tree_dump_index): Add lto_stream_out.
5192 * gimple-streamer-out.c: Include gimple-pretty-print.h
5193 (output_bb): Dump stmts streamed.
5194 * lto-section-out.c: Include print-tree.h
5195 (lto_begin_section): Dump sections created.
5196 (lto_output_decl_index): Dump decl encoded.
5197 * lto-streamer-out.c: Include print-tree.h
5198 (create_output_block): Dump output block created.
5199 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
5200 (output_function): Dump function output.
5201 (output_constructor): Dump constructor streamed.
5202 (write_global_stream): Output indexes encoded.
5203 (produce_asm_for_decls): Dump streams encoded.
5204 * lto-streamer.c (streamer_dump_file): New global var.
5205 * lto-streamer.h (streamer_dump_file): Declare.
5206 * passes.c (ipa_write_summaries): Initialize streamer dump.
5207 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
5210 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
5213 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
5214 offsets for register save directives. Emit a second batch of save
5215 directives, if need be, when the function accesses prior frames.
5217 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
5219 * config/arc/fpu.md (fmasf4): Force operand to register.
5220 (fnmasf4): Likewise.
5222 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
5224 * config/arc/arc-protos.h (arc_pad_return): Remove.
5225 * config/arc/arc.c (machine_function): Remove force_short_suffix
5227 (arc_print_operand): Adjust printing of '&'.
5228 (arc_verify_short): Remove conditional printing of short suffix.
5229 (arc_final_prescan_insn): Remove reference to size_reason.
5230 (pad_return): New function.
5231 (arc_reorg): Call pad_return.
5232 (arc_pad_return): Remove.
5233 (arc_init_machine_status): Remove reference to force_short_suffix.
5234 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
5235 (attr length): When attribute iscompact is true force to 2
5236 regardless; in the case of maybe check if we want to force the
5237 instruction to have 4 bytes length.
5238 (nopv): Change it to generate 4 byte long nop as well.
5239 (blockage): New pattern.
5240 (simple_return): Remove call to arc_pad_return.
5241 (p_return_i): Likewise.
5243 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
5245 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
5247 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
5249 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
5252 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
5254 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
5255 for ARC700 and ARCv2.
5257 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
5260 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
5261 operands[2] instead of operands[1].
5264 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
5266 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
5267 case, check whether the outer register overlaps an unallocatable
5268 register, not just whether it fits the required class.
5270 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
5272 * poly-int.h (can_div_trunc_p): Add new overload in which all values
5274 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
5275 (memrefs_conflict_p): Likewise.
5276 (init_alias_analysis): Likewise.
5277 * cfgexpand.c (expand_debug_expr): Likewise.
5278 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
5279 * cse.c (fold_rtx): Likewise.
5280 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
5281 * expr.c (emit_block_move_hints): Likewise.
5282 (clear_storage_hints, push_block, emit_push_insn): Likewise.
5283 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
5284 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
5285 (emit_group_store): Likewise.
5286 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
5287 to read the PRE/POST_MODIFY increment.
5288 * calls.c (store_one_arg): Use strip_offset.
5289 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
5291 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
5293 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
5294 (simplify_binary_operation_1): Extend CONST_INT handling to
5296 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
5297 than a HOST_WIDE_INT.
5298 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
5300 (adjust_mems, add_stores): Update accodingly.
5301 (vt_canonicalize_addr): Track polynomial offsets.
5302 (emit_note_insn_var_location): Likewise.
5303 (vt_add_function_parameter): Likewise.
5304 (vt_initialize): Likewise.
5306 2018-06-12 Jeff Law <law@redhat.com>
5308 * config.gcc (alpha*-*-freebsd*): Remove.
5309 * config/alpha/freebsd.h: Remove.
5311 2018-06-12 David Malcolm <dmalcolm@redhat.com>
5314 * spellcheck-tree.c (levenshtein_distance): Rename to...
5315 (get_edit_distance): ...this, and update for underlying renaming.
5316 * spellcheck-tree.h (levenshtein_distance): Rename to...
5317 (get_edit_distance): ...this.
5318 * spellcheck.c (levenshtein_distance): Rename to...
5319 (get_edit_distance): ...this. Convert from Levenshtein distance
5320 to Damerau-Levenshtein distance by supporting transpositions of
5321 adjacent characters. Rename "v1" to "v_next" and "v0" to
5323 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
5324 (selftest::test_edit_distance_unit_test_oneway): ...this, and
5325 update for underlying renaming.
5326 (selftest::levenshtein_distance_unit_test): Rename to...
5327 (selftest::test_get_edit_distance_unit): ...this, and update for
5328 underlying renaming.
5329 (selftest::test_find_closest_string): Add example from PR 69968
5330 where transposition helps
5331 (selftest::test_metric_conditions): Update for renaming.
5332 (selftest::test_metric_conditions): Likewise.
5333 (selftest::spellcheck_c_tests): Likewise.
5334 * spellcheck.h (levenshtein_distance): Rename both overloads to...
5335 (get_edit_distance): ...this.
5336 (best_match::consider): Update for renaming.
5338 2018-06-12 Martin Sebor <msebor@redhat.com>
5340 PR tree-optimization/85259
5341 * builtins.c (compute_objsize): Handle constant offsets.
5342 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
5343 true iff a warning has been issued.
5344 * gimple.h (gimple_nonartificial_location): New function.
5345 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
5346 gimple_nonartificial_location and handle -Wno-system-headers.
5347 (handle_builtin_stxncpy): Same.
5349 2018-06-12 Martin Sebor <msebor@redhat.com>
5352 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
5354 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
5356 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
5357 BUILTIN_VEC_XST entries for pointer to double and long long.
5359 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
5362 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
5364 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
5367 2018-06-12 Martin Liska <mliska@suse.cz>
5369 * doc/options.texi: Document IntegerRange.
5371 2018-06-12 Martin Liska <mliska@suse.cz>
5373 * config/i386/i386.opt: Make MPX-related options as Deprecated.
5374 * opt-functions.awk: Handle Deprecated flag.
5375 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
5377 (read_cmdline_option): Report warning for a deprecated option.
5378 * opts.h (struct cl_option): Add new field cl_deprecated.
5379 (CL_ERR_DEPRECATED): New.
5381 2018-06-12 Martin Liska <mliska@suse.cz>
5383 * doc/options.texi: Document Deprecated option flag.
5385 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
5387 * config/arc/arc-arch.h (arc_extras): New enum.
5388 (arc_cpu_t):Add field extra.
5389 (arc_cpu_types): Consider the extras.
5390 * config/arc/arc-cpus.def: Add extras info.
5391 * config/arc/arc-opts.h (processor_type): Consider extra field.
5392 * config/arc/arc.c (arc_override_options): Handle extra field.
5394 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
5396 * config/arc/arc-arch.h: Update ARC_OPTX macro.
5397 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
5399 * config/arc/arc.c (arc_init): Update pic warning.
5400 (irq_range): Update irq range parsing warnings.
5401 (arc_override_options): Update various warning messages.
5402 (arc_handle_aux_attribute): Likewise.
5404 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
5406 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
5408 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5410 * doc/sourcebuild.texi: Document usage of line number 0 in verify
5411 compiler messages directives.
5413 2018-06-12 Matthew Fortune <mfortune@gmail.com>
5415 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
5416 * config/mips/mips-tables.opt: Regenerate.
5417 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
5419 * doc/invoke.texi: Document -march=i6500.
5421 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
5423 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
5424 (i6400_gpmul): Add cpu_unit.
5425 (i6400_gpdiv): Likewise.
5426 (i6400_msa_add_d): Update reservations.
5427 (i6400_msa_int_add) Likewise.
5428 (i6400_msa_short_logic3) Likewise.
5429 (i6400_msa_short_logic2) Likewise.
5430 (i6400_msa_short_logic) Likewise.
5431 (i6400_msa_move) Likewise.
5432 (i6400_msa_cmp) Likewise.
5433 (i6400_msa_short_float2) Likewise.
5434 (i6400_msa_div_d) Likewise.
5435 (i6400_msa_long_logic1) Likewise.
5436 (i6400_msa_long_logic2) Likewise.
5437 (i6400_msa_mult) Likewise.
5438 (i6400_msa_long_float2) Likewise.
5439 (i6400_msa_long_float4) Likewise.
5440 (i6400_msa_long_float5) Likewise.
5441 (i6400_msa_long_float8) Likewise.
5442 (i6400_fpu_fadd): Include frint type.
5443 (i6400_fpu_store): New define_insn_reservation.
5444 (i6400_fpu_load): Likewise.
5445 (i6400_fpu_move): Likewise.
5446 (i6400_fpu_fcmp): Likewise.
5447 (i6400_fpu_fmadd): Likewise.
5448 (i6400_int_mult): Include imul3nc type and update reservation.
5449 (i6400_int_div): Include idiv3 type and update reservation.
5450 (i6400_int_load): Update to check type not move_type.
5451 (i6400_int_store): Likewise.
5452 (i6400_int_prefetch): Set zero latency.
5454 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
5456 * gcc.c: Document new %@{...} sequence.
5457 (LINK_COMMAND_SPEC): Use it for the -L switches.
5458 (cpp_unique_options): Use it for the -I switches.
5459 (at_file_argbuf): New global variable.
5460 (in_at_file): Likewise.
5461 (alloc_args): Create at_file_argbuf.
5462 (clear_args): Truncate at_file_argbuf.
5463 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
5464 (open_at_file): New function.
5465 (close_at_file): Likewise.
5466 (create_at_file): Delete.
5467 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
5470 (validate_switches_from_spec): Deal with %@{...} sequence.
5471 (validate_switches): Likewise.
5472 (driver::finalize): Call clear_args.
5474 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
5476 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
5478 2018-06-11 Martin Sebor <msebor@redhat.com>
5480 * doc/invoke.texi (-Wall): List -Wc++17-compat.
5481 (Wno-class-memaccess): Add @opindex.
5482 (Wno-templates, Wno-multiple-inheritance): Same.
5483 (Wno-virtual-inheritance, Wno-namespaces): Same.
5484 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
5485 (Wno-format-overflow, Wno-format-truncation): Same.
5486 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
5487 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
5488 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
5489 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
5490 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
5491 (Wno-misspelled-isr): Same.
5493 2018-06-11 Martin Sebor <msebor@redhat.com>
5495 * PR tree-optimization/86083
5496 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
5498 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
5500 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
5502 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
5505 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
5506 on the correct operand.
5507 (*movdi_internal64): Ditto.
5509 2018-06-11 Martin Liska <mliska@suse.cz>
5511 PR tree-optimization/86089
5512 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
5514 2018-06-11 Julia Koval <julia.koval@intel.com>
5516 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
5517 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
5518 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
5520 2018-06-11 Olivier Hainque <hainque@adacore.com>
5522 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
5523 for Ada with strict dwarf2.
5525 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
5528 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
5531 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
5533 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
5535 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
5537 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
5540 2018-06-08 Martin Liska <mliska@suse.cz>
5542 * tree-cfg.h (debug_function): Fix argument type to match
5545 2018-06-08 Martin Liska <mliska@suse.cz>
5547 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
5548 Remove usage of MPX-related (and removed) fields.
5549 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
5551 2018-06-08 David Malcolm <dmalcolm@redhat.com>
5553 * cfg.c (debug): Use TDF_NONE rather than 0.
5554 * cfghooks.c (debug): Likewise.
5555 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
5556 (struct dump_option_value_info): Convert to...
5557 (struct kv_pair): ...this template type.
5558 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
5560 (optinfo_verbosity_options): Likewise.
5561 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
5563 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
5564 than int for "optgroup_flags" param.
5565 (dump_generic_expr_loc): Use dump_flags_t rather than int for
5567 (dump_dec): Likewise.
5568 (dump_finish): Use TDF_NONE rather than 0.
5569 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
5570 rather than int for "optgroup_flags" param. Use TDF_NONE rather
5571 than 0. Update for change to option_ptr.
5572 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
5573 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
5574 0. Update for changes to optinfo_verbosity_options and
5576 (opt_info_switch_p): Convert optgroup_flags from int to
5578 (dump_basic_block): Use dump_flags_t rather than int
5579 for "dump_kind" param.
5580 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
5581 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
5582 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
5583 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
5584 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
5585 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
5586 TDF_NONE): Convert from macros to...
5587 (enum dump_flag): ...this new enum.
5588 (dump_flags_t): Update to use enum.
5589 (operator|, operator&, operator~, operator|=, operator&=):
5590 Implement for dump_flags_t.
5591 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
5592 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
5593 Convert from macros to...
5594 (enum optgroup_flag): ...this new enum.
5595 (optgroup_flags_t): New typedef.
5596 (operator|, operator|=): Implement for optgroup_flags_t.
5597 (struct dump_file_info): Convert field "alt_flags" to
5598 dump_flags_t. Convert field "optgroup_flags" to
5600 (dump_basic_block): Use dump_flags_t rather than int for param.
5601 (dump_generic_expr_loc): Likewise.
5602 (dump_dec): Likewise.
5603 (dump_register): Convert param "optgroup_flags" to
5605 (opt_info_enable_passes): Likewise.
5606 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
5608 * gimple-pretty-print.c (debug): Likewise.
5609 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
5610 (merged_store_group::apply_stores): Likewise.
5611 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
5612 * gimple.c (verify_gimple_pp): Likewise.
5613 * graphite-poly.c (print_pbb_body): Likewise.
5614 * passes.c (pass_manager::register_one_dump_file): Convert
5615 local "optgroup_flags" to optgroup_flags_t.
5616 * print-tree.c (print_node): Use TDF_NONE rather than 0.
5618 (debug_body): Likewise.
5619 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
5620 to optgroup_flags_t.
5621 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
5623 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
5624 (convert_mult_to_fma): Likewise.
5625 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
5626 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
5627 * tree-vect-data-refs.c (dump_lower_bound): Convert param
5628 "dump_kind" to dump_flags_t.
5630 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
5632 * config/rs6000/rs6000.c (min, max): Delete.
5634 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
5636 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
5637 -mabi=spe and -mabi=no-spe.
5639 2018-06-08 Martin Liska <mliska@suse.cz>
5641 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
5642 where we expect an existing summary.
5644 2018-06-08 Martin Liska <mliska@suse.cz>
5646 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
5647 * ipa-inline.h (estimate_edge_growth): Likewise.
5649 2018-06-08 Martin Liska <mliska@suse.cz>
5651 * cgraph.c (function_version_hasher::hash): Use
5652 cgraph_node::get_uid ().
5653 (function_version_hasher::equal):
5654 * cgraph.h (cgraph_node::get_uid): New method.
5655 * ipa-inline.c (update_caller_keys): Use
5656 cgraph_node::get_uid ().
5657 (update_callee_keys): Likewise.
5658 * ipa-utils.c (searchc): Likewise.
5659 (ipa_reduced_postorder): Likewise.
5660 * lto-cgraph.c (input_node): Likewise.
5661 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
5662 * symbol-summary.h (symtab_insertion): Likewise.
5663 (symtab_removal): Likewise.
5664 (symtab_duplication): Likewise.
5665 * tree-pretty-print.c (dump_function_header): Likewise.
5666 * tree-sra.c (convert_callers_for_node): Likewise.
5668 2018-06-08 Martin Liska <mliska@suse.cz>
5670 * cgraph.c (symbol_table::create_edge): Always assign a new
5672 (symbol_table::free_edge): Do not recycle numbers.
5673 * cgraph.h (cgraph_edge::get): New method.
5674 * symbol-summary.h (symtab_removal): Use it.
5675 (symtab_duplication): Likewise.
5676 (call_summary::hashable_uid): Remove.
5678 2018-06-08 Martin Liska <mliska@suse.cz>
5680 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
5681 (initialize_growth_caches): Remove.
5682 (free_growth_caches): Likewise.
5683 (do_estimate_edge_time): Use edge_growth_cache.
5684 (do_estimate_edge_size): Likewise.
5685 (do_estimate_edge_hints): Likewise.
5686 * ipa-inline.c (reset_edge_caches): Likewise.
5687 (recursive_inlining): Likewise.
5688 (inline_small_functions): Likewise.
5689 * ipa-inline.h (initialize_growth_caches): Remove.
5690 (estimate_edge_size): Likewise.
5691 (estimate_edge_time): Likewise.
5692 (estimate_edge_hints): Likewise.
5693 (reset_edge_growth_cache): Likewise.
5694 * symbol-summary.h (call_summary::remove): New method.
5696 2018-06-08 Martin Liska <mliska@suse.cz>
5698 * ipa-cp.c (class edge_clone_summary): New summary.
5699 (grow_edge_clone_vectors): Remove.
5700 (ipcp_edge_duplication_hook): Remove.
5701 (class edge_clone_summary_t): New call_summary class.
5702 (ipcp_edge_removal_hook): Remove.
5703 (edge_clone_summary_t::duplicate): New function.
5704 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
5705 (create_specialized_node): Likewise.
5706 (ipcp_driver): Initialize edge_clone_summaries and do not
5709 2018-06-08 Martin Liska <mliska@suse.cz>
5711 * symbol-summary.h (get): New function.
5712 (call_summary::m_initialize_when_cloning): New class member.
5714 2018-06-08 Martin Liska <mliska@suse.cz>
5716 * cgraph.c (cgraph_node::remove): Do not recycle uid.
5717 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
5718 (symbol_table::allocate_cgraph_symbol): Do not set uid.
5719 * passes.c (uid_hash_t): Record removed_nodes by their uids.
5720 (remove_cgraph_node_from_order): Use the removed_nodes set.
5721 (do_per_function_toporder): Likwise.
5722 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
5723 instead of summary_uid.
5724 (symtab_removal): Likewise.
5725 (symtab_duplication): Likewise.
5727 2018-06-08 Martin Liska <mliska@suse.cz>
5729 * ipa-cp.c (ipcp_store_bits_results): Use
5730 ipcp_transformation_sum.
5731 (ipcp_store_vr_results): Likewise.
5732 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
5734 (ipcp_transformation_initialize): ... this.
5735 (ipa_set_node_agg_value_chain):
5736 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
5737 (write_ipcp_transformation_info): Likewise.
5738 (read_ipcp_transformation_info): Likewise.
5739 (ipcp_update_bits): Likewise.
5740 (ipcp_update_vr): Likewise.
5741 (ipcp_transform_function): Likewise.
5742 * ipa-prop.h: Rename ipcp_transformation_summary to
5743 ipcp_transformation.
5744 (class ipcp_transformation_t): New function summary.
5745 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
5746 (ipa_get_agg_replacements_for_node): Likewise.
5748 2018-06-08 Martin Liska <mliska@suse.cz>
5750 * ipa-pure-const.c (struct funct_state_d): Do it class instead
5752 (class funct_state_summary_t): New function_summary class.
5753 (has_function_state): Remove.
5754 (get_function_state): Likewise.
5755 (set_function_state): Likewise.
5756 (add_new_function): Likewise.
5757 (funct_state_summary_t::insert): New function.
5758 (duplicate_node_data): Remove.
5759 (remove_node_data): Remove.
5760 (funct_state_summary_t::duplicate): New function.
5761 (register_hooks): Create new funct_state_summaries.
5762 (pure_const_generate_summary): Use it.
5763 (pure_const_write_summary): Likewise.
5764 (pure_const_read_summary): Likewise.
5765 (propagate_pure_const): Likewise.
5766 (propagate_nothrow): Likewise.
5767 (dump_malloc_lattice): Likewise.
5768 (propagate_malloc): Likewise.
5769 (execute): Do not register hooks, just remove summary
5771 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
5774 2018-06-08 Martin Liska <mliska@suse.cz>
5776 * ipa-reference.c (remove_node_data): Remove.
5777 (duplicate_node_data): Likewise.
5778 (class ipa_ref_var_info_summary_t): New class.
5779 (class ipa_ref_opt_summary_t): Likewise.
5780 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
5781 (get_reference_optimization_summary): Use
5782 ipa_ref_opt_sum_summaries.
5783 (set_reference_vars_info): Remove.
5784 (set_reference_optimization_summary): Likewise.
5785 (ipa_init): Create summaries.
5786 (init_function_info): Use function summary.
5787 (ipa_ref_opt_summary_t::duplicate): New function.
5788 (ipa_ref_opt_summary_t::remove): New function.
5789 (get_read_write_all_from_node): Fix GNU coding style.
5790 (propagate): Use function summary.
5791 (write_node_summary_p): Fix GNU coding style.
5792 (stream_out_bitmap): Likewise.
5793 (ipa_reference_read_optimization_summary): Use function summary.
5794 (ipa_reference_c_finalize): Do not release hooks.
5796 2018-06-08 Martin Liska <mliska@suse.cz>
5798 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
5799 (analyze_function_body): Extract multiple calls of get_create.
5800 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
5801 * ipa-inline.c (recursive_inlining): Use ::get method.
5802 * ipa-inline.h (estimate_edge_growth): Likewise.
5804 2018-06-08 Martin Liska <mliska@suse.cz>
5806 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
5808 (hsa_function_summary::hsa_function_summary): Use the new enum
5810 (hsa_gpu_implementation_p): Use hsa_summaries::get.
5811 * hsa-gen.c (hsa_get_host_function): Likewise.
5812 (get_brig_function_name): Likewise.
5813 * ipa-hsa.c (process_hsa_functions): Likewise.
5814 (ipa_hsa_write_summary): Likewise.
5815 * symbol-summary.h (symtab_duplication): Use ::get function/
5816 (get): New function.
5818 2018-06-08 Martin Liska <mliska@suse.cz>
5820 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
5822 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
5823 (hsa_register_kernel): Likewise.
5824 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
5825 * hsa-gen.c (hsa_get_host_function): Likewise.
5826 (get_brig_function_name): Likewise.
5827 (generate_hsa): Likewise.
5828 (pass_gen_hsail::execute): Likewise.
5829 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
5830 (devirtualization_time_bonus): Likewise.
5831 (ipcp_propagate_stage): Likewise.
5832 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
5833 (edge_set_predicate): Likewise.
5834 (evaluate_conditions_for_known_args): Likewise.
5835 (evaluate_properties_for_edge): Likewise.
5836 (ipa_fn_summary::reset): Likewise.
5837 (ipa_fn_summary_t::duplicate): Likewise.
5838 (dump_ipa_call_summary): Likewise.
5839 (ipa_dump_fn_summary): Likewise.
5840 (analyze_function_body): Likewise.
5841 (compute_fn_summary): Likewise.
5842 (estimate_edge_devirt_benefit): Likewise.
5843 (estimate_edge_size_and_time): Likewise.
5844 (estimate_calls_size_and_time): Likewise.
5845 (estimate_node_size_and_time): Likewise.
5846 (inline_update_callee_summaries): Likewise.
5847 (remap_edge_change_prob): Likewise.
5848 (remap_edge_summaries): Likewise.
5849 (ipa_merge_fn_summary_after_inlining): Likewise.
5850 (ipa_update_overall_fn_summary): Likewise.
5851 (read_ipa_call_summary): Likewise.
5852 (inline_read_section): Likewise.
5853 (write_ipa_call_summary): Likewise.
5854 (ipa_fn_summary_write): Likewise.
5855 (ipa_free_fn_summary): Likewise.
5856 * ipa-hsa.c (process_hsa_functions): Likewise.
5857 (ipa_hsa_write_summary): Likewise.
5858 (ipa_hsa_read_section): Likewise.
5859 * ipa-icf.c (sem_function::merge): Likewise.
5860 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
5861 (do_estimate_edge_time): Likewise.
5862 (estimate_size_after_inlining): Likewise.
5863 (estimate_growth): Likewise.
5864 (growth_likely_positive): Likewise.
5865 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
5866 (inline_call): Likewise.
5867 * ipa-inline.c (caller_growth_limits): Likewise.
5868 (can_inline_edge_p): Likewise.
5869 (can_inline_edge_by_limits_p): Likewise.
5870 (compute_uninlined_call_time): Likewise.
5871 (compute_inlined_call_time): Likewise.
5872 (want_inline_small_function_p): Likewise.
5873 (edge_badness): Likewise.
5874 (update_caller_keys): Likewise.
5875 (update_callee_keys): Likewise.
5876 (recursive_inlining): Likewise.
5877 (inline_small_functions): Likewise.
5878 (inline_to_all_callers_1): Likewise.
5879 (dump_overall_stats): Likewise.
5880 (early_inline_small_functions): Likewise.
5881 (early_inliner): Likewise.
5882 * ipa-inline.h (estimate_edge_growth): Likewise.
5883 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
5884 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
5885 * ipa-prop.h (IPA_NODE_REF): Likewise.
5886 (IPA_EDGE_REF): Likewise.
5887 * ipa-pure-const.c (malloc_candidate_p): Likewise.
5888 (propagate_malloc): Likewise.
5889 * ipa-split.c (execute_split_functions): Likewise.
5890 * symbol-summary.h: Rename get to get_create.
5892 (get_create): Likewise.
5893 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
5895 2018-06-08 Martin Liska <mliska@suse.cz>
5897 * symbol-summary.h (release): Move definition out of class
5899 (symtab_removal): Likewise.
5900 (symtab_duplication): Likewise.
5902 2018-06-08 Martin Liska <mliska@suse.cz>
5904 * symbol-summary.h (function_summary): Move constructor
5905 implementation out of class declaration.
5906 (release): Likewise.
5907 (symtab_insertion): Likewise.
5908 (symtab_removal): Likewise.
5909 (symtab_duplication): Likewise.
5912 2018-06-08 Martin Liska <mliska@suse.cz>
5914 * Makefile.in: Remove support for MPX (macros, related functions,
5915 fields in cgraph_node, ...).
5916 * builtin-types.def (BT_BND): Likewise.
5917 (BT_FN_BND_CONST_PTR): Likewise.
5918 (BT_FN_CONST_PTR_BND): Likewise.
5919 (BT_FN_VOID_PTR_BND): Likewise.
5920 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
5921 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
5922 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
5923 (expand_builtin_mempcpy_with_bounds): Likewise.
5924 (expand_builtin_memset_with_bounds): Likewise.
5925 (expand_builtin_memset_args): Likewise.
5926 (std_expand_builtin_va_start): Likewise.
5927 (expand_builtin): Likewise.
5928 (expand_builtin_with_bounds): Likewise.
5929 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
5930 (DEF_LIB_BUILTIN_CHKP): Likewise.
5931 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
5932 (DEF_CHKP_BUILTIN): Likewise.
5933 (BUILT_IN_MEMCPY): Likewise.
5934 (BUILT_IN_MEMMOVE): Likewise.
5935 (BUILT_IN_MEMPCPY): Likewise.
5936 (BUILT_IN_MEMSET): Likewise.
5937 (BUILT_IN_STPCPY): Likewise.
5938 (BUILT_IN_STRCAT): Likewise.
5939 (BUILT_IN_STRCHR): Likewise.
5940 (BUILT_IN_STRCPY): Likewise.
5941 (BUILT_IN_STRLEN): Likewise.
5942 (BUILT_IN_MEMCPY_CHK): Likewise.
5943 (BUILT_IN_MEMMOVE_CHK): Likewise.
5944 (BUILT_IN_MEMPCPY_CHK): Likewise.
5945 (BUILT_IN_MEMSET_CHK): Likewise.
5946 (BUILT_IN_STPCPY_CHK): Likewise.
5947 (BUILT_IN_STRCAT_CHK): Likewise.
5948 (BUILT_IN_STRCPY_CHK): Likewise.
5949 * calls.c (store_bounds): Likewise.
5950 (emit_call_1): Likewise.
5951 (special_function_p): Likewise.
5952 (maybe_warn_nonstring_arg): Likewise.
5953 (initialize_argument_information): Likewise.
5954 (finalize_must_preallocate): Likewise.
5955 (compute_argument_addresses): Likewise.
5956 (expand_call): Likewise.
5957 * cfgexpand.c (expand_call_stmt): Likewise.
5958 (expand_return): Likewise.
5959 (expand_gimple_stmt_1): Likewise.
5960 (pass_expand::execute): Likewise.
5961 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
5962 (cgraph_node::remove): Likewise.
5963 (cgraph_node::dump): Likewise.
5964 (cgraph_node::verify_node): Likewise.
5965 * cgraph.h (chkp_function_instrumented_p): Likewise.
5966 (symtab_node::get_alias_target): Likewise.
5967 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
5968 (cgraph_local_p): Likewise.
5969 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
5970 (cgraph_edge::rebuild_references): Likewise.
5971 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
5972 (walk_polymorphic_call_targets): Likewise.
5973 (cgraph_node::expand_thunk): Likewise.
5974 (symbol_table::output_weakrefs): Likewise.
5975 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
5976 (ix86_handle_option): Likewise.
5977 * config/i386/constraints.md: Likewise.
5978 * config/i386/i386-builtin-types.def (BND): Likewise.
5982 * config/i386/i386-builtin.def (BDESC_END): Likewise.
5983 (BDESC_FIRST): Likewise.
5985 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
5986 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
5987 * config/i386/i386.c (enum reg_class): Likewise.
5988 (ix86_target_string): Likewise.
5989 (ix86_option_override_internal): Likewise.
5990 (ix86_conditional_register_usage): Likewise.
5991 (ix86_valid_target_attribute_inner_p): Likewise.
5992 (ix86_set_indirect_branch_type): Likewise.
5993 (ix86_set_current_function): Likewise.
5994 (ix86_function_arg_regno_p): Likewise.
5995 (init_cumulative_args): Likewise.
5996 (ix86_function_arg_advance): Likewise.
5997 (ix86_function_arg): Likewise.
5998 (ix86_pass_by_reference): Likewise.
5999 (ix86_function_value_regno_p): Likewise.
6000 (ix86_function_value_1): Likewise.
6001 (ix86_function_value_bounds): Likewise.
6002 (ix86_return_in_memory): Likewise.
6003 (ix86_setup_incoming_vararg_bounds): Likewise.
6004 (ix86_va_start): Likewise.
6005 (indirect_thunk_need_prefix): Likewise.
6006 (print_reg): Likewise.
6007 (ix86_print_operand): Likewise.
6008 (ix86_expand_call): Likewise.
6009 (ix86_output_function_return): Likewise.
6010 (reg_encoded_number): Likewise.
6011 (BDESC_VERIFYS): Likewise.
6012 (ix86_init_mpx_builtins): Likewise.
6013 (ix86_init_builtins): Likewise.
6014 (ix86_emit_cmove): Likewise.
6015 (ix86_emit_move_max): Likewise.
6016 (ix86_expand_builtin): Likewise.
6017 (ix86_builtin_mpx_function): Likewise.
6018 (ix86_get_arg_address_for_bt): Likewise.
6019 (ix86_load_bounds): Likewise.
6020 (ix86_store_bounds): Likewise.
6021 (ix86_load_returned_bounds): Likewise.
6022 (ix86_store_returned_bounds): Likewise.
6023 (ix86_class_likely_spilled_p): Likewise.
6024 (ix86_hard_regno_mode_ok): Likewise.
6025 (x86_order_regs_for_local_alloc): Likewise.
6026 (ix86_mitigate_rop): Likewise.
6027 (ix86_bnd_prefixed_insn_p): Likewise.
6028 (ix86_mpx_bound_mode): Likewise.
6029 (ix86_make_bounds_constant): Likewise.
6030 (ix86_initialize_bounds): Likewise.
6031 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
6032 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
6033 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
6034 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
6035 (TARGET_CHKP_BOUND_MODE): Likewise.
6036 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
6037 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
6038 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
6039 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
6040 * config/i386/i386.h (TARGET_MPX): Likewise.
6041 (TARGET_MPX_P): Likewise.
6042 (VALID_BND_REG_MODE): Likewise.
6043 (FIRST_BND_REG): Likewise.
6044 (LAST_BND_REG): Likewise.
6045 (enum reg_class): Likewise.
6046 (BND_REG_P): Likewise.
6047 (BND_REGNO_P): Likewise.
6048 (BNDmode): Likewise.
6049 (ADJUST_INSN_LENGTH): Likewise.
6050 * config/i386/i386.md: Likewise.
6051 * config/i386/i386.opt: Likewise.
6052 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
6053 (defined): Likewise.
6054 (LINK_MPX): Likewise.
6055 (MPX_SPEC): Likewise.
6056 (LIBMPX_SPEC): Likewise.
6057 (LIBMPXWRAPPERS_SPEC): Likewise.
6058 (CHKP_SPEC): Likewise.
6059 * config/i386/predicates.md: Likewise.
6060 * dbxout.c (dbxout_type): Likewise.
6061 * doc/extend.texi: Likewise.
6062 * doc/invoke.texi: Likewise.
6063 * doc/md.texi: Likewise.
6064 * doc/tm.texi: Likewise.
6065 * doc/tm.texi.in: Likewise.
6066 * dwarf2out.c (is_base_type): Likewise.
6067 (gen_formal_types_die): Likewise.
6068 (gen_subprogram_die): Likewise.
6069 (gen_type_die_with_usage): Likewise.
6070 (gen_decl_die): Likewise.
6071 (dwarf2out_late_global_decl): Likewise.
6072 * expr.c (expand_assignment): Likewise.
6073 (emit_storent_insn): Likewise.
6074 (store_expr_with_bounds): Likewise.
6075 (store_expr): Likewise.
6076 (expand_expr_real_1): Likewise.
6077 * expr.h (store_expr_with_bounds): Likewise.
6078 * function.c (use_register_for_decl): Likewise.
6079 (struct bounds_parm_data): Likewise.
6080 (assign_parms_augmented_arg_list): Likewise.
6081 (assign_parm_find_entry_rtl): Likewise.
6082 (assign_parm_is_stack_parm): Likewise.
6083 (assign_parm_load_bounds): Likewise.
6084 (assign_bounds): Likewise.
6085 (assign_parms): Likewise.
6086 (expand_function_start): Likewise.
6087 * gcc.c (CHKP_SPEC): Likewise.
6088 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
6089 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
6090 (wrestrict_dom_walker::check_call): Likewise.
6091 * gimple.c (gimple_build_call_from_tree): Likewise.
6092 * gimple.h (enum gf_mask): Likewise.
6093 (gimple_call_with_bounds_p): Likewise.
6094 (gimple_call_set_with_bounds): Likewise.
6095 * gimplify.c (gimplify_init_constructor): Likewise.
6096 * ipa-cp.c (initialize_node_lattices): Likewise.
6097 (propagate_constants_across_call): Likewise.
6098 (find_more_scalar_values_for_callers_subset): Likewise.
6099 * ipa-hsa.c (process_hsa_functions): Likewise.
6100 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
6101 * ipa-icf.c (sem_function::merge): Likewise.
6102 * ipa-inline.c (early_inliner): Likewise.
6103 * ipa-pure-const.c (warn_function_noreturn): Likewise.
6104 (warn_function_cold): Likewise.
6105 (propagate_pure_const): Likewise.
6106 * ipa-ref.h (enum GTY): Likewise.
6107 * ipa-split.c (find_retbnd): Likewise.
6108 (consider_split): Likewise.
6109 (split_function): Likewise.
6110 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
6111 * ipa.c (walk_polymorphic_call_targets): Likewise.
6112 (symbol_table::remove_unreachable_nodes): Likewise.
6113 (process_references): Likewise.
6114 (cgraph_build_static_cdtor_1): Likewise.
6115 * lto-cgraph.c (lto_output_node): Likewise.
6116 (output_refs): Likewise.
6117 (compute_ltrans_boundary): Likewise.
6118 (input_overwrite_node): Likewise.
6119 (input_node): Likewise.
6120 (input_cgraph_1): Likewise.
6121 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
6122 * passes.c (pass_manager::execute_early_local_passes): Likewise.
6123 (class pass_chkp_instrumentation_passes): Likewise.
6124 (make_pass_chkp_instrumentation_passes): Likewise.
6125 * passes.def: Likewise.
6126 * rtl.h (struct GTY): Likewise.
6127 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
6128 * stor-layout.c (layout_type): Likewise.
6129 * symtab.c: Likewise.
6130 * target.def: Likewise.
6131 * targhooks.c (default_chkp_bound_type): Likewise.
6132 (default_chkp_bound_mode): Likewise.
6133 (default_builtin_chkp_function): Likewise.
6134 (default_chkp_function_value_bounds): Likewise.
6135 (default_chkp_make_bounds_constant): Likewise.
6136 (default_chkp_initialize_bounds): Likewise.
6137 * targhooks.h (default_chkp_bound_type): Likewise.
6138 (default_chkp_bound_mode): Likewise.
6139 (default_builtin_chkp_function): Likewise.
6140 (default_chkp_function_value_bounds): Likewise.
6141 (default_chkp_make_bounds_constant): Likewise.
6142 (default_chkp_initialize_bounds): Likewise.
6143 * toplev.c (compile_file): Likewise.
6144 (process_options): Likewise.
6145 * tree-core.h (DEF_BUILTIN): Likewise.
6146 (DEF_BUILTIN_CHKP): Likewise.
6147 * tree-inline.c (declare_return_variable): Likewise.
6148 (remap_gimple_stmt): Likewise.
6149 (copy_bb): Likewise.
6150 (initialize_inlined_parameters): Likewise.
6151 (expand_call_inline): Likewise.
6152 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
6153 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
6154 (make_pass_ipa_chkp_produce_thunks): Likewise.
6155 (make_pass_chkp): Likewise.
6156 (make_pass_chkp_opt): Likewise.
6157 (make_pass_chkp_instrumentation_passes): Likewise.
6158 * tree-pretty-print.c (dump_generic_node): Likewise.
6159 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
6160 * tree-ssa-dce.c (propagate_necessity): Likewise.
6161 (eliminate_unnecessary_stmts): Likewise.
6162 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
6163 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
6164 * tree-ssa-sccvn.h: Likewise.
6165 * tree-ssa-strlen.c (get_string_length): Likewise.
6166 (valid_builtin_call): Likewise.
6167 (adjust_last_stmt): Likewise.
6168 (handle_builtin_strchr): Likewise.
6169 (handle_builtin_strcpy): Likewise.
6170 (handle_builtin_stxncpy): Likewise.
6171 (handle_builtin_memcpy): Likewise.
6172 (handle_builtin_strcat): Likewise.
6173 (strlen_check_and_optimize_stmt): Likewise.
6174 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
6175 * tree-streamer-in.c: Likewise.
6176 * tree-streamer.c (record_common_node): Likewise.
6177 * tree.c (tree_code_size): Likewise.
6178 (wide_int_to_tree_1): Likewise.
6179 (type_contains_placeholder_1): Likewise.
6180 (build_common_tree_nodes): Likewise.
6181 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
6182 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
6183 (POINTER_BOUNDS_P): Likewise.
6184 (BOUNDED_TYPE_P): Likewise.
6185 (BOUNDED_P): Likewise.
6186 (CALL_WITH_BOUNDS_P): Likewise.
6187 (pointer_bounds_type_node): Likewise.
6188 * value-prof.c (gimple_ic): Likewise.
6189 * var-tracking.c (vt_add_function_parameters): Likewise.
6190 * varasm.c (make_decl_rtl): Likewise.
6191 (assemble_start_function): Likewise.
6192 (output_constant): Likewise.
6193 (maybe_assemble_visibility): Likewise.
6194 * varpool.c (ctor_for_folding): Likewise.
6195 * chkp-builtins.def: Remove.
6196 * ipa-chkp.c: Remove.
6197 * ipa-chkp.h: Remove.
6198 * rtl-chkp.c: Remove.
6199 * rtl-chkp.h: Remove.
6200 * tree-chkp-opt.c: Remove.
6201 * tree-chkp.c: Remove.
6202 * tree-chkp.h: Remove.
6204 2018-06-07 Carl Love <cel@us.ibm.com>
6206 * config/rs6000/vsx.md (vextract_fp_from_shorth,
6207 vextract_fp_from_shortl): Add BE support.
6209 2018-06-07 Paul Koning <ni1d@arrl.net>
6211 * compare-elim.c (try_merge_compare): Don't merge compare if
6212 address contains a side effect.
6213 (try_eliminate_compare): Likewise.
6215 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
6217 * config.gcc: Support "tremont".
6218 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
6219 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
6221 * config/i386/i386.c (m_TREMONT): Define.
6222 (processor_target_table): Add "tremont".
6223 (PTA_TREMONT): Define.
6224 (ix86_lea_outperforms): Add TARGET_TREMONT.
6225 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
6226 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
6227 and M_INTEL_GOLDMONT_PLUS.
6228 (fold_builtin_cpu): Add "tremont".
6229 (ix86_add_stmt_cost): Add TARGET_TREMONT.
6230 (ix86_option_override_internal): Add "tremont".
6231 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
6232 (processor_type): Add PROCESSOR_TREMONT.
6233 * config/i386/x86-tune.def: Add m_TREMONT.
6234 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
6236 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
6238 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
6239 symbol defined for msp430i* devices to be lower case.
6241 2018-06-07 Richard Biener <rguenther@suse.de>
6243 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
6244 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
6245 Properly wrap signed arithmetic if overflow wraps.
6247 2018-06-07 Jakub Jelinek <jakub@redhat.com>
6249 PR tree-optimization/69615
6250 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
6251 of a cast from a same precision integral SSA_NAME in a bb dominated
6252 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
6253 cast to utype if rhs2 has already a compatible type.
6255 2018-06-07 Richard Biener <rguenther@suse.de>
6257 PR tree-optimization/85935
6258 * graphite-scop-detection.c (find_params_in_bb): Analyze
6259 condition operands with respect to the correct loop. Assert
6260 the analysis doesn't fail.
6262 2018-06-04 Carl Love <cel@us.ibm.com>
6264 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
6265 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
6266 as it is slightly cheaper.
6267 (first_match_or_eos_index_<mode>):
6268 Calculate index using natural element order.
6269 (first_match_index_<mode>):
6270 Calculate index using natural element order.
6271 (first_match_or_eos_index_<mode>):
6272 Calculate index using natural order.
6273 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
6274 for BE and LE modes.
6275 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
6276 P9V_BUILTIN_VCLZLSBB_V16QI.
6277 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
6280 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
6282 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
6283 indentation and line wrap for many prototypes. Add missing
6284 @smallexample directives around block of prototypes for vec_xl and
6287 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
6289 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
6290 track if we pass or return IEEE 128-bit floating point.
6291 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
6292 C++ mangling that is compatible with GCC 8.1.
6293 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
6294 (init_cumulative_args): Note if we pass or return IEEE 128-bit
6295 floating point types.
6296 (rs6000_function_arg_advance_1): Likewise.
6297 (rs6000_mangle_type): Optionally generate mangled names that match
6298 what GCC 8.1 generated for IEEE 128-bit floating point types.
6299 (rs6000_globalize_decl_name): If we have an external function that
6300 passes or returns IEEE 128-bit types, generate a weak reference
6301 from the mangled name used in GCC 8.1 to the current mangled
6303 (rs6000_init_builtins): Make __ibm128 use the long double type if
6304 long double is IBM extended double. Make __float128 use the long
6305 double type if long double is IEEE 128-bit.
6308 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
6309 macro for __ibm128 built-in functions.
6310 (PACK_IF): Add __ibm128 pack/unpack functions.
6311 (UNPACK_IF): Likewise.
6312 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
6313 enable long double built-in functions if long double is IEEE
6314 128-bit floating point.
6315 (rs6000_invalid_builtin): Update long double built-in function
6317 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
6318 functions, adjust the built-in function to use the long double
6319 built-in function if __ibm128 and long double are the same type.
6320 * doc/extend.texi (PowerPC builtins): Update documention for
6321 __builtin_{,un}pack_longdouble. Add documentation for
6322 __builtin_{,un}pack_ibm128.
6324 2018-06-06 Jim Wilson <jimw@sifive.com>
6326 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
6327 (struct machine_function): New field interrupt_mode.
6328 (riscv_handle_type_attribute): New function. Add forward declaration.
6329 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
6330 (riscv_expand_epilogue): Check interrupt_mode field.
6331 (riscv_set_current_function): Check interrupt attribute args and
6332 set interrupt_mode field.
6333 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
6334 (riscv_sret, riscv_uret): New.
6335 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
6336 new arguments to interrupt attribute.
6338 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
6341 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
6342 Don't handle -mcpu=power8 if -mpower9-vector is also used.
6344 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
6346 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
6347 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
6348 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
6349 several redundant entries.
6351 2018-06-06 David Malcolm <dmalcolm@redhat.com>
6353 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
6354 type from "rtx" to "rtx_insn *".
6355 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
6356 for local "call_insn", removing cast.
6357 (ix86_expand_call): Likewise, introducing a "call_insn" local.
6359 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
6361 PR tree-optimization/86066
6362 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
6363 for BIT_INSERT_EXPR stores.
6365 2018-06-06 Richard Biener <rguenther@suse.de>
6367 PR tree-optimization/86062
6368 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
6369 component refs ontop
6370 of to be offsetted base.
6372 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
6374 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
6375 to be static and remove check on interrupt attribute name.
6377 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
6379 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
6380 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
6382 2018-06-05 Steve Ellcey <sellcey@cavium.com>
6385 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
6387 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
6388 Remove second argument, change how error is called.
6389 (aarch64_layout_arg): Remove second argument from
6390 aarch64_err_no_fpadvsimd call.
6391 (aarch64_init_cumulative_args): Ditto.
6392 (aarch64_gimplify_va_arg_expr): Ditto.
6393 * config/aarch64/aarch64.md (mov<mode>): Ditto.
6395 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
6397 * config/i386/i386.md (simple_return_indirect_internal): New expander.
6398 (*simple_return_indirect_internal<mode>): Rename from
6399 simple_return_indirect_internal. Use W mode iterator.
6400 (rstorssp): New expander.
6401 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
6402 (clrssbsy): New expander.
6403 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
6405 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
6407 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
6409 (cmse_check_pointed_object): Likewise.
6411 2018-06-05 Martin Liska <mliska@suse.cz>
6413 PR gcov-profile/47618
6414 * doc/invoke.texi: Document how -fprofile-dir format
6417 2018-06-05 Richard Biener <rguenther@suse.de>
6419 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
6420 removal pretend DOM info isn't available so we do not update
6421 it and only remove edges, not dominated blocks. Actually free
6422 DOM info in case we removed something. Remove unreachable blocks.
6423 (mfb_keep_latches): Work with either DOM info or marked backedges.
6424 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
6425 first. Mark backedges if DOM info isn't available.
6426 (Re-)compute DOM info after cleanup_control_flow_pre.
6428 2018-06-05 Richard Biener <rguenther@suse.de>
6430 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
6431 (locus_discrim_hasher::hash): Adjust.
6432 (locus_discrim_hasher::equal): Likewise.
6433 (next_discriminator_for_locus): Work on line directly.
6434 (same_line_p): Pass in expanded locus1 as well.
6435 (assign_discriminators): Avoid redundant location expansions.
6437 2018-06-05 Richard Biener <rguenther@suse.de>
6439 PR tree-optimization/86046
6440 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
6441 if required after clearing TREE_ADDRESSABLE.
6443 2018-06-05 Richard Biener <rguenther@suse.de>
6445 PR tree-optimization/86047
6446 * tree-ssa-loop.c (for_each_index): Glob handling of all
6447 decls and constants and really handle all of them.
6449 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6452 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
6453 qualifier_void_pointer and qualifier_const_void_pointer.
6454 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
6455 (arm_init_builtins): Handle the above.
6456 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
6457 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
6458 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
6461 2018-06-05 Martin Liska <mliska@suse.cz>
6463 * auto-profile.c (read_autofdo_file): Do not use
6464 gcov_ctr_summary struct.
6465 (afdo_callsite_hot_enough_for_early_inline): Likewise.
6466 * coverage.c (struct counts_entry): Likewise.
6467 (read_counts_file): Read just single summary entry.
6468 (get_coverage_counts): Use gcov_summary struct.
6469 * coverage.h (get_coverage_counts): Likewise.
6470 * gcov-dump.c (dump_working_sets): Likewise.
6471 (tag_summary): Dump just single summary.
6472 * gcov-io.c (gcov_write_summary): Write just histogram
6474 (gcov_read_summary): Read just single summary.
6475 (compute_working_sets): Use gcov_summary struct.
6476 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
6477 of GCOV_COUNTERS_SUMMABLE.
6478 (GCOV_COUNTERS_SUMMABLE): Remove.
6479 (GCOV_FIRST_VALUE_COUNTER): Replace with
6480 GCOV_COUNTER_V_INTERVAL.
6481 (struct gcov_ctr_summary): Remove.
6482 (struct gcov_summary): Directly use fields of former
6484 (compute_working_sets): Use gcov_summary struct.
6485 * gcov.c (read_count_file): Do not use ctrs fields.
6486 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
6488 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
6490 * profile.c: Likewise.
6491 * profile.h: Likewise.
6493 2018-06-05 Martin Liska <mliska@suse.cz>
6495 PR gcov-profile/84846
6496 * gcov.c (output_lines): Print working directory only
6497 in intermediate format.
6499 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
6501 * config/s390/s390-builtin-types.def: Add void function type.
6502 * config/s390/s390-builtins.def: Use the function type for the
6505 2018-06-04 Jim Wilson <jimw@sifive.com>
6507 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
6509 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
6510 and maybe_eh_return. Change regno to unsigned int. Use new args to
6511 handle EH_RETURN_DATA_REGNO registers properly.
6512 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
6513 (riscv_expand_epilogue): Update comment. Change argument name and
6514 type. Update code to use new name and type. Pass new args to
6515 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
6517 * config/riscv/riscv.md (NORMAL_RETURN): New.
6518 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
6519 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
6520 (eh_return): Call gen_eh_return_internal and emit barrier.
6521 (eh_return_internal): Call riscv_expand_epilogue.
6523 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
6525 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
6526 bit_insertion field and declare can_be_merged_into method.
6527 (merged_store_group::can_be_merged_into): New method.
6528 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
6529 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
6530 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
6532 2018-06-04 Richard Biener <rguenther@suse.de>
6534 PR tree-optimization/85955
6535 * builtins.c (fold_builtin_sincos): Convert pointers to
6536 destination to appropriate type before dereferencing.
6538 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
6540 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
6542 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
6544 * expr.c (expand_expr_real_1): Force the operand into memory if
6545 its TYPE_MODE is BLKmode and if there is no integer mode for
6546 the number of bits being extracted.
6548 2018-06-04 Jakub Jelinek <jakub@redhat.com>
6552 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
6553 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
6555 2018-06-04 Richard Biener <rguenther@suse.de>
6557 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
6558 (cleanup_tree_cfg_noloop): ... single caller. Do
6559 start_recording_case_labels later.
6561 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
6563 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
6564 to _IMMINTRIN_H_INCLUDED.
6565 * config/i386/pconfigintrin.h: Ditto.
6566 * config/i386/waitpkgintrin.h: Ditto.
6567 * config/i386/immintrin.h: Add includes for sgxintrin.h,
6568 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
6569 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
6570 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
6571 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
6572 waitpkgintrin.h and cldemoteintrin.h.
6574 2018-06-04 Richard Biener <rguenther@suse.de>
6576 PR tree-optimization/86038
6577 * tracer.c (find_best_successor): Check probability for
6578 being initialized, bail out if not.
6580 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
6583 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
6584 of bits to ignore when comparing architectures.
6586 2018-06-04 Jakub Jelinek <jakub@redhat.com>
6588 PR tree-optimization/69615
6589 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
6590 maximum or minimum of the type, try to merge it also as if
6591 range1 is + [-, x - 1] or + [x + 1, -].
6594 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
6596 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
6598 PR tree-optimization/86034
6599 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
6600 the unsigned bitfield type in a bit insertion sequence if it does not
6601 have a larger precision than the bitfield size.
6602 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
6604 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
6606 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
6608 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
6610 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
6611 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
6612 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
6613 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
6615 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
6617 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
6618 Disable -fdelete-null-pointer-checks for ELF toolchain.
6620 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
6621 Kito Cheng <kito.cheng@gmail.com>
6623 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
6624 (nds32le-*-*, nds32be-*-*): Integrate checking process.
6625 (nds32*-*-*): Add glibc and uclibc conditions.
6626 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
6627 (TARGET_EXCEPT_UNWIND_INFO): Define.
6628 * config/nds32/elf.h: New file.
6629 * config/nds32/linux.h: New file.
6630 * config/nds32/nds32-elf.opt: New file.
6631 * config/nds32/nds32-linux.opt: New file.
6632 * config/nds32/nds32-fp-as-gp.c
6633 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
6634 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
6636 (nds32_asm_file_end): Ditto.
6637 (nds32_print_operand): Ditto.
6638 (nds32_insert_attributes): Ditto.
6639 (nds32_init_libfuncs): New function.
6640 (TARGET_HAVE_TLS): Define.
6641 (TARGET_INIT_LIBFUNCS): Define.
6642 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
6644 (TARGET_ELF): Apply different mcmodel setting.
6645 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
6646 been migrated into elf.h and linux.h files.
6647 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
6648 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
6649 (mcmodel): The content has been migrated into nds32-elf.opt and
6650 nds32-linux.opt files.
6651 * config/nds32/t-elf: New file.
6652 * config/nds32/t-linux: New file.
6654 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
6655 Shiva Chen <shiva0217@gmail.com>
6657 * config/nds32/constants.md (unspec_volatile_element): Add
6658 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
6659 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
6661 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
6662 (make_pass_nds32_fp_as_gp): Declare.
6663 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
6665 (nds32_asm_function_end_prologue): Remove unused asm output.
6666 (nds32_asm_function_begin_epilogue): Remove unused asm output.
6667 (nds32_asm_file_start): Output necessary fp_as_gp information.
6668 (nds32_option_override): Adjust register usage.
6669 (nds32_expand_prologue): Consider fp_as_gp situation.
6670 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
6671 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
6674 (simple_return): Ditto.
6675 (omit_fp_begin): Output special directive for fp_as_gp.
6676 (omit_fp_end): Output special directive for fp_as_gp.
6677 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
6678 mforbid-fp-as-gp): New options.
6680 2018-06-01 Mark Wielaard <mark@klomp.org>
6682 * dwarf2out.c (dwarf2out_finish): Remove generation of
6683 DW_AT_loclists_base.
6685 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
6687 * gimple-ssa-store-merging.c: Include gimple-fold.h.
6688 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
6689 (struct merged_store_group): Add bit_insertion field.
6690 (dump_char_array): Use standard hexadecimal format.
6691 (merged_store_group::merged_store_group): Set bit_insertion to false.
6692 (merged_store_group::apply_stores): Use optimal buffer size. Deal
6693 with BIT_INSERT_EXPR stores. Move up code updating the mask and
6694 also print the mask in the dump file.
6695 (pass_store_merging::gate): Minor tweak.
6696 (imm_store_chain_info::coalesce_immediate): Fix wrong association
6697 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
6698 stores with INTEGER_CST stores.
6699 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
6700 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
6701 and use it throughout. Generate bit insertion sequences if need be.
6702 (pass_store_merging::process_store): Remove redundant condition.
6703 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
6705 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
6707 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
6708 the 128-bit floating point types. Fix function comment.
6710 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6712 * config/aarch64/aarch64-simd.md
6713 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
6715 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
6718 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
6720 PR tree-optimization/85989
6721 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
6723 (backprop::intersect_uses): Check it when deciding whether this
6724 is a backedge reference.
6725 (backprop::process_block): Add each phi to m_visited_phis
6726 after visiting it, then clear it at the end.
6728 2018-06-01 Richard Biener <rguenther@suse.de>
6730 * tree-vectorizer.h (vect_dr_stmt): New function.
6731 (vect_get_load_cost): Adjust.
6732 (vect_get_store_cost): Likewise.
6733 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
6734 Use vect_dr_stmt instead of DR_SMTT.
6735 (vect_record_base_alignments): Likewise.
6736 (vect_calculate_target_alignment): Likewise.
6737 (vect_compute_data_ref_alignment): Likewise and make static.
6738 (vect_update_misalignment_for_peel): Likewise.
6739 (vect_verify_datarefs_alignment): Likewise.
6740 (vector_alignment_reachable_p): Likewise.
6741 (vect_get_data_access_cost): Likewise. Pass down
6742 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
6743 (vect_get_peeling_costs_all_drs): Likewise.
6744 (vect_peeling_hash_get_lowest_cost): Likewise.
6745 (vect_enhance_data_refs_alignment): Likewise.
6746 (vect_find_same_alignment_drs): Likewise.
6747 (vect_analyze_data_refs_alignment): Likewise.
6748 (vect_analyze_group_access_1): Likewise.
6749 (vect_analyze_group_access): Likewise.
6750 (vect_analyze_data_ref_access): Likewise.
6751 (vect_analyze_data_ref_accesses): Likewise.
6752 (vect_vfa_segment_size): Likewise.
6753 (vect_small_gap_p): Likewise.
6754 (vectorizable_with_step_bound_p): Likewise.
6755 (vect_prune_runtime_alias_test_list): Likewise.
6756 (vect_analyze_data_refs): Likewise.
6757 (vect_supportable_dr_alignment): Likewise.
6758 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
6759 (vect_gen_prolog_loop_niters): Likewise.
6760 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
6761 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
6763 (vect_recog_mask_conversion_pattern): Likewise.
6764 (vect_try_gather_scatter_pattern): Likewise.
6765 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
6766 to vect_get_store_cost.
6767 (vect_get_store_cost): Get stmt_info instead of DR.
6768 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
6769 (vect_get_load_cost): Get stmt_info instead of DR.
6771 2018-06-01 Richard Biener <rguenther@suse.de>
6774 * gimple-fold.c (var_decl_component_p): Also allow offsetted
6775 vars wrapped in MEM_REFs.
6777 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
6779 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
6780 Fix subreg tests so that we only return a choice between
6781 GENERAL_REGS and FP_REGS if the original classes included both.
6783 2018-06-01 Richard Biener <rguenther@suse.de>
6786 * tree-ssa-structalias.c (get_function_part_constraint):
6787 Handle NULL fi->decl.
6788 (find_func_aliases_for_call): Properly handle indirect
6789 fi from direct call.
6790 (find_func_clobbers): Likewise.
6791 (ipa_pta_execute): Likewise.
6792 (create_variable_info_for): For functions that are ifunc_resolver
6793 resolve to a varinfo that contains the result of the resolver call.
6794 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
6797 2018-05-31 Michael Collison <michael.collison@arm.com>
6799 * config/aarch64/aarch64.md:
6800 (*fix_to_zero_extenddfdi2): New pattern.
6801 * gcc.target/aarch64/fix_extend1.c: New testcase.
6803 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
6807 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
6808 and BUILT_IN_STRNCMP_EQ.
6809 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
6810 BUILT_IN_STRNCMP_EQ.
6811 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
6812 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
6813 (gimple_fold_builtin): Likewise.
6814 * tree-ssa-strlen.c (compute_string_length): New function.
6815 (determine_min_obsize): New function.
6816 (handle_builtin_string_cmp): New function to handle calls to
6817 string compare functions.
6818 (strlen_optimize_stmt): Add handling to builtin string compare
6820 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
6821 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
6822 * tree.c (build_common_builtin_nodes): Add new defines of
6823 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
6825 2018-05-31 Jakub Jelinek <jakub@redhat.com>
6828 * bb-reorder.c (pass_partition_blocks::gate): Return false for
6829 functions with naked attribute.
6831 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
6833 * config/i386/sse.md (avx_vec_concat<mode>):
6834 Substitute concat_tg_mode mode attribute with xtg_mode.
6835 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
6836 (concat_tg_mode): Remove mode attribute.
6838 2018-05-31 Martin Sebor <msebor@redhat.com>
6841 * calls.c (alloc_max_size): Correct a logic error/typo.
6842 Treat excessive arguments as infinite. Warn for invalid arguments.
6843 * doc/invoke.texi (-Walloc-size-larger-than): Update.
6845 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
6848 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
6849 and movx for Haswell.
6851 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
6852 Cesar Philippidis <cesar@codesourcery.com>
6855 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
6856 when emitting error on private/firstprivate reductions.
6857 * omp-low.c (lower_omp_target): Avoid reference-type processing
6858 on pointers for firstprivate clause.
6860 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
6862 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
6865 * config/aarch64/aarch64-simd.md
6866 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
6867 (aarch64_ld1_x3_<mode>): Likewise
6868 (aarch64_st1x2<VALLDIF:mode>): Likewise
6869 (aarch64_st1_x2_<mode>): Likewise
6870 (aarch64_st1x3<VALLDIF:mode>): Likewise
6871 (aarch64_st1_x3_<mode>): Likewise
6872 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
6873 (vld1_s8_x3): Likewise.
6874 (vld1_u16_x3): Likewise.
6875 (vld1_s16_x3): Likewise.
6876 (vld1_u32_x3): Likewise.
6877 (vld1_s32_x3): Likewise.
6878 (vld1_u64_x3): Likewise.
6879 (vld1_s64_x3): Likewise.
6880 (vld1_f16_x3): Likewise.
6881 (vld1_f32_x3): Likewise.
6882 (vld1_f64_x3): Likewise.
6883 (vld1_p8_x3): Likewise.
6884 (vld1_p16_x3): Likewise.
6885 (vld1_p64_x3): Likewise.
6886 (vld1q_u8_x3): Likewise.
6887 (vld1q_s8_x3): Likewise.
6888 (vld1q_u16_x3): Likewise.
6889 (vld1q_s16_x3): Likewise.
6890 (vld1q_u32_x3): Likewise.
6891 (vld1q_s32_x3): Likewise.
6892 (vld1q_u64_x3): Likewise.
6893 (vld1q_s64_x3): Likewise.
6894 (vld1q_f16_x3): Likewise.
6895 (vld1q_f32_x3): Likewise.
6896 (vld1q_f64_x3): Likewise.
6897 (vld1q_p8_x3): Likewise.
6898 (vld1q_p16_x3): Likewise.
6899 (vld1q_p64_x3): Likewise.
6900 (vst1_s64_x2): Likewise.
6901 (vst1_u64_x2): Likewise.
6902 (vst1_f64_x2): Likewise.
6903 (vst1_s8_x2): Likewise.
6904 (vst1_p8_x2): Likewise.
6905 (vst1_s16_x2): Likewise.
6906 (vst1_p16_x2): Likewise.
6907 (vst1_s32_x2): Likewise.
6908 (vst1_u8_x2): Likewise.
6909 (vst1_u16_x2): Likewise.
6910 (vst1_u32_x2): Likewise.
6911 (vst1_f16_x2): Likewise.
6912 (vst1_f32_x2): Likewise.
6913 (vst1_p64_x2): Likewise.
6914 (vst1q_s8_x2): Likewise.
6915 (vst1q_p8_x2): Likewise.
6916 (vst1q_s16_x2): Likewise.
6917 (vst1q_p16_x2): Likewise.
6918 (vst1q_s32_x2): Likewise.
6919 (vst1q_s64_x2): Likewise.
6920 (vst1q_u8_x2): Likewise.
6921 (vst1q_u16_x2): Likewise.
6922 (vst1q_u32_x2): Likewise.
6923 (vst1q_u64_x2): Likewise.
6924 (vst1q_f16_x2): Likewise.
6925 (vst1q_f32_x2): Likewise.
6926 (vst1q_f64_x2): Likewise.
6927 (vst1q_p64_x2): Likewise.
6928 (vst1_s64_x3): Likewise.
6929 (vst1_u64_x3): Likewise.
6930 (vst1_f64_x3): Likewise.
6931 (vst1_s8_x3): Likewise.
6932 (vst1_p8_x3): Likewise.
6933 (vst1_s16_x3): Likewise.
6934 (vst1_p16_x3): Likewise.
6935 (vst1_s32_x3): Likewise.
6936 (vst1_u8_x3): Likewise.
6937 (vst1_u16_x3): Likewise.
6938 (vst1_u32_x3): Likewise.
6939 (vst1_f16_x3): Likewise.
6940 (vst1_f32_x3): Likewise.
6941 (vst1_p64_x3): Likewise.
6942 (vst1q_s8_x3): Likewise.
6943 (vst1q_p8_x3): Likewise.
6944 (vst1q_s16_x3): Likewise.
6945 (vst1q_p16_x3): Likewise.
6946 (vst1q_s32_x3): Likewise.
6947 (vst1q_s64_x3): Likewise.
6948 (vst1q_u8_x3): Likewise.
6949 (vst1q_u16_x3): Likewise.
6950 (vst1q_u32_x3): Likewise.
6951 (vst1q_u64_x3): Likewise.
6952 (vst1q_f16_x3): Likewise.
6953 (vst1q_f32_x3): Likewise.
6954 (vst1q_f64_x3): Likewise.
6955 (vst1q_p64_x3): Likewise.
6957 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
6959 * config/msp430/msp430.c (msp430_output_labelref): Prepend
6960 user_label_prefix to name.
6962 * tree-core.h: Update comment about the format of NAME string
6963 passed to handler in attribute_spec.
6965 * config/msp430/msp430.md: Remove erroneous subreg expression from
6966 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
6967 zero_extend{q,h}isi2.
6969 2018-05-30 Borislav Petkov <bp@suse.de>
6971 * doc/extend.texi: Document some architecture specific
6972 constraints and sort entries.
6974 2018-05-30 Martin Sebor <msebor@redhat.com>
6977 * builtins.c (expand_builtin_stpcpy_1): New function.
6978 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
6979 only if the former succeeds.
6981 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
6983 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
6986 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
6988 * doc/invoke.texi (-flinker-output): Document
6990 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
6992 * passes.c (ipa_write_summaries): Only modify statements if body
6994 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
6995 incrementally linking.
6996 (ipa_passes): Likewise.
6997 * lto-cgraph.c (lto_output_node): When incrementally linking do not
6998 pass down resolution info.
6999 * common.opt (flag_incremental_link): Update info.
7000 * gcc.c (plugin specs): Turn flinker-output=* to
7001 -plugin-opt=-linker-output-known
7002 * toplev.c (compile_file): Also cut compilation when doing incremental
7004 * flag-types. (enum lto_partition_model): Add
7005 LTO_LINKER_OUTPUT_NOLTOREL.
7006 (invoke.texi): Add -flinker-output docs.
7007 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
7008 link same way as WPA; do not stream in dead initializers.
7010 * dwarf2out.c (dwarf2out_die_ref_for_decl,
7011 darf2out_register_external_decl): Support incremental link.
7013 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
7015 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
7017 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
7019 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
7020 it down to simple_object_copy_lto_debug_sections.
7021 (run_gcc): Determine incremental LTO link time and configure
7022 lto1 into non-wpa mode, disable renaming of debug sections.
7024 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
7026 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
7027 descriptions of various incorrectly documented functions.
7029 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
7032 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
7033 address check not strict.
7035 2018-05-30 Richard Biener <rguenther@suse.de>
7037 PR tree-optimization/85964
7038 * tracer.c (better_p): Drop initialized count check, we only
7039 call the function with initialized counts now.
7040 (find_best_successor): Do find a best edge if one
7041 has uninitialized count.
7042 (find_best_predecessor): Likewise. Do BB frequency check only
7043 if count is initialized.
7045 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
7047 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
7048 (aarch64_ldrstr_offset_compare): New.
7049 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
7050 load/store orderings.
7051 (aarch64_gen_adjusted_ldpstp): Likewise.
7053 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
7055 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
7056 Check for subset of GENERAL_REGS and FP_REGS.
7057 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
7060 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
7062 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
7063 and wi::to_poly_offset. Add the current offset and then check
7064 whether the sum fits, rather than using an unchecked addition of
7065 a checked term. Check for a shwi rather than a uhwi.
7066 * expr.c (get_bit_range): Use tree_to_poly_uint64.
7067 (store_constructor): Use poly_int_tree_p.
7068 (expand_expr_real_1): Likewise.
7069 * function.c (assign_temp): Likewise.
7070 * fold-const.c (const_binop): Use poly_int_tree_p and
7072 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
7074 * ipa-icf-gimple.c (func_checker::compare_operand): Use
7075 to_poly_offset for MEM offsets.
7076 * ipa-icf.c (sem_variable::equals): Likewise.
7077 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
7078 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
7079 wi::to_poly_offset for BIT_FIELD_REF offsets.
7080 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
7082 * var-tracking.c (emit_note_insn_var_location): Use
7083 tree_to_poly_uint64.
7085 2018-05-29 Jim Wilson <jimw@sifive.com>
7087 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
7089 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
7092 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
7093 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
7095 (sse4_1_round<mode>2): Use nonimmediate_operand
7096 for operand 1 predicate.
7098 2018-05-29 Martin Sebor <msebor@redhat.com>
7099 Richard Biener <rguenther@suse.de>
7102 * calls.c (get_size_range): Call determine_value_range instead
7103 of get_value_range..
7104 * tree-vrp.h (determine_value_range): Declared new function.
7105 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
7107 2018-05-29 Richard Biener <rguenther@suse.de>
7109 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
7110 sure to use non-pattern stmts for get_earlier_stmt arguments.
7111 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
7112 called on pattern stmts.
7113 (get_later_stmt): Likewise.
7115 2018-05-29 Martin Liska <mliska@suse.cz>
7117 PR gcov-profile/85759
7118 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
7121 2018-05-29 Jakub Jelinek <jakub@redhat.com>
7123 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
7125 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
7129 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
7130 VEC_PACK_FLOAT_EXPR): New tree codes.
7131 * tree-pretty-print.c (op_code_prio): Handle
7132 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
7133 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
7134 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
7135 * tree-inline.c (estimate_operator_cost): Likewise.
7136 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
7137 * fold-const.c (const_binop): Likewise.
7138 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
7139 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
7140 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
7141 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
7142 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
7143 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
7144 * expr.c (expand_expr_real_2): Likewise.
7145 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
7146 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
7147 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
7149 * optabs.c (expand_widen_pattern_expr): For
7150 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
7151 sign from result type rather than operand's type.
7152 (expand_binop_directly): For vec_packu_float_optab and
7153 vec_packs_float_optab allow result type to be different from operand's
7155 * optabs-tree.c (optab_for_tree_code): Handle
7156 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
7157 VEC_PACK_FLOAT_EXPR. Formatting fixes.
7158 * tree-vect-generic.c (expand_vector_operations_1): Handle
7159 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
7160 VEC_PACK_FLOAT_EXPR.
7161 * tree-vect-stmts.c (supportable_widening_operation): Handle
7163 (supportable_narrowing_operation): Handle FLOAT_EXPR.
7164 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
7165 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
7166 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
7167 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
7169 (vec_pack<floatprefix>_float_<mode>): New expander.
7170 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
7172 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
7173 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
7174 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
7175 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
7176 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
7178 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
7179 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
7180 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
7181 VEC_PACK_FLOAT_EXPR): Document.
7183 2018-05-29 Richard Biener <rguenther@suse.de>
7185 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
7187 (stmt_vec_info_vec): Make pointer.
7188 (init_stmt_vec_info_vec): Remove.
7189 (free_stmt_vec_info_vec): Likewise.
7190 (set_stmt_vec_info_vec): New function.
7191 (free_stmt_vec_infos): Likewise.
7192 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
7193 (set_vinfo_for_stmt): Likewise.
7194 (get_earlier_stmt): Likewise.
7195 (get_later_stmt): Likewise.
7196 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
7197 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
7198 (vec_info::~vec_info): Free stmt_vec_infos.
7199 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
7200 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
7201 (pass_slp_vectorize::execute): Likewise.
7202 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
7203 (free_stmt_vec_info_vec): Likewise.
7204 (set_stmt_vec_info_vec): New function.
7205 (free_stmt_vec_infos): Likewise.
7206 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
7207 the global stmt_vec_info_vec.
7208 * tree-parloops.c (gather_scalar_reductions): Use
7209 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
7212 2018-05-29 Richard Biener <rguenther@suse.de>
7214 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
7216 2018-05-29 Martin Liska <mliska@suse.cz>
7217 David Malcolm <dmalcolm@redhat.com>
7219 * vec.c (test_reverse): New.
7220 (vec_c_tests): Add new test.
7221 * vec.h (vl_ptr>::reverse): New function.
7223 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
7225 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
7227 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
7230 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
7232 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
7234 2018-05-28 Richard Biener <rguenther@suse.de>
7236 PR tree-optimization/85933
7237 * tree-vect-data-refs.c (vect_record_base_alignments): Only
7238 look at stmts marked as vectorizable.
7240 2018-05-28 Richard Biener <rguenther@suse.de>
7242 PR tree-optimization/85934
7243 * tree-vect-generic.c (expand_vector_operations_1): Hoist
7244 vector boolean check before scalar optimization.
7246 2018-05-28 Jakub Jelinek <jakub@redhat.com>
7248 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
7251 2018-05-28 Mark Wielaard <mark@klomp.org>
7253 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
7254 if it is an expression containing a minus sign.
7256 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
7258 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
7260 2018-05-27 Paul Koning <ni1d@arrl.net>
7262 * config/pdp11/pdp11.md (truncsihi2): Remove.
7264 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
7265 Chung-Ju Wu <jasonwucj@gmail.com>
7267 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
7269 (unaligned_store_dw): Ditto.
7270 * config/nds32/nds32-memory-manipulation.c
7271 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
7272 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
7273 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
7274 (emit_setmem_word_loop): Rename to ...
7275 (emit_setmem_doubleword_loop): ... this.
7276 (nds32_gen_dup_4_byte_to_word_value): New function.
7277 (nds32_gen_dup_8_byte_to_double_word_value): New function.
7278 (nds32_expand_setmem_loop): Refine implementation.
7279 (nds32_expand_setmem_loop_v3m): Ditto.
7280 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
7283 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
7285 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
7287 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
7289 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
7290 (nds32_init_machine_status): Initialize machine->attr_naked_p and
7291 machine->attr_no_prologue_p.
7292 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
7293 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
7294 (nds32_expand_epilogue): Consider attr_naked_p.
7295 (nds32_expand_epilogue_v3pop): Likewise.
7296 (nds32_can_use_return_insn): Likewise.
7297 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
7298 attr_no_prologue_p fields.
7299 * config/nds32/nds32.opt (mret-in-naked-func): New option.
7301 2018-05-27 Jakub Jelinek <jakub@redhat.com>
7304 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
7306 * config/i386/sse.md
7307 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
7309 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
7311 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
7313 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
7315 (*<floatsuffix>floatv2div2sf2): Rename to ...
7316 (*float<floatunssuffix>v2div2sf2): ... this.
7317 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
7318 (float<floatunssuffix>v2div2sf2_mask): ... this.
7319 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
7320 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
7321 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
7323 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
7325 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
7327 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
7329 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
7331 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
7333 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
7334 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
7335 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
7336 gen_ufix_truncv8dfv8si2.
7337 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
7338 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
7339 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
7340 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
7341 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
7342 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
7343 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
7344 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
7346 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
7350 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
7352 2018-05-25 Jim Wilson <jimw@sifive.com>
7354 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
7355 * config/riscv/riscv.c (struct machine_function): Add
7356 interrupt_handler_p and attribute_checked_p fields.
7357 (riscv_attribute_table): Add interrupt.
7358 (riscv_interrupt_type_p): New.
7359 (riscv_save_reg_p): Save extra regs for interrupt handler.
7360 (riscv_use_save_libcall): Return false for interrupt handler.
7361 (riscv_first_stack_step): Add forward declaration.
7362 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
7363 for interrupt handler with large frame. Use it for saved reg list.
7364 (riscv_expand_prologue): Move flag_stack_usage_info support to
7365 eliminate duplication.
7366 (riscv_expand_epilogue): Generate mret for interrupt handler.
7367 (riscv_epilogue_uses): New.
7368 (riscv_can_use_return_insn): Return false for interrupt handler.
7369 (riscv_function_ok_for_sibcall): Likewise.
7370 (riscv_set_current_function): Add interrupt handler support.
7371 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
7372 * config/riscv/riscv.md (UNSPECV_MRET): New.
7374 (riscv_frflags, riscv_fsflags): Use tab after opcode.
7376 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
7378 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
7380 PR tree-optimization/85712
7381 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
7382 this candidate has already been replaced in-situ by a copy.
7384 2018-05-25 Jason Merrill <jason@redhat.com>
7386 PR c++/80485 - inline function non-zero address.
7387 * symtab.c (nonzero_address): Check DECL_COMDAT.
7389 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
7392 * config/alpha/alpha.md (ashlsi3): New insn pattern.
7393 (*ashlsi_se): Rename from *ashldi_se. Define as sign
7394 extension of SImode operation. Use const123_operand predicate.
7395 (*saddsi_1): Remove.
7396 (*saddl_se_1): Ditto.
7398 (*ssubl_se_1): Ditto.
7399 * config/alpha/predicates.md (const123_operand): New predicate.
7400 * config/alpha/constraints.md (P): Use IN_RANGE.
7402 2018-05-25 Richard Biener <rguenther@suse.de>
7404 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
7406 (ref_maybe_used_by_stmt_p): Likewise.
7407 (stmt_may_clobber_ref_p): Likewise.
7408 (stmt_may_clobber_ref_p_1): Likewise.
7409 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
7411 (ref_maybe_used_by_stmt_p): Likewise.
7412 (stmt_may_clobber_ref_p): Likewise.
7413 (stmt_may_clobber_ref_p_1): Likewise.
7414 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
7415 the alias oracle to disambiguate DRs with stmts DR analysis
7417 (vect_analyze_data_refs): Do not give up on not analyzable
7418 DRs for BB vectorization. Remove code truncating the dataref
7421 2018-05-25 Jakub Jelinek <jakub@redhat.com>
7424 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
7425 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
7426 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
7428 2018-05-25 Richard Biener <rguenther@suse.de>
7430 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
7431 function, combining stmt data ref gathering and fatal analysis
7433 (vect_analyze_data_refs): Remove now redudnant code and simplify.
7434 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
7435 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
7436 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
7437 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
7439 2018-05-25 Bin Cheng <bin.cheng@arm.com>
7441 PR tree-optimization/85720
7442 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
7443 SCC if all partitions are builtins.
7444 (version_loop_by_alias_check): New parameter. Generate cancelable
7445 runtime alias check if all partitions are builtins.
7446 (distribute_loop): Update call to above function.
7448 2018-05-25 Bin Cheng <bin.cheng@arm.com>
7450 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
7451 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
7452 (parm_default_def_partition_arg): Ditto.
7453 (set_parm_default_def_partition): Ditto.
7454 (get_parm_default_def_partitions): Ditto and make it static.
7455 (get_undefined_value_partitions): Ditto and make it static.
7456 (remove_ssa_form): Refactor call to init_var_map here.
7457 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
7458 computation for loop region.
7459 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
7460 (register_default_def): Delete.
7461 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
7462 (parm_default_def_partition_arg): Ditto.
7463 (set_parm_default_def_partition): Ditto.
7464 (get_parm_default_def_partitions): Ditto and make it static.
7465 (get_undefined_value_partitions): Ditto and make it static.
7466 (coalesce_with_default, coalesce_with_default): Update comment.
7467 (create_coalesce_list_for_region): New func factored out from
7468 create_outofssa_var_map.
7469 (populate_coalesce_list_for_outofssa): New func factored out from
7470 create_outofssa_var_map and coalesce_ssa_name.
7471 (create_outofssa_var_map): Delete.
7472 (coalesce_ssa_name): Refactor to support live range computation.
7473 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
7474 (get_parm_default_def_partitions): Delete.
7475 (get_undefined_value_partitions): Ditto.
7476 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
7477 computation for loop region.
7478 (new_tree_live_info, loe_visit_block): Ditto.
7479 (live_worklist, set_var_live_on_entry): Ditto.
7480 (calculate_live_on_exit, verify_live_on_entry): Ditto.
7481 * tree-ssa-live.h (struct _var_map): New fields.
7482 (init_var_map): Change decl.
7483 (region_contains_p): New.
7485 2018-05-25 Bin Cheng <bin.cheng@arm.com>
7487 * tree-ssa-live.h (live_merge_and_clear): Delete.
7489 2018-05-25 Richard Biener <rguenther@suse.de>
7492 * tree-dump.c (dequeue_and_dump): Remove access to removed
7493 operand 2 of a SWITCH_EXPR.
7495 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
7497 * doc/sourcebuild.texi (vect_double_cond_arith): Include
7498 multiplication and division.
7499 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
7500 (cond_udiv@var{m}, cond_umod@var{m}): Document.
7501 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
7502 (cond_udiv_optab, cond_umod_optab): New optabs.
7503 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
7504 (IFN_COND_RDIV): New internal functions.
7505 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
7506 TRUNC_MOD_EXPR and RDIV_EXPR.
7507 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
7508 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
7510 (SVE_INT_BINARY): Include mult.
7511 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
7512 (optab, sve_int_op): Handle mult.
7513 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
7515 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
7516 for SVE_INT_BINARY_SD.
7518 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
7520 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
7521 (optab, sve_int_op): Handle div and udiv.
7522 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
7523 for SVE_INT_BINARY_SD.
7524 (*<optab><mode>3): New insn for the same.
7526 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
7528 * tree-vect-patterns.c: Include predict.h.
7529 (vect_recog_divmod_pattern): Restrict check for division support
7530 to when optimizing for size.
7532 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
7534 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
7535 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
7536 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
7537 (gimple_match_op::set_op): Likewise.
7538 (gimple_resimplify4): Declare.
7539 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
7540 (expr::gen_transform): Likewise.
7541 (decision_tree::gen): Generate a simplification routine for 4 operands.
7542 * gimple-match-head.c (gimple_simplify): Add an overload for
7543 4 operands. In the top-level function, handle up to 4 call
7544 arguments and call gimple_resimplify4.
7545 (gimple_resimplify4): New function.
7546 (build_call_internal): Pass a fourth operand.
7547 (maybe_push_to_seq): Likewise.
7548 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
7549 Fold VEC_COND_EXPRs of an operation and a default value into
7550 an IFN_COND_* function if possible.
7551 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
7553 (SVE_COND_FP_BINARY): Include them.
7554 (optab, sve_fp_op): Handle them.
7555 (SVE_INT_BINARY_REV): New code iterator.
7556 (SVE_COND_FP_BINARY_REV): New int iterator.
7557 (commutative): New int attribute.
7558 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
7560 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
7562 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
7563 (*cond_<optab><mode>): New patterns for reversed operands.
7565 2018-05-25 Richard Biener <rguenther@suse.de>
7567 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
7568 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
7569 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
7570 (STMT_VINFO_GROUPED_ACCESS): Adjust.
7571 * tree-vect-data-refs.c (everywhere): Adjust users.
7572 * tree-vect-loop.c (everywhere): Likewise.
7573 * tree-vect-slp.c (everywhere): Likewise.
7574 * tree-vect-stmts.c (everywhere): Likewise.
7575 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
7577 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7579 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
7581 (gcc_cv_as_section_exclude): ... this.
7582 Try Solaris as #exclude syntax.
7583 * configure: Regenerate.
7584 * config.in: Regenerate.
7585 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
7587 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
7588 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
7590 * varasm.c (default_elf_asm_named_section): Don't check if
7591 HAVE_GAS_SECTION_EXCLUDE is defined.
7593 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
7595 * doc/md.texi: Update the documentation of the cond_* optabs
7596 to mention the new final operand. Fix GET_MODE_NUNITS call.
7597 Describe the scalar case too.
7598 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
7599 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
7601 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
7602 (get_conditional_internal_fn): Update comment.
7603 * tree-vect-loop.c (vectorizable_reduction): Pass the original
7604 accumulator value as a final argument to conditional functions.
7605 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
7606 a define_expand and add an "else" operand. Assert for now that
7607 the else operand is equal to operand 2. Use SVE_INT_BINARY and
7608 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
7609 (*cond_<optab><mode>): New patterns.
7610 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
7611 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
7612 (UNSPEC_COND_EOR): Delete.
7613 (optab): Remove associated mappings.
7614 (SVE_INT_BINARY): New code iterator.
7615 (sve_int_op): Remove int attribute and add "minus" to the code
7617 (SVE_COND_INT_OP): Delete.
7618 (SVE_COND_FP_OP): Rename to...
7619 (SVE_COND_FP_BINARY): ...this.
7621 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
7623 * optabs.c (can_reuse_operands_p): New function.
7624 (maybe_legitimize_operands): Try to reuse the results for
7627 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
7629 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
7630 Add {q} suffix to insn mnemonic.
7632 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
7634 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
7635 (msp430_warn_func_return): New.
7637 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
7639 * fold-const.c (tree_nonzero_bits): New function.
7640 * fold-const.h (tree_nonzero_bits): Likewise.
7641 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
7642 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
7644 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
7648 * varasm.c (assemble_alias): Check ifunc_resolver only on
7651 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
7654 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
7655 when memory input operand is handled.
7657 2018-05-24 Luis Machado <luis.machado@linaro.org>
7659 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
7661 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
7663 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
7665 * match.pd: Delay FMA folds until after vectorization.
7667 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
7670 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
7671 address check not strict.
7673 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
7675 * gimple-match.h (gimple_match_op): New class.
7676 (mprts_hook): Replace parameters with a gimple_match_op *.
7677 (maybe_build_generic_op): Likewise.
7678 (gimple_simplified_result_is_gimple_val): Replace parameters with
7679 a const gimple_match_op *.
7680 (gimple_simplify): Replace code_helper * and tree * parameters with
7681 a gimple_match_op * parameter.
7682 (gimple_resimplify1): Replace code_helper *, tree and tree *
7683 parameters with a gimple_match_op * parameter.
7684 (gimple_resimplify2): Likewise.
7685 (gimple_resimplify3): Likewise.
7686 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
7687 parameters with a gimple_match_op * parameter.
7688 * gimple-match-head.c (gimple_simplify): Change prototypes of
7689 auto-generated functions to take a gimple_match_op * instead of
7690 separate code_helper * and tree * parameters. Make the same
7691 change in the top-level overload and update calls to the
7692 gimple_resimplify routines. Update calls to the auto-generated
7693 functions and to maybe_push_res_to_seq in the publicly-facing
7694 operation-specific gimple_simplify overloads.
7695 (gimple_match_op::MAX_NUM_OPS): Define.
7696 (gimple_resimplify1): Replace rcode and ops with a single res_op
7697 parameter. Update call to gimple_simplify.
7698 (gimple_resimplify2): Likewise.
7699 (gimple_resimplify3): Likewise.
7700 (mprts_hook): Replace parameters with a gimple_match_op *.
7701 (maybe_build_generic_op): Likewise.
7702 (build_call_internal): Replace type, nargs and ops with
7703 a gimple_match_op *.
7704 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
7705 with a single gimple_match_op *. Update calls to mprts_hook,
7706 build_call_internal and gimple_simplified_result_is_gimple_val.
7707 Factor out code that is common to the tree_code and combined_fn cases.
7708 * genmatch.c (expr::gen_transform): Replace tem_code and
7709 tem_ops with a gimple_match_op called tem_op. Update calls
7710 to the gimple_resimplify functions and maybe_push_res_to_seq.
7711 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
7712 res_ops. Update call to the gimple_resimplify functions.
7713 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
7714 (decision_tree::gen): Make the functions take a gimple_match_op *
7715 called res_op instead of separate res_code and res_ops parameters.
7716 Update call accordingly.
7717 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
7718 and ops with a single res_op parameter. Update calls to
7719 maybe_build_generic_op and maybe_push_res_to_seq.
7720 (fold_stmt_1): Update calls to gimple_simplify and
7721 replace_stmt_with_simplification.
7722 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
7723 and gimple_simplified_result_is_gimple_val.
7724 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
7726 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
7727 with a gimple_match_op *.
7728 (vn_nary_build_or_lookup): Likewise. Update call to
7729 vn_nary_build_or_lookup_1.
7730 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
7731 gimple_match_op *. Update calls to the gimple_resimplify routines
7732 and to gimple_simplified_result_is_gimple_val.
7733 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
7734 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
7735 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
7736 (visit_nary_op): Likewise.
7737 (visit_reference_op_load): Likewise.
7739 2018-05-23 Luis Machado <luis.machado@linaro.org>
7741 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
7742 modifier for printing the step amount.
7744 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
7747 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
7750 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
7752 * doc/sourcebuild.texi (Endianness): New subsubsection.
7754 2018-05-23 Luis Machado <luis.machado@linaro.org>
7756 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7757 <prefetch_dynamic_strides>: New const bool field.
7758 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7759 prefetch_dynamic_strides.
7760 (exynosm1_prefetch_tune): Likewise.
7761 (thunderxt88_prefetch_tune): Likewise.
7762 (thunderx_prefetch_tune): Likewise.
7763 (thunderx2t99_prefetch_tune): Likewise.
7764 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
7766 (aarch64_override_options_internal): Update to set
7767 PARAM_PREFETCH_DYNAMIC_STRIDES.
7768 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
7769 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
7770 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
7771 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
7772 prefetch-dynamic-strides setting.
7774 2018-05-23 Luis Machado <luis.machado@linaro.org>
7776 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7777 <minimum_stride>: New const int field.
7778 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7779 minimum_stride field defaulting to -1.
7780 (exynosm1_prefetch_tune): Likewise.
7781 (thunderxt88_prefetch_tune): Likewise.
7782 (thunderx_prefetch_tune): Likewise.
7783 (thunderx2t99_prefetch_tune): Likewise.
7784 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
7785 <default_opt_level>: Set to 3.
7786 (aarch64_override_options_internal): Update to set
7787 PARAM_PREFETCH_MINIMUM_STRIDE.
7788 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
7789 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
7790 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
7791 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
7792 stride is constant and is below the minimum stride threshold.
7794 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7796 * config/arm/arm-cpus.in (mode26): Delete.
7797 (armv4): Delete mode26 reference.
7798 * config/arm/arm.c (arm_configure_build_target): Delete use of
7801 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
7803 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
7805 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
7806 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
7808 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
7809 (floatunsdidf2): Ditto.
7811 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
7813 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
7814 (fixuns_trunc<mode>si2_avx512f): Ditto.
7815 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
7816 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
7817 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
7819 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
7821 PR rtl-optimization/79985
7822 * df-scan.c (df_insn_refs_collect): Remove special case for
7823 global registers and asm statements.
7825 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
7827 * extend.texi (Global Register Variables): Rewrite the bullet list.
7828 Note that the register is available for allocation. Note that access
7829 via inline asm must use constraints. Add note about async-signal
7830 handlers. Remove paragraph about automagic register selection.
7832 2018-05-23 Richard Biener <rguenther@suse.de>
7834 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
7835 of fixed offset from memset VN.
7837 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
7839 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
7841 (alloc_cand_and_find_basis): Initialize first_interp field.
7842 (slsr_process_mul): Modify first_interp field.
7843 (slsr_process_add): Likewise.
7844 (slsr_process_cast): Modify first_interp field for each new
7846 (slsr_process_copy): Likewise.
7847 (dump_candidate): Dump first_interp field.
7848 (replace_mult_candidate): Process all interpretations, not just
7850 (replace_rhs_if_not_dup): Likewise.
7851 (replace_one_candidate): Likewise.
7853 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
7855 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
7857 (aarch64_needs_frame_chain): New function.
7858 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
7860 2018-05-23 Sudakshina Das <sudi.das@arm.com>
7863 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
7864 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
7865 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
7866 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
7867 as true for strict-align.
7868 (aarch64_can_inline_p): Perform checks even when callee has no
7869 attributes to check for strict alignment.
7870 * doc/extend.texi (AArch64 Function Attributes): Document
7872 * doc/invoke.texi: (AArch64 Options): Likewise.
7874 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
7876 PR tree-optimization/85853
7877 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
7878 the handling of the root of the node to...
7879 (vect_slp_analyze_node_operations_1): ...this new function,
7880 and run the whole thing with the child nodes' def types
7881 set according to their SLP node's def type.
7883 2018-05-23 Richard Biener <rguenther@suse.de>
7886 * tree-data-ref.c (create_runtime_alias_checks): Defer
7887 and ignore overflow warnings.
7889 2018-05-23 Yury Gribov <tetra2005@gmail.com>
7891 PR tree-optimization/85822
7892 * tree-vrp.c (is_masked_range_test): Fix handling of negative
7895 2018-05-23 Richard Biener <rguenther@suse.de>
7897 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
7898 memset constants via native_interpret_expr.
7900 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
7903 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
7905 (cgraph_node::create_alias): Likewise.
7906 (cgraph_node::get_availability): Check ifunc_resolver instead
7907 of looking up ifunc attribute.
7908 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
7909 * varasm.c (do_assemble_alias): Likewise.
7910 (assemble_alias): Likewise.
7911 (default_binds_local_p_3): Likewise.
7912 * cgraph.h (cgraph_node): Add ifunc_resolver.
7913 (cgraph_node::only_called_directly_or_aliased_p): Return false
7915 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
7917 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
7918 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
7919 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
7920 instead of looking up ifunc attribute.
7922 2018-05-22 Luis Machado <luis.machado@linaro.org>
7924 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
7926 2018-05-22 Martin Sebor <msebor@redhat.com>
7929 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
7930 only when expasion succeeds.
7931 (expand_builtin_strcmp): Same.
7932 (expand_builtin_strncmp): Same.
7934 2018-05-22 Martin Sebor <msebor@redhat.com>
7936 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
7938 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
7939 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7941 * config/aarch64/aarch64-ldpstp.md: Replace uses of
7942 aarch64_mem_pair_operand with memory_operand and delete operand swapping
7944 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
7945 Add check for legitimate_address.
7946 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
7947 (aarch64_swap_ldrstr_operands): New.
7948 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
7951 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
7952 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7954 * config/aarch64/aarch64.md: New patterns to generate stp
7956 (store_pair_sw, store_pair_dw): New patterns to generate stp for
7957 single words and double words.
7958 (load_pair_sw, load_pair_dw): Likewise.
7959 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
7961 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
7963 * config/aarch64/aarch64-ldpstp.md: Modify peephole
7964 for different mode ldpstp and add peephole for merged zero stores.
7966 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
7968 (aarch64_gen_store_pair): Rename calls to match new patterns.
7969 (aarch64_gen_load_pair): Rename calls to match new patterns.
7970 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
7971 (load_pair<DREG:mode><DREG2:mode>): ... This.
7972 (store_pair<mode>): Rename to...
7973 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
7974 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
7976 (V_INT_EQUIV): Handle SImode.
7977 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
7980 2018-05-22 Martin Sebor <msebor@redhat.com>
7983 * calls.c (maybe_warn_nonstring_arg): Use string length to set
7984 or ajust the presumed bound on an operation to avoid unnecessary
7987 2018-05-22 Martin Sebor <msebor@redhat.com>
7989 PR tree-optimization/85826
7990 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
7991 assuming that a DECL necesarily has a constant size.
7993 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
7996 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
7998 2018-05-22 Richard Biener <rguenther@suse.de>
8000 PR tree-optimization/85834
8001 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
8002 non-constant and non-zero memset arguments.
8004 2018-05-22 Martin Liska <mliska@suse.cz>
8007 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
8009 2018-05-22 Richard Biener <rguenther@suse.de>
8011 PR tree-optimization/85863
8012 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
8013 comparisons when vectype is specified.
8014 (vectorizable_condition): Do not specify vectype for
8015 vect_is_simple_cond when SLP vectorizing.
8017 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
8020 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
8021 define __ibm128 as long double.
8022 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
8023 as a distinct type when IEEE 128-bit support is enabled.
8024 (init_float128_ieee): Fix up conversions between IFmode and IEEE
8025 128-bit types to use the correct functions.
8026 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
8027 convert between 128-bit floating point types that have different
8028 modes but the same representation, instead of using gen_lowpart to
8030 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
8032 (IFKF_reg): New attributes to give the register constraints for
8034 (extend<mode>tf2_internal): New insns to mark an explicit
8035 conversion between 128-bit floating point types that have a
8036 different mode but share the same representation.
8038 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
8040 PR tree-optimization/85814
8041 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
8042 a null return from get_strinfo when unsharing the next
8043 strinfo in the chain.
8045 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
8048 * varasm.c (weak_finish): Clean up weak_decls.
8050 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8052 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
8053 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
8054 UNSPEC_UADALP values.
8055 * config/aarch64/iterators.md (ABAL): New int iterator.
8058 (sur): Add mappings for the above.
8059 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
8061 (aarch64_<sur>abal<mode>_4): Likewise.
8062 (aarch64_<sur>adalp<mode>_3): Likewise.
8063 (<sur>sadv16qi): New define_expand.
8065 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
8067 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
8068 (*movdf_internal): Ditto.
8069 (*rcpsf2_sse): Ditto.
8070 (*rsqrtsf2_sse): Ditto.
8071 (*sqrt<mode>2_sse): Ditto.
8073 2018-05-21 Tamar Christina <tamar.christina@arm.com>
8075 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
8077 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
8078 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
8079 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
8080 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
8082 * config/aarch64/arm_neon.h: Likewise.
8083 * config/aarch64/iterators.md (VQ_I): New.
8085 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
8087 * config.gcc: Add arc/t-multilib-linux to tmake_file for
8089 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
8092 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
8094 * config/nds32/constraints.md (S): New constraint.
8095 * config/nds32/nds32.md (call_internal): Use constraint S.
8096 (call_value_internal): Likewise.
8097 (sibcall_internal): Likewise.
8098 (sibcall_value_internal): Likewise.
8100 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
8101 Chung-Ju Wu <jasonwucj@gmail.com>
8103 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
8106 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
8107 Chung-Ju Wu <jasonwucj@gmail.com>
8109 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
8110 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
8111 (nds32_rtx_costs_impl): Simplify.
8112 (nds32_address_cost_impl): Simplify.
8113 (nds32_init_rtx_costs): New function.
8114 (nds32_rtx_costs_speed_prefer): Likewise.
8115 (nds32_rtx_costs_size_prefer): Likewise.
8116 (nds32_address_cost_speed_prefer): Likewise.
8117 (nds32_address_cost_speed_fwprop): Likewise.
8118 (nds32_address_cost_size_prefer): Likewise.
8119 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
8120 * config/nds32/nds32.c (nds32_option_override): Use
8121 nds32_init_rtx_costs function.
8123 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
8125 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
8126 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
8127 (TARGET_PIPELINE_N8): Likewise.
8128 (TARGET_PIPELINE_N10): Likewise.
8129 (TARGET_PIPELINE_N13): Likewise.
8130 (TARGET_PIPELINE_GRAYWOLF): Likewise.
8132 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
8134 * config/nds32/nds32-fpu.md: Update copyright year.
8136 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
8138 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
8140 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
8142 * config/nds32/nds32.c
8143 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
8144 * config/nds32/nds32.opt (minline-asm-r15): New option.
8146 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
8148 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
8150 * config/nds32/nds32.md (abssi2): New pattern.
8152 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
8154 * config/i386/i386.md (rex64namesuffix): New mode attribute.
8155 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
8156 Merge insn pattern from sse_cvtsi2ss<round_name> and
8157 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
8158 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
8159 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
8160 using SWI48 mode iterator.
8161 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
8162 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
8163 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
8164 pattern from sse_cvttss2si<round_saeonly_name>
8165 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
8166 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
8167 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
8168 using SWI48 mode iterator.
8169 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
8170 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
8171 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
8172 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
8173 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
8174 using SWI48 mode iterator.
8175 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
8176 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
8177 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
8178 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
8179 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
8180 SWI48 mode iterator.
8181 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
8182 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
8183 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
8184 pattern from sse_cvttsd2si<round_saeonly_name>
8185 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
8187 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
8189 * config/nds32/nds32-md-auxiliary.c
8190 (nds32_valid_smw_lwm_base_p): Refine.
8191 (nds32_output_smw_single_word): Refine.
8192 (nds32_output_smw_double_word): New.
8193 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
8195 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
8197 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
8198 (nds32_output_stack_pop): Refine.
8199 (nds32_expand_unaligned_load): Refine.
8200 (nds32_expand_unaligned_store): Refine.
8202 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
8203 Chung-Ju Wu <jasonwucj@gmail.com>
8205 * config/nds32/constants.md: Add TP_REGNUM constant.
8206 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
8207 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
8209 * config/nds32/nds32-doubleword.md: Consider flag_pic.
8210 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
8211 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
8212 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
8213 and PIC code generation.
8214 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
8216 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
8218 * config/nds32/nds32.md: Support TLS and PIC.
8219 * config/nds32/nds32.c: Support TLS and PIC.
8220 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
8221 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
8224 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
8226 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
8227 mode with E_ prefix.
8229 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
8230 Chung-Ju Wu <jasonwucj@gmail.com>
8232 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
8233 * config/nds32/nds32-md-auxiliary.c
8234 (symbolic_reference_mentioned_p): New.
8235 (nds32_legitimize_ict_address): New.
8236 (nds32_expand_ict_move): New.
8237 (nds32_indirect_call_referenced_p): New.
8238 (nds32_symbol_binds_local_p): Delete.
8239 (nds32_long_call_p): Modify.
8240 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
8241 * config/nds32/nds32-protos.h
8242 (symbolic_reference_mentioned_p): Declare.
8243 (nds32_legitimize_ict_address): Declare.
8244 (nds32_expand_ict_move): Declare.
8245 (nds32_indirect_call_referenced_p): Declare.
8246 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
8247 (nds32_relax_group): Use nds32_ict_const_p as condition.
8248 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
8249 (nds32_asm_file_start): Output ict_model directive in asm code.
8250 (nds32_legitimate_address_p): Consider indirect call.
8251 (nds32_print_operand): Consider indirect call.
8252 (nds32_print_operand_address): Consider indirect call.
8253 (nds32_insert_attributes): Handle "indirect_call" attribute.
8254 (TARGET_LEGITIMATE_ADDRESS_P): Define.
8255 (TARGET_LEGITIMATE_CONSTANT_P): Define.
8256 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
8257 (TARGET_DELEGITIMIZE_ADDRESS): Define.
8258 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
8259 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
8260 (TARGET_ICT_MODEL_SMALL): Define.
8261 (TARGET_ICT_MODEL_LARGE): Define.
8262 * config/nds32/nds32.md (movsi): Consider ict model.
8263 (call, call_value): Consider ict model.
8264 (sibcall, sibcall_value): Consider ict model.
8265 * config/nds32/nds32.opt (mict-model): New option.
8266 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
8269 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
8270 Monk Chiang <sh.chiang04@gmail.com>
8271 Jim Wilson <jimw@sifive.com>
8273 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
8274 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
8275 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
8276 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
8277 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
8278 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
8279 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
8280 compute save_libcall_adjustment properly.
8281 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
8282 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
8283 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
8284 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
8285 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
8286 (ABI_SPEC): Handle mabi=ilp32e.
8287 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
8288 (RVE): Add RVE mask.
8289 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
8290 <-march>: Add rv32e as an example.
8292 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
8295 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
8296 (intra_create_variable_infos): Handle C++ constructors.
8298 2018-05-18 Martin Liska <mliska@suse.cz>
8300 * passes.def: Remove a redundant pass.
8302 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
8305 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
8307 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8309 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
8311 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
8313 (armv2, armv2a, armv3, armv3m): Delete architectures.
8314 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
8315 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
8316 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
8318 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
8319 (*mulsidi3adddi): Likewise.
8320 (mulsidi3): Likewise.
8321 (*mulsidi3_nov6): Likewise.
8322 (umulsidi3): Likewise.
8323 (umulsidi3_nov6): Likewise.
8324 (umaddsidi4): Likewise.
8325 (*umulsidi3adddi): Likewise.
8326 (smulsi3_highpart): Likewise.
8327 (*smulsi3_highpart_nov6): Likewise.
8328 (umulsi3_highpart): Likewise.
8329 (*umulsi3_highpart_nov6): Likewise.
8330 * config/arm/arm.h (arm_arch3m): Delete.
8331 * config/arm/arm.c (arm_arch3m): Delete.
8332 (arm_option_override_internal): Update armv3-related comment.
8333 (arm_configure_build_target): Delete use of isa_bit_mode32.
8334 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
8335 (arm_rtx_costs_internal): Delete check of arm_arch3m.
8336 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
8338 (mulusa3): Likewise.
8339 * config/arm/arm-protos.h (arm_arch3m): Delete.
8340 * config/arm/arm-tables.opt: Regenerate.
8341 * config/arm/arm-tune.md: Likewise.
8342 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
8343 deleted architectures.
8345 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8347 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
8348 (armv5t, armv5te): New features.
8349 (ARMv5, ARMv5e): Delete fgroups.
8350 (ARMv5t, ARMv5te): Adjust for above changes.
8352 (armv5, armv5e): Delete arches.
8353 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
8355 (*call_reg_arm): Likewise.
8356 (*call_value_reg_armv5): Likewise.
8357 (*call_value_reg_arm): Likewise.
8358 (*call_symbol): Likewise.
8359 (*call_value_symbol): Likewise.
8360 (*sibcall_insn): Likewise.
8361 (*sibcall_value_insn): Likewise.
8363 (prefetch): Likewise.
8364 (define_split and define_peephole2 dependent on arm_arch5):
8366 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
8368 (TARGET_ARM_QBIT): Likewise.
8369 (TARGET_DSP_MULTIPLY): Likewise.
8370 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
8371 (arm_arch5, arm_arch5e): Delete.
8372 (arm_arch5t, arm_arch5te): Declare.
8373 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
8374 (arm_arch5t): Declare.
8375 (arm_option_reconfigure_globals): Update for the above.
8376 (arm_options_perform_arch_sanity_checks): Update comment, replace
8377 use of arm_arch5 with arm_arch5t.
8378 (use_return_insn): Likewise.
8379 (arm_emit_call_insn): Likewise.
8380 (output_return_instruction): Likewise.
8381 (arm_final_prescan_insn): Likewise.
8382 (arm_coproc_builtin_available): Likewise.
8383 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
8384 arm_arch5e with arm_arch5t and arm_arch5te.
8385 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
8386 (arm_arch5t, arm_arch5te): Declare.
8387 * config/arm/arm-tables.opt: Regenerate.
8388 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
8389 * config/arm/t-multilib: Likewise.
8390 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
8391 instead of arm_arch5.
8392 (*call_reg_thumb1): Likewise.
8393 (*call_value_reg_thumb1_v5): Likewise.
8394 (*call_value_reg_thumb1): Likewise.
8395 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
8397 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
8399 2018-05-18 Martin Liska <mliska@suse.cz>
8401 PR gcov-profile/84846
8402 * doc/gcov.texi: Document -t option of gcov tool.
8404 2018-05-18 Martin Liska <mliska@suse.cz>
8406 PR gcov-profile/84846
8407 * gcov.c (print_usage): Add new -t option.
8408 (process_args): Handle the option.
8409 (generate_results): Use stdout as output when requested by
8412 2018-05-18 Martin Liska <mliska@suse.cz>
8414 PR gcov-profile/84846
8415 * coverage.c (coverage_init): Write PWD to .gcno file.
8416 * doc/gcov.texi: Document how working directory is printed.
8417 * gcov-dump.c (dump_gcov_file): Print PWD.
8418 * gcov.c (output_intermediate_file): Likewise.
8419 (read_graph_file): Read PWD string.
8420 (output_lines): Print PWD.
8422 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8425 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
8426 for retval and return false if all args to phi are zero.
8428 2018-05-18 Richard Biener <rguenther@suse.de>
8430 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
8432 (evrp_dom_walker::before_dom_children): Call it.
8434 2018-05-18 Richard Biener <rguenther@suse.de>
8436 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
8437 results when processing array refs with variable index.
8439 2018-05-18 Toon Moene <toon@moene.org>
8441 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
8442 directly after that of -floop-interchange. Indicate that both
8443 options are enabled by default when specifying -O3.
8445 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8447 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
8448 iterator. Delete separate integer-mode vec_set<mode> expander.
8449 (aarch64_simd_vec_setv2di): Delete.
8450 (vec_setv2di): Delete.
8451 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
8452 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
8453 the "w, r" alternative.
8455 2018-05-18 Martin Liska <mliska@suse.cz>
8457 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
8458 * tree-pass.h (make_pass_lower_switch_O0): New function.
8459 * tree-switch-conversion.c (node_has_low_bound): Remove.
8460 (node_has_high_bound): Likewise.
8461 (node_is_bounded): Likewise.
8462 (class pass_lower_switch): Make it a template type and create
8464 (pass_lower_switch::execute): Add template argument.
8465 (make_pass_lower_switch): New function.
8466 (make_pass_lower_switch_O0): New function.
8467 (do_jump_if_equal): Remove.
8468 (emit_case_nodes): Simplify to just handle all 3 cases and leave
8469 all the hard work to tree optimization passes.
8471 2018-05-18 Martin Liska <mliska@suse.cz>
8473 * dbgcnt.c (limit_low): Renamed from limit.
8474 (limit_high): New variable.
8475 (dbg_cnt_is_enabled): Check for upper limit.
8476 (dbg_cnt): Adjust dumping.
8477 (dbg_cnt_set_limit_by_index): Add new argument for high
8479 (dbg_cnt_set_limit_by_name): Likewise.
8480 (dbg_cnt_process_single_pair): Parse new format.
8481 (dbg_cnt_process_opt): Use strtok.
8482 (dbg_cnt_list_all_counters): Remove 'value' and add
8484 * doc/invoke.texi: Document changes.
8486 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
8488 * doc/sourcebuild.texi (scalar_all_fma): Document.
8489 * tree.def (FMA_EXPR): Delete.
8490 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
8491 * internal-fn.c (ternary_direct): New macro.
8492 (expand_ternary_optab_fn): Likewise.
8493 (direct_ternary_optab_supported_p): Likewise.
8494 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
8495 * builtins.c (fold_builtin_fma): Delete.
8496 (fold_builtin_3): Don't call it.
8497 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
8498 * expr.c (expand_expr_real_2): Likewise.
8499 * fold-const.c (operand_equal_p): Likewise.
8500 (fold_ternary_loc): Likewise.
8501 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
8502 * gimple.c (DEFTREECODE): Likewise.
8503 * gimplify.c (gimplify_expr): Likewise.
8504 * optabs-tree.c (optab_for_tree_code): Likewise.
8505 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
8506 * tree-eh.c (operation_could_trap_p): Likewise.
8507 (stmt_could_throw_1_p): Likewise.
8508 * tree-inline.c (estimate_operator_cost): Likewise.
8509 * tree-pretty-print.c (dump_generic_node): Likewise.
8510 (op_code_prio): Likewise.
8511 * tree-ssa-loop-im.c (stmt_cost): Likewise.
8512 * tree-ssa-operands.c (get_expr_operands): Likewise.
8513 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
8514 * fold-const-call.h (fold_fma): Delete.
8515 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
8516 CFN_FNMA and CFN_FNMS.
8518 * genmatch.c (combined_fn): New enum.
8519 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
8520 (commutative_op): New function.
8521 (commutate): Use it. Handle more than 2 operands.
8522 (dt_operand::gen_gimple_expr): Use commutative_op.
8523 (parser::parse_expr): Allow :c to be used with non-binary
8524 operators if the commutative operand is known.
8525 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
8526 CFN_FMS, CFN_FNMA and CFN_FNMS.
8527 (backprop::process_assign_use): Remove FMA_EXPR handling.
8528 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
8529 (gen_hsa_fma): New function.
8530 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
8531 IFN_FNMA and IFN_FNMS.
8532 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
8533 * gimple-fold.h (follow_all_ssa_edges): Declare.
8534 * gimple-fold.c (follow_all_ssa_edges): New function.
8535 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
8536 gimple_build interface and use follow_all_ssa_edges to fold the result.
8537 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
8538 instead of checking for optabs directly.
8539 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
8540 rather than FMA_EXPRs.
8541 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
8542 call to IFN_FMA instead of an FMA_EXPR.
8544 2018-05-17 Jim Wilson <jimw@sifive.com>
8546 * expr.c (do_tablejump): When converting index to Pmode, if we have a
8547 sign extended promoted subreg, and the range does not have the sign bit
8548 set, then do a sign extend.
8550 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
8551 test, check for sign extended subreg and/or constant operands, and
8552 do a sign extend in that case.
8554 2018-05-17 Steve Ellcey <sellcey@cavium.com>
8556 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
8557 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
8559 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
8560 Change logics_shift_reg to logics_shift_imm.
8561 (thunderx2t99_fp_loadpair_basic): Delete.
8562 (thunderx2t99_fp_storepair_basic): Delete.
8563 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
8564 (thunderx2t99_asimd_polynomial): Delete.
8565 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
8566 and neon_fp_mul_d_scalar_q.
8567 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
8568 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
8569 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
8570 (thunderx2t99_asimd_lut): Add missing tbl types.
8571 (thunderx2t99_asimd_ext): Delete.
8572 (thunderx2t99_asimd_load1_1_mult): Delete.
8573 (thunderx2t99_asimd_load1_2_mult): Delete.
8574 (thunderx2t99_asimd_load1_ldp): New.
8575 (thunderx2t99_asimd_load1): New.
8576 (thunderx2t99_asimd_load2): Add missing *load2* types.
8577 (thunderx2t99_asimd_load3): New.
8578 (thunderx2t99_asimd_load4): New.
8579 (thunderx2t99_asimd_store1_1_mult): Delete.
8580 (thunderx2t99_asimd_store1_2_mult): Delete.
8581 (thunderx2t99_asimd_store2_mult): Delete.
8582 (thunderx2t99_asimd_store2_onelane): Delete.
8583 (thunderx2t99_asimd_store_stp): New.
8584 (thunderx2t99_asimd_store1): New.
8585 (thunderx2t99_asimd_store2): New.
8586 (thunderx2t99_asimd_store3): New.
8587 (thunderx2t99_asimd_store4): New.
8589 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
8591 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
8592 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
8594 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
8595 Segher Boessenkool <segher@kernel.crashing.org>
8598 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
8601 2018-05-17 Richard Biener <rguenther@suse.de>
8603 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
8604 for pruning loop and prune defs feeding only already visited PHIs.
8606 2018-05-17 Richard Biener <rguenther@suse.de>
8608 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
8610 2018-05-17 Bin Cheng <bin.cheng@arm.com>
8611 Richard Biener <rguenther@suse.de>
8613 PR tree-optimization/85793
8614 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
8615 for VMAT_ELEMENTWISE.
8617 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
8619 * internal-fn.h (lookup_internal_fn): Declare
8620 * internal-fn.c (lookup_internal_fn): New function.
8621 * gimple.c (gimple_build_call_from_tree): Handle calls to
8623 * gimple-pretty-print.c (dump_gimple_call): Print "." before
8624 internal function names.
8625 * tree-pretty-print.c (dump_generic_node): Likewise.
8626 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
8628 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
8630 * gimple-fold.h (gimple_build): Make the function forms take
8631 combined_fn rather than built_in_function.
8632 (gimple_simplify): Likewise.
8633 * gimple-match-head.c (gimple_simplify): Likewise.
8634 * gimple-fold.c (gimple_build): Likewise.
8635 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
8636 rather than gimple_build_call_internal.
8637 (get_initial_defs_for_reduction): Likewise.
8638 (vect_create_epilog_for_reduction): Likewise.
8639 (vectorizable_live_operation): Likewise.
8641 2018-05-17 Martin Liska <mliska@suse.cz>
8643 * gimple-ssa-sprintf.c (format_directive): Do not use
8644 space in between 'G_' and '('.
8646 2018-05-17 Jakub Jelinek <jakub@redhat.com>
8649 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
8650 even if the mask is not all ones.
8653 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
8655 (ix86_gimple_fold_builtin): Likewise.
8658 * config/i386/i386.c: Include tree-vector-builder.h.
8659 (ix86_vector_shift_count): New function.
8660 (ix86_fold_builtin): Fold shift builtins by scalar count.
8661 (ix86_gimple_fold_builtin): Likewise.
8663 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
8664 _mm512_setzero): New intrinsics.
8666 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
8667 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8669 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
8670 code generation for cases where splatting a value is not useful.
8671 * simplify-rtx.c (simplify_ternary_operation): Simplify
8672 vec_merge across a vec_duplicate and a paradoxical subreg forming
8673 a vector mode to a vec_concat.
8675 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
8677 * config.gcc: Support "goldmont-plus".
8678 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
8680 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
8681 PROCESSOR_GOLDMONT_PLUS.
8682 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
8683 (processor_target_table): Add "goldmont-plus".
8684 (PTA_GOLDMONT_PLUS): Define.
8685 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
8686 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
8687 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
8688 (fold_builtin_cpu): Add "goldmont-plus".
8689 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
8690 (ix86_option_override_internal): Add "goldmont-plus".
8691 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
8692 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
8693 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
8694 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
8696 2018-05-17 Richard Biener <rguenther@suse.de>
8698 PR tree-optimization/85757
8699 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
8700 remove defs that only feed that PHI from further processing.
8702 2018-05-16 Jim Wilson <jimw@sifive.com>
8704 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
8706 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
8708 2018-05-16 Mark Wielaard <mark@klomp.org>
8710 * dwarf2out.c (count_index_strings): New function.
8711 (output_indirect_strings): Call count_index_strings and generate
8712 header for dwarf_version >= 5.
8714 2018-05-16 Mark Wielaard <mark@klomp.org>
8716 * dwarf2out.c (dwarf_FORM): New function.
8717 (set_indirect_string): Use dwarf_FORM.
8718 (reset_indirect_string): Likewise.
8719 (size_of_die): Likewise.
8720 (value_format): Likewise.
8721 (output_die): Likewise.
8722 (add_skeleton_AT_string): Likewise.
8723 (output_macinfo_op): Likewise.
8724 (index_string): Likewise.
8725 (output_index_string_offset): Likewise.
8726 (output_index_string): Likewise.
8727 (count_index_strings): Likewise.
8729 2018-05-16 Carl Love <cel@us.ibm.com>
8731 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
8732 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
8734 2018-05-16 Martin Jambor <mjambor@suse.cz>
8736 * ipa-prop.c (ipa_free_all_edge_args): Remove.
8737 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
8739 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
8741 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
8742 (fnma<mode>4): Likewise.
8743 (fms<mode>4): Likewise.
8744 (fnms<mode>4): Likewise.
8745 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
8746 (aarch64_fnma<mode>4): Likewise.
8747 (aarch64_fms<mode>4): Likewise.
8748 (aarch64_fnms<mode>4): Likewise.
8749 (aarch64_fnmadd<mode>4): Likewise.
8751 2018-05-16 Jason Merrill <jason@redhat.com>
8753 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
8755 2018-05-16 Richard Biener <rguenther@suse.de>
8757 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
8758 (dump_stmt_cost): Declare.
8759 (add_stmt_cost): Dump cost we add.
8760 (add_stmt_costs): New function.
8761 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
8763 (vect_analyze_stmt): Adjust prototype.
8764 (vectorizable_condition): Likewise.
8765 (vectorizable_live_operation): Likewise.
8766 (vectorizable_reduction): Likewise.
8767 (vectorizable_induction): Likewise.
8768 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
8769 cost vector to pass to vectorizable_ and record afterwards.
8770 (vect_model_reduction_cost): Take cost vector argument and adjust.
8771 (vect_model_induction_cost): Likewise.
8772 (vectorizable_reduction): Likewise.
8773 (vectorizable_induction): Likewise.
8774 (vectorizable_live_operation): Likewise.
8775 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
8776 SLP_TREE_NUMBER_OF_VEC_STMTS.
8777 (vect_analyze_slp_cost_1): Remove.
8778 (vect_analyze_slp_cost): Likewise.
8779 (vect_slp_analyze_node_operations): Take visited args and
8780 a target cost vector. Avoid processing already visited stmt sets.
8781 (vect_slp_analyze_operations): Use a local cost vector to gather
8782 costs and register those of non-discarded instances.
8783 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
8784 (vect_schedule_slp_instance): Remove copying of
8785 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
8787 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
8788 adding cost. Record cost entry location.
8789 (vect_prologue_cost_for_slp_op): Function to compute cost of
8790 a constant or invariant generated for SLP vect in the prologue,
8791 split out from vect_analyze_slp_cost_1.
8792 (vect_model_simple_cost): Make static. Adjust for SLP costing.
8793 (vect_model_promotion_demotion_cost): Likewise.
8794 (vect_model_store_cost): Likewise, make static.
8795 (vect_model_load_cost): Likewise.
8796 (vectorizable_bswap): Add cost vector arg and adjust.
8797 (vectorizable_call): Likewise.
8798 (vectorizable_simd_clone_call): Likewise.
8799 (vectorizable_conversion): Likewise.
8800 (vectorizable_assignment): Likewise.
8801 (vectorizable_shift): Likewise.
8802 (vectorizable_operation): Likewise.
8803 (vectorizable_store): Likewise.
8804 (vectorizable_load): Likewise.
8805 (vectorizable_condition): Likewise.
8806 (vectorizable_comparison): Likewise.
8807 (can_vectorize_live_stmts): Likewise.
8808 (vect_analyze_stmt): Likewise.
8809 (vect_transform_stmt): Adjust calls to vectorizable_*.
8810 * tree-vectorizer.c: Include gimple-pretty-print.h.
8811 (dump_stmt_cost): New function.
8813 2018-05-16 Richard Biener <rguenther@suse.de>
8815 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
8816 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
8817 * tree-ssa-dse.c: Include tree-ssa-loop.h.
8818 (check_name): New callback.
8819 (dse_classify_store): Track cycles via a visited bitmap of PHI
8820 defs and simplify handling of in-loop and across loop dead stores
8821 and properly fail for loop-variant refs. Handle byte-tracking with
8822 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
8825 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
8827 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
8828 (vect_get_mask_type_for_stmt): Likewise.
8829 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
8831 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
8832 to determine the statement's vector type and the vector type that
8833 should be used for calculating nunits. Deal with cases in which
8834 the type has to be deferred.
8835 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
8836 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
8837 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
8838 (vect_determine_vf_for_stmt): New functions, split out from...
8839 (vect_determine_vectorization_factor): ...here.
8840 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
8841 (vect_get_mask_type_for_stmt): New functions, split out from
8842 vect_determine_vectorization_factor.
8844 2018-05-16 Richard Biener <rguenther@suse.de>
8846 * tree-cfg.c (verify_gimple_assign_ternary): Properly
8847 verify the [VEC_]COND_EXPR embedded comparison.
8849 2018-05-15 Martin Sebor <msebor@redhat.com>
8851 PR tree-optimization/85753
8852 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
8853 RECORD_TYPE in addition to ARRAY_TYPE.
8855 2018-05-15 Martin Sebor <msebor@redhat.com>
8858 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
8860 2018-05-15 Richard Biener <rguenther@suse.de>
8862 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
8863 add by_clobber_p one. Change algorithm to collect all defs
8864 representing uses we need to walk and try reducing them to
8865 a single one before failing.
8866 (dse_dom_walker::dse_optimize_stmt): Adjust.
8868 2018-05-13 Mark Wielaard <mark@klomp.org>
8870 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
8871 (size_of_loc_descr): Likewise.
8872 (output_loc_operands): Likewise.
8873 (output_loc_operands_raw): Likewise.
8874 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
8875 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
8876 (hash_loc_operands): Likewise.
8877 (compare_loc_operands): Likewise.
8879 2018-05-14 Mark Wielaard <mark@klomp.org>
8881 * dwarf2out.c (count_index_addrs): New function.
8882 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
8884 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8886 PR tree-optimization/83648
8887 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
8888 return value as malloc candidate.
8890 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8893 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
8894 param as true in call to suggest_attribute.
8896 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
8898 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
8899 -mreadonly-in-sdata.
8901 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8903 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
8905 (aarch64_crypto_aesd_fused): Likewise.
8907 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
8909 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
8910 (movsi_aarch64): Likewise.
8911 (load_pairsi): Likewise.
8912 (load_pairdi): Likewise.
8913 (store_pairsi): Likewise.
8914 (store_pairdi): Likewise.
8915 (load_pairsf): Likewise.
8916 (load_pairdf): Likewise.
8917 (store_pairsf): Likewise.
8918 (store_pairdf): Likewise.
8919 (zero_extend): Likewise.
8920 (trunc): Swap alternatives.
8921 (fcvt_target): Add '?' to prefer w over r.
8923 2018-05-14 Jakub Jelinek <jakub@redhat.com>
8926 * config/i386/i386.md: Disallow non-commutative arithmetics in
8927 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
8928 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
8929 in the peephole2 before it.
8931 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
8933 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
8934 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
8935 (ix86_handle_option): Handle -mcldemote.
8936 * config.gcc: New header.
8937 * config/i386/cldemoteintrin.h: New file.
8938 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
8939 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
8941 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
8942 OPTION_MASK_ISA_CLDEMOTE.
8943 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
8944 (ix86_valid_target_attribute_inner_p): Ditto.
8945 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
8946 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
8947 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
8948 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
8949 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
8951 * config/i386/i386.opt: Add -mcldemote.
8952 * config/i386/x86intrin.h: New header.
8953 * doc/invoke.texi: Add -mcldemote.
8955 2018-05-14 Richard Biener <rguenther@suse.de>
8957 * doc/match-and-simplify.texi: Adjust :s documentation.
8959 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
8961 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
8962 intended memcpy size.
8963 (REORDER_45): Likewise.
8965 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
8967 * sort.cc: New file.
8968 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
8969 * vec.c (qsort_chk): Use gcc_qsort.
8970 * Makefile.in (OBJS-libcommon): Add sort.o.
8971 (build/sort.o): New target. Use it...
8972 (BUILD_RTL): ... here, and...
8973 (build/gencfn-macros): ... here, and...
8974 (build/genmatch): ... here.
8976 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
8977 Chung-Ju Wu <jasonwucj@gmail.com>
8979 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
8980 * config/nds32/nds32-graywolf.md: New file.
8981 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
8982 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
8984 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
8985 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
8986 * config/nds32/nds32.md (pipeline_model): Add graywolf.
8987 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
8988 * config/nds32/pipelines.md: Include n15 settings.
8990 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
8991 Chung-Ju Wu <jasonwucj@gmail.com>
8993 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
8994 * config/nds32/nds32-n13.md: New file.
8995 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
8996 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
8998 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
8999 * config/nds32/nds32.md (pipeline_model): Add n13.
9000 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
9001 * config/nds32/pipelines.md: Include n13 settings.
9003 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
9004 Chung-Ju Wu <jasonwucj@gmail.com>
9006 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
9007 * config/nds32/nds32-n10.md: New file.
9008 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
9009 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
9011 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
9012 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
9013 * config/nds32/nds32.md (pipeline_model): Add n10.
9014 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
9015 * config/nds32/pipelines.md: Include n10 settings.
9017 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
9018 Kito Cheng <kito.cheng@gmail.com>
9019 Chung-Ju Wu <jasonwucj@gmail.com>
9021 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
9022 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
9023 Add enum values for DSP extension instructions.
9024 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
9026 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
9027 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
9029 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
9030 * config/nds32/nds32-dspext.md: New file for DSP implementation.
9031 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
9032 * config/nds32/nds32-intrinsic.md: Likewise.
9033 * config/nds32/nds32_intrinsic.h: Likewise.
9034 * config/nds32/nds32-md-auxiliary.c: Likewise.
9035 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
9036 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
9037 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
9038 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
9039 * config/nds32/nds32-protos.h: New declarations for DSP extension.
9040 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
9041 TYPE_DMAC in switch statement.
9042 * config/nds32/nds32.c: New checking and implementation for DSP
9043 extension instructions.
9044 * config/nds32/nds32.h: Likewise.
9045 * config/nds32/nds32.md: Likewise.
9046 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
9047 * config/nds32/predicates.md: Implement new predicates for DSP
9050 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
9052 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
9053 Reformat alternatives and attributes so it is easier to identify
9054 which constraints/attributes go with which instruction.
9055 (mov<mode>_hardfloat32, FMOVE64): Likewise.
9056 (mov<mode>_softfloat32, FMOVE64): Likewise.
9057 (mov<mode>_hardfloat64, FMOVE64): Likewise.
9058 (mov<mode>_softfloat64, FMOVE64): Likewise.
9060 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
9062 * doc/extend.texi (PowerPC Built-in Functions): Rename this
9064 (Basic PowerPC Built-in Functions): The new name of the
9065 subsection previously known as "PowerPC Built-in Functions".
9066 (Basic PowerPC Built-in Functions Available on all Configurations):
9068 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
9069 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
9070 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
9071 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
9073 2018-05-11 Martin Jambor <mjambor@suse.cz>
9076 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
9079 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
9082 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
9084 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
9086 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
9087 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
9088 (ix86_handle_option): Handle -mwaitpkg.
9089 * config.gcc: New header.
9090 * config/i386/cpuid.h (bit_WAITPKG): New bit.
9091 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
9092 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
9094 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
9095 OPTION_MASK_ISA_WAITPKG.
9096 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
9097 (ix86_option_override_internal): Add PTA_WAITPKG.
9098 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
9099 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
9100 IX86_BUILTIN_TPAUSE.
9101 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
9102 __builtin_ia32_umwait and __builtin_ia32_tpause.
9103 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
9104 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
9105 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
9106 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
9107 UNSPECV_TPAUSE): New.
9108 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
9109 * config/i386/i386.opt: Add -mwaitpkg.
9110 * config/i386/waitpkgintrin.h: New file.
9111 * config/i386/x86intrin.h: New header.
9112 * doc/invoke.texi: Add -mwaitpkg.
9114 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
9117 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
9119 (cortex-m0): Use armv6s-m isa.
9120 (cortex-m0plus): Likewise.
9121 (cortex-m1): Likewise.
9122 (cortex-m0.small-multiply): Likewise.
9123 (cortex-m0plus.small-multiply): Likewise.
9124 (cortex-m1.small-multiply): Likewise.
9126 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
9127 Jakub Jelinek <jakub@redhat.com>
9129 PR tree-optimization/85692
9130 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
9131 source permute as well.
9133 2018-05-11 Martin Liska <mliska@suse.cz>
9136 * doc/extend.texi: Document LLVM style format for no_sanitize
9139 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
9141 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
9142 mode_supports_vsx_dform_quad to mode_supports_dq_form.
9143 (mode_supports_vsx_dform_quad): Likewise.
9144 (mode_supports_vmx_dform): Move these functions to be next to the
9145 other mode_supports functions.
9146 (mode_supports_dq_form): Likewise.
9147 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
9148 mode_supports_dq_form.
9149 (reg_offset_addressing_ok_p): Likewise.
9150 (offsettable_ok_by_alignment): Likewise.
9151 (rs6000_legitimate_offset_address_p): Likewise.
9152 (legitimate_lo_sum_address_p): Likewise.
9153 (rs6000_legitimize_address): Likewise.
9154 (rs6000_legitimize_reload_address): Likewise.
9155 (rs6000_secondary_reload_inner): Likewise.
9156 (rs6000_preferred_reload_class): Likewise.
9157 (rs6000_output_move_128bit): Likewise.
9159 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
9161 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
9162 Generate SImode target register for null target.
9163 <case IX86_BUILTIN_XGETBV>: Ditto.
9164 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
9165 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
9167 2018-05-10 Carl Love <cel@us.ibm.com>
9169 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
9170 dcbtt and dcbtstt if operands[2] is 0.
9172 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
9175 * config/i386/sse.md (usadv64qi): New expander.
9177 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
9179 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
9180 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
9181 -maltivec=be support.
9182 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
9183 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
9184 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
9185 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
9186 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
9187 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
9188 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
9189 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
9190 altivec_vsumsws): Adjust.
9191 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
9192 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
9193 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
9195 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
9196 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
9197 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
9198 (altivec_lve<VI_char>x): Delete expand.
9199 (*altivec_lve<VI_char>x_internal): Rename to...
9200 (altivec_lve<VI_char>x): ... this.
9201 (altivec_lvxl_<mode>): Delete expand.
9202 (*altivec_lvxl_<mode>_internal): Rename to ...
9203 (altivec_lvxl_<mode>): ... this.
9204 (altivec_stvxl_<mode>): Delete expand.
9205 (*altivec_stvxl_<mode>_internal): Rename to ...
9206 (altivec_stvxl_<mode>): ... this.
9207 (altivec_stve<VI_char>x): Delete expand.
9208 (*altivec_stve<VI_char>x_internal): Rename to ...
9209 (altivec_stve<VI_char>x): ... this.
9210 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
9211 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
9212 reduc_plus_scal_<mode>): Adjust.
9213 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
9215 (rs6000_cpu_cpp_builtins): Adjust.
9216 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
9217 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
9218 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
9219 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
9220 -maltivec=be support.
9221 (rs6000_split_vec_extract_var): Adjust.
9222 (rs6000_split_v4si_init): Adjust.
9223 (swap_selector_for_mode): Delete.
9224 (altivec_expand_lvx_be, altivec_expand_stvx_be,
9225 altivec_expand_stvex_be): Delete.
9226 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
9227 -maltivec=be support.
9228 (rs6000_gimple_fold_builtin): Ditto.
9229 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
9231 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
9232 (TARGET_DIRECT_MOVE_64BIT): Adjust.
9233 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
9234 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
9235 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
9236 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
9237 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
9238 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
9239 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
9240 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
9241 anonymous split): Adjust.
9242 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
9243 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
9245 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
9247 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
9248 when --with-gxx-include-dir is also specified.
9249 * configure: Regenerate.
9251 2018-05-09 Jim Wilson <jimw@sifive.com>
9254 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
9255 * config/riscv/t-withmultilib: New.
9256 * config/riscv/withmultilib.h: New.
9257 * doc/install.texi: Document RISC-V --with-multilib-list support.
9259 2018-05-09 Richard Biener <rguenther@suse.de>
9261 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
9263 (vect_bb_vectorization_profitable_p): Adjust. Compute
9264 actual scalar cost using the cost vector and the add_stmt_cost
9267 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
9269 PR rtl-optimization/85645
9270 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
9271 in the REG_CFA_REGISTER note for LR, don't leave it empty.
9273 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
9275 PR rtl-optimization/85645
9276 * shrink-wrap.c (spread_components): Return a boolean saying if
9277 anything was changed.
9278 (try_shrink_wrapping_separate): Iterate spread_components until
9279 nothing changes anymore.
9281 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
9283 PR rtl-optimization/85645
9284 * regrename.c (build_def_use): Also kill the chains that include the
9285 destination of a REG_CFA_REGISTER note.
9287 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
9289 PR rtl-optimization/85645
9290 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
9291 insn that has a REG_CFA_REGISTER note.
9293 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
9295 * cfgexpand.c (expand_clobber): New function.
9296 (expand_gimple_stmt_1): Use it.
9297 * tree-vect-stmts.c (vect_clobber_variable): New function,
9299 (vectorizable_simd_clone_call): ...here.
9300 (vectorizable_store): Emit a clobber either side of an
9301 IFN_STORE_LANES sequence.
9302 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
9304 2018-05-09 Tom de Vries <tom@codesourcery.com>
9307 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
9308 (define_insn "trap_if_false"): Add exit after trap.
9310 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
9312 PR rtl-optimization/85638
9313 * bb-reorder.c: Include common/common-target.h.
9314 (create_forwarder_block): New function extracted from...
9315 (fix_up_crossing_landing_pad): ...here. Rename into...
9316 (dw2_fix_up_crossing_landing_pad): ...this.
9317 (sjlj_fix_up_crossing_landing_pad): New function.
9318 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
9319 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
9320 from both partitions and exit the loop after one iteration.
9322 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
9325 * doc/extend.texi (PowerPC Built-in Functions): Rename this
9327 (Basic PowerPC Built-in Functions): The new name of the
9328 subsection previously known as "PowerPC Built-in Functions".
9329 (Basic PowerPC Built-in Functions Available on all Configurations):
9331 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
9333 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
9335 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
9337 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
9340 2018-05-08 Jim Wilson <jimw@sifive.com>
9342 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
9343 (LD_EMUL_SUFFIX): New.
9344 (LINK_SPEC): Use it.
9346 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
9348 * doc/extend.texi (PowerPC Built-in Functions): Rename this
9350 (Basic PowerPC Built-in Functions): The new name of the
9351 subsection previously known as "PowerPC Built-in Functions".
9352 (Basic PowerPC Built-in Functions Available on all Configurations):
9354 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
9356 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
9358 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
9360 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
9363 2018-05-08 Jakub Jelinek <jakub@redhat.com>
9366 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
9367 after cmpelim optimization.
9369 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
9371 * config.gcc: Support "goldmont".
9372 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
9373 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
9375 * config/i386/i386.c (m_GOLDMONT): Define.
9376 (processor_target_table): Add "goldmont".
9377 (PTA_GOLDMONT): Define.
9378 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
9379 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
9380 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
9381 (fold_builtin_cpu): Add "goldmont".
9382 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
9383 (ix86_option_override_internal): Add "goldmont".
9384 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
9385 (processor_type): Add PROCESSOR_GOLDMONT.
9386 * config/i386/i386.md: Add CPU "glm".
9387 * config/i386/glm.md: New file.
9388 * config/i386/x86-tune.def: Add m_GOLDMONT.
9389 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
9391 2018-05-08 Jakub Jelinek <jakub@redhat.com>
9394 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
9396 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
9397 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
9398 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
9401 * config/i386/i386.c (ix86_fold_builtin): Handle
9402 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
9405 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
9406 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
9408 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
9411 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
9412 (check_arch): Likewise.
9413 (check_fpu): Return the result rather than printing it.
9414 (end arch): Fix operator precedence.
9415 (end cpu): Likewise.
9416 (END): Print the result from check_fpu.
9418 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
9419 Alan Hayward <alan.hayward@arm.com>
9420 David Sherwood <david.sherwood@arm.com>
9422 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
9423 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
9424 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
9425 (*fcmuo<mode>_and): New patterns.
9427 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
9429 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
9430 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
9431 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
9432 (cmp_op, sve_imm_con): New code attributes.
9433 (SVE_COND_INT_CMP, imm_con): Delete.
9434 (cmp_op): Remove above unspecs from int attribute.
9435 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
9437 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
9438 comparison-specific unspecs.
9439 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
9440 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
9441 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
9442 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
9443 (*vec_fcm<cmp_op><mode>): Rename to...
9444 (*fcm<cmp_op><mode>): ...this and adjust likewise.
9445 (*vec_fcmuo<mode>): Rename to...
9446 (*fcmuo<mode>): ...this and adjust likewise.
9447 (*pred_fcm<cmp_op><mode>): New pattern.
9448 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
9449 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
9451 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
9453 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
9454 (aarch64_emit_sve_predicated_cond): New function.
9455 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
9456 (aarch64_emit_unspec_cond_or): Replace with...
9457 (aarch64_emit_sve_or_conds): ...this new function. Use
9458 aarch64_emit_sve_ptrue_op for the individual comparisons and
9459 aarch64_emit_binop to OR them together.
9460 (aarch64_emit_inverted_unspec_cond): Replace with...
9461 (aarch64_emit_sve_inverted_cond): ...this new function. Use
9462 aarch64_emit_sve_ptrue_op for the comparison and
9463 aarch64_emit_unop to invert the result.
9464 (aarch64_expand_sve_vec_cmp_float): Update after the above
9465 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
9467 2018-05-07 Nathan Sidwell <nathan@acm.org>
9469 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
9470 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
9471 (Backwards Compatibility): Likewise.
9473 2018-05-07 Luis Machado <luis.machado@linaro.org>
9477 2018-05-07 Luis Machado <luis.machado@linaro.org>
9479 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
9480 <prefetch_dynamic_strides>: New const bool field.
9481 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
9482 prefetch_dynamic_strides.
9483 (exynosm1_prefetch_tune): Likewise.
9484 (thunderxt88_prefetch_tune): Likewise.
9485 (thunderx_prefetch_tune): Likewise.
9486 (thunderx2t99_prefetch_tune): Likewise.
9487 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
9489 (aarch64_override_options_internal): Update to set
9490 PARAM_PREFETCH_DYNAMIC_STRIDES.
9491 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
9492 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
9493 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
9494 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
9495 prefetch-dynamic-strides setting.
9497 2018-05-07 Luis Machado <luis.machado@linaro.org>
9499 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
9500 <minimum_stride>: New const int field.
9501 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
9502 minimum_stride field.
9503 (exynosm1_prefetch_tune): Likewise.
9504 (thunderxt88_prefetch_tune): Likewise.
9505 (thunderx_prefetch_tune): Likewise.
9506 (thunderx2t99_prefetch_tune): Likewise.
9507 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
9508 (aarch64_override_options_internal): Update to set
9509 PARAM_PREFETCH_MINIMUM_STRIDE.
9510 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
9511 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
9512 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
9513 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
9514 stride is constant and is below the minimum stride threshold.
9516 2018-05-07 Luis Machado <luis.machado@linaro.org>
9518 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
9521 2018-05-07 Luis Machado <luis.machado@linaro.org>
9523 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
9524 <prefetch_dynamic_strides>: New const bool field.
9525 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
9526 prefetch_dynamic_strides.
9527 (exynosm1_prefetch_tune): Likewise.
9528 (thunderxt88_prefetch_tune): Likewise.
9529 (thunderx_prefetch_tune): Likewise.
9530 (thunderx2t99_prefetch_tune): Likewise.
9531 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
9533 (aarch64_override_options_internal): Update to set
9534 PARAM_PREFETCH_DYNAMIC_STRIDES.
9535 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
9536 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
9537 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
9538 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
9539 prefetch-dynamic-strides setting.
9541 2018-05-07 Luis Machado <luis.machado@linaro.org>
9543 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
9544 <minimum_stride>: New const int field.
9545 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
9546 minimum_stride field.
9547 (exynosm1_prefetch_tune): Likewise.
9548 (thunderxt88_prefetch_tune): Likewise.
9549 (thunderx_prefetch_tune): Likewise.
9550 (thunderx2t99_prefetch_tune): Likewise.
9551 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
9552 (aarch64_override_options_internal): Update to set
9553 PARAM_PREFETCH_MINIMUM_STRIDE.
9554 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
9555 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
9556 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
9557 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
9558 stride is constant and is below the minimum stride threshold.
9560 2018-05-06 Jakub Jelinek <jakub@redhat.com>
9563 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
9564 the type is addressable. Don't force op into register if it has
9567 2018-05-05 Roland McGrath <mcgrathr@google.com>
9570 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
9571 any section for which we don't know a specific type it should have,
9572 regardless of name. Previously this was done only for the exact
9573 names ".init_array", ".fini_array", and ".preinit_array".
9574 (default_elf_asm_named_section): Add comment about
9575 relationship with default_section_type_flags and SECTION_NOTYPE.
9576 (get_section): Don't consider it a type conflict if one side has
9577 SECTION_NOTYPE and the other doesn't, as long as neither has the
9578 SECTION_BSS et al used in the default_section_type_flags logic.
9580 2018-05-05 Tom de Vries <tom@codesourcery.com>
9583 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
9584 (workaround_barsyncs): New function.
9585 (nvptx_reorg): Use workaround_barsyncs.
9586 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
9587 (define_expand "nvptx_membar_cta"): New define_expand.
9588 (define_insn "*nvptx_membar_cta"): New insn.
9590 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
9592 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
9593 To improve optimization opportunities.
9594 * builtin-types.def: The new needed builtin types for the above.
9596 2018-05-04 Richard Biener <rguenther@suse.de>
9598 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
9599 * gimple-ssa-store-merging.c
9600 (imm_store_chain_info::output_merged_store): Remove redundant create,
9601 release split_store vector contents on failure.
9602 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
9603 scalar stmt vector on cache hit.
9605 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
9607 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
9609 * config.gcc (powerpc-xilinx-eabi*): Remove.
9610 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
9612 (fusion_addis_mem_combo_load): Ditto.
9613 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
9615 (rs6000_cpu_cpp_builtins): Ditto.
9616 * config/rs6000/rs6000-linux.c
9617 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
9618 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
9619 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
9621 (rs6000_setup_reg_addr_masks): Ditto.
9622 (rs6000_init_hard_regno_mode_ok): Ditto.
9623 (rs6000_option_override_internal): Ditto.
9624 (legitimate_lo_sum_address_p): Ditto.
9625 (rs6000_legitimize_address): Ditto.
9626 (rs6000_legitimize_reload_address): Ditto.
9627 (rs6000_legitimate_address_p): Ditto.
9628 (abi_v4_pass_in_fpr): Ditto.
9629 (setup_incoming_varargs): Ditto.
9630 (rs6000_gimplify_va_arg): Ditto.
9631 (rs6000_split_multireg_move): Ditto.
9632 (rs6000_savres_strategy): Ditto.
9633 (rs6000_emit_prologue_components): Ditto.
9634 (rs6000_emit_epilogue_components): Ditto.
9635 (rs6000_emit_prologue): Ditto.
9636 (rs6000_emit_epilogue): Ditto.
9637 (rs6000_elf_file_end): Ditto.
9638 (rs6000_function_value): Ditto.
9639 (rs6000_libcall_value): Ditto.
9640 * config/rs6000/rs6000.h: Ditto.
9641 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
9642 (TARGET_MINMAX): ... this. New.
9643 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
9644 * config/rs6000/rs6000.md: Remove Xilinx FP support.
9645 (*movsi_internal1_single): Delete.
9646 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
9647 mfpu=, mxilinx-fpu): Delete.
9648 * config/rs6000/singlefp.h: Delete.
9649 * config/rs6000/sysv4.h: Remove Xilinx FP support.
9650 * config/rs6000/t-rs6000: Ditto.
9651 * config/rs6000/t-xilinx: Delete.
9652 * config/rs6000/titan.md: Adjust for fp_type removal.
9653 * config/rs6000/vsx.md: Remove Xilinx FP support.
9654 (VStype_simple): Delete.
9655 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
9656 * config/rs6000/xfpu.h: Delete.
9657 * config/rs6000/xfpu.md: Delete.
9658 * config/rs6000/xilinx.h: Delete.
9659 * config/rs6000/xilinx.opt: Delete.
9660 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
9661 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
9663 2018-05-04 Tom de Vries <tom@codesourcery.com>
9666 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
9669 2018-05-04 Richard Biener <rguenther@suse.de>
9672 * tree-complex.c (update_complex_assignment): We are always in SSA form.
9673 (expand_complex_div_wide): Likewise.
9674 (expand_complex_operations_1): Likewise.
9675 (expand_complex_libcall): Preserve EH info of the original stmt.
9676 (tree_lower_complex): Handle removed blocks.
9677 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
9678 on complex multiplication and division libcall builtins.
9680 2018-05-04 Richard Biener <rguenther@suse.de>
9683 * fold-const.c (negate_expr_p): Restrict negation of operand
9684 zero of a division to when we know that can happen without
9686 (fold_negate_expr_1): Likewise.
9688 2018-05-04 Jakub Jelinek <jakub@redhat.com>
9691 * real.h (real_nextafter): Declare.
9692 * real.c (real_nextafter): New function.
9693 * fold-const-call.c (fold_const_nextafter): New function.
9694 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
9695 CASE_CFN_NEXTTOWARD.
9696 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
9697 even when arg1_mode is different from arg0_mode.
9699 2018-05-03 Nathan Sidwell <nathan@acm.org>
9701 * doc/extend.texi (Deprecated Features): Remove
9703 (Backwards Compatibility): Likewise.
9704 * doc/invoke.texi (C++ Language Options): Likewise.
9705 (C++ Dialect Options): Likewise.
9707 2018-05-03 Jakub Jelinek <jakub@redhat.com>
9710 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
9711 _mm512_mask_mullox_epi64): New intrinsics.
9713 2018-05-03 Tom de Vries <tom@codesourcery.com>
9716 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
9717 dump files): Add offload-tree.
9719 2018-05-03 Richard Biener <rguenther@suse.de>
9721 PR tree-optimization/85615
9722 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
9723 to loops not nested in BBs loop father to avoid creating multi-entry
9726 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9728 PR tree-optimization/70291
9729 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
9730 arguments. Change return type to tree. Emit libcall as a new
9731 statement rather than replacing existing one when inplace_p is true.
9732 (expand_complex_multiplication_components): New function.
9733 (expand_complex_multiplication): Expand floating-point complex
9734 multiplication using the above.
9735 (expand_complex_division): Rename inner_type parameter to type.
9736 Update expand_complex_libcall call-site.
9737 (expand_complex_operations_1): Update expand_complex_multiplication
9738 and expand_complex_division call-sites.
9740 2018-05-02 Jakub Jelinek <jakub@redhat.com>
9743 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
9744 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
9745 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
9746 the highest significant bit of the shift count mask is clear. In
9747 check whether and[sq]i3 is needed verify that all significant bits
9748 of the shift count other than the highest are set.
9750 2018-05-02 Tom de Vries <tom@codesourcery.com>
9753 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
9754 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
9755 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
9756 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
9757 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
9758 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
9759 __builtin_goacc_parlevel_size.
9761 2018-05-02 Richard Biener <rguenther@suse.de>
9763 PR tree-optimization/85597
9764 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
9765 do not use split vect_get_vec_defs call but call vect_get_slp_defs
9768 2018-05-02 Tom de Vries <tom@codesourcery.com>
9771 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
9772 dump files): Add ltrans-tree.
9774 2018-05-02 Tom de Vries <tom@codesourcery.com>
9777 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
9778 dump files): Add wpa-ipa.
9780 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
9782 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
9783 powerpc*-*-linux*paired* target.
9784 * config/rs6000/750cl.h: Delete.
9785 * config/rs6000/paired.h: Delete.
9786 * config/rs6000/paired.md: Delete.
9787 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
9789 * config/rs6000/rs6000-builtin.def: Remove paired float support.
9790 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
9791 comment. Remove paired float support.
9792 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
9793 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
9795 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
9796 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
9798 * config/rs6000/rs6000.c: Remove paired float support.
9799 (paired_expand_vector_init, paired_expand_vector_move,
9800 paired_emit_vector_compare, paired_emit_vector_cond_expr,
9801 (paired_expand_lv_builtin, paired_expand_stv_builtin,
9802 paired_expand_builtin, paired_expand_predicate_builtin,
9803 paired_init_builtins): Delete.
9804 * config/rs6000/rs6000.h: Remove paired float support.
9805 * config/rs6000/rs6000.md: Remove paired float support.
9806 (move_from_CR_ov_bit): Delete.
9807 * config/rs6000/rs6000.opt (mpaired): Delete.
9808 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
9809 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
9811 2018-05-02 Richard Biener <rguenther@suse.de>
9814 * gimplify.c (gimplify_save_expr): When in SSA form allow
9815 SAVE_EXPRs to compute to SSA vars.
9817 2018-05-02 Jakub Jelinek <jakub@redhat.com>
9820 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
9821 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
9822 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
9823 clobber operands[2], instead use a new pseudo. Formatting fixes.
9825 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
9827 PR tree-optimization/85586
9828 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
9829 exit early for statements in the same group if the accesses are
9832 2018-05-02 Tom de Vries <tom@codesourcery.com>
9835 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
9838 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
9840 PR tree-optimization/85143
9841 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
9843 2018-05-01 Tom de Vries <tom@codesourcery.com>
9846 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
9847 not found" error message.
9849 2018-05-01 Tom de Vries <tom@codesourcery.com>
9852 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
9853 * vec.c (test_ordered_remove_if): New function.
9854 (vec_c_tests): Call test_ordered_remove_if.
9855 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
9856 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
9857 * tree-vect-patterns.c (vect_pattern_recog_1): Use
9858 VEC_ORDERED_REMOVE_IF.
9860 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9862 PR tree-optimization/82665
9863 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
9864 pointer subtraction where arguments come from a memchr call.
9866 2018-05-01 Jakub Jelinek <jakub@redhat.com>
9868 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
9869 --push-state --as-needed and --pop-state instead of --as-needed and
9870 --no-as-needed if ld supports it.
9871 * configure: Regenerated.
9874 * doc/install.texi2html: Replace _002d with - and _002a with * in
9875 generated html files using sed.
9877 2018-04-30 David Malcolm <dmalcolm@redhat.com>
9880 * gcc-rich-location.c (blank_line_before_p): New function.
9881 (use_new_line): New function.
9882 (gcc_rich_location::add_fixit_insert_formatted): New function.
9883 * gcc-rich-location.h
9884 (gcc_rich_location::add_fixit_insert_formatted): New function.
9886 2018-04-30 David Malcolm <dmalcolm@redhat.com>
9888 * selftest.c (assert_streq): Rename "expected" and "actual" to
9889 "val1" and "val2". Extend NULL-handling to cover both inputs
9890 symmetrically, while still requiring both to be non-NULL for a pass.
9891 * selftest.h (assert_streq): Rename "expected" and "actual" to
9893 (ASSERT_EQ): Likewise.
9894 (ASSERT_EQ_AT): Likewise.
9895 (ASSERT_KNOWN_EQ): Likewise.
9896 (ASSERT_KNOWN_EQ_AT): Likewise.
9897 (ASSERT_NE): Likewise.
9898 (ASSERT_MAYBE_NE): Likewise.
9899 (ASSERT_MAYBE_NE_AT): Likewise.
9900 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
9901 the assertion to pass.
9902 (ASSERT_STREQ_AT): Likewise.
9904 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
9906 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
9907 interaction with -pie.
9909 2018-04-30 David Malcolm <dmalcolm@redhat.com>
9911 * selftest.h: Fix alphabetization of per-source-file selftest
9914 2018-04-30 Jason Merrill <jason@redhat.com>
9916 PR c++/61982 - dead stores to destroyed objects.
9917 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
9920 2018-04-30 Jason Merrill <jason@redhat.com>
9922 * tree.c (build_clobber): New.
9923 * tree.h: Declare it.
9924 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
9926 2018-04-30 David Malcolm <dmalcolm@redhat.com>
9928 * diagnostic-show-locus.c (layout::layout): Update for
9929 location_get_source_line returning a char_span.
9930 (struct char_span): Move to input.h.
9931 (struct correction): Update for fields in char_span becoming
9933 (struct source_line): Update for location_get_source_line
9934 returning a char_span.
9935 (layout::print_line): Likewise.
9936 * edit-context.c (edited_file::print_content): Likewise.
9937 (edited_file::print_diff_hunk): Likewise.
9938 (edited_file::print_run_of_changed_lines): Likewise.
9939 (edited_file::get_num_lines): Likewise.
9940 (edited_line::edited_line): Likewise.
9941 * final.c (asm_show_source): Likewise.
9942 * input.c (location_get_source_line): Convert return type
9943 from const char * to char_span, losing the final "line_len"
9945 (dump_location_info): Update for the above.
9946 (get_substring_ranges_for_loc): Likewise. Use a char_span
9947 when handling the literal within the line.
9948 (test_reading_source_line): Update for location_get_source_line
9949 returning a char_span.
9950 * input.h (class char_span): Move here from
9951 diagnostic-show-locus.c, converting from a struct to a class.
9952 Make data members private.
9953 (char_span::operator bool): New.
9954 (char_span::length): New.
9955 (char_span::get_buffer): New.
9956 (char_span::operator[]): New.
9957 (char_span::subspan): Make const.
9958 (char_span::xstrdup): New.
9959 (location_get_source_line): Convert return type from const char *
9960 to char_span, losing the final "line_size" param.
9962 2018-04-30 Jan Hubicka <jh@suse.cz>
9964 * lto-wrapper.c (ltrans_priorities): New static var.
9965 (cmp_priority): New.
9966 (run_gcc): Read priorities and if doing parallel build order
9967 the Makefile by them.
9969 2018-04-30 David Malcolm <dmalcolm@redhat.com>
9971 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
9973 2018-04-30 Richard Biener <rguenther@suse.de>
9975 * tree-cfg.c (verify_address): Remove base argument, add
9976 flag whether to check TREE_ADDRESSABLE and do that.
9977 (verify_expr): Remove.
9978 (verify_types_in_gimple_reference): Add pieces from verify_expr.
9979 (verify_gimple_assign_single): Likewise.
9980 (verify_gimple_switch): Likewise.
9981 (verify_expr_location_1): Dereference tp once. Add (disabled)
9982 piece from verify_expr.
9983 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
9985 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
9987 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
9989 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
9991 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
9992 (small_data_pattern): Likewise.
9993 (arc_rewrite_small_data): Likewise.
9994 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
9995 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
9996 (get_symbol_alignment): New function.
9997 (legitimate_small_data_address_p): Likewise.
9998 (legitimate_scaled_address): Update, call
9999 legitimate_small_data_address_p.
10000 (output_sdata): New static variable.
10001 (arc_print_operand): Update how we handle small data operands.
10002 (arc_print_operand_address): Likewise.
10003 (arc_legitimate_address_p): Update, use
10004 legitimate_small_data_address_p.
10005 (arc_rewrite_small_data_p): Remove.
10006 (arc_rewrite_small_data_1): Likewise.
10007 (arc_rewrite_small_data): Likewise.
10008 (small_data_pattern): Likewise.
10009 (compact_sda_memory_operand): Update to use
10010 legitimate_small_data_address_p and get_symbol_alignment.
10011 (prepare_move_operands): Don't rewite sdata pattern.
10012 (prepare_extend_operands): Remove.
10013 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
10015 (zero_extendqisi2): Likewise.
10016 (zero_extendhisi2): Likewise.
10017 (extendqihi2): Likewise.
10018 (extendqisi2): Likewise.
10019 (extendhisi2): Likewise.
10020 (addsi3): Likewise.
10021 (subsi3): Likewise.
10022 (andsi3): Likewise.
10023 * config/arc/constraints.md (Usd): Change it to memory constraint.
10025 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
10027 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
10028 as source of std instructions.
10029 * config/arc/arc.md (movsi_insn): Update pattern predicate to
10030 allow 6-bit constants as source for store instructions.
10031 (movdi_insn): Update instruction pattern to allow 6-bit constants
10032 as source for store instructions.
10034 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
10036 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
10038 2018-04-30 Nathan Sidwell <nathan@acm.org>
10039 Sandra Loosemore <sandra@codesourcery.com>
10041 * dumpfile.c (dump_open): Allow '-' for stdout.
10042 * doc/invoke.texi (Developer Options): Document dump filename
10043 determination early. Document stdin/stdout selection.
10045 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
10047 Microblaze Target: PIC data text relative
10049 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
10050 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
10052 * config/microblaze/microblaze.h (microblaze_constant_address_p):
10053 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
10054 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
10055 New addressing mode for data-text relative position indepenedent code.
10056 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
10057 'ADDRESS_SYMBOLIC_TXT_REL'.
10058 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
10059 (microblaze_legitimate_pic_operand): Exclude function calls from
10060 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
10061 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
10063 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
10064 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
10065 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
10066 for 'address + offset'.
10067 (microblaze_expand_prologue): Add new function prologue call for
10069 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
10070 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
10071 table in case of TARGET_PIC_DATA_TEXT_REL.
10072 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
10073 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
10074 Add new macros 'UNSPEC_TEXT',
10075 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
10076 + exclude function calls from 'UNSPEC_PLT' in case of data text
10078 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
10079 new target hook for generating address diff vector tables in case of
10081 * doc/tm.texi : Regenerate.
10082 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
10083 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
10084 of addr diff vector generation.
10085 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
10086 target hook definition.
10087 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
10088 Add default function for generate_pic_addr_diff_vec -> flag_pic.
10089 * doc/invoke.texi (Add new pic option): Add new microblaze pic
10090 option for data text relative.
10092 2018-04-30 Richard Biener <rguenther@suse.de>
10094 * tree-chrec.h (evolution_function_is_constant_p): Remove
10096 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
10098 2018-04-30 Richard Biener <rguenther@suse.de>
10101 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
10103 2018-04-30 Richard Biener <rguenther@suse.de>
10105 PR tree-optimization/28364
10106 PR tree-optimization/85275
10107 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
10108 copying first exit test.
10110 2018-04-28 Mark Wielaard <mark@klomp.org>
10112 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
10113 dwarf_version >= 5.
10114 (dwarf_AT): Handle DW_AT_addr_base.
10115 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
10117 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
10120 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
10121 (*ashl<dwi>3_doubleword_mask_1): Ditto.
10122 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
10123 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
10125 2018-04-28 Richard Biener <rguenther@suse.de>
10127 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
10128 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
10129 to reflect use. Only add interesting stmts.
10131 2018-04-27 Martin Jambor <mjambor@suse.cz>
10134 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
10135 the jump function allows for passing through aggregate values.
10137 2018-04-27 David Malcolm <dmalcolm@redhat.com>
10139 * input.h (in_system_header_at): Convert from macro to inline
10141 (from_macro_expansion_at): Likewise.
10142 (from_macro_definition_at): Likewise.
10144 2018-04-27 Jeff Law <law@redhat.com>
10146 * config.gcc: Mark tile* targets as deprecated/obsolete.
10148 2018-04-27 Richard Biener <rguenther@suse.de>
10150 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
10153 2018-04-27 Richard Biener <rguenther@suse.de>
10155 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
10157 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
10159 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
10160 with Yd constraint. Set "preferred_for_speed" attribute from
10161 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
10162 with Yd constraint.
10163 (*movdi_internal): Ditto.
10164 (movti_interunit splitters): Remove
10165 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
10166 (movdi_interunit splitters): Ditto.
10167 * config/i386/constraints.md (Ye): Remove.
10168 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
10170 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10173 * config/aarch64/constraints.md (Usg): Limit to 31.
10174 (Usj): Limit to 63.
10176 2018-04-27 Jakub Jelinek <jakub@redhat.com>
10178 PR tree-optimization/85529
10179 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
10180 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
10181 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
10182 zero extension or masking of the MSB bit.
10183 (optimize_range_tests): Add FIRST_BB argument, pass it through
10184 to optimize_range_tests_var_bound.
10185 (maybe_optimize_range_tests, reassociate_bb): Adjust
10186 optimize_range_tests callers.
10188 2018-04-26 Richard Biener <rguenther@suse.de>
10189 Jakub Jelinek <jakub@redhat.com>
10191 * cgraph.h (symbol_table): Just declare debug method here.
10192 * symtab.c (symbol_table::debug): Define.
10194 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
10196 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
10198 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
10200 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
10201 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
10202 (*movdi_internal): Substitute Yi and Yj constraint with x
10203 and Ym and Yn constraint with y constraint. Update "isa"
10204 attribute and set "preferred_for_speed" attribute from
10205 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
10206 (*movsi_internal): Ditto.
10207 (*movdf_internal): Ditto.
10208 (*movsf_internal): Ditto.
10209 (*zero_extendsidi2): Ditto.
10210 * config/i386/sse.md (vec_set<mode>_0): Ditto.
10211 (sse2_loadld): Ditto.
10212 (*vec_extract<ssevecmodelower>_0): Ditto.
10213 (*vec_extractv4si_0_zext_sse4): Ditto.
10214 (vec_concatv2di): Ditto.
10215 (*vec_dup<mode>): Ditto.
10216 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
10217 * config/i386/constraints.md (Yi): Remove.
10222 2018-04-26 Nathan Sidwell <nathan@acm.org>
10224 * dumpfile.c (dump_open): New.
10225 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
10226 (dump_finish): Detect stdio/stderr by value not name.
10228 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
10230 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
10232 2018-04-26 Tom de Vries <tom@codesourcery.com>
10235 * config/nvptx/nvptx.c (verify_neutering_jumps)
10236 (verify_neutering_labels): New function
10237 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
10239 2018-04-26 Tom de Vries <tom@codesourcery.com>
10242 * config/nvptx/nvptx.c (needs_neutering_p): New function.
10243 (nvptx_single): Use needs_neutering_p to skip over insns that do not
10246 2018-04-26 Richard Biener <rguenther@suse.de>
10247 Tom de Vries <tom@codesourcery.com>
10250 * lto-streamer-out.c (output_function): Fixup loops if required to match
10251 discovery done in the reader.
10253 2018-04-26 Richard Biener <rguenther@suse.de>
10255 PR tree-optimization/85116
10256 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
10257 have a loop exit from the single latch predecessor. Remove
10258 case of header with just condition.
10259 (ch_base::copy_headers): Exclude infinite loops from any
10261 (pass_ch::execute): Record exits.
10263 2018-04-26 Richard Biener <rguenther@suse.de>
10265 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
10266 prologue cost vector and pass it to vect_get_load_cost.
10267 (vect_get_peeling_costs_all_drs): Likewise.
10268 (vect_peeling_hash_get_lowest_cost): Likewise.
10269 (vect_enhance_data_refs_alignment): Likewise.
10271 2018-04-26 Richard Biener <rguenther@suse.de>
10273 PR middle-end/85450
10274 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
10275 checking of integer<->pointer conversions.
10276 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
10277 sign-/zero-extending pointer types.
10278 (expand_omp_for_static_chunk): Likewise.
10280 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
10281 Jean Lee <xiaoyur347@gmail.com>
10283 * config/mips/mips.c (mips_asan_shadow_offset): New function.
10284 (TARGET_ASAN_SHADOW_OFFSET): Define.
10285 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
10286 true for -fsanitize=address.
10288 2018-04-25 Mark Wielaard <mark@klomp.org>
10290 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
10293 2018-04-25 Jakub Jelinek <jakub@redhat.com>
10295 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
10296 than "alu", remove explicit "memory" and "imm_disp" attributes.
10297 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
10299 PR middle-end/85414
10300 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
10301 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
10302 gen_lowpart_no_emit.
10304 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
10307 * config/i386/i386.c (ix86_expand_builtin): Change memory
10308 operand to XI, extend p0 to Pmode.
10309 * config/i386/i386.md: Change unspec volatile and operand
10310 1 mode to XI, change operand 0 mode to P.
10312 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
10314 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
10315 GET_MODE_MASK before any checking.
10316 (nds32_can_use_bset_p): Likewise.
10317 (nds32_can_use_btgl_p): Likewise.
10319 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
10321 * config/nds32/nds32-doubleword.md: New define_split pattern for
10322 illegal register number.
10324 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
10326 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
10328 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
10330 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
10332 2018-04-25 Richard Biener <rguenther@suse.de>
10334 * lto-streamer.h (LTO_major_version): Bump to 8.
10336 2018-04-25 Jakub Jelinek <jakub@redhat.com>
10338 * BASE-VER: Set to 9.0.0.
10340 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
10342 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
10343 in __abskf2 and __powikf2.
10345 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10348 * config/aarch64/constraints.md (Usg, Usj): New constraints.
10349 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
10350 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
10351 Use the above on operand 2. Reindent.
10352 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
10354 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
10357 * common/config/i386/i386-common.c (ix86_handle_option): Don't
10359 * config/i386/i386.opt (mcet): Removed.
10360 * doc/install.texi: Remove -mcet documentation.
10361 * doc/invoke.texi: Likewise.
10363 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
10366 * doc/install.texi: Remove -mcet from bootstrap-cet.
10368 2018-04-24 Jakub Jelinek <jakub@redhat.com>
10371 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
10372 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
10376 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
10377 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
10378 containing a CONST_VECTOR.
10380 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
10382 * doc/install.texi: Update newlib dependency for nvptx.
10384 2018-04-24 Jakub Jelinek <jakub@redhat.com>
10387 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
10388 instead of INTVAL when shifting x left.
10390 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
10392 PR tree-optimization/85478
10393 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
10394 vect_grouped_store_supported for single element vectors.
10396 2018-04-24 Richard Biener <rguenther@suse.de>
10399 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
10400 load cost increase to the case of non-constant step.
10402 2018-04-24 Jakub Jelinek <jakub@redhat.com>
10405 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
10406 destination if any_malformed_asm.
10408 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
10410 PR middle-end/85496
10411 * expr.c (store_field): In the bitfield case, if the value comes from
10412 a function call and is returned in registers by means of a PARALLEL,
10413 do not change the mode of the temporary unless BLKmode and VOIDmode.
10415 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
10417 PR rtl-optimization/85423
10418 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
10419 dependencies to debug insns when the previous insn is non-debug.
10421 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
10423 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
10424 enums into a single definition.
10425 (fls): Fix predicates and printing.
10428 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
10430 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
10431 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
10432 and short u6 immediate.
10433 (check_if_valid_sleep_operand): Remove.
10434 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
10436 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
10438 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
10439 flag_always_save_lp condition.
10440 * config/nds32/nds32.opt (malways-save-lp): New option.
10442 2018-04-22 Shiva Chen <shiva0217@gmail.com>
10444 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
10445 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
10446 * config/nds32/nds32.h
10447 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
10448 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
10450 2018-04-22 Shiva Chen <shiva0217@gmail.com>
10452 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
10453 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
10455 2018-04-22 Shiva Chen <shiva0217@gmail.com>
10456 Chung-Ju Wu <jasonwucj@gmail.com>
10458 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
10460 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
10461 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
10463 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
10465 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
10467 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
10469 * config/nds32/nds32-protos.h (nds32_data_alignment,
10470 nds32_local_alignment): Declare.
10471 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
10472 nds32_local_alignment): New functions.
10473 (TARGET_CONSTANT_ALIGNMENT): Define.
10474 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
10476 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
10478 * config/nds32/nds32.c
10479 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
10480 (TARGET_MODES_TIEABLE_P): Likewise.
10482 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
10484 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
10485 level Ofast and Og.
10487 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
10488 Chung-Ju Wu <jasonwucj@gmail.com>
10490 * config/nds32/constants.md (unspec_volatile_element): Add enum values
10491 for unaligned access.
10492 * config/nds32/nds32-intrinsic.c: Implementation of expanding
10494 * config/nds32/nds32-intrinsic.md: Likewise.
10495 * config/nds32/nds32_intrinsic.h: Likewise.
10496 * config/nds32/nds32.h (nds32_builtins): Likewise.
10497 * config/nds32/nds32.opt (munaligned-access): New option.
10498 * config/nds32/nds32.c (nds32_asm_file_start): Display
10499 flag_unaligned_access status.
10501 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
10503 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
10504 -mno-relax is present.
10505 * config/riscv/linux.h (LINK_SPEC): Ditto.
10507 2018-04-20 Martin Sebor <msebor@redhat.com>
10510 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
10512 (gimple_fold_builtin_stxcpy_chk): Same.
10513 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
10515 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
10518 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
10519 __powikf2 when long double is IEEE 128-bit.
10521 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
10523 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
10524 step to make sure stack always aligned.
10526 2018-04-20 Carl Love <cel@us.ibm.com>
10529 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
10530 size check for arg0.
10532 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
10533 Tom de Vries <tom@codesourcery.com>
10536 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
10537 Emit insns for calls too.
10538 (nvptx_find_par): Always look for worker-level predecessor insn.
10539 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
10541 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
10542 (nvptx_process_pars): Propagate frames for calls.
10544 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
10547 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
10549 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
10550 (ix86_handle_option): Don't handle OPT_mibt.
10551 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
10553 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
10555 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
10556 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
10557 (ix86_target_macros): Define __CET__ with flag_cf_protection
10558 for -fcf-protection.
10559 * config/i386/i386.c (isa2_opts): Remove -mibt.
10560 * config/i386/i386.h (TARGET_IBT): Removed.
10561 (TARGET_IBT_P): Likewise.
10562 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
10563 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
10564 * config/i386/i386.opt (mcet): Update help message.
10565 (mshstk): Likewise.
10567 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
10568 -mcet as an alias for -mshstk.
10570 2018-04-20 Richard Biener <rguenther@suse.de>
10572 PR middle-end/85475
10573 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
10574 complexity by forcing a single use of the multiply operand.
10576 2018-04-20 Martin Jambor <mjambor@suse.cz>
10579 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
10580 recursion dependency to only apply to non-clones.
10582 2018-04-20 Martin Jambor <mjambor@suse.cz>
10585 * ipa-cp.c (create_specialized_node): Check that clones of
10586 self-recursive edges exist during IPA-CP.
10588 2018-04-19 Toon Moene <toon@moene.org>
10590 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
10593 2018-04-19 Jakub Jelinek <jakub@redhat.com>
10595 PR tree-optimization/85467
10596 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
10597 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
10598 VECTOR_CST element to type.
10600 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
10603 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
10604 * config/i386/i386.md (builtin_setjmp_setup): Removed.
10605 (builtin_longjmp): Likewise.
10606 (save_stack_nonlocal): New pattern.
10607 (restore_stack_nonlocal): Likewise.
10609 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
10612 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
10613 Replace ASM_OUTPUT_LABEL with fprintf.
10615 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
10618 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
10619 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
10620 * config/i386/i386-c.c (ix86_target_macros_internal): Also
10621 define __IBT__ and __SHSTK__ for -fcf-protection.
10622 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
10624 (ix86_trampoline_init): Likewise.
10625 (x86_output_mi_thunk): Likewise.
10626 (ix86_notrack_prefixed_insn_p): Likewise.
10627 (ix86_option_override_internal): Don't disallow -fcf-protection.
10628 * config/i386/i386.md (rdssp<mode>): Also enable for
10630 (incssp<mode>): Likewise.
10631 (nop_endbr): Likewise.
10632 * config/i386/i386.opt (mcet): Change help message to built-in
10635 (mshstk): Likewise.
10636 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
10637 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
10638 enable CET built-in functions.
10640 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
10642 * common/config/i386/i386-common.c
10643 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
10644 OPTION_MASK_ISA_MOVDIRI_UNSET,
10645 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
10646 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
10647 * config.gcc (movdirintrin.h): New header.
10648 * config/i386/cpuid.h (bit_MOVDIRI,
10649 bit_MOVDIR64B): New bits.
10650 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
10652 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
10653 (VOID, PVOID, PCVOID)): New function types.
10654 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
10655 __builtin_ia32_directstoreu_u64,
10656 __builtin_ia32_movdir64b): New builtins.
10657 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
10658 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
10660 (ix86_valid_target_attribute_inner_p): Ditto.
10661 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
10662 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
10663 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
10664 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
10665 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
10666 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
10667 (movdiri<mode>, movdir64b_<mode>): New.
10668 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
10669 * config/i386/immintrin.h: Include movdirintrin.h.
10670 * config/i386/movdirintrin.h: New file.
10671 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
10673 2018-04-19 Richard Biener <rguenther@suse.de>
10675 PR middle-end/85455
10676 * cfg.c (clear_bb_flags): When loop state says we have
10677 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
10679 2018-04-19 Richard Biener <rguenther@suse.de>
10681 PR tree-optimization/84737
10682 * tree-vect-data-refs.c (vect_copy_ref_info): New function
10683 copying restrict info.
10684 (vect_setup_realignment): Use it.
10685 * tree-vectorizer.h (vect_copy_ref_info): Declare.
10686 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
10687 the first DR to all generated stores.
10688 (vectorizable_load): Likewise for loads.
10690 2018-04-19 Jakub Jelinek <jakub@redhat.com>
10692 PR tree-optimization/85446
10693 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
10694 the integral and pointer types to have the same precision.
10696 * doc/install.texi: Document --disable-cet being the default and
10699 2018-04-18 Martin Liska <mliska@suse.cz>
10701 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
10704 2018-04-18 Martin Liska <mliska@suse.cz>
10707 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
10710 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
10711 arguments if they are comparable.
10713 2018-04-18 Martin Liska <mliska@suse.cz>
10716 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
10719 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
10722 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
10725 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
10726 ENDBR after calling __morestack.
10728 2018-04-18 David Malcolm <dmalcolm@redhat.com>
10731 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
10732 by using gcc_base_ver to generate a gcc_driver_version, and use
10733 it when generating GCC_DRIVER_NAME.
10734 * configure: Regenerate.
10736 2018-04-18 Jakub Jelinek <jakub@redhat.com>
10739 * config.gcc: Obsolete powerpc*-*-*spe*.
10741 2018-04-17 Jakub Jelinek <jakub@redhat.com>
10744 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
10745 (stabstr_D): Change type of unum from unsigned int to
10746 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
10749 2018-04-17 Jim Wilson <jimw@sifive.com>
10752 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
10753 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
10754 Set arg_pointer_offset after using pretend_args_size.
10756 2018-04-17 Jakub Jelinek <jakub@redhat.com>
10758 PR rtl-optimization/85431
10759 * dse.c (record_store): Ignore zero width stores.
10762 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
10763 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
10764 __builtin_stack_restore rather than after it.
10765 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
10766 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
10767 argument instead of virtual_dynamic_stack_rtx.
10769 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
10771 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
10773 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
10774 Add note to error message to explain internal mapping of overloaded
10775 built-in function name to non-overloaded built-in function name.
10776 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
10779 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
10782 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
10783 where the inputs overlap with the output.
10785 2018-04-17 Jakub Jelinek <jakub@redhat.com>
10787 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
10788 (=v, v) alternative and explicit "memory" attribute.
10789 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
10790 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
10792 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
10793 "sselog1" type instead of "sselog".
10794 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
10795 "sselog". Remove explicit "memory" attribute.
10796 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
10797 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
10799 (vec_extract_hi_v32hi): Merge all alternatives into one, use
10800 "sselog1" type instead of "sselog". Remove explicit "memory"
10802 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
10803 use "sselog1" type instead of "sselog". Remove explicit "memory"
10805 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
10806 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
10808 (vec_extract_hi_v64qi): Merge all alternatives into one, use
10809 "sselog1" type instead of "sselog". Remove explicit "memory"
10811 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
10812 use "sselog1" type instead of "sselog". Remove explicit "memory"
10816 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
10818 PR middle-end/85414
10819 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
10822 2018-04-17 Martin Jambor <mjambor@suse.cz>
10825 * ipa-cp.c (create_specialized_node): Call
10826 expand_all_artificial_thunks if necessary.
10828 2018-04-17 Martin Liska <mliska@suse.cz>
10831 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
10832 in message, remote space in between '_G' and '('.
10834 2018-04-17 Jakub Jelinek <jakub@redhat.com>
10837 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
10838 avx512f_vmcmp<mode>3<round_saeonly_name>,
10839 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
10840 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
10841 avx512f_rndscale<mode><round_saeonly_name>,
10842 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
10843 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
10844 Use %<iptr>2 instead of %2 for -masm=intel.
10845 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
10846 avx512f_vcvttss2usi<round_saeonly_name>,
10847 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
10849 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
10850 avx512f_vcvttsd2usi<round_saeonly_name>,
10851 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
10852 Use %q1 instead of %1 for -masm=intel.
10853 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
10854 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
10855 of %3 for -masm=intel.
10856 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
10858 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
10860 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
10862 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
10863 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
10865 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
10867 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
10868 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
10869 %g1 and one with %0 and %1.
10870 (avx512er_vmrcp28<mode><round_saeonly_name>,
10871 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
10872 %1 for -masm=intel.
10873 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
10874 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
10875 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
10876 of %0 and %{%4%} for -masm=intel.
10877 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
10878 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
10879 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
10880 order of %0 and %{%5%}%{z%} for -masm=intel.
10882 2018-04-17 Jan Hubicka <jh@suse.cz>
10885 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
10887 2018-04-17 Martin Liska <mliska@suse.cz>
10890 * multiple_target.c (create_dispatcher_calls): Set apostrophes
10891 for target_clone error message. Make default implementation
10892 clone to be a local declaration.
10893 (separate_attrs): Add new argument and check for an empty
10895 (expand_target_clones): Handle it.
10896 (ipa_target_clone): Make redirection just for target_clones
10899 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
10900 Tom de Vries <tom@codesourcery.com>
10902 PR middle-end/84955
10903 * omp-expand.c (expand_oacc_for): Add dummy false branch for
10904 tiled basic blocks without omp continue statements.
10906 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
10909 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
10910 vec_extract expression as having side effects to make sure it gets
10913 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
10916 * config/i386/i386.c (get_builtin_code_for_version): Check
10919 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
10922 * config.gcc: Support "skylake".
10923 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10925 * config/i386/i386.c (m_SKYLAKE): Define.
10926 (processor_target_table): Add "skylake".
10927 (ix86_option_override_internal): Add "skylake".
10928 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
10929 PROCESSOR_CANNONLAKE.
10930 (get_builtin_code_for_version): Fix priority for
10931 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
10932 PROCESSOR_SKYLAKE-AVX512.
10933 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
10934 (processor_type): Add PROCESSOR_SKYLAKE.
10936 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
10937 Jason Merrill <jason@redhat.com>
10940 * convert.c (convert_to_integer_1): Use direct recursion for
10941 enumeral types and types with a precision less than the number
10942 of bits in their mode.
10944 2018-04-16 Julia Koval <julia.koval@intel.com>
10947 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
10948 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
10950 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
10953 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
10954 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
10955 and -mno-direct-move.
10957 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
10960 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
10961 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
10962 Ensure negative shifts result in {0}.
10964 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
10966 PR rtl-optimization/79916
10967 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
10968 regs (if any) to define how to gnerate SD moves when LRA is in
10971 2018-04-13 Jakub Jelinek <jakub@redhat.com>
10973 PR rtl-optimization/85393
10974 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
10975 * except.c (expand_dw2_landing_pad_for_region): Make static.
10976 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
10977 a label and unconditional jump to old_bb, rather than
10978 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
10981 PR rtl-optimization/85376
10982 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
10983 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
10984 instead of a specific value.
10986 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
10987 Bin Cheng <bin.cheng@arm.com>
10989 PR tree-optimization/82965
10990 PR tree-optimization/83991
10991 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
10992 by_profile_only parameter.
10993 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
10994 information if the loop was predicted to iterate too many times.
10995 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
10997 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
11000 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
11003 2018-04-13 Martin Liska <mliska@suse.cz>
11004 Jakub Jelinek <jakub@redhat.com>
11006 PR middle-end/81657
11007 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
11008 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
11009 * builtins.c (expand_builtin_memory_copy_args): Use
11010 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
11011 handle dest_addr == pc_rtx.
11013 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
11016 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
11017 asked to not generate direct moves.
11018 (fix_trunc<mode>si2_stfiwx): Similar.
11019 (fix_trunc<mode>si2_internal): Similar.
11021 2018-04-12 Jakub Jelinek <jakub@redhat.com>
11024 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
11025 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
11026 lookup if dest in some wider mode is known to be const0_rtx and
11027 if so, record permanent equivalence for it to be ZERO_EXTEND of
11028 the narrower mode destination.
11030 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
11032 * lto-streamer-out.c (output_function): Revert 259346.
11033 * omp-expand.c (expand_oacc_for): Likewise.
11035 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
11037 PR rtl-optimization/85354
11038 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
11039 * sel-sched.c (sel_global_init): ... here.
11041 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
11044 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
11045 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
11046 mode for PE-COFF targets.
11047 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
11048 (i386_pe_asm_lto_end): Likewise.
11049 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
11050 (TARGET_ASM_LTO_END): Likewise.
11051 * config/i386/winnt.c (saved_debug_info_level): New static variable.
11052 (i386_pe_asm_lto_start): New function.
11053 (i386_pe_asm_lto_end): Likewise.
11055 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
11056 Richard Biener <rguenther@suse.de>
11058 PR middle-end/84955
11059 * lto-streamer-out.c (output_function): Fix CFG loop state before
11061 * omp-expand.c (expand_oacc_for): Handle calls to internal
11062 functions like regular functions.
11064 2018-04-12 Richard Biener <rguenther@suse.de>
11067 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
11068 for the early LTO debug to properly generate references to it
11069 during DIE emission. Do not re-use that for the skeleton for
11071 (dwarf2out_early_finish): Likewise.
11073 2018-04-12 Jakub Jelinek <jakub@redhat.com>
11076 * config/i386/sse.md
11077 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
11078 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
11079 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
11080 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
11081 and output is a reg, avoid creating invalid lowpart subreg, but
11082 instead split into a 512-bit move. Don't split if not AVX512VL,
11083 input is xmm16+ reg and output is a mem.
11084 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
11085 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
11086 xmm16+ reg and output is a mem.
11088 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11090 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
11091 also for flag_dwarf2_cfi_asm.
11093 2018-04-12 Jakub Jelinek <jakub@redhat.com>
11095 PR rtl-optimization/85342
11096 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
11097 a bool scalar var inside of the loop instead. Don't try to update
11098 recog_data.operand after failed apply_change_group.
11100 2018-04-12 Tom de Vries <tom@codesourcery.com>
11103 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
11104 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
11105 array with flexible array member as array without given dimension.
11106 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
11107 argument for undefined param to true.
11109 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
11112 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
11113 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
11114 from PowerPC section.
11115 * config/rs6000/sysv4.opt (mcall-): Improve help text.
11116 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
11117 help text that is too long.
11118 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
11119 help text that is too long.
11120 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
11121 help text that is too long.
11123 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
11125 * config/alpha/alpha.md (stack_probe_internal): Rename
11126 from "probe_stack". Update all callers.
11128 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
11130 PR rtl-optimization/84566
11131 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
11132 sched_macro_fuse_insns.
11134 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
11137 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
11138 (compute_block_dependences): ... from here.
11140 2018-04-11 Jakub Jelinek <jakub@redhat.com>
11142 PR tree-optimization/85331
11143 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
11144 from int to HOST_WIDE_INT.
11146 2018-04-11 Martin Jambor <mjambor@suse.cz>
11149 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
11150 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
11151 not the same as the source val.
11152 (cgraph_edge_brings_value_p): New parameter.
11153 (gather_edges_for_value): Pass destination value to
11154 cgraph_edge_brings_value_p.
11155 (perhaps_add_new_callers): Likewise.
11156 (get_info_about_necessary_edges): Likewise and exclude values brought
11157 only by self-recursive edges.
11158 (create_specialized_node): Redirect only clones of self-calling edges.
11159 (+self_recursive_pass_through_p): New function.
11160 (find_more_scalar_values_for_callers_subset): Use it.
11161 (find_aggregate_values_for_callers_subset): Likewise.
11162 (known_aggs_to_agg_replacement_list): Removed.
11163 (decide_whether_version_node): Re-calculate known constants for all
11164 remaining context clones.
11166 2018-04-11 Richard Biener <rguenther@suse.de>
11169 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
11170 from early DWARF output.
11171 (dwarf2out_early_finish): Output line info unconditionally into
11172 early DWARF and add reference to it.
11174 2018-04-11 Jakub Jelinek <jakub@redhat.com>
11177 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
11178 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
11179 other than V2DFmode using iptr mode attribute.
11180 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
11182 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
11184 PR rtl-optimization/84659
11185 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
11187 2018-04-11 Jakub Jelinek <jakub@redhat.com>
11190 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
11192 (output_loc_list): Pass address of a dummy size variable even in the
11193 locview handling loop.
11194 (index_location_lists): Add comment on why skip_loc_list_entry can't
11197 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
11200 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
11203 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
11206 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
11207 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
11208 and -mstring-compare-inline-limit.
11210 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
11213 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
11214 for stack clash protection in a register whenever we need it to be in
11217 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
11219 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
11220 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
11222 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
11225 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
11227 (mlong-double-): Ditto.
11228 * config/rs6000/sysv4.opt (msdata=): Ditto.
11229 (mtls-size=): Ditto.
11231 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
11233 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
11234 erroneous entries for
11235 "vector int vec_ldl (int, long int *)", and
11236 "vector unsigned int vec_ldl (int, unsigned long int *)".
11237 Add comments and entries for
11238 "vector bool char vec_ldl (int, bool char *)",
11239 "vector bool short vec_ldl (int, bool short *)",
11240 "vector bool int vec_ldl (int, bool int *)",
11241 "vector bool long long vec_ldl (int, bool long long *)",
11242 "vector pixel vec_ldl (int, pixel *)",
11243 "vector long long vec_ldl (int, long long *)",
11244 "vector unsigned long long vec_ldl (int, unsigned long long *)".
11245 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
11246 type tree bool_long_long_type_node and correct definition of
11247 bool_V2DI_type_node to make reference to this new type tree.
11248 (rs6000_mangle_type): Replace erroneous reference to
11249 bool_long_type_node with bool_long_long_type_node.
11250 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
11251 comments to emphasize sign distinctions for char and int types and
11252 replace RS6000_BTI_bool_long constant with
11253 RS6000_BTI_bool_long_long constant. Also add comment to restrict
11254 use of RS6000_BTI_pixel.
11255 (bool_long_type_node): Remove this macro definition.
11256 (bool_long_long_type_node): New macro definition
11258 2018-04-10 Jakub Jelinek <jakub@redhat.com>
11260 PR rtl-optimization/85300
11261 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
11262 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
11263 simplify_unary_operation fails.
11265 2018-04-10 Martin Liska <mliska@suse.cz>
11267 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
11268 cgraph_edge and ipa_ref.
11270 2018-04-10 Jakub Jelinek <jakub@redhat.com>
11274 * config/i386/sse.md
11275 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
11276 computation of the VEC_MERGE selector from mask.
11277 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
11278 Fix decoding of the VEC_MERGE selector into mask.
11280 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
11282 PR tree-optimization/85286
11283 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
11285 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
11287 * final.c (final_1): Set insn_last_address as well as
11288 insn_current_address.
11290 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11293 * explow.c (emit_stack_probe): Call validize_mem on memory location
11294 before passing it to gen_probe_stack. Create address operand and
11295 legitimize it for the probe_stack_address case.
11297 2018-04-09 Jan Hubicka <jh@suse.cz>
11300 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
11301 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
11302 * tree.c (free_lang_data_in_type): Fix handling of binfos;
11304 (free_lang_data): Rebuild type inheritance graph.
11306 2018-04-09 Martin Sebor <msebor@redhat.com>
11308 * invoke.texi (-finline-small-functions): Mention other optimization
11310 (-findirect-inlining, -fpartial-inlining): Same.
11311 (-finline-functions-called-once): Same.
11312 (-freorder-blocks-and-partition): Same.
11314 2018-04-09 Jan Hubicka <jh@suse.cz>
11317 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
11318 jumps; choose last target that matches the criteria (i.e.
11319 no partition changes for non-crossing jumps).
11320 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
11321 support for redirecting crossing jumps to non-crossing.
11323 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
11325 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
11326 also for naked functions.
11328 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
11330 * config/arc/arc.md (add_shift): New pattern.
11331 (add_shift2): Likewise.
11332 (sub_shift): Likewise.
11333 (sub_shift_cmp0_noout): Likewise.
11334 (compare_si_ashiftsi): Likewise.
11335 (xbfu_cmp0_noout): New combine pattern.
11336 (xbfu_cmp0"): Likewise.
11337 (movsi_set_cc_insn): Place the predicable variant first.
11338 (commutative_binary_cmp0_noout): Remove clobber.
11339 (commutative_binary_cmp0): New pattern.
11340 (noncommutative_binary_cmp0): Likewise.
11341 (noncommutative_binary_cmp0_noout): Likewise.
11342 (noncommutative_binary_comparison_result_used): Removed.
11343 (rsub_cmp0): New pattern.
11344 (rsub_cmp0_noout): Likewise.
11345 (extzvsi): Changed, keep only meaningful variants.
11346 (SQH, SEZ): New iterators.
11347 (SQH_postfix): New mode attribute.
11348 (SEZ_prefix): New code attribute.
11349 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
11350 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
11351 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
11352 of numerical value.
11353 (noncommutative_operator): Check the availability of barrel
11356 2018-04-09 Richard Biener <rguenther@suse.de>
11358 PR tree-optimization/85284
11359 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
11360 Only use the niter constraining form of simple_iv when the exit
11361 is always executed.
11363 2018-04-09 Tom de Vries <tom@codesourcery.com>
11366 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
11367 (define_expand "*memory_barrier"): New define_expand.
11368 (define_insn "memory_barrier"): New insn.
11370 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
11372 PR rtl-optimization/80463
11373 PR rtl-optimization/83972
11374 PR rtl-optimization/83480
11376 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
11377 correct producer for the insn.
11378 (tidy_control_flow): Fixup seqnos in case of debug insns.
11380 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
11382 PR rtl-optimization/83913
11384 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
11385 different sched-times when merging exprs.
11387 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
11389 PR rtl-optimization/83962
11391 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
11392 tidy_fallthru_edge and tidy_control_flow.
11394 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
11396 PR rtl-optimization/83530
11398 * sel-sched.c (force_next_insn): New global variable.
11399 (remove_insn_for_debug): When force_next_insn is true, also leave only
11400 next insn in the ready list.
11401 (sel_sched_region): When the region wasn't scheduled, make another pass
11402 over it with force_next_insn set to 1.
11404 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
11406 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
11408 * config/nds32/constants.md (unspec_volatile_element): Add enum values
11409 for interrupt control.
11410 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
11411 functions for interrupt control.
11412 * config/nds32/nds32-intrinsic.md: Likewise.
11413 * config/nds32/nds32_intrinsic.h: Likewise.
11414 * config/nds32/nds32.h (nds32_builtins): Likewise.
11416 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
11418 * config/nds32/nds32.c (nds32_init_machine_status,
11419 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
11420 strict_aligned_p field.
11421 (nds32_expand_to_rtl_hook): New function.
11422 (TARGET_EXPAND_TO_RTL_HOOK): Define.
11423 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
11425 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
11426 Chung-Ju Wu <jasonwucj@gmail.com>
11428 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
11429 * config/nds32/nds32-n7.md: New file.
11430 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
11431 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
11433 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
11434 * config/nds32/nds32.md (pipeline_model): Add n7.
11435 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
11436 * config/nds32/pipelines.md: Include n7 settings.
11438 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
11439 Chung-Ju Wu <jasonwucj@gmail.com>
11441 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
11442 * config/nds32/nds32-e8.md: New file.
11443 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
11444 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
11446 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
11447 * config/nds32/nds32.md (pipeline_model): Add e8.
11448 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
11449 * config/nds32/pipelines.md: Include e8 settings.
11451 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
11452 Chung-Ju Wu <jasonwucj@gmail.com>
11454 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
11455 * config/nds32/nds32-n8.md: New file.
11456 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
11457 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
11459 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
11460 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
11461 * config/nds32/nds32.md (pipeline_model): Add n8.
11462 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
11463 * config/nds32/pipelines.md: Include n8 settings.
11465 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
11466 Chung-Ju Wu <jasonwucj@gmail.com>
11468 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
11469 * config/nds32/nds32-n9-2r1w.md: New file.
11470 * config/nds32/nds32-n9-3r2w.md: New file.
11471 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
11472 nds32_register_ports): New or modify for cpu n9.
11473 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
11475 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
11476 * config/nds32/nds32-utils.c: New file.
11477 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
11478 TARGET_MUL_SLOW): Define.
11479 * config/nds32/nds32.md (pipeline_model): New attribute.
11480 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
11481 New options that support cpu n9.
11482 * config/nds32/pipelines.md: Include n9 settings.
11483 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
11485 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
11487 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
11488 information if necessary.
11489 (output_cond_branch_compare_zero): Likewise.
11490 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
11491 (nds32_target_alignment): Refine for alignment.
11492 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
11493 (FUNCTION_BOUNDARY): Modify.
11494 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
11496 * config/nds32/nds32.opt (malways-align, malign-functions): New.
11498 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
11500 * config/nds32/constants.md (unspec_volatile_element): Add values for
11501 TLB operation and data prefetch.
11502 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
11503 functions for TLB operation and data prefetch.
11504 * config/nds32/nds32-intrinsic.md: Likewise.
11505 * config/nds32/nds32_intrinsic.h: Likewise.
11506 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
11507 (nds32_print_operand): Likewise.
11508 * config/nds32/nds32.h (nds32_builtins): Likewise.
11510 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
11511 Andrew Pinski <pinsika@gcc.gnu.org>
11513 PR middle-end/82976
11514 * match.pd: Use constant_boolean_node of correct type instead of
11515 boolean_true_node or boolean_false_node for simplifying
11516 pointer comparisons to zero.
11518 2018-04-07 Jakub Jelinek <jakub@redhat.com>
11520 PR tree-optimization/80021
11521 * tree.c (verify_type_variant): Make error call in verify_variant_match
11522 translatable and remove final full stop.
11524 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
11526 * config/nds32/constants.md (unspec_volatile_element): Add
11527 UNSPEC_VOLATILE_EH_RETURN.
11528 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
11529 nds32_output_stack_pop): Support dwarf exception handling process.
11530 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
11531 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
11532 exception handling process.
11533 (nds32_compute_stack_frame): Likewise.
11534 (nds32_return_addr_rtx): Likewise.
11535 (nds32_initial_elimination_offset): Likewise.
11536 (nds32_expand_prologue): Likewise.
11537 (nds32_expand_epilogue): Likewise.
11538 (nds32_dynamic_chain_address): New function.
11539 * config/nds32/nds32.h (machine_function): Add fields for dwarf
11540 exception handling.
11541 (DYNAMIC_CHAIN_ADDRESS): Define.
11542 (EH_RETURN_DATA_REGNO): Define.
11543 (EH_RETURN_STACKADJ_RTX): Define.
11544 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
11545 patterns for dwarf exception handling.
11547 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
11549 * config/nds32/nds32.h: Clean up obsolete macros.
11551 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
11553 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
11554 Add enum values for particular instructions.
11555 * config/nds32/nds32-intrinsic.c: Implementation of expanding
11556 particular intrinsic functions.
11557 * config/nds32/nds32-intrinsic.md: Likewise.
11558 * config/nds32/nds32_intrinsic.h: Likewise.
11559 * config/nds32/nds32.h (nds32_builtins): Likewise.
11560 * config/nds32/nds32.md (type): Add pbsad and pbsada.
11561 (btst, ave): New patterns for particular instructions.
11563 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
11565 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
11566 Add enum values for atomic load/store and memory sync.
11567 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
11569 * config/nds32/nds32-intrinsic.md: Likewise.
11570 * config/nds32/nds32_intrinsic.h: Likewise.
11571 * config/nds32/nds32.h (nds32_builtins): Likewise.
11573 2018-04-07 Jakub Jelinek <jakub@redhat.com>
11575 PR tree-optimization/85257
11576 * fold-const.c (native_encode_vector): If not all elts could fit
11577 and off is -1, return 0 rather than offset.
11578 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
11579 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
11580 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
11581 adjust buffer in native_interpret_expr call.
11583 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
11585 * config/nds32/constants.md (unspec_volatile_element): Add cache
11586 control enum values.
11587 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
11588 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
11589 * config/nds32/nds32.c (nds32_cctl_names): New.
11590 (nds32_print_operand): Handle cache control register names.
11591 * config/nds32/nds32.h (nds32_builtins): New enum values.
11592 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
11594 * config/nds32/nds32.md (type): Add mmu.
11595 * config/nds32/pipelines.md (simple_insn): Add mmu.
11597 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
11599 * config/nds32/nds32.md (type): Remove call.
11600 * config/nds32/pipelines.md (simple_insn): Likewise.
11602 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
11604 * config/nds32/constants.md (unspec_volatile_element): Add
11605 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
11606 UNSPEC_VOLATILE_FMFCFG.
11607 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
11608 description for fmfcfg and fmfcsr.
11609 (bdesc_1arg): Add fmtcsr.
11610 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
11611 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
11612 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
11613 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
11614 unspec_fmfcfg): New patterns.
11615 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
11616 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
11617 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
11618 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
11619 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
11620 __nds32__fmfcfg): Define.
11622 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
11624 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
11625 intrinsic register names.
11626 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
11627 intrinsic register enum values and macros.
11629 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
11631 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
11632 for load/store addressing form.
11633 (nds32_print_operand_address): Likewise.
11635 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
11638 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
11639 based on LABEL_REF. Remove useless assertion.
11640 (pic_address_needs_scratch): Fix formatting.
11641 (sparc_legitimize_pic_address): Minor tweaks.
11642 (sparc_delegitimize_address): Adjust assertion accordingly.
11643 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
11644 into symbolic_operand.
11645 (movsi_high_pic_label_ref): Likewise.
11646 (movsi_lo_sum_pic_label_ref): Likewise.
11647 (movdi_pic_label_ref): Likewise.
11648 (movdi_high_pic_label_ref): Likewise.
11649 (movdi_lo_sum_pic_label_ref): Likewise.
11651 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
11653 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
11654 custom LIB_SPEC setup.
11656 2018-04-06 Ruslan Bukin <br@bsdpad.com>
11657 Kito Cheng <kito.cheng@gmail.com>
11659 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
11660 * config/riscv/freebsd.h: New.
11662 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
11664 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
11665 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
11668 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
11669 Kito Cheng <kito.cheng@gmail.com>
11671 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
11672 nds32_output_call, nds32_symbol_binds_local_p): New functions.
11673 * config/nds32/nds32-protos.h (nds32_output_call,
11674 nds32_output_return): Declare.
11675 * config/nds32/nds32.md: Refine all the call and return patterns.
11677 2018-04-06 Jakub Jelinek <jakub@redhat.com>
11680 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
11681 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
11683 PR rtl-optimization/84872
11684 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
11685 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
11686 EDGE_CROSSING edge.
11688 2018-04-06 Tamar Christina <tamar.christina@arm.com>
11690 * expr.c (copy_blkmode_to_reg): Revert 254862.
11691 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
11693 2018-04-06 Richard Biener <rguenther@suse.de>
11695 PR middle-end/85244
11696 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
11697 after seeing a component reference with an adjacent field. Treat
11698 refs to arrays at struct end of external decls similar to
11699 refs to unconstrained commons.
11701 2018-04-06 Jakub Jelinek <jakub@redhat.com>
11704 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
11705 look through SAVE_EXPRs with non-side-effects argument. Adjust
11707 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
11710 2018-04-06 Richard Biener <rguenther@suse.de>
11712 PR middle-end/85180
11713 * alias.c (find_base_term): New wrapper around find_base_term
11714 unwinding CSELIB_VAL_PTR changes.
11715 (find_base_term): Do not restore CSELIB_VAL_PTR during the
11718 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11720 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
11722 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
11723 constant definitions.
11724 ("nop"): lr 0,0 -> nopr r0
11725 ("nop_lr0", "nop_lr1"): New insn definitions.
11727 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
11729 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
11730 NDS32_V3PUSH_AVAILABLE_P macro.
11732 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
11733 Chung-Ju Wu <jasonwucj@gmail.com>
11735 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
11736 (nds32*-*-*): Add float and fpu_config into supported_defaults.
11737 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
11738 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
11739 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
11740 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
11741 * config/nds32/constraints.md: New constraints and checking for hard
11742 float configuration.
11743 * config/nds32/iterators.md: New mode iterator and attribute for hard
11744 float configuration.
11745 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
11747 * config/nds32/nds32-fpu.md: New file.
11748 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
11749 deal with hard float code generation.
11750 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
11752 (abi_type, float_reg_number): New enum type.
11753 * config/nds32/nds32-predicates.c: New predicates for hard float.
11754 * config/nds32/nds32-protos.h: Declare functions for hard float.
11755 * config/nds32/nds32.c: Implementation for hard float configuration.
11756 * config/nds32/nds32.h: Definitions for hard float configuration.
11757 * config/nds32/nds32.md: Include hard float machine description and
11758 modify patterns for hard float configuration.
11759 * config/nds32/nds32.opt: New options for hard float configuration.
11760 * config/nds32/predicates.md: New predicates for hard float
11763 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
11765 * common/config/nds32/nds32-common.c
11766 (nds32_option_optimization_table): Enable -mreleax-hint by default.
11768 2018-04-05 Jakub Jelinek <jakub@redhat.com>
11770 PR middle-end/85195
11771 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
11772 CONSTRUCTOR_ELT (ctor, ...)->value.
11774 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
11777 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
11779 2018-04-05 Tom de Vries <tom@codesourcery.com>
11782 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
11785 2018-04-05 Shiva Chen <shiva0217@gmail.com>
11786 Kito Cheng <kito.cheng@gmail.com>
11788 * config/nds32/constraints.md (U33): Fine-tune checking condition.
11789 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
11790 * config/nds32/nds32.h (nds32_16bit_address_type): Add
11791 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
11793 2018-04-05 Shiva Chen <shiva0217@gmail.com>
11794 Kito Cheng <kito.cheng@gmail.com>
11796 * config/nds32/constraints.md (Ufe): New memory constraint.
11797 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
11798 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
11799 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
11801 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
11802 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
11804 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11806 * config/nds32/nds32.md: Use optimize_size in the condition for
11807 alu-shift instructions.
11809 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11811 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
11813 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11815 * config/nds32/nds32.md (negsi2): Refine pattern.
11817 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
11818 Chung-Ju Wu <jasonwucj@gmail.com>
11820 * config/nds32/iterators.md (shift_rotate): New code iterator.
11821 (shift): New code attribute.
11822 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
11823 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
11824 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
11825 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
11826 bit-wise operations.
11827 (andsi3, *andsi3): Ditto.
11828 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
11829 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
11830 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
11831 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
11832 nds32_ior_operand, nds32_xor_operand): New predicates.
11834 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11836 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
11837 (addsi3, subsi3): ... this.
11839 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11841 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
11843 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11845 * config/nds32/nds32.md: Adjust indention.
11847 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
11849 * config/nds32/nds32.md (feature): New attribute.
11851 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11853 * config/nds32/nds32.md (subtype): New attribute.
11855 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
11858 * config/arm/arm-builtins.c (arm_expand_builtin): Change
11859 expansion to perform a bitwise AND of the argument followed by a
11860 boolean negation of the result.
11862 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
11864 PR rtl-optimization/84878
11865 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
11866 the basic block. Assert the use reference is not artificial and that
11867 it has an associated insn.
11869 2018-04-04 Michael Matz <matz@suse.de>
11871 * builtins.c (compute_objsize): Pass correct operand
11872 to array_at_struct_end_p.
11874 2018-04-04 Richard Biener <rguenther@suse.de>
11877 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
11878 from contexts for DINFO_LEVEL_TERSE and below.
11880 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
11882 * config/nds32/nds32-doubleword.md (move_<mode>): Require
11883 resiter_operand condition.
11884 * config/nds32/nds32.md (*move<mode>): Ditto.
11886 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
11887 Monk Chiang <sh.chiang04@gmail.com>
11889 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
11891 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
11893 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
11895 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
11896 Kito Cheng <kito.cheng@gmail.com>
11898 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
11899 nds32_cond_code_str, output_cond_branch,
11900 output_cond_branch_compare_zero, nds32_expand_cbranch,
11901 nds32_expand_cstore, nds32_expand_movcc,
11902 nds32_output_cbranchsi4_equality_zero,
11903 nds32_output_cbranchsi4_equality_reg,
11904 nds32_output_cbranchsi4_equality_reg_or_const_int,
11905 nds32_output_cbranchsi4_greater_less_zero: New functions.
11906 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
11907 nds32_expand_cstore, nds32_expand_movcc,
11908 nds32_output_cbranchsi4_equality_zero,
11909 nds32_output_cbranchsi4_equality_reg,
11910 nds32_output_cbranchsi4_equality_reg_or_const_int,
11911 nds32_output_cbranchsi4_greater_less_zero): Declare.
11912 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
11913 nds32_rimm11s_operand): New predicates.
11914 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
11915 * config/nds32/nds32.md: Rewrite all the branch and conditional move
11918 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
11920 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
11921 * config/nds32/nds32.md: Ditto.
11922 * config/nds32/pipelines.md: Ditto.
11924 2018-04-04 Richard Biener <rguenther@suse.de>
11926 PR tree-optimization/85168
11927 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
11928 propagating abnormals.
11930 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
11932 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
11934 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
11935 Kito Cheng <kito.cheng@gmail.com>
11937 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
11938 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
11939 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
11940 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
11941 * config/nds32/nds32.md (sibcall_internal): New.
11942 (sibcall_register): Remove.
11943 (sibcall_immediate): Remove.
11944 (sibcall_value_internal): New.
11945 (sibcall_value_register): Remove.
11946 (sibcall_value_immediate): Remove.
11947 * config/nds32/predicates.md (nds32_general_register_operand): New.
11948 (nds32_call_address_operand): New.
11950 2018-04-03 Jakub Jelinek <jakub@redhat.com>
11952 PR rtl-optimization/85167
11953 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
11954 bb_defs if *split_p, instead preinitialize it to NULL.
11956 PR tree-optimization/85156
11957 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
11958 evaluating the argument multiple times.
11960 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
11962 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
11964 (_mm_cvtpd_ps): Likewise.
11965 (_mm_cvttpd_epi32): Likewise.
11966 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
11967 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
11968 vector, pixel, and bool following altivec.h include.
11970 2018-04-03 Martin Sebor <msebor@redhat.com>
11972 * doc/extend.texi (Common Function Attributes): Clarify.
11973 (const attribute): Likewise.
11974 (pure attribute): Likewise.
11976 2018-04-03 Jakub Jelinek <jakub@redhat.com>
11979 * config/i386/i386.c (ix86_expand_vector_set): Use
11980 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
11982 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
11984 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
11985 instructions when changing rounding bits to preserve precision bits
11986 in the x87 control word.
11988 2018-04-03 Martin Liska <mliska@suse.cz>
11990 PR tree-optimization/82491
11991 * rtl.h (strip_offset_and_add): Replace += suboffset with
11992 poly_uint64 () + suboffset.
11994 2018-03-29 Martin Liska <mliska@suse.cz>
11995 Martin Jambor <mjambor@suse.cz>
11998 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
11999 param_type is not an integral or pointer type.
12001 2018-04-03 Richard Biener <rguenther@suse.de>
12003 * sese.h (recompute_all_dominators): Remove.
12005 2018-04-02 Martin Sebor <msebor@redhat.com>
12007 * doc/invoke.texi (-Wrestrict): Fix typos.
12009 2018-04-02 Jim Wilson <jimw@sifive.com>
12011 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
12012 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
12013 (<optab>di3, <optab>si3_extend): Likewise.
12014 (<optab>si3_mask, <optab>si3_mask_1): New.
12015 (<optab>di3_mask, <optab>di3_mask_1): New.
12016 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
12017 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
12018 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
12020 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
12022 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
12025 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
12027 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
12028 (nds32_canonicalize_comparison): New function.
12030 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
12031 Kito Cheng <kito.cheng@gmail.com>
12032 Kuan-Lin Chen <kuanlinchentw@gmail.com>
12034 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
12035 * config/nds32/constants.md (unspec_volatile_element): Add
12036 UNSPEC_VOLATILE_RELAX_GROUP.
12037 * config/nds32/nds32-relax-opt.c: New file.
12038 * config/nds32/nds32-predicates.c
12039 (nds32_symbol_load_store_p): New function.
12040 * config/nds32/nds32-protos.h
12041 (nds32_symbol_load_store_p): Declare function.
12042 (make_pass_nds32_relax_opt): Declare new rtl pass function.
12043 * config/nds32/nds32.c
12044 (nds32_register_pass): New function to register pass.
12045 (nds32_register_passes): New function to register passes.
12046 * config/nds32/nds32.md (relax_group): New pattern.
12047 * config/nds32/nds32.opt (mrelax-hint): New option.
12048 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
12050 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
12052 * config/nds32/t-nds32: Modify files dependency.
12054 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
12056 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
12057 (PROFILE_HOOK): Define its implementation.
12059 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
12061 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
12062 type and 32-bit size.
12064 2018-04-01 Jakub Jelinek <jakub@redhat.com>
12066 PR middle-end/85090
12067 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
12068 (V_128_256): New mode iterator.
12069 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
12070 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
12071 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
12073 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
12074 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
12076 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
12079 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
12080 NaN inputs correctly.
12082 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
12085 * config/rs6000/vsx.md (??r): New mode attribute.
12086 (*vsx_mov<mode>_64bit): Use it.
12087 (*vsx_mov<mode>_32bit): Likewise.
12089 2018-03-30 Martin Sebor <msebor@redhat.com>
12091 PR tree-optimization/84818
12092 * builtins.c (check_access): Use warning_n.
12094 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12097 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
12099 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
12102 2018-03-30 Julia Koval <julia.koval@intel.com>
12105 * x86-tune.def (movx, partial_reg_dependency): Enable for
12108 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
12110 PR inline-asm/84985
12111 * lra-constraints.c (process_alt_operands): Move setting
12112 this_alternative_matches below.
12114 2018-03-29 Martin Liska <mliska@suse.cz>
12117 * doc/invoke.texi: Document how LTO works with debug info.
12118 Describe auto-load support of binutils. Mention 'x86-64'
12119 as valid option value of -march option.
12121 2018-03-29 Jakub Jelinek <jakub@redhat.com>
12123 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
12126 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
12127 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
12128 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
12131 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
12134 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
12135 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
12136 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
12137 for RS6000_BTM_POWERPC64.
12138 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
12139 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
12140 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
12143 (DIVDEU): Likewise.
12145 2018-03-28 Carl Love <cel@us.ibm.com>
12148 2017-09-27 Carl Love <cel@us.ibm.com>
12150 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
12151 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
12152 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
12155 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12157 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
12158 instead of __vector bool.
12159 (_mm_max_pu8): Likewise.
12160 (_mm_min_pi16): Likewise.
12162 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
12165 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
12166 (DIVWEUO): Likewise.
12167 (DIVDEO): Likewise.
12168 (DIVDEUO): Likewise.
12169 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
12170 DIVWEUO and DIVDEUO.
12171 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
12172 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
12173 (div_extend): Likewise.
12174 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
12176 (__builtin_divweuo): Likewise.
12177 (__builtin_divdeo): Likewise.
12178 (__builtin_divdeuo): Likewise.
12180 2018-03-28 Jakub Jelinek <jakub@redhat.com>
12183 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
12184 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
12186 PR tree-optimization/82004
12187 * gimple-match-head.c (optimize_pow_to_exp): New function.
12188 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
12189 Don't fold to exp if optimize_pow_to_exp is false.
12191 2018-03-28 Martin Liska <mliska@suse.cz>
12194 * calls.c (initialize_argument_information): Fix trailing space.
12195 * common.opt: Fix typo and provide better explanation for
12196 -fsanitize-coverage option.
12197 * config/i386/i386.opt: Fix typo.
12199 2018-03-28 Jakub Jelinek <jakub@redhat.com>
12200 Martin Liska <mliska@suse.cz>
12203 * gimplify.c (asan_poison_variable): Don't do the check for
12204 gimplify_omp_ctxp here.
12205 (gimplify_decl_expr): Do it here.
12206 (gimplify_target_expr): Likewise.
12208 2018-03-28 Martin Liska <mliska@suse.cz>
12211 * config/i386/i386.c (ix86_function_arg_advance): Do not call
12212 chkp_type_bounds_count if MPX is not enabled.
12214 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
12216 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
12218 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
12221 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
12222 function to create the function decl for complex long double
12223 multiply and divide for -mabi=ieeelongdouble.
12224 (init_float128_ieee): Call it.
12226 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
12229 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
12230 -fcf-protection=branch -mibt.
12231 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
12233 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
12236 * config/arm/arm.c (arm_valid_symbolic_address): Handle
12237 arm_word_relocations.
12239 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
12242 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
12243 extern array declarations.
12245 2018-03-27 Richard Biener <rguenther@suse.de>
12247 PR middle-end/84067
12248 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
12249 explicit single_use checks.
12251 2018-03-27 Richard Biener <rguenther@suse.de>
12253 PR tree-optimization/85082
12254 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
12257 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
12259 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
12260 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
12261 Turn on fasynchronous-unwind-tables and funwind-tables.
12263 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
12266 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
12267 (*bmi_blsr_<mode>_ccz): Ditto.
12269 2018-03-26 Tom de Vries <tom@codesourcery.com>
12271 PR tree-optimization/85063
12272 * omp-general.c (offloading_function_p): New function. Factor out
12274 * omp-offload.c (pass_omp_target_link::gate): ... here.
12275 * omp-general.h (offloading_function_p): Declare.
12276 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
12277 with attribute omp declare target for offloading functions.
12279 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
12281 PR tree-optimization/84005
12282 * tree-data-ref.h (get_base_for_alignment): Declare.
12283 * tree-data-ref.c (get_base_for_alignment_1): New function.
12284 (get_base_for_alignment): Likewise.
12285 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
12286 get_base_for_alignment to find a suitable base object, instead
12287 of always using drb->base_address.
12289 2018-03-23 Jakub Jelinek <jakub@redhat.com>
12291 PR inline-asm/85022
12292 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
12293 known size by default.
12295 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
12297 PR inline-asm/85030
12298 * lra-constraints.c (process_alt_operands): Don't match BLKmode
12299 and non BLKmode operands.
12301 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12304 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
12305 Clean up attributes.
12307 2018-03-23 Richard Biener <rguenther@suse.de>
12310 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
12311 we are going to emit early debug for LTO.
12313 2018-03-23 Jakub Jelinek <jakub@redhat.com>
12315 PR inline-asm/85034
12316 * function.c (match_asm_constraints_1): Don't optimize if input
12317 doesn't satisfy general_operand predicate for output's mode.
12319 PR inline-asm/85022
12320 * alias.c (write_dependence_p): Don't require for x_canonicalized
12321 non-VOIDmode if x has VOIDmode.
12324 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
12325 just don't try to optimize it rather than assert it never happens.
12327 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
12329 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
12330 macro expansions for definition of ST_INTERNAL_<mode> and
12331 LD_INTERNAL_<mode> builtins.
12332 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
12334 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
12336 (altivec_expand_st_builtin): Likewise.
12337 (altivec_expand_builtin): Remove calls to deleted functions.
12338 (rs6000_address_for_altivec): Delete this function.
12339 * config/rs6000/vector.md: Remove expands for
12340 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
12342 2018-03-22 Sudakshina Das <sudi.das@arm.com>
12345 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
12346 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
12347 re-computing once computed.
12348 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
12349 (arm_init_machine_status): Initialize
12350 machine->static_chain_stack_bytes.
12352 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
12355 * doc/extend.texi: Add four new prototypes for vec_ld.
12356 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
12357 definitions for more logical presentation.
12358 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
12359 entries for V1TI variants of __builtin_altivec_ld builtin.
12360 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
12361 handling of V1TI variant of LVX icode pattern.
12362 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
12363 (rs6000_gimple_fold_builtin): Likewise.
12364 (altivec_init_builtins): Add code to define
12365 __builtin_altivec_lvx_v1ti function.
12367 2018-03-22 Jakub Jelinek <jakub@redhat.com>
12369 PR inline-asm/84941
12370 * function.c (match_asm_constraints_1): Don't do the optimization
12371 if input isn't a REG, SUBREG, MEM or constant.
12373 2018-03-22 Tom de Vries <tom@codesourcery.com>
12375 PR tree-optimization/84956
12376 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
12377 bb_has_abnormal_pred.
12379 2018-03-22 Jakub Jelinek <jakub@redhat.com>
12382 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
12383 DECL_INITIAL (decl) to decl at the end.
12384 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
12385 adjust the comment.
12387 2018-03-21 Joseph Myers <joseph@codesourcery.com>
12389 * doc/extend.texi (__builtin_tgmath): Document when complex
12390 integer types are treated as _Complex _Float64.
12392 2018-03-21 Tom de Vries <tom@codesourcery.com>
12394 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
12396 2018-03-21 Jakub Jelinek <jakub@redhat.com>
12398 PR tree-optimization/84960
12399 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
12400 if it is ENTRY block, move them into single succ of ENTRY in that case.
12402 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
12404 PR tree-optimization/84811
12405 * poly-int.h (poly_span_traits): Remove the T3 parameter and
12406 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
12407 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
12408 (known_subrange_p): Update accordingly. Cast each value involved
12409 in the size comparison, rather than casting the result of the
12412 2018-03-21 Jakub Jelinek <jakub@redhat.com>
12414 PR tree-optimization/84982
12415 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
12416 by flipping the least significant bit rather than all bits from
12417 bitpos to bitpos + bitsize - 1.
12419 2018-03-21 Nathan Sidwell <nathan@acm.org>
12421 * doc/extend.texi (Deprecated Features): Remove mention of
12422 long-deleted deprecations.
12424 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12427 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
12428 * configure: Regenerate.
12430 2018-03-21 Tom de Vries <tom@codesourcery.com>
12432 PR tree-optimization/83126
12433 * tree-parloops.c (num_phis): New function.
12434 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
12436 2018-03-21 Nathan Sidwell <nathan@acm.org>
12438 * doc/extend.texi (Deprecated Features): Update deprecated flags,
12439 mention anon-struct/union members and trailing attributes.
12441 2018-03-21 Bin Cheng <bin.cheng@arm.com>
12443 PR tree-optimization/84969
12444 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
12445 builtin memset partitions if they set different rhs values.
12447 2018-03-21 Jakub Jelinek <jakub@redhat.com>
12449 PR rtl-optimization/84989
12450 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
12451 VEC_DUPLICATE with scalar result mode.
12453 2018-03-21 Martin Liska <mliska@suse.cz>
12456 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
12457 not intended return statement.
12459 2018-03-21 Martin Liska <mliska@suse.cz>
12462 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
12463 (chkp_find_bound_slots_1): Limit number of iterations.
12465 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
12468 * Minor grammar fixes for x86 options.
12470 2018-03-20 Jakub Jelinek <jakub@redhat.com>
12473 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
12474 holding REG_CFA_RESTORE notes, instead turn them into a USE.
12476 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
12479 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
12480 (altivec_lvx_<mode>_1op): Likewise.
12481 (altivec_stvx_<mode>_2op): Likewise.
12482 (altivec_stvx_<mode>_1op): Likewise.
12483 (altivec_lvx_<VM2:mode>): New define_expand.
12484 (altivec_stvx_<VM2:mode>): Likewise.
12485 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
12486 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
12487 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
12488 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
12489 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
12490 (rs6000_gen_lvx): Likewise.
12491 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
12492 (altivec_expand_stv_builtin): Likewise.
12493 (altivec_expand_builtin): Likewise.
12494 * config/rs6000/vector.md: Likewise.
12496 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12499 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
12502 2018-03-20 Richard Biener <rguenther@suse.de>
12505 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
12506 sign-conversions as zero, fall back to standard scalar_stmt
12509 2018-03-20 Martin Liska <mliska@suse.cz>
12512 * predict.c (rebuild_frequencies): Handle case when we have
12513 PROFILE_ABSENT, but flag_guess_branch_prob is false.
12515 2018-03-20 Jakub Jelinek <jakub@redhat.com>
12518 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
12519 flag_section_anchors.
12520 * varasm.c (use_blocks_for_decl_p): Remove hack for
12521 dw2_force_const_mem.
12524 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
12526 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
12527 be created, use lowpart_subreg of operands[0] rather than operands[0]
12529 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
12530 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
12531 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
12532 and n constraint instead of aarch64_shift_imm_di and Usd.
12533 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
12534 (*aarch64_<optab>_reg_minus<mode>3): ... this.
12536 2018-03-20 Sudakshina Das <sudi.das@arm.com>
12539 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
12540 to favor GPR over NEON registers.
12541 (<shift>di3_neon): Likewise.
12543 2018-03-20 Tom de Vries <tom@codesourcery.com>
12546 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
12547 (nvptx_process_pars): Emit bar.sync asap and alap.
12549 2018-03-20 Tom de Vries <tom@codesourcery.com>
12552 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
12553 seen_label if seen_label is already set.
12555 2018-03-20 Jakub Jelinek <jakub@redhat.com>
12558 * config/i386/i386.c (fold_builtin_cpu): For features above 31
12559 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
12560 Use 1U instead of 1. Formatting fixes.
12563 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
12564 instead of TREE_TYPE (s1) for the return value.
12566 2018-03-19 Jakub Jelinek <jakub@redhat.com>
12568 PR tree-optimization/84946
12569 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
12570 bitsize + bitsize in poly_uint64 rather than poly_int64.
12573 * dwarf2asm.c: Include fold-const.c.
12574 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
12575 of decl rather than decl itself.
12577 PR rtl-optimization/84643
12578 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
12580 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
12583 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
12584 calling assemble_variable.
12586 2018-03-19 Sudakshina Das <sudi.das@arm.com>
12589 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
12590 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
12592 2018-03-19 Jim Wilson <jimw@sifive.com>
12595 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
12596 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
12597 (riscv_first_stack_step): Likewise.
12598 (riscv_option_override): Use STACK_BOUNDARY instead of
12599 MIN_STACK_BOUNDARY.
12600 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
12601 MIN_STACK_BOUNDARY.
12602 (BIGGEST_ALIGNMENT): Set to 128.
12603 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
12604 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
12607 2018-03-19 Richard Biener <rguenther@suse.de>
12609 PR tree-optimization/84933
12610 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
12611 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
12613 2018-03-19 Richard Biener <rguenther@suse.de>
12615 PR tree-optimization/84859
12616 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
12617 (cond_if_else_store_replacement): Perform sinking operation on
12618 single-store BBs regardless of MAX_STORES_TO_SINK setting.
12619 Generalize what a BB with a single eligible store is.
12621 2018-03-19 Richard Biener <rguenther@suse.de>
12623 PR tree-optimization/84929
12624 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
12625 chrec_is_positive against non-chrec arg.
12627 2018-03-19 Tamar Christina <tamar.christina@arm.com>
12630 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
12632 2018-03-18 Martin Liska <mliska@suse.cz>
12634 PR rtl-optimization/84635
12635 * regrename.c (build_def_use): Use matches_mode only when
12638 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
12640 PR tree-optimization/84913
12641 * tree-vect-loop.c (vectorizable_reduction): Don't try to
12642 vectorize chains of COND_EXPRs.
12644 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
12646 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
12648 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
12650 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
12652 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
12654 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
12656 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
12657 Kito Cheng <kito.cheng@gmail.com>
12659 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
12660 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
12661 (nds32_adjust_reg_alloc_order): New function.
12662 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
12664 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
12666 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
12667 nds32_print_operand, nds32_print_operand_address): Use
12668 HOST_WIDE_INT_PRINT_DEC instead.
12670 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
12672 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
12674 2018-03-17 Jakub Jelinek <jakub@redhat.com>
12677 * config/i386/i386.c (initial_ix86_tune_features,
12678 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
12679 unsigned long long.
12680 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
12681 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
12682 rather than 1u << ix86_tune. Formatting fix.
12683 (ix86_option_override_internal): Change ix86_arch_mask from
12684 unsigned int to unsigned HOST_WIDE_INT, initialize to
12685 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
12686 (ix86_function_specific_restore): Likewise.
12688 2018-03-16 Jakub Jelinek <jakub@redhat.com>
12691 * postreload.c (reload_combine_recognize_pattern): Perform
12692 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
12693 truncate_int_for_mode the result for the destination's mode.
12696 * hsa-gen.c (mem_type_for_type): Fix comment typo.
12697 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
12699 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
12702 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
12705 * lra-assigns.c (lra_split_hard_reg_for): Don't use
12706 regno_allocno_class_array and sorted_pseudos.
12707 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
12708 insns where regno is used.
12710 2018-03-16 Martin Liska <mliska@suse.cz>
12713 * multiple_target.c (create_dispatcher_calls): Redirect
12714 reference in the symbol table.
12716 2018-03-16 Martin Liska <mliska@suse.cz>
12719 * multiple_target.c (create_dispatcher_calls): Redirect also
12722 2018-03-16 Jakub Jelinek <jakub@redhat.com>
12726 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
12727 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
12728 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
12730 2018-03-16 Julia Koval <julia.koval@intel.com>
12732 * doc/invoke.texi (Skylake Server): Add CLWB.
12733 Cannonlake): Remove CLWB.
12735 2018-03-16 Jakub Jelinek <jakub@redhat.com>
12737 PR tree-optimization/84841
12738 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
12740 (FLOAT_ONE_CONST_TYPE): Define.
12741 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
12742 (sort_by_operand_rank): Put entries with higher constant_type last
12743 rather than first to match comments.
12745 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
12747 * config/nios2/nios2.md (movsi_internal): Fix thinko in
12750 2018-03-15 Jakub Jelinek <jakub@redhat.com>
12753 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
12754 check and use address of target always.
12756 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
12759 * config/i386/i386.c (indirect_thunk_needed): Update comments.
12760 (indirect_thunk_bnd_needed): Likewise.
12761 (indirect_thunks_used): Likewise.
12762 (indirect_thunks_bnd_used): Likewise.
12763 (indirect_return_needed): New.
12764 (indirect_return_bnd_needed): Likewise.
12765 (output_indirect_thunk_function): Add a bool argument for
12767 (output_indirect_thunk_function): Don't generate alias for
12768 function return thunk.
12769 (ix86_code_end): Call output_indirect_thunk_function to generate
12770 function return thunks.
12771 (ix86_output_function_return): Set indirect_return_bnd_needed
12772 and indirect_return_needed instead of indirect_thunk_bnd_needed
12773 and indirect_thunk_needed.
12775 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
12777 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
12778 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
12779 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
12781 2018-03-15 David Malcolm <dmalcolm@redhat.com>
12782 Paul Hua <paul.hua.gm@gmail.com>
12785 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
12787 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
12789 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
12790 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
12791 resp. SFmode cases.
12793 2018-03-15 Tamar Christina <tamar.christina@arm.com>
12796 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
12797 instead of GET_MODE_SIZE when comparing Units.
12799 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
12802 * varasm.c (hash_section): Return an unchangeble hash value
12803 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
12804 Return !aarch64_can_use_per_function_literal_pools_p ().
12806 2018-03-15 Jakub Jelinek <jakub@redhat.com>
12809 * optabs.c (emit_conditional_move): Pass address of cmode's copy
12810 rather than address of cmode as last argument to prepare_cmp_insn.
12812 2018-03-15 Julia Koval <julia.koval@intel.com>
12814 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
12815 F_AVX512VNNI, F_AVX512BITALG): New.
12817 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
12820 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
12821 insn for floating-point loads and stores.
12823 2018-03-14 Carl Love <cel@us.ibm.com>
12825 * config/rs6000/rs6000-c.c: Add macro definitions for
12826 ALTIVEC_BUILTIN_VEC_PERMXOR.
12827 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
12828 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
12829 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
12830 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
12832 * config/doc/extend.texi: Add prototypes for vec_permxor.
12834 2018-03-14 David Malcolm <dmalcolm@redhat.com>
12837 * diagnostic-show-locus.c (class layout_point): Convert m_line
12838 from int to linenum_type.
12839 (line_span::comparator): Use linenum "compare" function when
12840 comparing line numbers.
12841 (test_line_span): New function.
12842 (layout_range::contains_point): Convert param "row" from int to
12844 (layout_range::intersects_line_p): Likewise.
12845 (layout::will_show_line_p): Likewise.
12846 (layout::print_source_line): Likewise.
12847 (layout::should_print_annotation_line_p): Likewise.
12848 (layout::print_annotation_line): Likewise.
12849 (layout::print_leading_fixits): Likewise.
12850 (layout::annotation_line_showed_range_p): Likewise.
12851 (struct line_corrections): Likewise for field m_row.
12852 (line_corrections::line_corrections): Likewise for param "row".
12853 (layout::print_trailing_fixits): Likewise.
12854 (layout::get_state_at_point): Likewise.
12855 (layout::get_x_bound_for_row): Likewise.
12856 (layout::print_line): Likewise.
12857 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
12858 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
12859 * input.c (selftest::test_linenum_comparisons): New function.
12860 (selftest::input_c_tests): Call it.
12861 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
12862 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
12863 * selftest.h (ASSERT_GT): New macro.
12864 (ASSERT_GT_AT): New macro.
12865 (ASSERT_LT): New macro.
12866 (ASSERT_LT_AT): New macro.
12868 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
12870 PR rtl-optimization/84780
12871 * combine.c (distribute_links): Don't make a link based on pc_rtx.
12873 2018-03-14 Martin Liska <mliska@suse.cz>
12875 * tree.c (record_node_allocation_statistics): Use
12876 get_stats_node_kind.
12877 (get_stats_node_kind): New function extracted from
12878 record_node_allocation_statistics.
12879 (free_node): Use get_stats_node_kind.
12881 2018-03-14 Richard Biener <rguenther@suse.de>
12883 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
12884 that the value-set of ANTIC_IN doesn't grow.
12887 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
12889 (BB_VISITED_WITH_VISITED_SUCCS): New define.
12890 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
12892 2018-03-14 Julia Koval <julia.koval@intel.com>
12894 * config.gcc (icelake-client, icelake-server): New.
12896 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
12897 (initial_ix86_arch_features): Ditto.
12898 (PTA_SKYLAKE): Add SGX.
12899 (PTA_ICELAKE): Remove.
12900 (PTA_ICELAKE_CLIENT): New.
12901 (PTA_ICELAKE_SERVER): New.
12902 (ix86_option_override_internal): Split up icelake on icelake client and
12904 (get_builtin_code_for_version): Ditto.
12905 (fold_builtin_cpu): Ditto.
12906 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
12907 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
12908 * config/i386/i386.h (processor_type): Ditto.
12909 * doc/invoke.texi: Ditto.
12911 2018-03-14 Jakub Jelinek <jakub@redhat.com>
12914 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
12915 INTEGER_CST offset, add it together with bitpos / 8 and
12916 sign extend based on POINTER_SIZE.
12920 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
12923 * config/i386/constraints.md (Yc): New register constraint.
12924 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
12925 Use Yc constraint for alternative 2 of operand 0. Remove
12926 preferred_for_speed attribute.
12928 2018-03-14 Richard Biener <rguenther@suse.de>
12930 PR tree-optimization/84830
12931 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
12932 with the old one to avoid oscillations.
12934 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
12937 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
12939 (assign_by_spills): Return a flag of reload assignment failure.
12940 Do not process the reload assignment failures. Do not spill other
12941 reload pseudos if they has the same reg class. Update n if
12943 (lra_assign): Add a return arg. Set up from the result of
12944 assign_by_spills call.
12945 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
12946 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
12947 usage_insns if it is not NULL.
12948 (spill_hard_reg_in_range): New function.
12949 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
12950 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
12951 function prototypes.
12952 (lra_assign): Change prototype.
12953 * lra.c (lra): Add code to deal with fails by splitting hard reg
12956 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
12958 * config/riscv/riscv.opt (mrelax): New option.
12959 * config/riscv/riscv.c (riscv_file_start): Emit ".option
12960 "norelax" when riscv_mrelax is disabled.
12961 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
12963 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12966 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
12967 reassociation for int modes.
12969 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
12971 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
12972 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
12974 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
12975 * config/aarch64/aarch64-sve.md
12976 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
12977 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
12978 (*extend<mode><Vwide>2): Rename to...
12979 (aarch64_sve_extend<mode><Vwide>2): ...this.
12980 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
12981 renaming the old pattern to...
12982 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
12984 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
12985 define_expand, renaming the old pattern to...
12986 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
12987 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
12988 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
12989 account when deciding which SVE instruction the optab should use.
12990 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
12992 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
12994 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
12995 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
12996 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
12997 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
12998 (tlsdesc_small_<mode>): Turn a define_expand and use
12999 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
13000 (tlsdesc_small_advsimd_<mode>): ...this.
13001 (tlsdesc_small_sve_<mode>): New pattern.
13003 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
13005 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
13006 (UNSPEC_UMUL_HIGHPART): New constants.
13007 (MUL_HIGHPART): New int iteraor.
13008 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
13009 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
13011 (*<su>mul<mode>3_highpart): New define_insn.
13013 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
13016 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
13019 2018-03-13 Martin Liska <mliska@suse.cz>
13022 * (sem_item_optimizer::sem_item_optimizer): Initialize new
13024 (sem_item_optimizer::~sem_item_optimizer): Release it.
13025 (sem_item_optimizer::merge_classes): Register variable aliases.
13026 (sem_item_optimizer::fixup_pt_set): New function.
13027 (sem_item_optimizer::fixup_points_to_sets): Likewise.
13028 * ipa-icf.h: Declare new variables and functions.
13030 2018-03-13 Jakub Jelinek <jakub@redhat.com>
13032 PR middle-end/84834
13033 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
13034 integer_pow2p@2 and test integer_pow2p in condition.
13035 (A < 0 ? C : 0): Similarly for @1.
13037 PR middle-end/84831
13038 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
13039 characters starting at p contain '\0' character, don't look beyond
13043 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
13044 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
13047 * reg-stack.c (change_stack): Change update_end var from int to
13048 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
13049 also call set_block_for_insn on the newly added insns and rescan.
13052 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
13053 on the last operand.
13056 * tree.c (stabilize_reference_1): Return save_expr (e) for
13057 STATEMENT_LIST even if it doesn't have side-effects.
13059 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
13061 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
13063 2018-03-12 Renlin Li <renlin.li@arm.com>
13065 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
13066 aarch64_output_scalar_simd_mov_immediate.
13068 2018-03-12 Martin Sebor <msebor@redhat.com>
13070 PR tree-optimization/83456
13071 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
13072 for perfectly overlapping calls to memcpy.
13073 (gimple_fold_builtin_memory_chk): Same.
13074 (gimple_fold_builtin_strcpy): Handle no-warning.
13075 (gimple_fold_builtin_stxcpy_chk): Same.
13076 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
13078 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
13080 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
13081 parameter. Use it for SFmode.
13082 (rs6000_function_arg_advance_1): Adjust.
13083 (rs6000_function_arg): Adjust.
13084 (rs6000_gimplify_va_arg): Pass false for that new parameter.
13086 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
13088 PR rtl-optimization/84169
13089 PR rtl-optimization/84780
13090 * combine.c (can_combine_p): Check for a 2-insn combination whether
13091 the destination register is used between the two insns, too.
13093 2018-03-12 Richard Biener <rguenther@suse.de>
13095 PR tree-optimization/84803
13096 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
13097 for refs DR analysis didn't process.
13099 2018-03-12 Richard Biener <rguenther@suse.de>
13101 PR tree-optimization/84777
13102 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
13103 force-vectorize loops ignore whether we are optimizing for size.
13105 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
13107 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
13108 (TARGET_MD_ASM_ADJUST): Define.
13110 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
13111 Kito Cheng <kito.cheng@gmail.com>
13112 Chung-Ju Wu <jasonwucj@gmail.com>
13114 * config/nds32/nds32.c (nds32_compute_stack_frame,
13115 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
13116 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
13117 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
13118 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
13119 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
13120 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
13121 * config/nds32/nds32.md (prologue, epilogue): Use macro
13122 NDS32_V3PUSH_AVAILABLE_P to do checking.
13124 2018-03-11 Jakub Jelinek <jakub@redhat.com>
13127 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
13128 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
13129 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
13130 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
13131 addition of most attributes on !orig_type_die or the attribute not
13132 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
13134 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
13135 Chung-Ju Wu <jasonwucj@gmail.com>
13137 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
13138 __NDS32_VH__ macro.
13139 * config/nds32/nds32.opt (mvh): New option.
13141 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
13142 Chung-Ju Wu <jasonwucj@gmail.com>
13144 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
13146 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
13147 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
13150 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
13151 Chung-Ju Wu <jasonwucj@gmail.com>
13153 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
13155 * config/nds32/nds32-multiple.md (strlensi): New pattern.
13156 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
13158 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
13159 Kito Cheng <kito.cheng@gmail.com>
13160 Chung-Ju Wu <jasonwucj@gmail.com>
13162 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
13163 UNSPEC_FFMISM and UNSPEC_FLMISM.
13164 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
13165 for ffb, ffmism and flmism.
13166 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
13167 (unspec_ffmism): Ditto.
13168 (unspec_flmism): Ditto.
13169 (nds32_expand_builtin_impl): Check if string extension is available.
13170 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
13171 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
13173 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
13176 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
13179 * lra-assigns.c (assign_by_spills): Return a flag of reload
13180 assignment failure. Do not process the reload assignment
13181 failures. Do not spill other reload pseudos if they has the same
13183 (lra_assign): Add a return arg. Set up from the result of
13184 assign_by_spills call.
13185 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
13186 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
13187 usage_insns if it is not NULL.
13188 (spill_hard_reg_in_range): New function.
13189 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
13190 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
13191 function prototypes.
13192 (lra_assign): Change prototype.
13193 * lra.c (lra): Add code to deal with fails by splitting hard reg
13196 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
13199 * config/i386/i386.opt: Replace Enforcment with Enforcement.
13201 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
13204 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
13205 (dw_val_node): Add val_symbolic_view.
13206 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
13207 (symview_upper_bound): New.
13208 (new_line_info_table): Initialize symviews_since_reset.
13209 (dwarf2out_source_line): Count symviews_since_reset and set
13210 symview_upper_bound.
13211 (dw_val_equal_p): Handle symview.
13212 (add_AT_symview): New.
13213 (print_dw_val): Handle symview.
13214 (attr_checksum, attr_checksum_ordered): Likewise.
13215 (same_dw_val_p, size_of_die): Likewise.
13216 (value_format, output_die): Likewise.
13217 (add_high_low_attributes): Use add_AT_symview for entry_view.
13218 (dwarf2out_finish): Reset symview_upper_bound, clear
13221 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
13224 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
13225 Add strict argument and use it.
13226 (rs6000_split_multireg_move): Update for new strict argument.
13227 (mem_operand_gpr): Disallow all non-offsettable addresses.
13228 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
13230 2018-03-09 Jakub Jelinek <jakub@redhat.com>
13233 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
13234 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
13235 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
13238 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
13239 decl, use remap_type if we want to use the type.
13241 2018-03-09 Martin Sebor <msebor@redhat.com>
13243 PR tree-optimization/84526
13244 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
13246 (builtin_access::generic_overlap): Be prepared to handle non-array
13249 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
13251 PR rtl-optimization/84682
13252 * lra-constraints.c (process_address_1): Check is_address flag
13253 for address constraints.
13254 (process_alt_operands): Likewise.
13255 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
13256 preprocess_constraints.
13257 * recog.h (preprocess_constraints): Add oploc parameter.
13259 * recog.c (preprocess_constraints): Test address_operand for
13260 CT_ADDRESS constraints.
13262 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
13265 * lra-assigns.c (assign_by_spills): Return a flag of reload
13266 assignment failure. Do not process the reload assignment
13267 failures. Do not spill other reload pseudos if they has the same
13269 (lra_assign): Add a return arg. Set up from the result of
13270 assign_by_spills call.
13271 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
13272 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
13273 usage_insns if it is not NULL.
13274 (spill_hard_reg_in_range): New function.
13275 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
13276 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
13277 function prototypes.
13278 (lra_assign): Change prototype.
13279 * lra.c (lra): Add code to deal with fails by splitting hard reg
13282 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13285 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
13286 Accept complain bool parameter. Only emit errors if it is true.
13287 (arm_parse_cpu_option_name): Likewise.
13288 (arm_target_thumb_only): Adjust callers of the above.
13289 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
13290 prototype to take a default true bool parameter.
13291 (arm_parse_arch_option_name): Likewise.
13293 2018-03-09 David Malcolm <dmalcolm@redhat.com>
13294 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
13298 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
13299 * configure: Regenerate.
13300 * configure.ac ("linker --version-script option"): New.
13301 ("linker soname option"): New.
13303 2018-03-09 Richard Biener <rguenther@suse.de>
13305 PR tree-optimization/84775
13306 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
13307 immediate uses of predicate stmts and mark them modified.
13310 PR tree-optimization/84178
13311 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
13313 (version_loop_for_if_conversion): Delay update_ssa call.
13314 (tree_if_conversion): Delay update_ssa until after predicate
13317 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
13320 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
13321 when the function accesses prior frames.
13323 2018-03-08 Jakub Jelinek <jakub@redhat.com>
13326 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
13327 gen_llsym, otherwise call maybe_gen_llsym.
13329 PR inline-asm/84742
13330 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
13331 has ',' character inside of it.
13333 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13336 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
13337 as clobbering CC_REGNUM.
13339 2018-03-08 Richard Biener <rguenther@suse.de>
13341 PR middle-end/84552
13342 * tree-scalar-evolution.c: Include tree-into-ssa.h.
13343 (follow_copies_to_constant): Do not follow SSA names registered
13346 2018-03-08 Richard Biener <rguenther@suse.de>
13348 PR tree-optimization/84178
13349 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
13351 (version_loop_for_if_conversion): Delay update_ssa call.
13352 (tree_if_conversion): Delay update_ssa until after predicate
13355 2018-03-08 David Malcolm <dmalcolm@redhat.com>
13357 PR tree-optimization/84178
13358 * tree-if-conv.c (release_bb_predicate): Remove the
13359 the assertion that the stmts have NULL use_ops.
13360 Discard the statements, asserting that they haven't
13361 yet been added to a BB.
13363 2018-03-08 Richard Biener <rguenther@suse.de>
13365 PR tree-optimization/84746
13366 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
13367 (phi_translate): Pass in destination ANTIC_OUT set.
13368 (phi_translate_1): Likewise. For a simplified result lookup
13369 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
13370 (phi_translate_set): Adjust.
13371 (do_pre_regular_insertion): Likewise.
13372 (do_pre_partial_partial_insertion): Likewise.
13374 2018-03-08 Martin Liska <mliska@suse.cz>
13376 PR gcov-profile/84735
13377 * doc/gcov.texi: Document usage of profile files.
13378 * gcov-io.h: Document changes in the format.
13380 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
13384 * dwarf2out.c (struct dw_line_info_table): Update comments for
13386 (FORCE_RESET_NEXT_VIEW): New.
13387 (FORCE_RESETTING_VIEW_P): New.
13388 (RESETTING_VIEW_P): Check for -1 too.
13389 (ZERO_VIEW_P): Likewise.
13390 (new_line_info_table): Force-reset next view.
13391 (dwarf2out_begin_function): Likewise.
13392 (dwarf2out_source_line): Simplify zero_view_p initialization.
13393 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
13394 view directly. Omit view when omitting .loc at line 0.
13396 2018-03-08 Jakub Jelinek <jakub@redhat.com>
13398 PR tree-optimization/84740
13399 * tree-switch-conversion.c (process_switch): Call build_constructors
13400 only if info.phi_count is non-zero.
13402 PR tree-optimization/84739
13403 * tree-tailcall.c (find_tail_calls): Check call arguments against
13404 DECL_ARGUMENTS (current_function_decl) rather than
13405 DECL_ARGUMENTS (func) when checking for tail recursion.
13407 2018-03-07 Jakub Jelinek <jakub@redhat.com>
13409 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
13410 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
13411 Volker Reichelt's entry and add entries for people that perform
13412 GCC fuzzy testing and report numerous bugs.
13414 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
13417 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
13418 readonly data in sdata, if that is disabled.
13419 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
13420 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
13421 -mreadonly-in-sdata option.
13423 2018-03-07 Martin Sebor <msebor@redhat.com>
13425 PR tree-optimization/84468
13426 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
13427 basic block when looking for nul assignment.
13429 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
13432 * except.h (output_function_exception_table): Adjust prototype.
13433 * except.c (output_function_exception_table): Remove FNNAME parameter
13434 and add SECTION parameter. Ouput one part of the table at a time.
13435 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
13436 the first part of the exception table and emit unwind directives.
13437 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
13438 (i386_pe_seh_cold_init): Likewise.
13439 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
13440 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
13441 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
13442 (ix86_output_call_insn): Emit a nop in one more case for SEH.
13443 * config/i386/winnt.c: Include except.h.
13444 (struct seh_frame_state): Add reg_offset, after_prologue and
13445 in_cold_section fields.
13446 (i386_pe_seh_end_prologue): Set seh->after_prologue.
13447 (i386_pe_seh_cold_init): New function.
13448 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
13449 to seh->in_cold_section.
13450 (seh_emit_push): Record the offset of the push.
13451 (seh_emit_save): Record the offet of the save.
13452 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
13453 Test seh->after_prologue to disregard the epilogue.
13454 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
13455 (i386_pe_end_cold_function): New function.
13457 2018-03-07 Jakub Jelinek <jakub@redhat.com>
13460 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
13461 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
13464 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
13466 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
13467 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
13469 PR middle-end/84723
13470 * multiple_target.c: Include tree-inline.h and intl.h.
13471 (expand_target_clones): Diagnose and fail if node->definition and
13472 !tree_versionable_function_p (node->decl).
13474 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
13476 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
13478 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
13479 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
13480 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
13482 2018-03-06 Jakub Jelinek <jakub@redhat.com>
13485 * combine.c (try_combine): Use reg_or_subregno instead of handling
13486 just paradoxical SUBREGs and REGs.
13488 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
13490 * config/arc/arc.c (arc_finalize_pic): Remove function.
13491 (arc_must_save_register): We use single base PIC register, remove
13492 checks to save/restore the PIC register.
13493 (arc_expand_prologue): Likewise.
13494 * config/arc/arc-protos.h (arc_set_default_type_attributes):
13496 (arc_verify_short): Likewise.
13497 (arc_attr_type): Likewise.
13498 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
13499 (walk_stores): Likewise.
13500 (arc_address_cost): Make it static.
13501 (arc_verify_short): Likewise.
13502 (branch_dest): Likewise.
13503 (arc_attr_type): Likewise.
13504 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
13505 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
13506 (arc_final_prescan_insn): Remove inserting the nops due to
13507 hardware hazards. It is done in reorg step.
13508 (insn_length_variant_t): Remove.
13509 (insn_length_parameters_t): Likewise.
13510 (arc_insn_length_parameters): Likewise.
13511 (arc_get_insn_variants): Likewise.
13512 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
13514 2018-03-06 Jakub Jelinek <jakub@redhat.com>
13516 PR inline-asm/84683
13517 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
13520 PR tree-optimization/84687
13521 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
13523 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
13525 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13527 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
13528 Rename to ppc_speculation_barrier.
13529 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
13530 __builtin_ppc_speculation_barrier.
13532 2018-03-05 Jakub Jelinek <jakub@redhat.com>
13535 * combine.c (combine_simplify_rtx): Don't try to simplify if
13536 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
13539 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
13541 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
13542 to 32 bytes when compiling for POWER9.
13544 2018-03-05 Jakub Jelinek <jakub@redhat.com>
13547 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
13548 regparm >= 3 with no arg reg available also for calls with
13549 flag_force_indirect_call. Pass decl to ix86_function_regparm.
13552 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
13554 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
13556 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
13559 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
13561 2018-03-05 Richard Biener <rguenther@suse.de>
13563 PR tree-optimization/84486
13564 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
13565 When inserting a __builtin_assume_aligned call set the LHS
13566 SSA name alignment info accordingly.
13568 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
13570 PR tree-optimization/84114
13571 * config/aarch64/aarch64.c (aarch64_reassociation_width)
13572 Avoid reassociation of FLOAT_MODE addition.
13574 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
13576 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
13577 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
13578 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
13579 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
13580 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
13581 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
13582 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
13584 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
13585 __builtin_ia32_wbinvd): New builtins.
13586 (SPECIAL_ARGS2): New.
13587 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
13588 (SPECIAL_ARGS2): New.
13589 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
13590 (ix86_valid_target_attribute_inner_p): Ditto.
13591 (ix86_init_mmx_sse_builtins): Add special_args2.
13592 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
13593 TARGET_WBNOINVD_P): New.
13594 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
13595 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
13596 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
13597 * config/i386/immintrin.h (_wbinvd): New intrinsic.
13598 * config/i386/pconfigintrin.h: New file.
13599 * config/i386/wbnoinvdintrin.h: Ditto.
13600 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
13602 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
13604 2018-03-05 Richard Biener <rguenther@suse.de>
13606 PR tree-optimization/84670
13607 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
13609 (BB_VISITED_WITH_VISITED_SUCCS): New define.
13610 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
13611 (compute_antic_aux): Only assert the number of values in ANTIC_IN
13612 doesn't grow if all successors (recursively) were visited at least
13615 2018-03-05 Richard Biener <rguenther@suse.de>
13617 PR tree-optimization/84650
13618 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
13619 if executed in the loop pipeline.
13621 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
13623 * doc/configfiles.texi (Configuration Files): Move info about
13624 conditionalizing $target-protos.h to...
13625 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
13626 differs from $target-protos.h.
13628 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
13629 Chung-Ju Wu <jasonwucj@gmail.com>
13631 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
13632 * config/nds32/nds32-multiple.md (setmemsi): Define.
13633 * config/nds32/nds32-memory-manipulation.c
13634 (nds32_gen_dup_4_byte_to_word_value): New.
13635 (emit_setmem_word_loop): New.
13636 (emit_setmem_byte_loop): New.
13637 (nds32_expand_setmem_loop): New.
13638 (nds32_expand_setmem_loop_v3m): New.
13639 (nds32_expand_setmem_unroll): New.
13640 (nds32_expand_setmem): New.
13642 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
13643 Chung-Ju Wu <jasonwucj@gmail.com>
13645 * config/nds32/nds32-memory-manipulation.c
13646 (nds32_emit_load_store): New.
13647 (nds32_emit_post_inc_load_store): New.
13648 (nds32_emit_mem_move): New.
13649 (nds32_emit_mem_move_block): New.
13650 (nds32_expand_movmemsi_loop_unknown_size): New.
13651 (nds32_expand_movmemsi_loop_known_size): New.
13652 (nds32_expand_movmemsi_loop): New.
13653 (nds32_expand_movmemsi_unroll): New.
13654 (nds32_expand_movmemqi): Rename ...
13655 (nds32_expand_movmemsi): ... to this.
13656 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
13657 (movmemsi): ... to this.
13658 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
13659 (nds32_expand_movmemsi): ... to this.
13661 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
13662 Monk Chiang <sh.chiang04@gmail.com>
13663 Chung-Ju Wu <jasonwucj@gmail.com>
13665 * config/nds32/nds32-protos.h
13666 (nds32_expand_load_multiple): New arguments.
13667 (nds32_expand_store_multiple): Ditto.
13668 (nds32_valid_multiple_load_store): Rename ...
13669 (nds32_valid_multiple_load_store_p): ... to this.
13670 * config/nds32/nds32-memory-manipulation.c
13671 (nds32_expand_load_multiple): Refine implementation.
13672 (nds32_expand_store_multiple): Ditto.
13673 * config/nds32/nds32-multiple.md
13674 (load_multiple): Update nds32_expand_load_multiple interface.
13675 (store_multiple): Update nds32_expand_store_multiple interface.
13676 * config/nds32/nds32-predicates.c
13677 (nds32_valid_multiple_load_store): Rename ...
13678 (nds32_valid_multiple_load_store_p): ... to this and refine
13680 * config/nds32/predicates.md
13681 (nds32_load_multiple_and_update_address_operation): New predicate.
13682 (nds32_store_multiple_and_update_address_operation): New predicate.
13684 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
13685 Chung-Ju Wu <jasonwucj@gmail.com>
13687 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
13688 (combo): New attribute.
13689 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
13691 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
13693 * config/nds32/nds32.opt: Change -mcmodel= default value.
13695 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
13696 Monk Chiang <sh.chiang04@gmail.com>
13697 Chung-Ju Wu <jasonwucj@gmail.com>
13699 * config/nds32/constants.md (unspec_element): New enum.
13700 * config/nds32/constraints.md (Umw): New constraint.
13701 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
13702 * config/nds32/nds32-intrinsic.md: Likewise.
13703 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
13704 (nds32_valid_smw_lwm_base_p): New.
13705 (nds32_output_smw_single_word): New.
13706 (nds32_output_lmw_single_word): New.
13707 (nds32_expand_unaligned_load): New.
13708 (nds32_expand_unaligned_store): New.
13709 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
13710 (nds32_output_smw_single_word): Declare.
13711 (nds32_output_lmw_single_word): Declare.
13712 (nds32_expand_unaligned_load): Declare.
13713 (nds32_expand_unaligned_store): Declare.
13714 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
13715 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
13716 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
13717 NDS32_BUILTIN_UASTORE_DW.
13718 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
13721 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
13722 Kito Cheng <kito.cheng@gmail.com>
13723 Chung-Ju Wu <jasonwucj@gmail.com>
13725 * config/nds32/nds32-intrinsic.c
13726 (nds32_expand_builtin_null_ftype_reg): Delete.
13727 (nds32_expand_builtin_reg_ftype_imm): Ditto.
13728 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
13729 (nds32_read_argument): New.
13730 (nds32_legitimize_target): Ditto.
13731 (nds32_legitimize_argument): Ditto.
13732 (nds32_check_constant_argument): Ditto.
13733 (nds32_expand_unop_builtin): Ditto.
13734 (nds32_expand_unopimm_builtin): Ditto.
13735 (nds32_expand_binop_builtin): Ditto.
13736 (nds32_builtin_decl_impl): Ditto.
13737 (builtin_description): Ditto.
13738 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
13739 (nds32_init_builtins_impl): Ditto.
13740 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
13741 (nds32_builtin_decl): New.
13742 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
13743 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
13745 2018-03-02 Jeff Law <law@redhat.com>
13747 * reorg.c (stop_search_p): Handle DEBUG_INSN.
13748 (redundant_insn, fill_simple_delay_slots): Likewise.
13749 (fill_slots_from_thread): Likewise.
13750 * resource.c (mark_referenced_resources): Likewise.
13751 (mark_set_resources, find_dead_or_set_registers): Likewise.
13753 2018-03-02 Jakub Jelinek <jakub@redhat.com>
13755 * substring-locations.h (format_warning_va): Formatting fix for
13756 ATTRIBUTE_GCC_DIAG.
13757 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
13759 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
13760 * substring-locations.c: Include intl.h.
13761 (format_warning_va): Turned into small wrapper around
13762 format_warning_n_va, renamed to ...
13763 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
13764 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
13766 (format_warning_at_substring_n): New function.
13767 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
13768 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
13769 format_warning_at_substring with just a shorter name instead of
13770 const function pointer.
13771 (fmtwarn_n): New function.
13772 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
13773 appropriate, get rid of all the fmtstr temporaries, move conditionals
13774 with G_() wrapped string literals directly into fmtwarn arguments,
13775 cast dir.len to (int), formatting fixes.
13777 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
13779 * doc/invoke.texi: Remove "Cilk Plus" references.
13781 2018-03-02 Jakub Jelinek <jakub@redhat.com>
13782 Richard Biener <rguenther@suse.de>
13785 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
13786 for error or warning attributes if CALL_FROM_THUNK_P is set.
13789 2018-03-02 Jakub Jelinek <jakub@redhat.com>
13792 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
13793 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
13796 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
13797 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
13799 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
13800 instead of -1U in last predictors element's probability member.
13802 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
13805 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
13806 arguments if they are comparable.
13808 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
13810 PR tree-optimization/84634
13811 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
13812 masks and masked_loop_p with a single loop_masks, making sure it's
13813 null for bb vectorization.
13815 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
13817 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
13818 (vect_analyze_data_ref_access): Use loop->safe_len rather than
13819 loop->force_vectorize to check whether there is no alias.
13821 2018-03-02 Jakub Jelinek <jakub@redhat.com>
13824 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
13826 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
13828 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
13829 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
13830 instead of a loop around prev_real_insn.
13831 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
13834 PR inline-asm/84625
13835 * config/i386/i386.c (ix86_print_operand): Use conditional
13836 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
13839 2018-03-02 Richard Biener <rguenther@suse.de>
13841 PR tree-optimization/84427
13842 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
13843 (bitmap_set_subtract_values): Rewrite to handle multiple
13846 (prune_clobbered_mems): Likewise.
13847 (phi_translate): Take edge instead of pred/phiblock.
13848 (phi_translate_1): Likewise.
13849 (phi_translate_set): Likewise. Insert all translated
13850 exprs for a value into the set, keeping possibly multiple
13851 expressions per value.
13852 (compute_antic_aux): Adjust for phi_translate changes.
13853 When intersecting union the expressions and prune those
13854 not in the final value set, keeping possibly multiple
13855 expressions per value. Do not use value-insertion
13856 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
13857 all expressions. Add verification that the value-sets
13858 only shrink during iteration.
13859 (compute_partial_antic_aux): Adjust for the phi_translate changes.
13860 (do_pre_regular_insertion): Likewise.
13861 (do_pre_partial_partial_insertion): Likewise.
13863 2018-03-02 Richard Biener <rguenther@suse.de>
13866 * config/darwin.c (saved_debug_info_level): New static global.
13867 (darwin_asm_lto_start): Disable debug info generation for LTO out.
13868 (darwin_asm_lto_end): Restore debug info generation settings.
13870 2018-03-01 Martin Liska <mliska@suse.cz>
13873 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
13874 volatile arguments.
13876 2018-03-01 Richard Biener <rguenther@suse.de>
13879 * dwarf2out.c (gen_variable_die): Properly handle late VLA
13880 type annotation with LTO when debug was disabled at compile-time.
13882 2018-03-01 Matthew Fortune <mfortune@gmail.com>
13884 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
13886 (mips_final_postscan_insn): Likewise.
13888 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
13890 PR rtl-optimization/84528
13891 * alias.c (init_alias_target): Add commentary.
13892 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
13893 a unique base value if the frame pointer is not eliminated
13894 to the stack pointer.
13896 2018-03-01 Tom de Vries <tom@codesourcery.com>
13898 PR rtl-optimization/83327
13899 * lra-int.h (hard_regs_spilled_into): Declare.
13900 * lra.c (hard_regs_spilled_into): Define.
13901 (init_reg_info): Init hard_regs_spilled_into.
13902 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
13903 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
13904 (process_bb_lives): Handle hard_regs_spilled_into.
13905 (lra_create_live_ranges_1): Before doing liveness propagation, clear
13906 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
13908 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
13910 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
13911 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
13912 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
13913 * config/rs6000/aix72.h: New file.
13915 2018-02-28 Jakub Jelinek <jakub@redhat.com>
13917 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
13918 instead of warning_at with conditional singular and plural messages
13922 * stor-layout.c (update_alignment_for_field): For
13923 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
13924 && !DECL_PACKED (field), do the alignment update, just use
13925 only desired_align instead of MAX (type_align, desired_align)
13927 (place_field): Don't do known_align < desired_align handling
13928 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
13929 is non-NULL, instead do it after rli->prev_field handling and
13930 only if not within a bitfield word. For DECL_PACKED (field)
13931 use type_align of BITS_PER_UNIT.
13933 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
13935 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
13936 superfluous parentheses and trailing spaces.
13938 2018-02-28 Richard Biener <rguenther@suse.de>
13940 PR tree-optimization/84584
13941 * graphite-scop-detection.c (scop_detection::add_scop): Discard
13942 SCoPs with fake exit edge.
13944 2018-02-28 Martin Liska <mliska@suse.cz>
13947 * timevar.c (timer::print): Fix format to properly print 100%
13950 2018-02-28 Richard Biener <rguenther@suse.de>
13952 PR middle-end/84607
13953 * genmatch.c (capture_info::walk_match): Do not mark
13954 captured expressions without operands as expr_p given
13955 they act more like predicates and should be subject to
13956 "lost tail" side-effect preserving.
13958 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
13960 PR rtl-optimization/81611
13961 * auto-inc-dec.c (attempt_change): Move dead note from
13962 mem_insn if it's the next use of regno
13963 (find_address): Take address use of reg holding
13964 non-incremented value. Add parm to limit search to the named
13966 (merge_in_block): Attempt to use a mem insn that is the next
13967 use of the original regno.
13969 2018-02-27 Martin Sebor <msebor@redhat.com>
13972 * doc/invoke.texi (-Wmissing-attributes): New option.
13973 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
13975 2018-02-27 Martin Sebor <msebor@redhat.com>
13977 PR translation/84207
13978 * diagnostic-core.h (warning_n, error_n, inform_n): Change
13979 n argument to unsigned HOST_WIDE_INT.
13980 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
13981 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
13982 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
13983 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
13985 2018-02-27 Richard Biener <rguenther@suse.de>
13987 PR tree-optimization/84512
13988 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
13989 Do not use the estimate returned from record_stmt_cost for
13990 the scalar iteration cost but sum properly using add_stmt_cost.
13992 2018-02-27 Richard Biener <rguenther@suse.de>
13994 PR tree-optimization/84466
13995 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
13996 Adjust last change to less strictly validate use operands.
13998 2018-02-27 Martin Liska <mliska@suse.cz>
14000 PR gcov-profile/84548
14001 * gcov.c (process_file): Allow partial overlap and consider it
14002 also as group functions.
14003 (output_lines): Properly calculate range of lines for a group.
14005 2018-02-27 Martin Liska <mliska@suse.cz>
14007 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
14008 'ggc' suffixes. Change first column width.
14009 (timer::print): Fix formatting of the column.
14011 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
14013 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
14014 preserve inline entry blocks for the sake of debug inline
14015 entry point markers alone.
14016 (remove_unused_locals): Suggest in comments a better place to
14017 force the preservation of inline entry blocks that are
14018 otherwise unused, but do not preserve them.
14020 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
14022 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
14024 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
14027 * config/i386/constraints.md (Bs): Replace
14028 ix86_indirect_branch_register with
14029 TARGET_INDIRECT_BRANCH_REGISTER.
14031 * config/i386/i386.md (indirect_jump): Likewise.
14032 (tablejump): Likewise.
14033 (*sibcall_memory): Likewise.
14034 (*sibcall_value_memory): Likewise.
14035 Peepholes of indirect call and jump via memory: Likewise.
14036 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
14037 (*sibcall_value_GOT_32): Likewise.
14038 * config/i386/predicates.md (indirect_branch_operand): Likewise.
14039 (GOT_memory_operand): Likewise.
14040 (call_insn_operand): Likewise.
14041 (sibcall_insn_operand): Likewise.
14042 (GOT32_symbol_operand): Likewise.
14043 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
14045 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
14047 PR rtl-optimization/83496
14048 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
14049 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
14050 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
14051 redundant insn, if any.
14052 (relax_delay_slots): Likewise.
14053 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
14055 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
14057 PR tree-optimization/83965
14058 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
14059 that grouped statements are part of a reduction chain. Return
14060 true if the statement is not marked as a reduction itself but
14061 is part of a group.
14062 (vect_recog_dot_prod_pattern): Don't check whether the statement
14063 is part of a group here.
14064 (vect_recog_sad_pattern): Likewise.
14065 (vect_recog_widen_sum_pattern): Likewise.
14067 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
14070 * final.c (rest_of_clean_state): Also look for calls inside sequences.
14072 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
14075 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
14077 (ix86_output_indirect_function_return): New prototype.
14078 (ix86_split_simple_return_pop_internal): Likewise.
14079 * config/i386/i386.c (indirect_return_via_cx): New.
14080 (indirect_return_via_cx_bnd): Likewise.
14081 (indirect_thunk_name): Handle return va CX_REG.
14082 (output_indirect_thunk_function): Create alias for
14083 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
14084 (ix86_output_indirect_jmp): Remove the bool argument.
14085 (ix86_output_indirect_function_return): New function.
14086 (ix86_split_simple_return_pop_internal): Likewise.
14087 * config/i386/i386.md (*indirect_jump): Don't pass false
14088 to ix86_output_indirect_jmp.
14089 (*tablejump_1): Likewise.
14090 (simple_return_pop_internal): Change it to define_insn_and_split.
14091 Call ix86_split_simple_return_pop_internal to split it for
14092 -mfunction-return=.
14093 (simple_return_indirect_internal): Call
14094 ix86_output_indirect_function_return instead of
14095 ix86_output_indirect_jmp.
14097 2018-02-26 Jakub Jelinek <jakub@redhat.com>
14100 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
14101 memset and value initialization afterwards.
14103 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
14105 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
14107 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
14110 * common/config/aarch64/aarch64-common.c
14111 (aarch_option_optimization_table[]): Switch
14112 off fomit-frame-pointer
14114 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
14115 Chung-Ju Wu <jasonwucj@gmail.com>
14117 * config/nds32/nds32-multiple.md (load_multiple): Disallow
14119 (store_multiple): Ditto.
14121 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
14123 * config.gcc: Add --with-cpu support for nds32 target.
14124 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
14125 * config/nds32/nds32.opt: Add -mcpu= option.
14127 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
14129 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
14130 isel=yes): Warn for these deprecated options.
14132 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
14134 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
14135 ISA_2_5_MASKS_EMBEDDED.
14137 2018-02-23 Jakub Jelinek <jakub@redhat.com>
14139 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
14140 p->max as pointers rather than using iterative_hash_expr.
14142 2018-02-23 Carl Love <cel@us.ibm.com>
14144 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
14145 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
14147 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
14148 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
14149 P8V_BUILTIN_VEC_VUNSIGNED2.
14151 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
14154 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
14155 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
14157 * lra-constraints.c (curr_insn_transform): Set up
14158 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
14160 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
14161 * lra-eliminations.c (spill_pseudos): Ditto.
14162 (process_insn_for_elimination): Ditto.
14163 * lra-lives.c (reg_early_clobber_p): Use the new macros.
14164 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
14165 LRA_NON_CLOBBERED_ALT.
14167 2018-02-22 Martin Sebor <msebor@redhat.com>
14169 PR tree-optimization/84480
14170 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
14171 to maybe_diag_stxncpy_trunc. Call it.
14172 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
14173 from gimple_fold_builtin_strcpy. Print inlining stack.
14174 (handle_builtin_stxncpy): Print inlining stack.
14175 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
14177 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
14180 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
14181 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
14182 and -fcheck-pointer-bounds are used together.
14183 (indirect_thunk_prefix): New enum.
14184 (indirect_thunk_need_prefix): New function.
14185 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
14186 "_nt" instead of "_bnd" for NOTRACK prefix.
14187 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
14188 (output_indirect_thunk_function): Likewise.
14190 (ix86_code_end): Update output_indirect_thunk_function calls.
14191 (ix86_output_indirect_branch_via_reg): Replace
14192 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
14193 (ix86_output_indirect_branch_via_push): Likewise.
14194 (ix86_output_function_return): Likewise.
14195 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
14196 incompatible with -fcf-protection=branch and
14197 -fcheck-pointer-bounds.
14199 2018-02-22 Steve Ellcey <sellcey@cavium.com>
14202 * config/aarch64/aarch64.c (aarch64_print_address_internal):
14203 Change gcc_assert call to output_operand_lossage.
14205 2018-02-22 Steve Ellcey <sellcey@cavium.com>
14207 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
14209 2018-02-22 DJ Delorie <dj@redhat.com>
14210 Sebastian Perta <sebastian.perta@renesas.com>
14211 Oleg Endo <olegendo@gcc.gnu.org>
14213 * config/rx/rx.c (rx_rtx_costs): New function.
14214 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
14216 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
14218 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
14220 2018-02-22 Martin Liska <mliska@suse.cz>
14223 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
14224 Add "native" as a possible value.
14226 2018-02-22 Martin Liska <mliska@suse.cz>
14229 * config/i386/i386.c (ix86_option_override_internal):
14230 Add "native" as a possible value for -march and -mtune.
14232 2018-02-22 Jakub Jelinek <jakub@redhat.com>
14235 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
14236 to all type variants.
14238 PR tree-optimization/84503
14239 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
14240 width as info->bitpos + info->bitsize - start.
14241 (merged_store_group::merge_overlapping): Simplify width computation.
14242 (check_no_overlap): New function.
14243 (imm_store_chain_info::try_coalesce_bswap): Compute expected
14244 start + width and last_order of the group, fail if check_no_overlap
14246 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
14247 to group if check_no_overlap fails.
14249 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
14251 * config/rs6000/altivec.md: Delete contraint arguments to
14252 define_expand, define_split, and define_peephole2, and in
14253 define_insn_and_split if always unused.
14254 * config/rs6000/darwin.md: Ditto.
14255 * config/rs6000/dfp.md: Ditto.
14256 * config/rs6000/rs6000.md: Ditto.
14257 * config/rs6000/sync.md: Ditto.
14258 * config/rs6000/vector.md: Ditto.
14259 * config/rs6000/vsx.md: Ditto.
14261 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
14263 * config/rs6000/altivec.md: Write output control strings as braced
14264 blocks instead of double-quoted strings.
14265 * config/rs6000/darwin.md: Ditto.
14266 * config/rs6000/rs6000.md: Ditto.
14267 * config/rs6000/vector.md: Ditto.
14268 * config/rs6000/vsx.md: Ditto.
14270 2018-02-21 Jason Merrill <jason@redhat.com>
14272 PR c++/84314 - ICE with templates and fastcall attribute.
14273 * attribs.c (build_type_attribute_qual_variant): Remove assert.
14275 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
14277 * ipa-cp.c (determine_versionability): Fix comment typos.
14279 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
14282 * ipa-cp.c (determine_versionability): Do not version functions caling
14285 2018-02-21 Martin Liska <mliska@suse.cz>
14288 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
14289 Add "native" as a possible value.
14290 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
14291 the macro when native cpu detection is available.
14293 2018-02-21 Martin Liska <mliska@suse.cz>
14296 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
14297 Add "native" as a possible value.
14298 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
14299 when native cpu detection is available.
14301 2018-02-21 Jakub Jelinek <jakub@redhat.com>
14302 Martin Sebor <msebor@redhat.com>
14304 PR tree-optimization/84478
14305 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
14307 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
14308 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
14309 support which is conservatively correct, for 2 only stay conservative
14310 for maxlen. Formatting and comment capitalization fixes. Add STRICT
14311 argument to the 2 argument get_range_strlen, adjust 6 arg
14312 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
14314 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
14315 (gimple_fold_builtin_strlen): Pass true as last argument to
14318 2018-02-20 Martin Sebor <msebor@redhat.com>
14320 PR middle-end/84095
14321 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
14322 (builtin_memref::set_base_and_offset): Same. Handle inner references.
14323 (builtin_memref::builtin_memref): Factor out parts into
14324 set_base_and_offset and call it.
14326 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
14328 PR middle-end/84406
14329 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
14330 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
14331 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
14332 search at the associated MODE_INT.
14334 2018-02-20 Jeff Law <law@redhat.com>
14336 PR middle-end/82123
14337 PR tree-optimization/81592
14338 PR middle-end/79257
14339 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
14340 for range data rather than using global data.
14341 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
14342 range data rather than using global data.
14343 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
14344 pass it to children as needed.
14345 (struct directive::fmtresult): Similarly.
14346 (struct directive::set_width): Similarly.
14347 (struct directive::set_precision): Similarly.
14348 (format_integer, format_directive, parse_directive): Similarly.
14349 (format_none): Accept unnamed vr_values parameter.
14350 (format_percent, format_floating, format_character): Similarly.
14351 (format_string, format_plain): Similarly.
14352 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
14353 the EVRP range analyzer for range data rather than using global data.
14354 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
14355 gimple-ssa-evrp-analyze.h
14356 (class sprintf_dom_walker): Add after_dom_children member function.
14357 Add evrp_range_analyzer member.
14358 (sprintf_dom_walker::before_dom_children): Call into the EVRP
14359 range analyzer as needed.
14360 (sprintf_dom_walker::after_dom_children): New member function.
14361 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
14363 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
14364 (evrp_range_analyzer::pop_to_marker): Likewise.
14366 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
14368 PR tree-optimization/84419
14369 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
14370 with the required type if its current type is compatible but
14373 2018-02-20 Jakub Jelinek <jakub@redhat.com>
14375 PR middle-end/82004
14376 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
14377 after vectorization.
14379 2018-02-20 Martin Liska <mliska@suse.cz>
14382 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
14383 possible values if we don't have a hint.
14385 2018-02-20 Martin Liska <mliska@suse.cz>
14389 * final.c (shorten_branches): Build align_tab array with one
14391 * opts.c (finish_options): Add alignment option limit check.
14392 (MAX_CODE_ALIGN): Likewise.
14393 (MAX_CODE_ALIGN_VALUE): Likewise.
14394 * doc/invoke.texi: Document maximum allowed option value for
14395 all -falign-* options.
14397 2018-02-19 Jakub Jelinek <jakub@redhat.com>
14400 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
14401 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
14402 * var-tracking.c (emit_note_insn_var_location): Remove all references
14403 to NOTE_INSN_CALL_ARG_LOCATION.
14404 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
14405 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
14406 Use copy_rtx_if_shared.
14407 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
14408 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
14409 (dwarf2out_var_location): Remove handling of
14410 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
14412 * final.c (final_scan_insn): Remove all references to
14413 NOTE_INSN_CALL_ARG_LOCATION.
14414 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
14415 before dumping final insns.
14416 * except.c (emit_note_eh_region_end): Remove all references to
14417 NOTE_INSN_CALL_ARG_LOCATION.
14418 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
14419 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
14420 * config/arc/arc.c (hwloop_optimize): Likewise.
14421 * config/arm/arm.c (create_fix_barrier): Likewise.
14422 * config/s390/s390.c (s390_chunkify_start): Likewise.
14423 * config/sh/sh.c (find_barrier): Likewise.
14424 * config/i386/i386.c (rest_of_insert_endbranch,
14425 ix86_seh_fixup_eh_fallthru): Likewise.
14426 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
14427 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
14428 * config/frv/frv.c (frv_function_prologue): Likewise.
14429 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
14431 (note_outside_basic_block_p): Remove all references to
14432 NOTE_INSN_CALL_ARG_LOCATION.
14433 * gengtype.c (adjust_field_rtx_def): Likewise.
14434 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
14436 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
14437 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
14440 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
14443 PR tree-optimization/84452
14444 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
14445 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
14448 2018-02-19 Martin Liska <mliska@suse.cz>
14451 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
14453 2018-02-19 Martin Liska <mliska@suse.cz>
14454 Richard Sandiford <richard.sandiford@linaro.org>
14456 PR tree-optimization/82491
14457 * gimple-fold.c (get_base_constructor): Make earlier bail out
14460 2018-02-19 Carl Love <cel@us.ibm.com>
14462 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
14463 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
14465 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
14466 P8V_BUILTIN_VEC_NEG.
14468 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
14470 * config/rl78/rl78.md (movdf): New define expand.
14472 2018-02-19 Martin Liska <mliska@suse.cz>
14475 * doc/invoke.texi: Fix typo.
14476 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
14478 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
14480 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
14481 handle rs6000_single_float and rs6000_double_float specially for
14484 2018-02-16 Jeff Law <law@redhat.com>
14486 * config/rx/rx.c (add_pop_cfi_notes): New function.;
14487 (pop_regs): Use it.
14489 2018-02-16 Jakub Jelinek <jakub@redhat.com>
14492 * ipa-inline.c (inline_small_functions): Fix a typo.
14494 2018-02-16 Nathan Sidwell <nathan@acm.org>
14496 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
14498 2018-02-16 Carl Love <cel@us.ibm.com>
14500 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
14501 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
14502 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
14503 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
14504 expansion to P8V_BUILTIN_VEC_FLOAT2.
14506 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
14508 PR rtl-optimization/70023
14509 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
14510 src_regno into account.
14512 2018-02-16 Carl Love <cel@us.ibm.com>
14514 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
14515 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
14516 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
14517 * config/rs6000/rs6000.c: Remove case statements for
14518 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
14519 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
14520 and P9V_BUILTIN_VEC_VINSERT4B.
14521 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
14522 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
14523 * config/rs6000/vsx.md:
14524 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
14527 2018-02-16 Carl Love <cel@us.ibm.com>
14529 * config/rs6000/altivec.h: Add builtin names vec_extract4b
14531 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
14533 * config/rs6000/rs6000-c.c: Add the definitions for
14534 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
14535 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
14536 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
14537 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
14538 definition for insert4b and define insn *insert3b_internal.
14539 * doc/extend.texi: Add documentation for vec_extract4b.
14541 2018-02-16 Nathan Sidwell <nathan@acm.org>
14543 * doc/extend.texi (Backwards Compatibility): Mention friend
14544 injection. Note for-scope is deprecated.
14545 * doc/invoke.texi (-ffriend-injection): Deprecate.
14547 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
14549 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
14550 that moved to I2, also allow destinations that are a paradoxical
14551 subreg (instead of a normal reg).
14553 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
14556 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
14559 2018-02-16 Richard Biener <rguenther@suse.de>
14561 PR tree-optimization/84037
14562 PR tree-optimization/84016
14564 * config/i386/i386.c (ix86_builtin_vectorization_cost):
14565 Adjust vec_construct for the fact we need additional higher latency
14566 128bit inserts for AVX256 and AVX512 vector builds.
14567 (ix86_add_stmt_cost): Scale vector construction cost for
14570 2018-02-16 Richard Biener <rguenther@suse.de>
14572 PR tree-optimization/84417
14573 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
14574 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
14575 (non_rewritable_lvalue_p): Likewise, use poly-ints.
14577 2018-02-16 Martin Liska <mliska@suse.cz>
14580 * internal-fn.def (ASAN_CHECK): Set proper flags.
14581 (ASAN_MARK): Likewise.
14583 2018-02-16 Julia Koval <julia.koval@intel.com>
14585 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
14586 from PTA_CANNONLAKE.
14588 2018-02-16 Jakub Jelinek <jakub@redhat.com>
14591 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
14592 Use ++iter rather than iter++ for std::list iterators.
14593 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
14594 defer deleting them until all nodes in the forest are processed. Do
14595 free even leaf nodes. Change to_process into auto_vec.
14598 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
14599 * vec.h (vec_default_construct): Use memset instead of placement new
14600 if BROKEN_VALUE_INITIALIZATION is defined.
14601 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
14602 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
14605 PR rtl-optimization/83723
14606 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
14607 * lra.c (lra_substitute_pseudo): Likewise. If true, use
14608 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
14610 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
14612 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
14614 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
14616 PR rtl-optimization/81443
14617 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
14618 from inner REGs to paradoxical SUBREGs.
14620 2018-02-16 Richard Biener <rguenther@suse.de>
14622 PR tree-optimization/84399
14623 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
14624 For operands we can analyze at their definition make sure we can
14625 analyze them at each use as well.
14627 2018-02-16 Richard Biener <rguenther@suse.de>
14629 PR tree-optimization/84190
14630 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
14631 volatile accesses if the decl isn't volatile.
14633 2018-02-15 Jason Merrill <jason@redhat.com>
14635 PR c++/84314 - ICE with templates and fastcall attribute.
14636 * attribs.c (build_type_attribute_qual_variant): Don't clobber
14637 TYPE_CANONICAL on an existing type.
14639 2018-02-15 Jakub Jelinek <jakub@redhat.com>
14641 PR tree-optimization/84383
14642 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
14643 dstoff nor call operand_equal_p if dstbase is NULL.
14645 PR tree-optimization/84334
14646 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
14647 also a CONSTANT_CLASS_P, punt.
14649 2018-02-14 Jim Wilson <jimw@sifive.com>
14651 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
14652 first SMALL_OPERAND check. New local min_second_step. Move assert
14653 to where locals are set. Add TARGET_RVC support.
14654 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
14656 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
14658 * doc/invoke.texi: Correct -Wformat-overflow code sample.
14660 2018-02-14 Martin Sebor <msebor@redhat.com>
14662 PR tree-optimization/83698
14663 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
14664 arrays constrain the offset range to their bounds.
14665 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
14666 (builtin_access::overlap): Avoid setting the size of overlap if it's
14668 (maybe_diag_overlap): Also consider arrays when deciding what values
14669 of offsets to include in diagnostics.
14671 2018-02-14 Martin Sebor <msebor@redhat.com>
14674 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
14675 that correspond to the kind of a declaration.
14677 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
14680 * config/pa/pa.md: Load address of PIC label using the linkage table
14681 if the label is nonlocal.
14683 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
14685 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
14686 warning message if user requests -maltivec=be.
14687 * doc/invoke.texi: Document deprecation of -maltivec=be.
14689 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
14692 * config/rs6000/rs6000-c.c: Update definitions for
14693 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
14694 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
14696 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
14699 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
14700 add _get_ssp intrinsics. Remove argument from
14701 __builtin_ia32_rdssp[d|q].
14702 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
14703 * config/i386/i386-builtin.def: Remove argument from
14704 __builtin_ia32_rdssp[d|q].
14705 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
14706 ix86_expand_special_args_builtin for _rdssp[d|q].
14707 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
14708 Clear register before usage.
14709 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
14710 Add documentation for new _get_ssp and _inc_ssp intrinsics.
14712 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
14714 PR tree-optimization/84357
14715 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
14716 operand 1 of an ARRAY_REF too.
14718 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
14721 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
14722 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
14724 (set_of_reg): New struct.
14725 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
14726 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
14727 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
14729 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
14730 Split into bitclr, bitset, bitinvert patterns if appropriate.
14731 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
14732 use rx_fuse_in_memory_bitop.
14733 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
14734 to named insn, correct maximum insn length.
14736 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
14739 * machmode.def: Define a complex mode for PARTIAL_INT.
14740 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
14742 * doc/rtl.texi: Document CSPImode.
14743 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
14745 (msp430_hard_regno_nregs_with_padding): Likewise.
14747 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
14750 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
14752 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
14754 PR rtl-optimization/84169
14755 * combine.c (try_combine): New variable split_i2i3. Set it to true if
14756 we generated a parallel as new i3 and we split that to new i2 and i3
14757 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
14758 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
14759 those to i2, not i1. Partially rewrite this scan code.
14761 2018-02-13 Jakub Jelinek <jakub@redhat.com>
14764 * stor-layout.c (place_field): For variable length fields, adjust
14765 offset_align afterwards not just based on the field's alignment,
14766 but also on the size.
14768 PR middle-end/84309
14769 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
14770 of exps and logs in the use_exp2 case.
14772 2018-02-13 Jeff Law <law@redhat.com>
14774 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
14775 entry for "vector".
14777 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
14780 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
14784 * common.opt (gas-loc-support, gas-locview-support): New.
14785 (ginline-points, ginternal-reset-location-views): New.
14786 * doc/invoke.texi: Document them. Use @itemx where intended.
14787 (gvariable-location-views): Adjust.
14788 * target.def (reset_location_view): New.
14789 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
14790 (TARGET_RESET_LOCATION_VIEW): New.
14791 * doc/tm.texi: Rebuilt.
14792 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
14793 (dwarf2out_default_as_locview_support): New.
14794 (output_asm_line_debug_info): Use option variables.
14795 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
14796 (output_loc_list): Likewise.
14797 (add_high_low_attributes): Check option variables.
14798 Don't output entry view attribute in strict mode.
14799 (gen_inlined_subroutine_die): Check option variables.
14800 (dwarf2out_inline_entry): Likewise.
14801 (init_sections_and_labels): Likewise.
14802 (dwarf2out_early_finish): Likewise.
14803 (maybe_reset_location_view): New, from...
14804 (dwarf2out_var_location): ... here. Call it.
14805 * debug.h (dwarf2out_default_as_loc_support): Declare.
14806 (dwarf2out_default_as_locview_support): Declare.
14807 * hooks.c (hook_int_rtx_insn_0): New.
14808 * hooks.h (hook_int_rtx_insn_0): Declare.
14809 * toplev.c (process_options): Take -gas-loc-support and
14810 -gas-locview-support from dwarf2out. Enable
14811 -gvariable-location-views by default only with locview
14812 assembler support. Enable -ginternal-reset-location-views by
14813 default only if the target defines the corresponding hook.
14814 Enable -ginline-points by default if location views are
14815 enabled; force it disabled if statement frontiers are
14817 * tree-inline.c (expand_call_inline): Check option variables.
14818 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
14820 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
14822 PR tree-optimization/84321
14823 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
14824 handling. Also check whether the anti-range contains any values
14825 that satisfy the mask; switch to a VR_RANGE if not.
14827 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
14830 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
14832 2018-02-13 Martin Jambor <mjambor@suse.cz>
14835 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
14836 of call statements, also set location of a load to a temporary.
14838 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
14840 * config/rl78/rl78.c (add_vector_labels): New function.
14841 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
14842 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
14843 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
14844 which checks that no arguments are passed.
14845 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
14846 * doc/extend.texi: Documentation for the new attribute.
14848 2018-02-13 Andreas Schwab <schwab@suse.de>
14850 * config/riscv/linux.h (CPP_SPEC): Define.
14852 2018-02-13 Jakub Jelinek <jakub@redhat.com>
14855 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
14856 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
14857 OPTION_MASK_ISA_AES as first argument to def_builtin_const
14858 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
14859 instead of OPTION_MASK_ISA_PCLMUL as first argument to
14860 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
14861 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
14862 temporarily for AES and PCLMUL builtins.
14864 PR tree-optimization/84339
14865 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
14866 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
14869 PR middle-end/84309
14870 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
14871 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
14872 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
14874 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
14876 * omp-simd-clone.h: New file.
14877 * omp-simd-clone.c: Include omp-simd-clone.h.
14878 (expand_simd_clones): No longer static.
14879 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
14880 cgraph.h and omp-simd-clone.h.
14881 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
14882 (vect_recog_widen_shift_pattern): Formatting fix.
14883 (vect_pattern_recog_1): Don't check optab for calls.
14886 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
14887 operands[2] into a REG before using gen_lowpart on it.
14889 2018-02-12 Jeff Law <law@redhat.com>
14892 * config/sh/sh.c (find_barrier): Consider a sibling call
14895 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
14896 successfully back substituting a reg.
14898 2018-02-12 Richard Biener <rguenther@suse.de>
14900 PR tree-optimization/84037
14901 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
14902 parameter, move visited init to caller.
14903 (vect_slp_analyze_operations): Separate cost from validity
14904 check, initialize visited once for all instances.
14905 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
14907 * tree-vect-stmts.c (vect_model_simple_cost): Make early
14909 (vect_model_promotion_demotion_cost): Likewise.
14910 (vectorizable_bswap): Guard cost modeling with !slp_node
14911 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
14913 (vectorizable_call): Likewise.
14914 (vectorizable_conversion): Likewise.
14915 (vectorizable_assignment): Likewise.
14916 (vectorizable_shift): Likewise.
14917 (vectorizable_operation): Likewise.
14918 (vectorizable_store): Likewise.
14919 (vectorizable_load): Likewise.
14920 (vectorizable_condition): Likewise.
14921 (vectorizable_comparison): Likewise.
14923 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
14926 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
14927 (ASAN_MARK): Fix fnspec to account for return value, change pointer
14928 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
14930 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
14932 PR middle-end/83665
14933 * params.def (inline-min-speedup): Increase from 8 to 15.
14934 (max-inline-insns-auto): Decrease from 40 to 30.
14935 * ipa-split.c (consider_split): Add some buffer for function to
14936 be considered inlining candidate.
14937 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
14940 2018-02-12 Richard Biener <rguenther@suse.de>
14942 PR tree-optimization/84037
14943 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
14944 matched stmts if we cannot swap the non-matched ones.
14946 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
14948 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
14949 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
14950 _mm_maskz_scalef_round_ss): New intrinsics.
14951 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
14952 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
14953 __builtin_ia32_scalefss_round): Remove.
14954 (__builtin_ia32_scalefsd_mask_round,
14955 __builtin_ia32_scalefss_mask_round): New intrinsics.
14956 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
14957 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
14958 ((match_operand:VF_128 2 "<round_nimm_predicate>"
14959 "<round_constraint>")): Changed to ...
14960 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
14961 "<round_scalar_constraint>")): ... this.
14962 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
14963 %0, %1, %2<round_op3>}"): Changed to ...
14964 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
14965 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
14966 %2<round_scalar_mask_op3>}"): ... this.
14967 * config/i386/subst.md (round_scalar_nimm_predicate): New.
14969 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
14971 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
14972 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
14973 (_mm_maskz_sqrt_round_ss): New intrinsics.
14974 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
14975 (__builtin_ia32_sqrtsd_mask_round)
14976 (__builtin_ia32_sqrtss_mask_round): New builtins.
14977 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
14978 (__builtin_ia32_sqrtss_round): Remove.
14979 (__builtin_ia32_sqrtsd_mask_round)
14980 (__builtin_ia32_sqrtss_mask_round): New builtins.
14981 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
14982 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
14983 ((match_operand:VF_128 1 "vector_operand"
14984 "xBm,<round_constraint>")): Changed to ...
14985 ((match_operand:VF_128 1 "vector_operand"
14986 "xBm,<round_scalar_constraint>")): ... this.
14987 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
14988 %0, %2, %<iptr>1<round_op3>}): Changed to ...
14989 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
14990 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
14991 %<iptr>1<round_scalar_mask_op3>}): ... this.
14992 ((set_attr "prefix" "<round_prefix>")): Changed to ...
14993 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
14995 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
14998 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
14999 Cast vec_cmpeq result to correct type.
15000 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
15001 Cast vec_cmpgt result to correct type.
15003 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
15005 * final.c (final_scan_insn_1): Renamed from...
15006 (final_scan_insn): ... this. New wrapper, to recover
15007 seen from the outermost call in recursive ones.
15008 * config/sparc/sparc.c (output_return): Drop seen from call.
15009 (output_sibcall): Likewise.
15010 * config/visium/visium.c (output_branch): Likewise.
15012 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
15014 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
15017 2018-02-10 Alan Modra <amodra@gmail.com>
15020 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
15021 Specify LR as an input.
15023 2018-02-10 Jakub Jelinek <jakub@redhat.com>
15026 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
15027 remove_member_access_dummy_vars): New functions.
15028 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
15029 lower_omp_1, execute_lower_omp): Use them.
15031 PR rtl-optimization/84308
15032 * shrink-wrap.c (spread_components): Release todo vector.
15034 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
15036 PR rtl-optimization/57193
15037 * ira-color.c (struct allocno_color_data): Add member
15038 conflict_allocno_hard_prefs.
15039 (update_conflict_allocno_hard_prefs): New.
15040 (bucket_allocno_compare_func): Add a preference based on
15041 conflict_allocno_hard_prefs.
15042 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
15043 (color_allocnos): Remove a dead code. Initiate
15044 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
15046 2018-02-09 Jakub Jelinek <jakub@redhat.com>
15049 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
15050 constraint from =wa to wa. Avoid a subreg on the output operand,
15051 instead use a pseudo and subreg it in a move.
15052 (p9_xxbrd_<mode>): Changed to ...
15053 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
15054 (p9_xxbrd_v2df): New expander.
15055 (p9_xxbrw_<mode>): Changed to ...
15056 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
15057 (p9_xxbrw_v4sf): New expander.
15059 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
15061 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
15063 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
15066 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
15067 multiply in 32-bit mode.
15068 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
15069 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
15072 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
15074 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
15075 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
15076 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
15077 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
15079 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
15082 * dwarf2out.c (is_trivial_indirect_ref): New function.
15083 (dwarf2out_late_global_decl): Do not generate a location
15084 attribute for variables that have a non-trivial DECL_VALUE_EXPR
15085 and that are not defined in the current unit.
15087 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
15089 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
15090 instead of a libcall for UNORDERED.
15092 2018-02-09 Tamar Christina <tamar.christina@arm.com>
15095 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
15096 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
15098 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
15101 * config/s390/s390.c (s390_set_current_function): Invoke
15102 s390_indirect_branch_settings also if fndecl didn't change.
15104 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
15106 * config/rs6000/rs6000.md (blockage): Set length to zero.
15108 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
15110 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
15112 2018-02-09 Jakub Jelinek <jakub@redhat.com>
15115 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
15116 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
15120 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
15121 PARALLEL incoming that failed vt_get_decl_and_offset check.
15123 PR middle-end/84237
15124 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
15125 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
15127 (get_variable_section): For decls in named .bss* sections pass true as
15128 second argument to bss_initializer_p.
15130 2018-02-09 Marek Polacek <polacek@redhat.com>
15131 Jakub Jelinek <jakub@redhat.com>
15134 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
15135 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
15136 Sync some changes from cxx_fold_indirect_ref.
15138 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
15140 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
15142 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
15143 (BLOCK_INLINE_ENTRY_LABEL): New.
15144 (dwarf2out_var_location): Disregard inline entry markers.
15145 (inline_entry_data): New struct.
15146 (inline_entry_data_hasher): New hashtable type.
15147 (inline_entry_data_hasher::hash): New.
15148 (inline_entry_data_hasher::equal): New.
15149 (inline_entry_data_table): New variable.
15150 (add_high_low_attributes): Add DW_AT_entry_pc and
15151 DW_AT_GNU_entry_view attributes if a pending entry is found
15152 in inline_entry_data_table. Add old entry_pc attribute only
15153 if debug nonbinding markers are disabled.
15154 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
15155 markers are enabled.
15156 (block_within_block_p, dwarf2out_inline_entry): New.
15157 (dwarf2out_finish): Check that no entries remained in
15158 inline_entry_data_table.
15159 * final.c (reemit_insn_block_notes): Handle inline entry notes.
15160 (final_scan_insn, notice_source_line): Likewise.
15161 (rest_of_clean_state): Skip inline entry markers.
15162 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
15164 * gimple.c (gimple_build_debug_inline_entry): New.
15165 * gimple.h (enum gimple_debug_subcode): Add
15166 GIMPLE_DEBUG_INLINE_ENTRY.
15167 (gimple_build_debug_inline_entry): Declare.
15168 (gimple_debug_inline_entry_p): New.
15169 (gimple_debug_nonbind_marker_p): Adjust.
15170 * insn-notes.def (INLINE_ENTRY): New.
15171 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
15172 inline entry marker notes.
15173 (print_insn): Likewise.
15174 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
15175 (INSN_DEBUG_MARKER_KIND): Likewise.
15176 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
15177 * tree-inline.c (expand_call_inline): Build and insert
15178 debug_inline_entry stmt.
15179 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
15180 inline entry blocks early, if nonbind markers are enabled.
15181 (dump_scope_block): Dump fragment info.
15182 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
15183 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
15184 (gimple_build_debug_inline_entry): New.
15185 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
15186 Enable/disable inline entry points too.
15187 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
15188 (DEBUG_INSN): Describe inline entry markers.
15190 * common.opt (gvariable-location-views): New.
15191 (gvariable-location-views=incompat5): New.
15192 * config.in: Rebuilt.
15193 * configure: Rebuilt.
15194 * configure.ac: Test assembler for view support.
15195 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
15196 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
15197 * dwarf2out.c (var_loc_view): New typedef.
15198 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
15199 (dwarf2out_locviews_in_attribute): New.
15200 (dwarf2out_locviews_in_loclist): New.
15201 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
15202 (enum dw_line_info_opcode): Add LI_adv_address.
15203 (struct dw_line_info_table): Add view.
15204 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
15205 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
15206 (zero_view_p): New variable.
15207 (ZERO_VIEW_P): New macro.
15208 (output_asm_line_debug_info): New.
15209 (struct var_loc_node): Add view.
15210 (add_AT_view_list, AT_loc_list): New.
15211 (add_var_loc_to_decl): Add view param. Test it against last.
15212 (new_loc_list): Add view params. Record them.
15213 (AT_loc_list_ptr): Handle loc and view lists.
15214 (view_list_to_loc_list_val_node): New.
15215 (print_dw_val): Handle dw_val_class_view_list.
15216 (size_of_die): Likewise.
15217 (value_format): Likewise.
15218 (loc_list_has_views): New.
15219 (gen_llsym): Set vl_symbol too.
15220 (maybe_gen_llsym, skip_loc_list_entry): New.
15221 (dwarf2out_maybe_output_loclist_view_pair): New.
15222 (output_loc_list): Output view list or entries too.
15223 (output_view_list_offset): New.
15224 (output_die): Handle dw_val_class_view_list.
15225 (output_dwarf_version): New.
15226 (output_compilation_unit_header): Use it.
15227 (output_skeleton_debug_sections): Likewise.
15228 (output_rnglists, output_line_info): Likewise.
15229 (output_pubnames, output_aranges): Update version comments.
15230 (output_one_line_info_table): Output view numbers in asm comments.
15231 (dw_loc_list): Determine current endview, pass it to new_loc_list.
15232 Call maybe_gen_llsym.
15233 (loc_list_from_tree_1): Adjust.
15234 (add_AT_location_description): Create view list attribute if
15235 needed, check it's absent otherwise.
15236 (convert_cfa_to_fb_loc_list): Adjust.
15237 (maybe_emit_file): Call output_asm_line_debug_info for test.
15238 (dwarf2out_var_location): Reset views as needed. Precompute
15239 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
15240 attribute. Set view.
15241 (new_line_info_table): Reset next view.
15242 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
15243 (dwarf2out_source_line): Likewise. Output view resets and labels to
15244 the assembler, or select appropriate line info opcodes.
15245 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
15246 (optimize_string_length): Catch it. Adjust.
15247 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
15248 dw_val_class_view_list, and remove it if no longer needed.
15249 (hash_loc_list): Hash view numbers.
15250 (loc_list_hasher::equal): Compare them.
15251 (optimize_location_lists): Check whether a view list symbol is
15252 needed, and whether the locview attribute is present, and
15253 whether they match. Remove the locview attribute if no longer
15255 (index_location_lists): Call skip_loc_list_entry for test.
15256 (dwarf2out_finish): Call output_asm_line_debug_info for test.
15257 Use output_dwarf_version.
15258 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
15259 (struct dw_val_node): Add val_view_list.
15260 * final.c (SEEN_NEXT_VIEW): New.
15261 (set_next_view_needed): New.
15262 (clear_next_view_needed): New.
15263 (maybe_output_next_view): New.
15264 (final_start_function): Rename to...
15265 (final_start_function_1): ... this. Take pointer to FIRST,
15266 add SEEN parameter. Emit param bindings in the initial view.
15267 (final_start_function): Reintroduce SEEN-less interface.
15268 (final): Rename to...
15269 (final_1): ... this. Take SEEN parameter. Output final pending
15270 next view at the end.
15271 (final): Reintroduce seen-less interface.
15272 (final_scan_insn): Output pending next view before switching
15273 sections or ending a block. Mark the next view as needed when
15274 outputting variable locations. Notify debug backend of section
15275 changes, and of location view changes.
15276 (rest_of_handle_final): Adjust.
15277 * toplev.c (process_options): Autodetect value for debug variable
15278 location views option. Warn on incompat5 without -gdwarf-5.
15279 * doc/invoke.texi (gvariable-location-views): New.
15280 (gvariable-location-views=incompat5): New.
15281 (gno-variable-location-views): New.
15283 2018-02-08 David Malcolm <dmalcolm@redhat.com>
15285 PR tree-optimization/84136
15286 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
15287 that the result of find_edge is non-NULL.
15289 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
15292 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
15293 storing integer register in SImode. Fix cost of 256 and 512
15294 byte aligned SSE register store.
15296 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
15298 * config/i386/i386.c (ix86_multiplication_cost): Fix
15299 multiplication cost for TARGET_AVX512DQ.
15301 2018-02-08 Marek Polacek <polacek@redhat.com>
15303 PR tree-optimization/84238
15304 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
15307 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
15309 PR tree-optimization/84265
15310 * tree-vect-stmts.c (vectorizable_store): Don't treat
15311 VMAT_CONTIGUOUS accesses as grouped.
15312 (vectorizable_load): Likewise.
15314 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
15316 PR tree-optimization/81635
15317 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
15318 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
15319 (test_round_for_mask): New functions.
15320 (wide_int_cc_tests): Call test_round_for_mask.
15321 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
15322 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
15323 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
15324 range returned by get_range_info.
15326 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
15329 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
15330 * symtab.c: Include builtins.h
15331 (symtab_node::output_to_lto_symbol_table_p): Move here
15332 from lto-streamer-out.c:output_symbol_p.
15333 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
15334 (output_symbol_p): Move all logic to symtab.c
15335 (produce_symtab): Update.
15337 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
15339 * config/s390/s390-opts.h (enum indirect_branch): Define.
15340 * config/s390/s390-protos.h (s390_return_addr_from_memory)
15341 (s390_indirect_branch_via_thunk)
15342 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
15343 (enum s390_indirect_branch_type): Define.
15344 * config/s390/s390.c (struct s390_frame_layout, struct
15345 machine_function): Remove.
15346 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
15347 (indirect_branch_table_label_no, indirect_branch_table_name):
15349 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
15350 (enum s390_indirect_branch_option): Define.
15351 (s390_return_addr_from_memory): New function.
15352 (s390_handle_string_attribute): New function.
15353 (s390_attribute_table): Add new attribute handler.
15354 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
15355 (s390_indirect_branch_via_thunk): New function.
15356 (s390_indirect_branch_via_inline_thunk): New function.
15357 (s390_function_ok_for_sibcall): When jumping via thunk disallow
15358 sibling call optimization for non z10 compiles.
15359 (s390_emit_call): Force indirect branch target to be a single
15360 register. Add r1 clobber for non-z10 compiles.
15361 (s390_emit_epilogue): Emit return jump via return_use expander.
15362 (s390_reorg): Handle JUMP_INSNs as execute targets.
15363 (s390_option_override_internal): Perform validity checks for the
15364 new command line options.
15365 (s390_indirect_branch_attrvalue): New function.
15366 (s390_indirect_branch_settings): New function.
15367 (s390_set_current_function): Invoke s390_indirect_branch_settings.
15368 (s390_output_indirect_thunk_function): New function.
15369 (s390_code_end): Implement target hook.
15370 (s390_case_values_threshold): Implement target hook.
15371 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
15373 * config/s390/s390.h (struct s390_frame_layout)
15374 (struct machine_function): Move here from s390.c.
15375 (TARGET_INDIRECT_BRANCH_NOBP_RET)
15376 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
15377 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
15378 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
15379 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
15380 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
15381 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
15382 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
15383 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
15384 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
15385 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
15386 (mnemonic attribute): Add values which aren't recognized
15388 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
15389 pattern for branch conversion. Fix mnemonic attribute.
15390 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
15391 indirect branch via thunk if requested.
15392 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
15393 ("*indirect_jump"): Disable for branch conversion using out of
15395 ("indirect_jump_via_thunk<mode>_z10")
15396 ("indirect_jump_via_thunk<mode>")
15397 ("indirect_jump_via_inlinethunk<mode>_z10")
15398 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
15399 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
15400 ("casesi_jump_via_inlinethunk<mode>_z10")
15401 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
15402 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
15403 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
15404 ("*indirect2_jump"): Disable for branch conversion.
15405 ("casesi_jump"): Turn into expander and expand patterns for branch
15407 ("return_use"): New expander.
15408 ("*return"): Emit return via thunk and rename it to ...
15409 ("*return<mode>"): ... this one.
15410 * config/s390/s390.opt: Add new options and and enum for the
15413 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
15415 * lra-constraints.c (match_reload): Unconditionally use
15416 gen_lowpart_SUBREG, rather than selecting between that
15417 and equivalent gen_rtx_SUBREG code.
15419 2018-02-08 Richard Biener <rguenther@suse.de>
15421 PR tree-optimization/84233
15422 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
15423 changed flag instead of boguously re-using phi_inserted.
15425 2018-02-08 Martin Jambor <mjambor@suse.cz>
15427 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
15428 static local variables.
15430 2018-02-08 Richard Biener <rguenther@suse.de>
15432 PR tree-optimization/84278
15433 * tree-vect-stmts.c (vectorizable_store): When looking for
15434 smaller vector types to perform grouped strided loads/stores
15435 make sure the mode is supported by the target.
15436 (vectorizable_load): Likewise.
15438 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
15440 * config/aarch64/aarch64.c (aarch64_components_for_bb):
15441 Increase LDP/STP opportunities by adding adjacent callee-saves.
15443 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
15445 PR rtl-optimization/84068
15446 PR rtl-optimization/83459
15447 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
15449 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
15451 PR tree-optimization/84224
15452 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
15453 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
15454 non-zero arguments.
15456 2018-02-07 Iain Sandoe <iain@codesourcery.com>
15459 * config/rs6000/altivec.md (*restore_world): Remove LR use.
15460 * config/rs6000/predicates.md (restore_world_operation): Adjust op
15461 count, remove one USE.
15463 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
15465 * doc/install.texi (Configuration): Document the
15466 --with-long-double-format={ibm,ieee} PowerPC configuration
15470 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
15471 Convert from define_expand to be define_insn_and_split. Rework
15472 float/double/_Float128 conversions to QI/HI/SImode to work with
15473 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
15474 conversions to QI/HImode types did a store and then a load to
15475 truncate the value. For conversions to VSX registers, don't split
15476 the insn, instead emit the code directly. Use the code iterator
15477 any_fix to combine signed and unsigned conversions.
15478 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
15479 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
15480 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
15481 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
15482 (fix_<mode>di2_hw): Likewise.
15483 (fixuns_<mode>di2_hw): Likewise.
15484 (fix_<mode>si2_hw): Likewise.
15485 (fixuns_<mode>si2_hw): Likewise.
15486 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
15487 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
15488 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
15489 fix<uns>_trunc<SFDF:mode>si2_p8.
15490 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
15491 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
15492 (fix<uns>_<mode>_mem): Likewise.
15493 (fctiw<u>z_<mode>_mem): Likewise.
15494 (fix<uns>_<mode>_mem): Likewise.
15495 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
15496 the register allocator from doing a direct move to the GPRs to do
15497 a store, and instead use the ISA 3.0 store byte/half-word from
15498 vector register instruction. For IEEE 128-bit floating point,
15499 also optimize stores of 32-bit ints.
15500 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
15502 2018-02-07 Alan Hayward <alan.hayward@arm.com>
15504 * genextract.c (push_pathstr_operand): New function to support
15506 (walk_rtx): Call push_pathstr_operand.
15507 (print_path): Support [a-zA-Z].
15509 2018-02-07 Richard Biener <rguenther@suse.de>
15511 PR tree-optimization/84037
15512 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
15513 (cse_and_gimplify_to_preheader): Declare.
15514 (vect_get_place_in_interleaving_chain): Likewise.
15515 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15517 (_loop_vec_info::~_loop_vec_info): Delete it.
15518 (cse_and_gimplify_to_preheader): New function.
15519 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
15520 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
15521 (vectorizable_load): Likewise. For grouped stores always base
15522 the IV on the first element.
15523 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
15524 condition before gimplifying.
15526 2018-02-07 Jakub Jelinek <jakub@redhat.com>
15528 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
15529 *DIV_EXPR and *MOD_EXPR.
15531 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
15534 * config/i386/i386.c (ix86_option_override_internal): Mask out
15535 the CF_SET bit when checking -fcf-protection.
15537 2018-02-07 Tom de Vries <tom@codesourcery.com>
15540 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
15543 2018-02-07 Richard Biener <rguenther@suse.de>
15545 PR tree-optimization/84204
15546 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
15549 PR tree-optimization/84205
15550 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
15551 special-case isl_ast_op_zdiv_r.
15553 PR tree-optimization/84223
15554 * graphite-scop-detection.c (gather_bbs::before_dom_children):
15555 Only add conditions from within the region.
15556 (gather_bbs::after_dom_children): Adjust.
15558 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
15561 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
15562 * config/avr/avr.md: Only post-reload split REG-REG moves if
15563 either register is GENERAL_REG_P.
15565 2018-02-07 Jakub Jelinek <jakub@redhat.com>
15567 PR tree-optimization/84235
15568 * tree-ssa-scopedtables.c
15569 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
15570 if the subtraction is performed in floating point type where NaNs are
15571 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
15572 build 1. Formatting fix.
15574 2018-02-06 Jakub Jelinek <jakub@redhat.com>
15577 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
15578 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
15579 and skip it regardless of bb boundaries. Use CALL_P macro,
15580 don't test INSN_P (insn) together with CALL_P or JUMP_P check
15581 unnecessarily, formatting fix.
15583 2018-02-06 Michael Collison <michael.collison@arm.com>
15585 * config/arm/thumb2.md:
15586 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
15587 (*thumb_mov_notscc): Ditto.
15589 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
15592 * config/rs6000/rs6000.md (su code attribute): Use "u" for
15593 unsigned_fix, not "s".
15595 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15597 * configure.ac (gcc_fn_eh_frame_ro): New function.
15598 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
15599 correct .eh_frame permissions.
15600 * configure: Regenerate.
15602 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
15604 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
15605 irrelevant options.
15607 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15609 * config/rs6000/rs6000.c (rs6000_option_override_internal):
15610 Display warning message for -mno-speculate-indirect-jumps.
15612 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
15614 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
15616 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
15618 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
15620 PR tree-optimization/84225
15621 * tree-eh.c (find_trapping_overflow): Only call
15622 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
15624 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
15627 * config/i386/i386.c: Reimplement the check of possible options
15628 -mibt/-mshstk conbination. Change error messages.
15629 * doc/invoke.texi: Fix a typo: remove extra '='.
15631 2018-02-06 Marek Polacek <polacek@redhat.com>
15633 PR tree-optimization/84228
15634 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
15636 2018-02-06 Tamar Christina <tamar.christina@arm.com>
15639 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
15640 emitted arch directives.
15641 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
15642 __ARM_FEATURE_COPROC before changing architectures.
15644 2018-02-06 Richard Biener <rguenther@suse.de>
15646 * config/i386/i386.c (print_reg): Fix typo.
15647 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
15649 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
15651 * configure: Regenerate.
15653 2018-02-05 Martin Sebor <msebor@redhat.com>
15655 PR tree-optimization/83369
15656 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
15659 2018-02-05 Martin Liska <mliska@suse.cz>
15661 * doc/invoke.texi: Cherry-pick upstream r323995.
15663 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
15665 * ira.c (ira_init_register_move_cost): Adjust comment.
15667 2018-02-05 Martin Liska <mliska@suse.cz>
15669 PR gcov-profile/84137
15670 * doc/gcov.texi: Fix typo in documentation.
15672 2018-02-05 Martin Liska <mliska@suse.cz>
15674 PR gcov-profile/83879
15675 * doc/gcov.texi: Document necessity of --dynamic-list-data when
15676 using dlopen functionality.
15678 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
15680 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
15681 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
15682 _mm_maskz_range_ss, _mm_mask_range_round_ss,
15683 _mm_maskz_range_round_ss): New intrinsics.
15684 (__builtin_ia32_rangesd128_round)
15685 (__builtin_ia32_rangess128_round): Remove.
15686 (__builtin_ia32_rangesd128_mask_round,
15687 __builtin_ia32_rangess128_mask_round): New builtins.
15688 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
15689 __builtin_ia32_rangess128_round): Remove.
15690 (__builtin_ia32_rangesd128_mask_round,
15691 __builtin_ia32_rangess128_mask_round): New builtins.
15692 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
15693 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
15694 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
15695 "<round_saeonly_constraint>")): Changed to ...
15696 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
15697 "<round_saeonly_scalar_constraint>")): ... this.
15698 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
15699 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
15700 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
15701 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
15702 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
15704 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
15706 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
15708 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
15709 Remove all values except native, 8540 and 8548.
15711 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
15713 * config/i386/i386.c (ix86_output_function_return): Pass
15714 INVALID_REGNUM, instead of -1, as invalid register number to
15715 indirect_thunk_name and output_indirect_thunk.
15717 2018-02-02 Julia Koval <julia.koval@intel.com>
15719 * config.gcc: Add -march=icelake.
15720 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
15721 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
15722 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
15723 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
15724 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
15725 (processor_target_table): Add icelake.
15726 (ix86_option_override_internal): Handle new PTAs.
15727 (get_builtin_code_for_version): Handle icelake.
15728 (M_INTEL_COREI7_ICELAKE): New.
15729 (fold_builtin_cpu): Handle icelake.
15730 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
15731 * doc/invoke.texi: Add -march=icelake.
15733 2018-02-02 Julia Koval <julia.koval@intel.com>
15735 * config/i386/i386.c (ix86_option_override_internal): Change flags type
15736 to wide_int_bitmask.
15737 * wide-int-bitmask.h: New.
15739 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
15742 * config/i386/i386.md: Replace Pmode with word_mode in
15743 builtin_setjmp_setup and builtin_longjmp to support x32.
15745 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
15749 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
15751 (rs6000_supported_cpu_names): New static variable.
15752 (linux_cpu_translation_table): Likewise.
15753 (elf_platform) <cpu>: Define new static variable and use it.
15754 Translate kernel AT_PLATFORM name to canonical name if needed.
15755 Error if platform name is unknown.
15757 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
15760 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
15762 2018-02-01 Jeff Law <law@redhat.com>
15765 * config/i386/i386.c (release_scratch_register_on_entry): Add new
15766 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
15767 the scratch if RELEASE_VIA_POP is false.
15768 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
15769 If we have to save a temporary register, decrement SIZE appropriately.
15770 Pass new arguments to release_scratch_register_on_entry.
15771 (ix86_adjust_stack_and_probe): Likewise.
15772 (ix86_emit_probe_stack_range): Pass new arguments to
15773 release_scratch_register_on_entry.
15775 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
15777 PR rtl-optimization/84157
15778 * combine.c (change_zero_ext): Use REG_P predicate in
15779 front of HARD_REGISTER_P predicate.
15781 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
15783 * config/avr/avr.c (avr_option_override): Move disabling of
15784 -fdelete-null-pointer-checks to...
15785 * common/config/avr/avr-common.c (avr_option_optimization_table):
15788 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
15790 PR tree-optimization/81635
15791 * tree-data-ref.c (split_constant_offset_1): For types that
15792 wrap on overflow, try to use range info to prove that wrapping
15795 2018-02-01 Renlin Li <renlin.li@arm.com>
15798 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
15799 TAILCALL_ADDR_REGS.
15800 (aarch64_register_move_cost): Likewise.
15801 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
15802 TAILCALL_ADDR_REGS.
15803 (REG_CLASS_NAMES): Likewise.
15804 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
15805 TAILCALL_ADDR_REGS. Remove IP registers.
15806 * config/aarch64/aarch64.md (Ucs): Update register constraint.
15808 2018-02-01 Richard Biener <rguenther@suse.de>
15810 * domwalk.h (dom_walker::dom_walker): Add additional constructor
15811 for specifying RPO order and allow NULL for that.
15812 * domwalk.c (dom_walker::dom_walker): Likewise.
15813 (dom_walker::walk): Handle NULL RPO order.
15814 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
15816 (rewrite_update_dom_walker): Likewise.
15817 (mark_def_dom_walker): Likewise.
15819 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
15821 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
15822 (aarch64_maybe_expand_sve_subreg_move): Declare.
15823 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
15824 * config/aarch64/predicates.md (aarch64_any_register_operand): New
15826 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
15827 that are semantically a reverse operation.
15828 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
15829 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
15830 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
15832 (aarch64_can_change_mode_class): For big-endian, forbid changes
15833 between two SVE modes if they have different element sizes.
15835 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
15837 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
15838 the TImode handling for big-endian targets.
15840 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
15842 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
15843 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
15845 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
15846 Remove BSWAP handing for big-endian targets and use the form of
15847 LD1RQ appropariate for the mode.
15849 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
15851 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
15852 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
15853 duplicated element.
15855 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
15858 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
15859 check for operands that need to go through aarch64_sve_reload_be.
15861 2018-02-01 Jakub Jelinek <jakub@redhat.com>
15863 PR tree-optimization/81661
15864 PR tree-optimization/84117
15865 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
15866 * tree-eh.c: Include gimplify.h.
15867 (find_trapping_overflow, replace_trapping_overflow,
15868 rewrite_to_non_trapping_overflow): New functions.
15869 * tree-vect-loop.c: Include tree-eh.h.
15870 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
15871 * tree-data-ref.c: Include tree-eh.h.
15872 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
15874 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
15876 PR rtl-optimization/84123
15877 * combine.c (change_zero_ext): Check if hard register satisfies
15878 can_change_dest_mode before calling gen_lowpart_SUBREG.
15880 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
15883 * ira.c (ira_init_register_move_cost): Remove assert.
15885 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
15887 PR rtl-optimization/84071
15888 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
15889 * doc/tm.texi: Regenerate.
15891 2018-01-31 Richard Biener <rguenther@suse.de>
15893 PR tree-optimization/84132
15894 * tree-data-ref.c (analyze_miv_subscript): Properly
15895 check whether evolution_function_is_affine_multivariate_p
15896 before calling gcd_of_steps_may_divide_p.
15898 2018-01-31 Julia Koval <julia.koval@intel.com>
15901 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
15902 * config/i386/i386.md (rdpid_rex64) New.
15903 (rdpid): Make 32bit only.
15905 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
15908 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
15909 an IDENTIFIER_NODE for FUNCTION_TYPE's.
15911 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
15914 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
15916 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
15918 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
15920 PR rtl-optimization/84071
15921 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
15922 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
15924 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
15926 * config/arc/arc.c (arc_handle_aux_attribute): New function.
15927 (arc_attribute_table): Add 'aux' attribute.
15928 (arc_in_small_data_p): Consider aux like variables.
15929 (arc_is_aux_reg_p): New function.
15930 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
15931 (arc_get_aux_arg): New function.
15932 (prepare_move_operands): Handle aux-register access.
15933 (arc_handle_aux_attribute): New function.
15934 * doc/extend.texi (ARC Variable attributes): Add subsection.
15936 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
15938 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
15939 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
15940 (arc_attribute_table): Add 'uncached' attribute.
15941 (arc_print_operand): Print '.di' flag for uncached memory
15943 (arc_in_small_data_p): Do not consider for small data the uncached
15945 (arc_is_uncached_mem_p): New function.
15946 * config/arc/predicates.md (compact_store_memory_operand): Check
15947 for uncached memory accesses.
15948 (nonvol_nonimm_operand): Likewise.
15949 * doc/extend.texi (ARC Type Attribute): New subsection.
15951 2018-01-31 Jakub Jelinek <jakub@redhat.com>
15954 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
15955 falign-loops=): Add Optimization flag.
15957 2018-01-30 Jeff Law <law@redhat.com>
15960 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
15961 INT_REGISTERS_SAVED. Check it prior to calling
15962 get_scratch_register_on_entry.
15963 (ix86_adjust_stack_and_probe): Similarly.
15964 (ix86_emit_probe_stack_range): Similarly.
15965 (ix86_expand_prologue): Corresponding changes.
15967 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15970 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
15971 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
15973 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
15976 * lra-constraints.c (curr_insn_transform): Process AND in the
15979 2018-01-30 Jakub Jelinek <jakub@redhat.com>
15981 PR rtl-optimization/83986
15982 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
15983 dependence against last_pending_memory_flush in addition to
15984 pending_jump_insns.
15986 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
15988 PR tree-optimization/81611
15989 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
15992 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15995 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
15998 2018-01-30 Richard Biener <rguenther@suse.de>
15999 Jakub Jelinek <jakub@redhat.com>
16001 PR tree-optimization/84111
16002 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
16003 inner loops added during recursion, as they don't have up-to-date
16006 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
16009 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
16010 (can_inline_edge_by_limits_p): ... here.
16011 (can_early_inline_edge_p, check_callers,
16012 update_caller_keys, update_callee_keys, recursive_inlining,
16013 add_new_edges_to_heap, speculation_useful_p,
16014 inline_small_functions,
16015 inline_small_functions, flatten_function,
16016 inline_to_all_callers_1): Update.
16018 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
16020 * profile-count.c (profile_count::combine_with_ipa_count): Handle
16023 2018-01-30 Richard Biener <rguenther@suse.de>
16025 PR tree-optimization/83008
16026 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
16027 invariant and constant vector uses in stmts when they need
16028 more than one stmt.
16030 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16033 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
16034 * configure: Regenerate.
16036 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
16038 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
16040 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
16041 Use gen_rtx_REG rather than gen_lowpart.
16043 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
16045 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
16046 rather than 0 when creating partial subregs.
16048 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
16050 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
16053 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
16056 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
16057 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
16058 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
16059 flags. This restores the settings used before the 2017-07-24.
16060 Turning off pre increment/decrement/modify allows IVOPTS to
16061 optimize DF/SF loops where the index is an int.
16063 2018-01-29 Richard Biener <rguenther@suse.de>
16064 Kelvin Nilsen <kelvin@gcc.gnu.org>
16067 * tree-vect-stmts.c (vectorizable_call): Don't call
16068 targetm.vectorize_builtin_md_vectorized_function if callee is
16071 2018-01-22 Carl Love <cel@us.ibm.com>
16073 * doc/extend.tex: Fix typo in second arg in
16074 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
16076 2018-01-29 Richard Biener <rguenther@suse.de>
16078 PR tree-optimization/84086
16079 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
16080 (flush_ssaname_freelist): When SSA names were released reset
16081 the SCEV hash table.
16083 2018-01-29 Richard Biener <rguenther@suse.de>
16085 PR tree-optimization/84057
16086 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
16087 removed paths when removing edges.
16089 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
16091 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
16092 -mfunction-return=@var{choice}.
16094 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
16096 PR diagnostic/84034
16097 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
16098 Handle CR like TAB.
16099 (layout::print_source_line): Likewise.
16100 (test_get_line_width_without_trailing_whitespace): Add test cases.
16102 2018-01-27 Jakub Jelinek <jakub@redhat.com>
16104 PR middle-end/84040
16105 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
16108 2018-01-26 Jim Wilson <jimw@sifive.com>
16110 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
16112 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
16115 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16117 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
16118 and CMP + SUB-immediate -> SUBS.
16120 2018-01-26 Martin Sebor <msebor@redhat.com>
16122 PR tree-optimization/83896
16123 * tree-ssa-strlen.c (get_string_len): Rename...
16124 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
16125 Avoid assuming length is constant.
16126 (handle_char_store): Use HOST_WIDE_INT for string length.
16128 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
16131 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
16132 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
16134 2018-01-26 Richard Biener <rguenther@suse.de>
16136 PR rtl-optimization/84003
16137 * dse.c (record_store): Only record redundant stores when
16138 the earlier store aliases at least all accesses the later one does.
16140 2018-01-26 Jakub Jelinek <jakub@redhat.com>
16142 PR rtl-optimization/83985
16143 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
16144 REG_CFA_RESTORE insns.
16145 (delete_unmarked_insns): Don't ignore separate shrink wrapping
16146 REG_CFA_RESTORE insns here.
16149 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
16150 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
16152 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
16154 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
16155 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
16156 (arc_init): Likewise.
16157 (arc_override_options): Likewise.
16158 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
16160 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
16162 * config/arc/arc.h (TARGET_DBNZ): Define.
16163 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
16164 properly set the tune attribute.
16165 (dbnz): Use TARGET_DBNZ guard.
16166 * config/arc/arc.opt (mtune): Add core3 option.
16168 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
16170 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
16171 recognize new pic like addresses.
16172 (arc_delegitimize_address): Clean up.
16174 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
16176 * config/arc/arc-arches.def: Option mrf16 valid for all
16178 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
16179 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
16180 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
16181 * config/arc/arc-tables.opt: Regenerate.
16182 * config/arc/arc.c (arc_conditional_register_usage): Handle
16183 reduced register file case.
16184 (arc_file_start): Set must have build attributes.
16185 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
16186 mrf16 option value.
16187 * config/arc/arc.opt (mrf16): Add new option.
16188 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
16189 * config/arc/genmultilib.awk: Handle new mrf16 option.
16190 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
16191 * config/arc/t-multilib: Regenerate.
16192 * doc/invoke.texi (ARC Options): Document mrf16 option.
16194 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
16196 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
16197 * config/arc/arc.c (arc_handle_secure_attribute): New function.
16198 (arc_attribute_table): Add 'secure_call' attribute.
16199 (arc_print_operand): Print secure call operand.
16200 (arc_function_ok_for_sibcall): Don't optimize tail calls when
16202 (arc_is_secure_call_p): New function. * config/arc/arc.md
16203 (call_i): Add support for sjli instruction.
16204 (call_value_i): Likewise.
16205 * config/arc/constraints.md (Csc): New constraint.
16207 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
16208 John Eric Martin <John.Martin@emmicro-us.com>
16210 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
16211 * config/arc/arc.c (_arc_jli_section): New struct.
16212 (arc_jli_section): New type.
16213 (rc_jli_sections): New static variable.
16214 (arc_handle_jli_attribute): New function.
16215 (arc_attribute_table): Add jli_always and jli_fixed attribute.
16216 (arc_file_end): New function.
16217 (TARGET_ASM_FILE_END): Define.
16218 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
16219 (arc_add_jli_section): New function.
16220 (jli_call_scan): Likewise.
16221 (arc_reorg): Call jli_call_scan.
16222 (arc_output_addsi): Remove 'S' from printing asm operand.
16223 (arc_is_jli_call_p): New function.
16224 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
16226 (movhi_insn): Likewise.
16227 (movsi_insn): Likewise.
16228 (movsi_set_cc_insn): Likewise.
16229 (loadqi_update): Likewise.
16230 (load_zeroextendqisi_update): Likewise.
16231 (load_signextendqisi_update): Likewise.
16232 (loadhi_update): Likewise.
16233 (load_zeroextendhisi_update): Likewise.
16234 (load_signextendhisi_update): Likewise.
16235 (loadsi_update): Likewise.
16236 (loadsf_update): Likewise.
16237 (movsicc_insn): Likewise.
16238 (bset_insn): Likewise.
16239 (bxor_insn): Likewise.
16240 (bclr_insn): Likewise.
16241 (bmsk_insn): Likewise.
16242 (bicsi3_insn): Likewise.
16243 (cmpsi_cc_c_insn): Likewise.
16244 (movsi_ne): Likewise.
16245 (movsi_cond_exec): Likewise.
16246 (clrsbsi2): Likewise.
16247 (norm_f): Likewise.
16257 (call_i): Remove 'S' asm letter, add jli instruction.
16258 (call_value_i): Likewise.
16259 * config/arc/arc.op (mjli-always): New option.
16260 * config/arc/constraints.md (Cji): New constraint.
16261 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
16263 (subsf3_fpx): Likewise.
16264 (mulsf3_fpx): Likewise.
16265 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
16267 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
16268 function attrbutes.
16269 * doc/invoke.texi (ARC): Document mjli-always option.
16271 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
16273 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
16274 avoid addition with 0 and use incw and decw where possible.
16276 2018-01-26 Richard Biener <rguenther@suse.de>
16278 PR tree-optimization/81082
16279 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
16280 association if it requires casting to unsigned.
16281 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
16282 from fold_plusminus_mult_expr to catch important cases late when
16283 range info is available.
16285 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16287 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
16288 * configure.ac (hidden_linkonce): New test.
16289 * configure: Regenerate.
16290 * config.in: Regenerate.
16292 2018-01-26 Julia Koval <julia.koval@intel.com>
16294 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
16295 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
16296 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
16297 _mm_mask_bitshuffle_epi64_mask): Fix type.
16298 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
16299 USI_FTYPE_V4DI_V4DI_USI): Remove.
16300 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
16301 __builtin_ia32_vpshufbitqmb256_mask,
16302 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
16303 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
16304 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
16306 2018-01-26 Alan Modra <amodra@gmail.com>
16309 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
16310 UNSPEC_VBPERMQ. Sort other unspecs.
16312 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
16314 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
16316 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
16318 PR middle-end/83055
16319 * predict.c (drop_profile): Do not push/pop cfun; update also
16321 (handle_missing_profiles): Fix logic looking for zero profiles.
16323 2018-01-25 Jakub Jelinek <jakub@redhat.com>
16325 PR middle-end/83977
16326 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
16327 on functions with #pragma omp declare simd or functions with simd
16329 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
16330 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
16331 Remove trailing \n from warning_at calls.
16333 2018-01-25 Tom de Vries <tom@codesourcery.com>
16336 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
16337 for neutered workers.
16339 2018-01-24 Joseph Myers <joseph@codesourcery.com>
16342 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
16343 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
16345 2018-01-24 Jeff Law <law@redhat.com>
16348 * i386.c (get_probe_interval): Move to earlier point.
16349 (ix86_compute_frame_layout): If -fstack-clash-protection and
16350 the frame is larger than the probe interval, then use pushes
16351 to save registers rather than reg->mem moves.
16352 (ix86_expand_prologue): Remove conditional for int_registers_saved
16355 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
16358 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
16359 min/max for never referenced object.
16361 2018-01-24 Jakub Jelinek <jakub@redhat.com>
16363 PR middle-end/83977
16364 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
16366 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
16367 attributes from DECL_ATTRIBUTES (decl) without affecting
16368 DECL_ATTRIBUTES (current_function_decl).
16369 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
16370 functions with non-NULL DECL_ABSTRACT_ORIGIN.
16372 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
16374 PR tree-optimization/83979
16375 * fold-const.c (fold_comparison): Use constant_boolean_node
16376 instead of boolean_{true,false}_node.
16378 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
16380 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
16383 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16385 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
16386 Simplify the clause that sets the length attribute.
16387 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
16388 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
16389 clause that sets the length attribute.
16390 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
16392 2018-01-24 Tom de Vries <tom@codesourcery.com>
16395 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
16396 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
16397 Add strict parameter.
16398 (prevent_branch_around_nothing): Insert dummy insn between branch to
16399 label and label with no ptx insn inbetween.
16400 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
16402 2018-01-24 Tom de Vries <tom@codesourcery.com>
16405 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
16406 for neutered threads in warp.
16407 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
16409 2018-01-24 Richard Biener <rguenther@suse.de>
16411 PR tree-optimization/83176
16412 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
16415 2018-01-24 Richard Biener <rguenther@suse.de>
16417 PR tree-optimization/82819
16418 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
16419 code generating pluses that are no-ops in the target precision.
16421 2018-01-24 Richard Biener <rguenther@suse.de>
16423 PR middle-end/84000
16424 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
16426 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
16428 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
16429 to merge probabilities.
16430 * predict.c (probably_never_executed): Also mark as cold functions
16431 with global 0 profile and guessed local profile.
16432 * profile-count.c (profile_probability::combine_with_count): New
16434 * profile-count.h (profile_probability::operator*,
16435 profile_probability::operator*=, profile_probability::operator/,
16436 profile_probability::operator/=): Reduce precision to adjusted
16437 and set value to guessed on contradictory divisions.
16438 (profile_probability::combine_with_freq): Remove.
16439 (profile_probability::combine_wiht_count): Declare.
16440 (profile_count::force_nonzero):: Set to adjusted.
16441 (profile_count::probability_in):: Set quality to adjusted.
16442 * tree-ssa-tail-merge.c (replace_block_by): Use
16443 combine_with_count.
16445 2018-01-23 Andrew Waterman <andrew@sifive.com>
16446 Jim Wilson <jimw@sifive.com>
16448 * config/riscv/riscv.c (riscv_stack_boundary): New.
16449 (riscv_option_override): Set riscv_stack_boundary. Handle
16450 riscv_preferred_stack_boundary_arg.
16451 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
16452 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
16453 (STACK_BOUNDARY): Set to riscv_stack_boundary.
16454 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
16455 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
16456 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
16458 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
16461 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
16462 of struct ix86_frame.
16463 (ix86_expand_epilogue): Likewise. Add a local variable for
16464 the reg_save_offset field in struct ix86_frame.
16466 2018-01-23 Bin Cheng <bin.cheng@arm.com>
16468 PR tree-optimization/82604
16469 * tree-loop-distribution.c (enum partition_kind): New enum item
16470 PKIND_PARTIAL_MEMSET.
16471 (partition_builtin_p): Support above new enum item.
16472 (generate_code_for_partition): Ditto.
16473 (compute_access_range): Differentiate cases that equality can be
16474 proven at all loops, the innermost loops or no loops.
16475 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
16476 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
16477 (finalize_partitions, distribute_loop): Don't fuse partition of
16478 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
16479 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
16480 parloop is enabled.
16482 2018-01-23 Martin Liska <mliska@suse.cz>
16484 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
16485 order to ignore the predictor.
16486 (PRED_POLYMORPHIC_CALL): Likewise.
16487 (PRED_RECURSIVE_CALL): Likewise.
16489 2018-01-23 Martin Liska <mliska@suse.cz>
16491 * tree-profile.c (tree_profiling): Print function header to
16492 aware reader which function we are working on.
16493 * value-prof.c (gimple_find_values_to_profile): Do not print
16494 not interesting value histograms.
16496 2018-01-23 Martin Liska <mliska@suse.cz>
16498 * profile-count.h (enum profile_quality): Add
16499 profile_uninitialized as the first value. Do not number values
16500 as they are zero based.
16501 (profile_count::verify): Update sanity check.
16502 (profile_probability::verify): Likewise.
16504 2018-01-23 Nathan Sidwell <nathan@acm.org>
16506 * doc/invoke.texi (ffor-scope): Deprecate.
16508 2018-01-23 David Malcolm <dmalcolm@redhat.com>
16510 PR tree-optimization/83510
16511 * domwalk.c (set_all_edges_as_executable): New function.
16512 (dom_walker::dom_walker): Convert bool param
16513 "skip_unreachable_blocks" to enum reachability. Move setup of
16514 edge flags to set_all_edges_as_executable and only do it when
16515 reachability is REACHABLE_BLOCKS.
16516 * domwalk.h (enum dom_walker::reachability): New enum.
16517 (dom_walker::dom_walker): Convert bool param
16518 "skip_unreachable_blocks" to enum reachability.
16519 (set_all_edges_as_executable): New decl.
16520 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
16521 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
16523 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
16524 but converting true to REACHABLE_BLOCKS.
16525 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
16527 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
16528 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
16529 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
16531 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
16532 if check_all_array_refs will be called.
16534 2018-01-23 David Malcolm <dmalcolm@redhat.com>
16536 * tree.c (selftest::test_location_wrappers): Add more test
16539 2018-01-23 David Malcolm <dmalcolm@redhat.com>
16541 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
16542 (selftest::test_bit_in_range): Likewise.
16544 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
16547 * doc/sourcebuild.texi (vect_float): Say that the selector
16548 only describes the situation when -funsafe-math-optimizations is on.
16549 (vect_float_strict): Document.
16551 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
16553 PR tree-optimization/83965
16554 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
16555 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
16556 instead of checking only for a reduction.
16557 (vect_recog_widen_sum_pattern): Likewise.
16559 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
16561 * predict.c (probably_never_executed): Only use precise profile info.
16562 (compute_function_frequency): Skip after inlining hack since we now
16563 have quality checking.
16565 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
16567 * profile-count.h (profile_probability::very_unlikely,
16568 profile_probability::unlikely, profile_probability::even): Set
16569 precision to guessed.
16571 2018-01-23 Richard Biener <rguenther@suse.de>
16573 PR tree-optimization/83963
16574 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
16575 Properly terminate dominator walk when crossing the exit edge not
16576 when visiting its source block.
16578 2018-01-23 Jakub Jelinek <jakub@redhat.com>
16581 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
16582 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
16584 2018-01-22 Jakub Jelinek <jakub@redhat.com>
16586 PR tree-optimization/83957
16587 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
16588 semicolon after for body surrounded by braces.
16590 PR tree-optimization/83081
16591 * profile-count.h (profile_probability::split): New method.
16592 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
16593 Use profile_probability::split.
16594 (do_compare_rtx_and_jump): Fix adjustment of probabilities
16595 when splitting a single conditional jump into 2.
16597 2018-01-22 David Malcolm <dmalcolm@redhat.com>
16599 PR tree-optimization/69452
16600 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
16603 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16605 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
16606 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
16607 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
16609 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16611 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
16613 * config/rl78/rl78.md (movdi): New define_expand.
16614 * config/rl78/rl78.c (rl78_split_movdi): New function.
16616 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
16619 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
16621 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
16622 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
16623 128-bit to produce an UNSPEC move to get the double word with the
16624 signbit and then a shift directly to do signbit.
16625 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
16626 implementation with a new version that just does either a direct
16627 move or a regular move. Move memory interface to separate insns.
16628 Move insns so they are next to the expander.
16629 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
16630 with signbit move. Split big and little endian case.
16631 (signbit<mode>2_dm_mem_le): Likewise.
16632 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
16633 (signbit<mode>2_dm2): Likewise.
16635 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16637 * config/rl78/rl78.md (anddi3): New define_expand.
16639 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16641 * config/rl78/rl78.md (umindi3): New define_expand.
16643 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16645 * config/rl78/rl78.md (smindi3): New define_expand.
16647 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16649 * config/rl78/rl78.md (smaxdi3): New define_expand.
16651 2018-01-22 Carl Love <cel@us.ibm.com>
16653 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
16654 LVX_V1TI): Add macro expansion.
16655 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
16656 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
16657 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
16658 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
16659 Change check to determine if the instruction is a byte reversing
16660 entry. Fix typo in comment.
16661 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
16662 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
16663 Add def_builtin calls for new builtins.
16664 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
16665 Add define_insn expansion.
16667 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16669 * config/rl78/rl78.md (umaxdi3): New define_expand.
16671 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16673 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
16674 for non-QImode registers.
16676 2018-01-22 Richard Biener <rguenther@suse.de>
16678 PR tree-optimization/83963
16679 * graphite-scop-detection.c (scop_detection::get_sese): Delay
16680 including the loop exit block.
16681 (scop_detection::merge_sese): Likewise.
16682 (scop_detection::add_scop): Do it here instead.
16684 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16686 * doc/sourcebuild.texi (arm_softfloat): Document.
16688 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
16691 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
16692 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
16693 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
16695 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16696 David Edelsohn <dje.gcc@gmail.com>
16699 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
16700 Change "crset eq" to "crset 2".
16701 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
16702 (*call_indirect_aix<mode>_nospec): Likewise.
16703 (*call_value_indirect_aix<mode>_nospec): Likewise.
16704 (*call_indirect_elfv2<mode>_nospec): Likewise.
16705 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
16706 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
16707 change assembly output from . to $.
16708 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
16709 (indirect_jump<mode>_nospec): Change assembly output from . to $.
16710 (*tablejump<mode>_internal1_nospec): Likewise.
16712 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
16715 * config/sh/sh_optimize_sett_clrt.cc:
16716 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
16718 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
16720 PR tree-optimization/83940
16721 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
16722 offset_dt to vect_constant_def rather than vect_unknown_def_type.
16723 (vect_check_load_store_mask): Add a mask_dt_out parameter and
16724 use it to pass back the definition type.
16725 (vect_check_store_rhs): Likewise rhs_dt_out.
16726 (vect_build_gather_load_calls): Add a mask_dt argument and use
16727 it instead of a call to vect_is_simple_use.
16728 (vectorizable_store): Update calls to vect_check_load_store_mask
16729 and vect_check_store_rhs. Use the dt returned by the latter instead
16730 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
16731 instead of calls to vect_is_simple_use. Pass the scalar rather
16732 than the vector operand to vect_is_simple_use when handling
16733 second and subsequent copies of an rhs value.
16734 (vectorizable_load): Update calls to vect_check_load_store_mask
16735 and vect_build_gather_load_calls. Use the cached mask_dt and
16736 gs_info.offset_dt instead of calls to vect_is_simple_use.
16738 2018-01-20 Jakub Jelinek <jakub@redhat.com>
16740 PR middle-end/83945
16741 * tree-emutls.c: Include gimplify.h.
16742 (lower_emutls_2): New function.
16743 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
16744 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
16745 it before further processing.
16748 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
16749 UINTVAL (trueop1) instead of INTVAL (op1).
16751 2018-01-19 Jakub Jelinek <jakub@redhat.com>
16755 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
16756 INCOMING_FRAME_SP_OFFSET if not defined.
16757 (scan_trace): Add ENTRY argument. If true and
16758 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
16759 emit a note to adjust the CFA offset.
16760 (create_cfi_notes): Adjust scan_trace callers.
16761 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
16762 INCOMING_FRAME_SP_OFFSET in the CIE.
16763 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
16764 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
16766 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
16767 * doc/tm.texi: Regenerated.
16769 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
16771 PR rtl-optimization/83147
16772 * lra-constraints.c (remove_inheritance_pseudos): Use
16773 lra_substitute_pseudo_within_insn.
16775 2018-01-19 Tom de Vries <tom@codesourcery.com>
16776 Cesar Philippidis <cesar@codesourcery.com>
16779 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
16781 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
16784 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
16785 spaces for function labels.
16787 2018-01-19 Martin Liska <mliska@suse.cz>
16789 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
16790 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
16791 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
16792 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
16793 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
16794 (PRED_CONST_RETURN): Change from 69 to 65.
16795 (PRED_NULL_RETURN): Change from 91 to 71.
16796 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
16797 (PRED_LOOP_GUARD): Change from 66 to 73.
16799 2018-01-19 Martin Liska <mliska@suse.cz>
16801 * predict.c (predict_insn_def): Add new assert.
16802 (struct branch_predictor): Change type to signed integer.
16803 (test_prediction_value_range): Amend test to cover
16804 PROB_UNINITIALIZED.
16805 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
16806 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
16807 (PRED_LOOP_ITERATIONS_MAX): Likewise.
16808 (PRED_LOOP_IV_COMPARE): Likewise.
16809 * predict.h (PROB_UNINITIALIZED): Define new constant.
16811 2018-01-19 Martin Liska <mliska@suse.cz>
16813 * predict.c (dump_prediction): Add new format for
16814 analyze_brprob.py script which is enabled with -details
16816 * profile-count.h (precise_p): New function.
16818 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
16820 PR tree-optimization/83922
16821 * tree-vect-loop.c (vect_verify_full_masking): Return false if
16822 there are no statements that need masking.
16823 (vect_active_double_reduction_p): New function.
16824 (vect_analyze_loop_operations): Use it when handling phis that
16825 are not in the loop header.
16827 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
16829 PR tree-optimization/83914
16830 * tree-vect-loop.c (vectorizable_induction): Don't convert
16831 init_expr or apply the peeling adjustment for inductions
16832 that are nested within the vectorized loop.
16834 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16836 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
16839 2018-01-18 Jakub Jelinek <jakub@redhat.com>
16843 * function.h (gimplify_parameters): Add gimple_seq * argument.
16844 * function.c: Include gimple.h and options.h.
16845 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
16846 for the added local temporaries if needed.
16847 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
16848 if there are any parameter cleanups, wrap whole body into a
16849 try/finally with the cleanups.
16851 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
16854 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
16855 Use GET_MODE_CLASS for scalar floating point.
16857 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
16861 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
16862 Fix call of call_cgraph_insertion_hooks.
16864 2018-01-18 Martin Sebor <msebor@redhat.com>
16866 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
16868 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
16871 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
16874 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
16877 * common.opt: (-ffile-prefix-map): New option.
16878 * opts.c (common_handle_option): Defer it.
16879 * opts-global.c (handle_common_deferred_options): Handle it.
16880 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
16881 * file-prefix-map.h: New file.
16882 (remap_debug_filename, add_debug_prefix_map): ...here.
16883 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
16884 * final.c (debug_prefix_map, add_debug_prefix_map
16885 remap_debug_filename): Move to...
16886 * file-prefix-map.c: New file.
16887 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
16888 generalize, get rid of alloca(), use strrchr() instead of strchr().
16889 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
16890 Implement in terms of add_prefix_map().
16891 (remap_macro_filename, remap_debug_filename): Implement in term of
16893 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
16894 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
16895 * dbxout.c: Include file-prefix-map.h.
16896 * varasm.c: Likewise.
16897 * vmsdbgout.c: Likewise.
16898 * xcoffout.c: Likewise.
16899 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
16900 * doc/cppopts.texi (-fmacro-prefix-map): Document.
16901 * doc/invoke.texi (-ffile-prefix-map): Document.
16902 (-fdebug-prefix-map): Update description.
16904 2018-01-18 Martin Liska <mliska@suse.cz>
16906 * config/i386/i386.c (indirect_thunk_name): Document that also
16908 (output_indirect_thunk): Document why both instructions
16909 (pause and lfence) are generated.
16911 2018-01-18 Richard Biener <rguenther@suse.de>
16913 PR tree-optimization/83887
16914 * graphite-scop-detection.c
16915 (scop_detection::get_nearest_dom_with_single_entry): Remove.
16916 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
16917 (scop_detection::merge_sese): Re-implement with a flood-fill
16918 algorithm that properly finds a SESE region if it exists.
16920 2018-01-18 Jakub Jelinek <jakub@redhat.com>
16923 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
16924 pointer_diff optimizations use view_convert instead of convert.
16926 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16928 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
16929 Generate different code for -mno-speculate-indirect-jumps.
16930 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
16931 (*call_indirect_aix<mode>): Disable for
16932 -mno-speculate-indirect-jumps.
16933 (*call_indirect_aix<mode>_nospec): New define_insn.
16934 (*call_value_indirect_aix<mode>): Disable for
16935 -mno-speculate-indirect-jumps.
16936 (*call_value_indirect_aix<mode>_nospec): New define_insn.
16937 (*sibcall_nonlocal_sysv<mode>): Generate different code for
16938 -mno-speculate-indirect-jumps.
16939 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
16941 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
16943 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
16944 long double type, set the flags for noting the default long double
16945 type, even if we don't pass or return a long double type.
16947 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
16950 * ipa-inline.c (flatten_function): Do not overwrite final inlining
16953 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
16955 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
16956 support for merge[hl].
16957 (fold_mergehl_helper): New helper function.
16958 (tree-vector-builder.h): New #include for tree_vector_builder usage.
16959 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
16960 (altivec_vmrglw_direct): Add xxmrglw insn.
16962 2018-01-17 Andrew Waterman <andrew@sifive.com>
16964 * config/riscv/riscv.c (riscv_conditional_register_usage): If
16965 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
16967 2018-01-17 David Malcolm <dmalcolm@redhat.com>
16970 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
16971 call the lto_location_cache before reading the
16972 DECL_SOURCE_LOCATION of the types.
16974 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
16975 Richard Sandiford <richard.sandiford@linaro.org>
16977 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
16978 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
16979 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
16980 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
16981 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
16983 * config/aarch64/constraints.md (aarch64_movti_operand):
16985 * config/aarch64/predicates.md (Uti): Add new constraint.
16987 2018-01-17 Carl Love <cel@us.ibm.com>
16989 * config/rs6000/vsx.md (define_expand xl_len_r,
16990 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
16991 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
16993 (define_expand, define_insn): Move the shift left from the
16994 define_insn to the define_expand for lxvl and stxvl instructions.
16995 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
16996 and XL_LEN_R definitions to PURE.
16998 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
17000 * config/i386/i386.c (indirect_thunk_name): Declare regno
17001 as unsigned int. Compare regno with INVALID_REGNUM.
17002 (output_indirect_thunk): Ditto.
17003 (output_indirect_thunk_function): Ditto.
17004 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
17005 in the call to output_indirect_thunk_function.
17007 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
17009 PR middle-end/83884
17010 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
17011 rather than the size of inner_type to determine the stack slot size
17012 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
17014 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
17017 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
17020 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
17022 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
17023 endian Linux systems to optionally enable multilibs for selecting
17024 the long double type if the user configured an explicit type.
17025 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
17026 have no long double multilibs if not defined.
17027 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
17028 warn if the user used -mabi={ieee,ibm}longdouble and we built
17029 multilibs for long double.
17030 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
17031 appropriate multilib option.
17032 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
17034 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
17035 for building long double multilibs.
17036 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
17038 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
17040 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
17043 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
17045 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
17048 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
17051 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
17054 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
17056 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
17057 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
17059 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
17061 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
17062 different rtl trees depending on TARGET_64BIT.
17063 (rs6000_gen_lvx): Likewise.
17065 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
17067 * config/visium/visium.md (nop): Tweak comment.
17068 (hazard_nop): Likewise.
17070 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17072 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
17073 -mspeculate-indirect-jumps.
17074 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
17075 for -mno-speculate-indirect-jumps.
17076 (*call_indirect_elfv2<mode>_nospec): New define_insn.
17077 (*call_value_indirect_elfv2<mode>): Disable for
17078 -mno-speculate-indirect-jumps.
17079 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
17080 (indirect_jump): Emit different RTL for
17081 -mno-speculate-indirect-jumps.
17082 (*indirect_jump<mode>): Disable for
17083 -mno-speculate-indirect-jumps.
17084 (*indirect_jump<mode>_nospec): New define_insn.
17085 (tablejump): Emit different RTL for
17086 -mno-speculate-indirect-jumps.
17087 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
17088 (tablejumpsi_nospec): New define_expand.
17089 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
17090 (tablejumpdi_nospec): New define_expand.
17091 (*tablejump<mode>_internal1): Disable for
17092 -mno-speculate-indirect-jumps.
17093 (*tablejump<mode>_internal1_nospec): New define_insn.
17094 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
17097 2018-01-16 Artyom Skrobov tyomitch@gmail.com
17099 * caller-save.c (insert_save): Drop unnecessary parameter. All
17102 2018-01-16 Jakub Jelinek <jakub@redhat.com>
17103 Richard Biener <rguenth@suse.de>
17106 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
17107 return early, inline manually is_gimple_sizepos. Make sure if we
17108 call gimplify_expr we don't end up with a gimple constant.
17109 * tree.c (variably_modified_type_p): Don't return true for
17110 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
17111 * gimplify.h (is_gimple_sizepos): Remove.
17113 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
17115 PR tree-optimization/83857
17116 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
17117 vectorizable_live_operation for pure SLP statements.
17118 (vectorizable_live_operation): Handle PHIs.
17120 2018-01-16 Richard Biener <rguenther@suse.de>
17122 PR tree-optimization/83867
17123 * tree-vect-stmts.c (vect_transform_stmt): Precompute
17124 nested_in_vect_loop_p since the scalar stmt may get invalidated.
17126 2018-01-16 Jakub Jelinek <jakub@redhat.com>
17129 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
17130 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
17131 If off is not INTEGER_CST, issue a may not be aligned warning
17132 rather than isn't aligned. Use isn%'t rather than isn't.
17133 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
17135 <case MULT_EXPR>: Improve the case when bottom and one of the
17136 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
17137 operand, in that case check if the other operand is multiple of
17138 bottom divided by the INTEGER_CST operand.
17140 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
17143 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
17144 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
17145 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
17146 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
17147 * config/pa/pa.c (pa_function_arg_advance): Likewise.
17148 (pa_function_arg, pa_arg_partial_bytes): Likewise.
17149 (pa_function_arg_size): New function.
17151 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
17153 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
17154 in a separate statement.
17156 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
17158 PR tree-optimization/83847
17159 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
17160 group gathers and scatters.
17162 2018-01-16 Jakub Jelinek <jakub@redhat.com>
17164 PR rtl-optimization/86620
17165 * params.def (max-sched-ready-insns): Bump minimum value to 1.
17167 PR rtl-optimization/83213
17168 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
17169 to last if both are JUMP_INSNs.
17171 PR tree-optimization/83843
17172 * gimple-ssa-store-merging.c
17173 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
17174 store_immediate_info for bswap/nop orig_stores.
17176 2018-01-15 Andrew Waterman <andrew@sifive.com>
17178 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
17180 <UDIV>: Increase cost if !TARGET_DIV.
17182 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
17184 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
17185 (define_attr "cr_logical_3op"): New.
17186 (cceq_ior_compare): Adjust.
17187 (cceq_ior_compare_complement): Adjust.
17188 (*cceq_rev_compare): Adjust.
17189 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
17190 (is_cracked_insn): Adjust.
17191 (insn_must_be_first_in_group): Adjust.
17192 * config/rs6000/40x.md: Adjust.
17193 * config/rs6000/440.md: Adjust.
17194 * config/rs6000/476.md: Adjust.
17195 * config/rs6000/601.md: Adjust.
17196 * config/rs6000/603.md: Adjust.
17197 * config/rs6000/6xx.md: Adjust.
17198 * config/rs6000/7450.md: Adjust.
17199 * config/rs6000/7xx.md: Adjust.
17200 * config/rs6000/8540.md: Adjust.
17201 * config/rs6000/cell.md: Adjust.
17202 * config/rs6000/e300c2c3.md: Adjust.
17203 * config/rs6000/e500mc.md: Adjust.
17204 * config/rs6000/e500mc64.md: Adjust.
17205 * config/rs6000/e5500.md: Adjust.
17206 * config/rs6000/e6500.md: Adjust.
17207 * config/rs6000/mpc.md: Adjust.
17208 * config/rs6000/power4.md: Adjust.
17209 * config/rs6000/power5.md: Adjust.
17210 * config/rs6000/power6.md: Adjust.
17211 * config/rs6000/power7.md: Adjust.
17212 * config/rs6000/power8.md: Adjust.
17213 * config/rs6000/power9.md: Adjust.
17214 * config/rs6000/rs64.md: Adjust.
17215 * config/rs6000/titan.md: Adjust.
17217 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
17219 * config/i386/predicates.md (indirect_branch_operand): Rewrite
17220 ix86_indirect_branch_register logic.
17222 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
17224 * config/i386/constraints.md (Bs): Update
17225 ix86_indirect_branch_register check. Don't check
17226 ix86_indirect_branch_register with GOT_memory_operand.
17228 * config/i386/predicates.md (GOT_memory_operand): Don't check
17229 ix86_indirect_branch_register here.
17230 (GOT32_symbol_operand): Likewise.
17232 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
17234 * config/i386/predicates.md (constant_call_address_operand):
17235 Rewrite ix86_indirect_branch_register logic.
17236 (sibcall_insn_operand): Likewise.
17238 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
17240 * config/i386/constraints.md (Bs): Replace
17241 ix86_indirect_branch_thunk_register with
17242 ix86_indirect_branch_register.
17244 * config/i386/i386.md (indirect_jump): Likewise.
17245 (tablejump): Likewise.
17246 (*sibcall_memory): Likewise.
17247 (*sibcall_value_memory): Likewise.
17248 Peepholes of indirect call and jump via memory: Likewise.
17249 * config/i386/i386.opt: Likewise.
17250 * config/i386/predicates.md (indirect_branch_operand): Likewise.
17251 (GOT_memory_operand): Likewise.
17252 (call_insn_operand): Likewise.
17253 (sibcall_insn_operand): Likewise.
17254 (GOT32_symbol_operand): Likewise.
17256 2018-01-15 Jakub Jelinek <jakub@redhat.com>
17258 PR middle-end/83837
17259 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
17260 type rather than type addr's type points to.
17261 (expand_omp_atomic_mutex): Likewise.
17262 (expand_omp_atomic): Likewise.
17264 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
17267 * config/i386/i386.c (output_indirect_thunk_function): Use
17268 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
17269 for __x86_return_thunk.
17271 2018-01-15 Richard Biener <rguenther@suse.de>
17273 PR middle-end/83850
17274 * expmed.c (extract_bit_field_1): Fix typo.
17276 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17279 * config/arm/iterators.md (VF): New mode iterator.
17280 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
17281 Remove integer-related logic from pattern.
17282 (neon_vabd<mode>_3): Likewise.
17284 2018-01-15 Jakub Jelinek <jakub@redhat.com>
17286 PR middle-end/82694
17287 * common.opt (fstrict-overflow): No longer an alias.
17288 (fwrapv-pointer): New option.
17289 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
17290 also for pointer types based on flag_wrapv_pointer.
17291 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
17292 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
17293 opts->x_flag_wrapv got set.
17294 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
17295 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
17296 POINTER_TYPE_OVERFLOW_UNDEFINED.
17297 * match.pd: Likewise in address comparison pattern.
17298 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
17300 2018-01-15 Richard Biener <rguenther@suse.de>
17303 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
17304 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
17305 Reset type names to their identifier if their TYPE_DECL doesn't
17306 have linkage (and thus is used for ODR and devirt).
17307 (save_debug_info_for_decl): Remove.
17308 (save_debug_info_for_type): Likewise.
17309 (add_tree_to_fld_list): Adjust.
17310 * tree-pretty-print.c (dump_generic_node): Make dumping of
17311 type names more robust.
17313 2018-01-15 Richard Biener <rguenther@suse.de>
17315 * BASE-VER: Bump to 8.0.1.
17317 2018-01-14 Martin Sebor <msebor@redhat.com>
17320 * builtins.c (check_access): Avoid warning when the no-warning bit
17323 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
17325 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
17326 * ira-color (allocno_hard_regs_compare): Likewise.
17328 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
17331 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
17332 Use .pushsection/.popsection.
17334 2018-01-14 Martin Sebor <msebor@redhat.com>
17337 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
17339 2018-01-14 Jakub Jelinek <jakub@redhat.com>
17341 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
17342 entry from extra_headers.
17343 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
17344 extra_headers, make the list bitwise identical to the i?86-*-* one.
17346 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
17348 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
17349 -mcmodel=large with -mindirect-branch=thunk,
17350 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
17351 -mfunction-return=thunk-extern.
17352 * doc/invoke.texi: Document -mcmodel=large is incompatible with
17353 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
17354 -mfunction-return=thunk and -mfunction-return=thunk-extern.
17356 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
17358 * config/i386/i386.c (print_reg): Print the name of the full
17359 integer register without '%'.
17360 (ix86_print_operand): Handle 'V'.
17361 * doc/extend.texi: Document 'V' modifier.
17363 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
17365 * config/i386/constraints.md (Bs): Disallow memory operand for
17366 -mindirect-branch-register.
17368 * config/i386/predicates.md (indirect_branch_operand): Likewise.
17369 (GOT_memory_operand): Likewise.
17370 (call_insn_operand): Likewise.
17371 (sibcall_insn_operand): Likewise.
17372 (GOT32_symbol_operand): Likewise.
17373 * config/i386/i386.md (indirect_jump): Call convert_memory_address
17374 for -mindirect-branch-register.
17375 (tablejump): Likewise.
17376 (*sibcall_memory): Likewise.
17377 (*sibcall_value_memory): Likewise.
17378 Disallow peepholes of indirect call and jump via memory for
17379 -mindirect-branch-register.
17380 (*call_pop): Replace m with Bw.
17381 (*call_value_pop): Likewise.
17382 (*sibcall_pop_memory): Replace m with Bs.
17383 * config/i386/i386.opt (mindirect-branch-register): New option.
17384 * doc/invoke.texi: Document -mindirect-branch-register option.
17386 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
17388 * config/i386/i386-protos.h (ix86_output_function_return): New.
17389 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
17390 set function_return_type.
17391 (indirect_thunk_name): Add ret_p to indicate thunk for function
17393 (output_indirect_thunk_function): Pass false to
17394 indirect_thunk_name.
17395 (ix86_output_indirect_branch_via_reg): Likewise.
17396 (ix86_output_indirect_branch_via_push): Likewise.
17397 (output_indirect_thunk_function): Create alias for function
17398 return thunk if regno < 0.
17399 (ix86_output_function_return): New function.
17400 (ix86_handle_fndecl_attribute): Handle function_return.
17401 (ix86_attribute_table): Add function_return.
17402 * config/i386/i386.h (machine_function): Add
17403 function_return_type.
17404 * config/i386/i386.md (simple_return_internal): Use
17405 ix86_output_function_return.
17406 (simple_return_internal_long): Likewise.
17407 * config/i386/i386.opt (mfunction-return=): New option.
17408 (indirect_branch): Mention -mfunction-return=.
17409 * doc/extend.texi: Document function_return function attribute.
17410 * doc/invoke.texi: Document -mfunction-return= option.
17412 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
17414 * config/i386/i386-opts.h (indirect_branch): New.
17415 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
17416 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
17417 with local indirect jump when converting indirect call and jump.
17418 (ix86_set_indirect_branch_type): New.
17419 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
17420 (indirectlabelno): New.
17421 (indirect_thunk_needed): Likewise.
17422 (indirect_thunk_bnd_needed): Likewise.
17423 (indirect_thunks_used): Likewise.
17424 (indirect_thunks_bnd_used): Likewise.
17425 (INDIRECT_LABEL): Likewise.
17426 (indirect_thunk_name): Likewise.
17427 (output_indirect_thunk): Likewise.
17428 (output_indirect_thunk_function): Likewise.
17429 (ix86_output_indirect_branch_via_reg): Likewise.
17430 (ix86_output_indirect_branch_via_push): Likewise.
17431 (ix86_output_indirect_branch): Likewise.
17432 (ix86_output_indirect_jmp): Likewise.
17433 (ix86_code_end): Call output_indirect_thunk_function if needed.
17434 (ix86_output_call_insn): Call ix86_output_indirect_branch if
17436 (ix86_handle_fndecl_attribute): Handle indirect_branch.
17437 (ix86_attribute_table): Add indirect_branch.
17438 * config/i386/i386.h (machine_function): Add indirect_branch_type
17439 and has_local_indirect_jump.
17440 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
17442 (tablejump): Likewise.
17443 (*indirect_jump): Use ix86_output_indirect_jmp.
17444 (*tablejump_1): Likewise.
17445 (simple_return_indirect_internal): Likewise.
17446 * config/i386/i386.opt (mindirect-branch=): New option.
17447 (indirect_branch): New.
17450 (thunk-inline): Likewise.
17451 (thunk-extern): Likewise.
17452 * doc/extend.texi: Document indirect_branch function attribute.
17453 * doc/invoke.texi: Document -mindirect-branch= option.
17455 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
17458 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
17460 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
17462 * ipa-inline.c (want_inline_small_function_p): Return false if
17463 inlining has already failed with CIF_FINAL_ERROR.
17464 (update_caller_keys): Call want_inline_small_function_p before
17466 (update_callee_keys): Likewise.
17468 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
17470 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
17472 (rs6000_quadword_masked_address_p): Likewise.
17473 (quad_aligned_load_p): Likewise.
17474 (quad_aligned_store_p): Likewise.
17475 (const_load_sequence_p): Add comment to describe the outer-most loop.
17476 (mimic_memory_attributes_and_flags): New function.
17477 (rs6000_gen_stvx): Likewise.
17478 (replace_swapped_aligned_store): Likewise.
17479 (rs6000_gen_lvx): Likewise.
17480 (replace_swapped_aligned_load): Likewise.
17481 (replace_swapped_load_constant): Capitalize argument name in
17482 comment describing this function.
17483 (rs6000_analyze_swaps): Add a third pass to search for vector loads
17484 and stores that access quad-word aligned addresses and replace
17485 with stvx or lvx instructions when appropriate.
17486 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
17487 New function prototype.
17488 (rs6000_quadword_masked_address_p): Likewise.
17489 (rs6000_gen_lvx): Likewise.
17490 (rs6000_gen_stvx): Likewise.
17491 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
17492 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
17493 when memory address is aligned.
17494 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
17495 this split to select lvx instruction when memory address is aligned.
17496 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
17497 instruction when memory address is aligned.
17498 (*vsx_le_perm_load_v16qi): Likewise.
17499 (four unnamed splitters): Modify to select the stvx instruction
17500 when memory is aligned.
17502 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
17504 * predict.c (determine_unlikely_bbs): Handle correctly BBs
17505 which appears in the queue multiple times.
17507 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17508 Alan Hayward <alan.hayward@arm.com>
17509 David Sherwood <david.sherwood@arm.com>
17511 * tree-vectorizer.h (vec_lower_bound): New structure.
17512 (_loop_vec_info): Add check_nonzero and lower_bounds.
17513 (LOOP_VINFO_CHECK_NONZERO): New macro.
17514 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
17515 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
17516 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
17517 fields. Make seg_len the distance travelled, not including the
17519 (dr_direction_indicator): Declare.
17520 (dr_zero_step_indicator): Likewise.
17521 (dr_known_forward_stride_p): Likewise.
17522 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
17524 (runtime_alias_check_p): Allow runtime alias checks with
17526 (operator ==): Compare access_size and align.
17527 (prune_runtime_alias_test_list): Rework for new distinction between
17528 the access_size and seg_len.
17529 (create_intersect_range_checks_index): Likewise. Cope with polynomial
17531 (get_segment_min_max): New function.
17532 (create_intersect_range_checks): Use it.
17533 (dr_step_indicator): New function.
17534 (dr_direction_indicator): Likewise.
17535 (dr_zero_step_indicator): Likewise.
17536 (dr_known_forward_stride_p): Likewise.
17537 * tree-loop-distribution.c (data_ref_segment_size): Return
17538 DR_STEP * (niters - 1).
17539 (compute_alias_check_pairs): Update call to the dr_with_seg_len
17541 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
17542 (vect_preserves_scalar_order_p): New function, split out from...
17543 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
17544 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
17545 (vect_vfa_access_size): New function.
17546 (vect_vfa_align): Likewise.
17547 (vect_compile_time_alias): Take access_size_a and access_b arguments.
17548 (dump_lower_bound): New function.
17549 (vect_check_lower_bound): Likewise.
17550 (vect_small_gap_p): Likewise.
17551 (vectorizable_with_step_bound_p): Likewise.
17552 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
17553 depencies if the vectorization factor is 1. Convert the checks
17554 for nonzero steps into checks on the bounds of DR_STEP. Try using
17555 a bunds check for variable steps if the minimum required step is
17556 relatively small. Update calls to the dr_with_seg_len
17557 constructor and to vect_compile_time_alias.
17558 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
17560 (vect_loop_versioning): Call it.
17561 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
17563 (vect_estimate_min_profitable_iters): Account for any bounds checks.
17565 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17566 Alan Hayward <alan.hayward@arm.com>
17567 David Sherwood <david.sherwood@arm.com>
17569 * doc/sourcebuild.texi (vect_scatter_store): Document.
17570 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
17572 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
17574 * genopinit.c (main): Add supports_vec_scatter_store and
17575 supports_vec_scatter_store_cached to target_optabs.
17576 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
17577 IFN_MASK_SCATTER_STORE.
17578 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
17580 * internal-fn.h (internal_store_fn_p): Declare.
17581 (internal_fn_stored_value_index): Likewise.
17582 * internal-fn.c (scatter_store_direct): New macro.
17583 (expand_scatter_store_optab_fn): New function.
17584 (direct_scatter_store_optab_supported_p): New macro.
17585 (internal_store_fn_p): New function.
17586 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
17587 IFN_MASK_SCATTER_STORE.
17588 (internal_fn_mask_index): Likewise.
17589 (internal_fn_stored_value_index): New function.
17590 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
17591 for scatter stores.
17592 * optabs-query.h (supports_vec_scatter_store_p): Declare.
17593 * optabs-query.c (supports_vec_scatter_store_p): New function.
17594 * tree-vectorizer.h (vect_get_store_rhs): Declare.
17595 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
17596 true for scatter stores.
17597 (vect_gather_scatter_fn_p): Handle scatter stores too.
17598 (vect_check_gather_scatter): Consider using scatter stores if
17599 supports_vec_scatter_store_p.
17600 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
17601 scatter stores too.
17602 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
17603 internal_fn_stored_value_index.
17604 (check_load_store_masking): Handle scatter stores too.
17605 (vect_get_store_rhs): Make public.
17606 (vectorizable_call): Use internal_store_fn_p.
17607 (vectorizable_store): Handle scatter store internal functions.
17608 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
17609 when deciding whether the end of the group has been reached.
17610 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
17611 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
17612 (mask_scatter_store<mode>): New insns.
17614 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17615 Alan Hayward <alan.hayward@arm.com>
17616 David Sherwood <david.sherwood@arm.com>
17618 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
17619 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
17620 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
17622 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
17623 Use vect_truncate_gather_scatter_offset if we can't treat the
17624 operation as a normal gather load or scatter store.
17625 (get_group_load_store_type): Take the gather_scatter_info
17626 as argument. Try using a gather load or scatter store for
17627 single-element groups.
17628 (get_load_store_type): Update calls to get_group_load_store_type
17629 and vect_use_strided_gather_scatters_p.
17631 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17632 Alan Hayward <alan.hayward@arm.com>
17633 David Sherwood <david.sherwood@arm.com>
17635 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
17636 optional tree argument.
17637 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
17639 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
17640 but continue to use the current value as a fallback.
17641 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
17642 to compare the updates.
17643 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
17644 (get_load_store_type): Use it when handling a strided access.
17645 (vect_get_strided_load_store_ops): New function.
17646 (vect_get_data_ptr_increment): Likewise.
17647 (vectorizable_load): Handle strided gather loads. Always pass
17648 a step to vect_create_data_ref_ptr and bump_vector_ptr.
17650 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17651 Alan Hayward <alan.hayward@arm.com>
17652 David Sherwood <david.sherwood@arm.com>
17654 * doc/md.texi (gather_load@var{m}): Document.
17655 (mask_gather_load@var{m}): Likewise.
17656 * genopinit.c (main): Add supports_vec_gather_load and
17657 supports_vec_gather_load_cached to target_optabs.
17658 * optabs-tree.c (init_tree_optimization_optabs): Use
17659 ggc_cleared_alloc to allocate target_optabs.
17660 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
17661 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
17663 * internal-fn.h (internal_load_fn_p): Declare.
17664 (internal_gather_scatter_fn_p): Likewise.
17665 (internal_fn_mask_index): Likewise.
17666 (internal_gather_scatter_fn_supported_p): Likewise.
17667 * internal-fn.c (gather_load_direct): New macro.
17668 (expand_gather_load_optab_fn): New function.
17669 (direct_gather_load_optab_supported_p): New macro.
17670 (direct_internal_fn_optab): New function.
17671 (internal_load_fn_p): Likewise.
17672 (internal_gather_scatter_fn_p): Likewise.
17673 (internal_fn_mask_index): Likewise.
17674 (internal_gather_scatter_fn_supported_p): Likewise.
17675 * optabs-query.c (supports_at_least_one_mode_p): New function.
17676 (supports_vec_gather_load_p): Likewise.
17677 * optabs-query.h (supports_vec_gather_load_p): Declare.
17678 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
17679 and memory_type field.
17680 (NUM_PATTERNS): Bump to 15.
17681 * tree-vect-data-refs.c: Include internal-fn.h.
17682 (vect_gather_scatter_fn_p): New function.
17683 (vect_describe_gather_scatter_call): Likewise.
17684 (vect_check_gather_scatter): Try using internal functions for
17685 gather loads. Recognize existing calls to a gather load function.
17686 (vect_analyze_data_refs): Consider using gather loads if
17687 supports_vec_gather_load_p.
17688 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
17689 (vect_get_gather_scatter_offset_type): Likewise.
17690 (vect_convert_mask_for_vectype): Likewise.
17691 (vect_add_conversion_to_patterm): Likewise.
17692 (vect_try_gather_scatter_pattern): Likewise.
17693 (vect_recog_gather_scatter_pattern): New pattern recognizer.
17694 (vect_vect_recog_func_ptrs): Add it.
17695 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
17696 internal_fn_mask_index and internal_gather_scatter_fn_p.
17697 (check_load_store_masking): Take the gather_scatter_info as an
17698 argument and handle gather loads.
17699 (vect_get_gather_scatter_ops): New function.
17700 (vectorizable_call): Check internal_load_fn_p.
17701 (vectorizable_load): Likewise. Handle gather load internal
17703 (vectorizable_store): Update call to check_load_store_masking.
17704 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
17705 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
17706 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
17707 (aarch64_gather_scale_operand_d): New predicates.
17708 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
17709 (mask_gather_load<mode>): New insns.
17711 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17712 Alan Hayward <alan.hayward@arm.com>
17713 David Sherwood <david.sherwood@arm.com>
17715 * optabs.def (fold_left_plus_optab): New optab.
17716 * doc/md.texi (fold_left_plus_@var{m}): Document.
17717 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
17718 * internal-fn.c (fold_left_direct): Define.
17719 (expand_fold_left_optab_fn): Likewise.
17720 (direct_fold_left_optab_supported_p): Likewise.
17721 * fold-const-call.c (fold_const_fold_left): New function.
17722 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
17723 * tree-parloops.c (valid_reduction_p): New function.
17724 (gather_scalar_reductions): Use it.
17725 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
17726 (vect_finish_replace_stmt): Declare.
17727 * tree-vect-loop.c (fold_left_reduction_fn): New function.
17728 (needs_fold_left_reduction_p): New function, split out from...
17729 (vect_is_simple_reduction): ...here. Accept reductions that
17730 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
17731 (vect_force_simple_reduction): Also store the reduction type in
17732 the assignment's STMT_VINFO_REDUC_TYPE.
17733 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
17734 (merge_with_identity): New function.
17735 (vect_expand_fold_left): Likewise.
17736 (vectorize_fold_left_reduction): Likewise.
17737 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
17738 scalar phi in place for it. Check for target support and reject
17739 cases that would reassociate the operation. Defer the transform
17740 phase to vectorize_fold_left_reduction.
17741 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
17742 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
17743 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
17745 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17747 * tree-if-conv.c (predicate_mem_writes): Remove redundant
17748 call to ifc_temp_var.
17750 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17751 Alan Hayward <alan.hayward@arm.com>
17752 David Sherwood <david.sherwood@arm.com>
17754 * target.def (legitimize_address_displacement): Take the original
17755 offset as a poly_int.
17756 * targhooks.h (default_legitimize_address_displacement): Update
17758 * targhooks.c (default_legitimize_address_displacement): Likewise.
17759 * doc/tm.texi: Regenerate.
17760 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
17761 as an argument, moving assert of ad->disp == ad->disp_term to...
17762 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
17763 Try calling targetm.legitimize_address_displacement before expanding
17764 the address rather than afterwards, and adjust for the new interface.
17765 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
17766 Match the new hook interface. Handle SVE addresses.
17767 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
17768 new hook interface.
17770 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17772 * Makefile.in (OBJS): Add early-remat.o.
17773 * target.def (select_early_remat_modes): New hook.
17774 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
17775 * doc/tm.texi: Regenerate.
17776 * targhooks.h (default_select_early_remat_modes): Declare.
17777 * targhooks.c (default_select_early_remat_modes): New function.
17778 * timevar.def (TV_EARLY_REMAT): New timevar.
17779 * passes.def (pass_early_remat): New pass.
17780 * tree-pass.h (make_pass_early_remat): Declare.
17781 * early-remat.c: New file.
17782 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
17784 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
17786 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17787 Alan Hayward <alan.hayward@arm.com>
17788 David Sherwood <david.sherwood@arm.com>
17790 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
17791 vfm1 with a bound_epilog parameter.
17792 (vect_do_peeling): Update calls accordingly, and move the prologue
17793 call earlier in the function. Treat the base bound_epilog as 0 for
17794 fully-masked loops and retain vf - 1 for other loops. Add 1 to
17795 this base when peeling for gaps.
17796 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
17797 with fully-masked loops.
17798 (vect_estimate_min_profitable_iters): Handle the single peeled
17799 iteration in that case.
17801 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17802 Alan Hayward <alan.hayward@arm.com>
17803 David Sherwood <david.sherwood@arm.com>
17805 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
17806 single-element interleaving even if the size is not a power of 2.
17807 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
17808 accesses for single-element interleaving if the group size is
17811 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17812 Alan Hayward <alan.hayward@arm.com>
17813 David Sherwood <david.sherwood@arm.com>
17815 * doc/md.texi (fold_extract_last_@var{m}): Document.
17816 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
17817 * optabs.def (fold_extract_last_optab): New optab.
17818 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
17819 * internal-fn.c (fold_extract_direct): New macro.
17820 (expand_fold_extract_optab_fn): Likewise.
17821 (direct_fold_extract_optab_supported_p): Likewise.
17822 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
17823 * tree-vect-loop.c (vect_model_reduction_cost): Handle
17824 EXTRACT_LAST_REDUCTION.
17825 (get_initial_def_for_reduction): Do not create an initial vector
17826 for EXTRACT_LAST_REDUCTION reductions.
17827 (vectorizable_reduction): Leave the scalar phi in place for
17828 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
17829 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
17830 epilogue code for EXTRACT_LAST_REDUCTION and defer the
17831 transform phase to vectorizable_condition.
17832 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
17834 (vect_finish_stmt_generation): ...here.
17835 (vect_finish_replace_stmt): New function.
17836 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
17837 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
17839 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
17841 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17842 Alan Hayward <alan.hayward@arm.com>
17843 David Sherwood <david.sherwood@arm.com>
17845 * doc/md.texi (extract_last_@var{m}): Document.
17846 * optabs.def (extract_last_optab): New optab.
17847 * internal-fn.def (EXTRACT_LAST): New internal function.
17848 * internal-fn.c (cond_unary_direct): New macro.
17849 (expand_cond_unary_optab_fn): Likewise.
17850 (direct_cond_unary_optab_supported_p): Likewise.
17851 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
17852 loops using EXTRACT_LAST.
17853 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
17854 (extract_last_<mode>): ...this optab.
17855 (vec_extract<mode><Vel>): Update accordingly.
17857 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17858 Alan Hayward <alan.hayward@arm.com>
17859 David Sherwood <david.sherwood@arm.com>
17861 * target.def (empty_mask_is_expensive): New hook.
17862 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
17863 * doc/tm.texi: Regenerate.
17864 * targhooks.h (default_empty_mask_is_expensive): Declare.
17865 * targhooks.c (default_empty_mask_is_expensive): New function.
17866 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
17867 if the target says that empty masks are expensive.
17868 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
17870 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
17872 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17873 Alan Hayward <alan.hayward@arm.com>
17874 David Sherwood <david.sherwood@arm.com>
17876 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
17877 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
17878 (vect_use_loop_mask_for_alignment_p): New function.
17879 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
17880 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
17881 niters_skip argument. Make sure that the first niters_skip elements
17882 of the first iteration are inactive.
17883 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
17884 Update call to vect_set_loop_masks_directly.
17885 (get_misalign_in_elems): New function, split out from...
17886 (vect_gen_prolog_loop_niters): ...here.
17887 (vect_update_init_of_dr): Take a code argument that specifies whether
17888 the adjustment should be added or subtracted.
17889 (vect_update_init_of_drs): Likewise.
17890 (vect_prepare_for_masked_peels): New function.
17891 (vect_do_peeling): Skip prologue peeling if we're using a mask
17892 instead. Update call to vect_update_inits_of_drs.
17893 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
17895 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
17896 alignment. Do not include the number of peeled iterations in
17897 the minimum threshold in that case.
17898 (vectorizable_induction): Adjust the start value down by
17899 LOOP_VINFO_MASK_SKIP_NITERS iterations.
17900 (vect_transform_loop): Call vect_prepare_for_masked_peels.
17901 Take the number of skipped iterations into account when calculating
17903 * tree-vect-stmts.c (vect_gen_while_not): New function.
17905 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17906 Alan Hayward <alan.hayward@arm.com>
17907 David Sherwood <david.sherwood@arm.com>
17909 * doc/sourcebuild.texi (vect_fully_masked): Document.
17910 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
17911 default value to 0.
17912 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
17914 (vect_analyze_loop_2): ...here. Don't check the vectorization
17915 factor against the number of loop iterations if the loop is
17918 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17919 Alan Hayward <alan.hayward@arm.com>
17920 David Sherwood <david.sherwood@arm.com>
17922 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
17923 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
17924 (dump_groups): Update accordingly.
17925 (iv_use::mem_type): New member variable.
17926 (address_p): New function.
17927 (record_use): Add a mem_type argument and initialize the new
17929 (record_group_use): Add a mem_type argument. Use address_p.
17930 Remove obsolete null checks of base_object. Update call to record_use.
17931 (find_interesting_uses_op): Update call to record_group_use.
17932 (find_interesting_uses_cond): Likewise.
17933 (find_interesting_uses_address): Likewise.
17934 (get_mem_type_for_internal_fn): New function.
17935 (find_address_like_use): Likewise.
17936 (find_interesting_uses_stmt): Try find_address_like_use before
17937 calling find_interesting_uses_op.
17938 (addr_offset_valid_p): Use the iv mem_type field as the type
17939 of the addressed memory.
17940 (add_autoinc_candidates): Likewise.
17941 (get_address_cost): Likewise.
17942 (split_small_address_groups_p): Use address_p.
17943 (split_address_groups): Likewise.
17944 (add_iv_candidate_for_use): Likewise.
17945 (autoinc_possible_for_pair): Likewise.
17946 (rewrite_groups): Likewise.
17947 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
17948 (determine_group_iv_cost): Update after split of USE_ADDRESS.
17949 (get_alias_ptr_type_for_ptr_address): New function.
17950 (rewrite_use_address): Rewrite address uses in calls that were
17951 identified by find_address_like_use.
17953 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17954 Alan Hayward <alan.hayward@arm.com>
17955 David Sherwood <david.sherwood@arm.com>
17957 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
17959 * gimple-expr.h (is_gimple_addressable: Likewise.
17960 * gimple-expr.c (is_gimple_address): Likewise.
17961 * internal-fn.c (expand_call_mem_ref): New function.
17962 (expand_mask_load_optab_fn): Use it.
17963 (expand_mask_store_optab_fn): Likewise.
17965 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17966 Alan Hayward <alan.hayward@arm.com>
17967 David Sherwood <david.sherwood@arm.com>
17969 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
17970 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
17971 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
17972 (cond_umax@var{mode}): Document.
17973 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
17974 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
17975 (cond_umin_optab, cond_umax_optab): New optabs.
17976 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
17977 (COND_IOR, COND_XOR): New internal functions.
17978 * internal-fn.h (get_conditional_internal_fn): Declare.
17979 * internal-fn.c (cond_binary_direct): New macro.
17980 (expand_cond_binary_optab_fn): Likewise.
17981 (direct_cond_binary_optab_supported_p): Likewise.
17982 (get_conditional_internal_fn): New function.
17983 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
17984 Cope with reduction statements that are vectorized as calls rather
17986 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
17987 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
17988 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
17989 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
17990 (UNSPEC_COND_EOR): New unspecs.
17991 (optab): Add mappings for them.
17992 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
17993 (sve_int_op, sve_fp_op): New int attributes.
17995 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17996 Alan Hayward <alan.hayward@arm.com>
17997 David Sherwood <david.sherwood@arm.com>
17999 * optabs.def (while_ult_optab): New optab.
18000 * doc/md.texi (while_ult@var{m}@var{n}): Document.
18001 * internal-fn.def (WHILE_ULT): New internal function.
18002 * internal-fn.h (direct_internal_fn_supported_p): New override
18003 that takes two types as argument.
18004 * internal-fn.c (while_direct): New macro.
18005 (expand_while_optab_fn): New function.
18006 (convert_optab_supported_p): Likewise.
18007 (direct_while_optab_supported_p): New macro.
18008 * wide-int.h (wi::udiv_ceil): New function.
18009 * tree-vectorizer.h (rgroup_masks): New structure.
18010 (vec_loop_masks): New typedef.
18011 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
18012 and fully_masked_p.
18013 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
18014 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
18015 (vect_max_vf): New function.
18016 (slpeel_make_loop_iterate_ntimes): Delete.
18017 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
18018 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
18019 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
18020 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
18021 internal-fn.h, stor-layout.h and optabs-query.h.
18022 (vect_set_loop_mask): New function.
18023 (add_preheader_seq): Likewise.
18024 (add_header_seq): Likewise.
18025 (interleave_supported_p): Likewise.
18026 (vect_maybe_permute_loop_masks): Likewise.
18027 (vect_set_loop_masks_directly): Likewise.
18028 (vect_set_loop_condition_masked): Likewise.
18029 (vect_set_loop_condition_unmasked): New function, split out from
18030 slpeel_make_loop_iterate_ntimes.
18031 (slpeel_make_loop_iterate_ntimes): Rename to..
18032 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
18033 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
18034 (vect_do_peeling): Update call accordingly.
18035 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
18037 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
18038 mask_compare_type, can_fully_mask_p and fully_masked_p.
18039 (release_vec_loop_masks): New function.
18040 (_loop_vec_info): Use it to free the loop masks.
18041 (can_produce_all_loop_masks_p): New function.
18042 (vect_get_max_nscalars_per_iter): Likewise.
18043 (vect_verify_full_masking): Likewise.
18044 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
18045 retries, and free the mask rgroups before retrying. Check loop-wide
18046 reasons for disallowing fully-masked loops. Make the final decision
18047 about whether use a fully-masked loop or not.
18048 (vect_estimate_min_profitable_iters): Do not assume that peeling
18049 for the number of iterations will be needed for fully-masked loops.
18050 (vectorizable_reduction): Disable fully-masked loops.
18051 (vectorizable_live_operation): Likewise.
18052 (vect_halve_mask_nunits): New function.
18053 (vect_double_mask_nunits): Likewise.
18054 (vect_record_loop_mask): Likewise.
18055 (vect_get_loop_mask): Likewise.
18056 (vect_transform_loop): Handle the case in which the final loop
18057 iteration might handle a partial vector. Call vect_set_loop_condition
18058 instead of slpeel_make_loop_iterate_ntimes.
18059 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
18060 (check_load_store_masking): New function.
18061 (prepare_load_store_mask): Likewise.
18062 (vectorizable_store): Handle fully-masked loops.
18063 (vectorizable_load): Likewise.
18064 (supportable_widening_operation): Use vect_halve_mask_nunits for
18066 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
18067 (vect_gen_while): New function.
18068 * config/aarch64/aarch64.md (umax<mode>3): New expander.
18069 (aarch64_uqdec<mode>): New insn.
18071 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18072 Alan Hayward <alan.hayward@arm.com>
18073 David Sherwood <david.sherwood@arm.com>
18075 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
18076 (reduc_xor_scal_optab): New optabs.
18077 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
18078 (reduc_xor_scal_@var{m}): Document.
18079 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
18080 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
18081 internal functions.
18082 * fold-const-call.c (fold_const_call): Handle them.
18083 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
18084 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
18085 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
18086 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
18087 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
18088 (UNSPEC_XORV): New unspecs.
18089 (optab): Add entries for them.
18090 (BITWISEV): New int iterator.
18091 (bit_reduc_op): New int attributes.
18093 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18094 Alan Hayward <alan.hayward@arm.com>
18095 David Sherwood <david.sherwood@arm.com>
18097 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
18098 * internal-fn.def (VEC_SHL_INSERT): New internal function.
18099 * optabs.def (vec_shl_insert_optab): New optab.
18100 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
18101 (duplicate_and_interleave): Likewise.
18102 * tree-vect-loop.c: Include internal-fn.h.
18103 (neutral_op_for_slp_reduction): New function, split out from
18104 get_initial_defs_for_reduction.
18105 (get_initial_def_for_reduction): Handle option 2 for variable-length
18106 vectors by loading the neutral value into a vector and then shifting
18107 the initial value into element 0.
18108 (get_initial_defs_for_reduction): Replace the code argument with
18109 the neutral value calculated by neutral_op_for_slp_reduction.
18110 Use gimple_build_vector for constant-length vectors.
18111 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
18112 but the first group_size elements have a neutral value.
18113 Use duplicate_and_interleave otherwise.
18114 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
18115 Update call to get_initial_defs_for_reduction. Handle SLP
18116 reductions for variable-length vectors by creating one vector
18117 result for each scalar result, with the elements associated
18118 with other scalar results stubbed out with the neutral value.
18119 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
18120 Require IFN_VEC_SHL_INSERT for double reductions on
18121 variable-length vectors, or SLP reductions that have
18122 a neutral value. Require can_duplicate_and_interleave_p
18123 support for variable-length unchained SLP reductions if there
18124 is no neutral value, such as for MIN/MAX reductions. Also require
18125 the number of vector elements to be a multiple of the number of
18126 SLP statements when doing variable-length unchained SLP reductions.
18127 Update call to vect_create_epilog_for_reduction.
18128 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
18129 and remove initial values.
18130 (duplicate_and_interleave): Make public.
18131 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
18132 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
18134 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18135 Alan Hayward <alan.hayward@arm.com>
18136 David Sherwood <david.sherwood@arm.com>
18138 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
18139 (can_duplicate_and_interleave_p): New function.
18140 (vect_get_and_check_slp_defs): Take the vector of statements
18141 rather than just the current one. Remove excess parentheses.
18142 Restriction rejectinon of vect_constant_def and vect_external_def
18143 for variable-length vectors to boolean types, or types for which
18144 can_duplicate_and_interleave_p is false.
18145 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
18146 (duplicate_and_interleave): New function.
18147 (vect_get_constant_vectors): Use gimple_build_vector for
18148 constant-length vectors and suitable variable-length constant
18149 vectors. Use duplicate_and_interleave for other variable-length
18150 vectors. Don't defer the update when inserting new statements.
18152 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18153 Alan Hayward <alan.hayward@arm.com>
18154 David Sherwood <david.sherwood@arm.com>
18156 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
18157 min_profitable_iters doesn't go negative.
18159 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18160 Alan Hayward <alan.hayward@arm.com>
18161 David Sherwood <david.sherwood@arm.com>
18163 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
18164 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
18165 * optabs.def (vec_mask_load_lanes_optab): New optab.
18166 (vec_mask_store_lanes_optab): Likewise.
18167 * internal-fn.def (MASK_LOAD_LANES): New internal function.
18168 (MASK_STORE_LANES): Likewise.
18169 * internal-fn.c (mask_load_lanes_direct): New macro.
18170 (mask_store_lanes_direct): Likewise.
18171 (expand_mask_load_optab_fn): Handle masked operations.
18172 (expand_mask_load_lanes_optab_fn): New macro.
18173 (expand_mask_store_optab_fn): Handle masked operations.
18174 (expand_mask_store_lanes_optab_fn): New macro.
18175 (direct_mask_load_lanes_optab_supported_p): Likewise.
18176 (direct_mask_store_lanes_optab_supported_p): Likewise.
18177 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
18179 (vect_load_lanes_supported): Likewise.
18180 * tree-vect-data-refs.c (strip_conversion): New function.
18181 (can_group_stmts_p): Likewise.
18182 (vect_analyze_data_ref_accesses): Use it instead of checking
18183 for a pair of assignments.
18184 (vect_store_lanes_supported): Take a masked_p parameter.
18185 (vect_load_lanes_supported): Likewise.
18186 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
18187 vect_store_lanes_supported and vect_load_lanes_supported.
18188 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
18189 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
18190 parameter. Don't allow gaps for masked accesses.
18191 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
18192 and vect_load_lanes_supported.
18193 (get_load_store_type): Take a masked_p parameter and update
18194 call to get_group_load_store_type.
18195 (vectorizable_store): Update call to get_load_store_type.
18196 Handle IFN_MASK_STORE_LANES.
18197 (vectorizable_load): Update call to get_load_store_type.
18198 Handle IFN_MASK_LOAD_LANES.
18200 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18201 Alan Hayward <alan.hayward@arm.com>
18202 David Sherwood <david.sherwood@arm.com>
18204 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
18206 * config/aarch64/aarch64-protos.h
18207 (aarch64_sve_struct_memory_operand_p): Declare.
18208 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
18209 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
18210 (VPRED, vpred): Handle SVE structure modes.
18211 * config/aarch64/constraints.md (Utx): New constraint.
18212 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
18213 (aarch64_sve_struct_nonimmediate_operand): New predicates.
18214 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
18215 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
18216 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
18217 structure modes. Split into pieces after RA.
18218 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
18219 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
18221 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
18222 SVE structure modes.
18223 (aarch64_classify_address): Likewise.
18224 (sizetochar): Move earlier in file.
18225 (aarch64_print_operand): Handle SVE register lists.
18226 (aarch64_array_mode): New function.
18227 (aarch64_sve_struct_memory_operand_p): Likewise.
18228 (TARGET_ARRAY_MODE): Redefine.
18230 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18231 Alan Hayward <alan.hayward@arm.com>
18232 David Sherwood <david.sherwood@arm.com>
18234 * target.def (array_mode): New target hook.
18235 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
18236 * doc/tm.texi: Regenerate.
18237 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
18238 * hooks.c (hook_optmode_mode_uhwi_none): New function.
18239 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
18240 targetm.array_mode.
18241 * stor-layout.c (mode_for_array): Likewise. Support polynomial
18244 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18245 Alan Hayward <alan.hayward@arm.com>
18246 David Sherwood <david.sherwood@arm.com>
18248 * fold-const.c (fold_binary_loc): Check the argument types
18249 rather than the result type when testing for a vector operation.
18251 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18253 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
18254 * doc/tm.texi: Regenerate.
18256 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18257 Alan Hayward <alan.hayward@arm.com>
18258 David Sherwood <david.sherwood@arm.com>
18260 * doc/invoke.texi (-msve-vector-bits=): Document new option.
18261 (sve): Document new AArch64 extension.
18262 * doc/md.texi (w): Extend the description of the AArch64
18263 constraint to include SVE vectors.
18264 (Upl, Upa): Document new AArch64 predicate constraints.
18265 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
18267 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
18268 (msve-vector-bits=): New option.
18269 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
18270 SVE when these are disabled.
18271 (sve): New extension.
18272 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
18273 modes. Adjust their number of units based on aarch64_sve_vg.
18274 (MAX_BITSIZE_MODE_ANY_MODE): Define.
18275 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
18276 aarch64_addr_query_type.
18277 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
18278 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
18279 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
18280 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
18281 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
18282 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
18283 (aarch64_simd_imm_zero_p): Delete.
18284 (aarch64_check_zero_based_sve_index_immediate): Declare.
18285 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
18286 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
18287 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
18288 (aarch64_sve_float_mul_immediate_p): Likewise.
18289 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
18290 rather than an rtx.
18291 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
18292 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
18293 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
18294 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
18295 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
18296 (aarch64_regmode_natural_size): Likewise.
18297 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
18298 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
18300 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
18301 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
18302 for VG and the SVE predicate registers.
18303 (V_ALIASES): Add a "z"-prefixed alias.
18304 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
18305 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
18306 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
18307 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
18308 (REG_CLASS_NAMES): Add entries for them.
18309 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
18310 and the predicate registers.
18311 (aarch64_sve_vg): Declare.
18312 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
18313 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
18314 (REGMODE_NATURAL_SIZE): Define.
18315 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
18317 * config/aarch64/aarch64.c: Include cfgrtl.h.
18318 (simd_immediate_info): Add a constructor for series vectors,
18319 and an associated step field.
18320 (aarch64_sve_vg): New variable.
18321 (aarch64_dbx_register_number): Handle VG and the predicate registers.
18322 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
18323 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
18324 (VEC_ANY_DATA, VEC_STRUCT): New constants.
18325 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
18326 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
18327 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
18328 (aarch64_get_mask_mode): New functions.
18329 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
18330 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
18331 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
18332 predicate modes and predicate registers. Explicitly restrict
18333 GPRs to modes of 16 bytes or smaller. Only allow FP registers
18334 to store a vector mode if it is recognized by
18335 aarch64_classify_vector_mode.
18336 (aarch64_regmode_natural_size): New function.
18337 (aarch64_hard_regno_caller_save_mode): Return the original mode
18339 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
18340 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
18341 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
18342 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
18344 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
18345 does not overlap dest if the function is frame-related. Handle
18347 (aarch64_split_add_offset): New function.
18348 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
18349 them aarch64_add_offset.
18350 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
18351 and update call to aarch64_sub_sp.
18352 (aarch64_add_cfa_expression): New function.
18353 (aarch64_expand_prologue): Pass extra temporary registers to the
18354 functions above. Handle the case in which we need to emit new
18355 DW_CFA_expressions for registers that were originally saved
18356 relative to the stack pointer, but now have to be expressed
18357 relative to the frame pointer.
18358 (aarch64_output_mi_thunk): Pass extra temporary registers to the
18360 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
18361 IP0 and IP1 values for SVE frames.
18362 (aarch64_expand_vec_series): New function.
18363 (aarch64_expand_sve_widened_duplicate): Likewise.
18364 (aarch64_expand_sve_const_vector): Likewise.
18365 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
18366 Handle SVE constants. Use emit_move_insn to move a force_const_mem
18367 into the register, rather than emitting a SET directly.
18368 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
18369 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
18370 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
18371 (offset_9bit_signed_scaled_p): New functions.
18372 (aarch64_replicate_bitmask_imm): New function.
18373 (aarch64_bitmask_imm): Use it.
18374 (aarch64_cannot_force_const_mem): Reject expressions involving
18375 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
18376 (aarch64_classify_index): Handle SVE indices, by requiring
18377 a plain register index with a scale that matches the element size.
18378 (aarch64_classify_address): Handle SVE addresses. Assert that
18379 the mode of the address is VOIDmode or an integer mode.
18380 Update call to aarch64_classify_symbol.
18381 (aarch64_classify_symbolic_expression): Update call to
18382 aarch64_classify_symbol.
18383 (aarch64_const_vec_all_in_range_p): New function.
18384 (aarch64_print_vector_float_operand): Likewise.
18385 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
18386 "vN" for FP registers with SVE modes. Handle (const ...) vectors
18387 and the FP immediates 1.0 and 0.5.
18388 (aarch64_print_address_internal): Handle SVE addresses.
18389 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
18390 (aarch64_regno_regclass): Handle predicate registers.
18391 (aarch64_secondary_reload): Handle big-endian reloads of SVE
18393 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
18394 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
18395 (aarch64_convert_sve_vector_bits): New function.
18396 (aarch64_override_options): Use it to handle -msve-vector-bits=.
18397 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
18398 rather than an rtx.
18399 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
18400 Handle SVE vector and predicate modes. Accept VL-based constants
18401 that need only one temporary register, and VL offsets that require
18402 no temporary registers.
18403 (aarch64_conditional_register_usage): Mark the predicate registers
18404 as fixed if SVE isn't available.
18405 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
18406 Return true for SVE vector and predicate modes.
18407 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
18408 rather than an unsigned int. Handle SVE modes.
18409 (aarch64_preferred_simd_mode): Update call accordingly. Handle
18411 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
18413 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
18414 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
18415 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
18416 (aarch64_sve_float_mul_immediate_p): New functions.
18417 (aarch64_sve_valid_immediate): New function.
18418 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
18419 Explicitly reject structure modes. Check for INDEX constants.
18420 Handle PTRUE and PFALSE constants.
18421 (aarch64_check_zero_based_sve_index_immediate): New function.
18422 (aarch64_simd_imm_zero_p): Delete.
18423 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
18424 vector modes. Accept constants in the range of CNT[BHWD].
18425 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
18426 ask for an Advanced SIMD mode.
18427 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
18428 (aarch64_simd_vector_alignment): Handle SVE predicates.
18429 (aarch64_vectorize_preferred_vector_alignment): New function.
18430 (aarch64_simd_vector_alignment_reachable): Use it instead of
18432 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
18433 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
18435 (MAX_VECT_LEN): Delete.
18436 (expand_vec_perm_d): Add a vec_flags field.
18437 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
18438 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
18439 (aarch64_evpc_ext): Don't apply a big-endian lane correction
18441 (aarch64_evpc_rev): Rename to...
18442 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
18443 (aarch64_evpc_rev_global): New function.
18444 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
18445 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
18447 (aarch64_evpc_sve_tbl): New function.
18448 (aarch64_expand_vec_perm_const_1): Update after rename of
18449 aarch64_evpc_rev. Handle SVE permutes too, trying
18450 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
18451 than aarch64_evpc_tbl.
18452 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
18453 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
18454 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
18455 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
18456 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
18457 (aarch64_expand_sve_vcond): New functions.
18458 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
18459 of aarch64_vector_mode_p.
18460 (aarch64_dwarf_poly_indeterminate_value): New function.
18461 (aarch64_compute_pressure_classes): Likewise.
18462 (aarch64_can_change_mode_class): Likewise.
18463 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
18464 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
18465 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
18466 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
18467 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
18468 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
18469 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
18470 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
18472 (Dn, Dl, Dr): Accept const as well as const_vector.
18473 (Dz): Likewise. Compare against CONST0_RTX.
18474 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
18475 of "vector" where appropriate.
18476 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
18477 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
18478 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
18479 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
18480 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
18481 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
18482 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
18483 (v_int_equiv): Extend to SVE modes.
18484 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
18486 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
18487 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
18488 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
18489 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
18490 (SVE_COND_FP_CMP): New int iterators.
18491 (perm_hilo): Handle the new unpack unspecs.
18492 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
18494 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
18495 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
18496 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
18497 (aarch64_equality_operator, aarch64_constant_vector_operand)
18498 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
18499 (aarch64_sve_nonimmediate_operand): Likewise.
18500 (aarch64_sve_general_operand): Likewise.
18501 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
18502 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
18503 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
18504 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
18505 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
18506 (aarch64_sve_float_arith_immediate): Likewise.
18507 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
18508 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
18509 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
18510 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
18511 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
18512 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
18513 (aarch64_sve_float_arith_operand): Likewise.
18514 (aarch64_sve_float_arith_with_sub_operand): Likewise.
18515 (aarch64_sve_float_mul_operand): Likewise.
18516 (aarch64_sve_vec_perm_operand): Likewise.
18517 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
18518 (aarch64_mov_operand): Accept const_poly_int and const_vector.
18519 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
18520 as well as const_vector.
18521 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
18522 in file. Use CONST0_RTX and CONSTM1_RTX.
18523 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
18524 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
18525 Use aarch64_simd_imm_zero.
18526 * config/aarch64/aarch64-sve.md: New file.
18527 * config/aarch64/aarch64.md: Include it.
18528 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
18529 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
18530 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
18531 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
18532 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
18533 (sve): New attribute.
18534 (enabled): Disable instructions with the sve attribute unless
18536 (movqi, movhi): Pass CONST_POLY_INT operaneds through
18537 aarch64_expand_mov_immediate.
18538 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
18539 CNT[BHSD] immediates.
18540 (movti): Split CONST_POLY_INT moves into two halves.
18541 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
18542 Split additions that need a temporary here if the destination
18543 is the stack pointer.
18544 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
18545 (*add<mode>3_poly_1): New instruction.
18546 (set_clobber_cc): New expander.
18548 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18550 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
18551 parameter and use it instead of GET_MODE_SIZE (innermode). Use
18552 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
18553 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
18554 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
18555 Change innermode from fixed_mode_size to machine_mode.
18556 (simplify_subreg): Update call accordingly. Handle a constant-sized
18557 subreg of a variable-length CONST_VECTOR.
18559 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18560 Alan Hayward <alan.hayward@arm.com>
18561 David Sherwood <david.sherwood@arm.com>
18563 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
18564 (add_offset_to_base): New function, split out from...
18565 (create_mem_ref): ...here. When handling a scale other than 1,
18566 check first whether the address is valid without the offset.
18567 Add it into the base if so, leaving the index and scale as-is.
18569 2018-01-12 Jakub Jelinek <jakub@redhat.com>
18572 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
18573 fold_for_warn before checking if arg2 is INTEGER_CST.
18575 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
18577 * config/rs6000/predicates.md (load_multiple_operation): Delete.
18578 (store_multiple_operation): Delete.
18579 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
18580 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
18581 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
18582 guarded by TARGET_STRING.
18583 (rs6000_output_load_multiple): Delete.
18584 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
18585 OPTION_MASK_STRING / TARGET_STRING handling.
18586 (print_operand) <'N', 'O'>: Add comment that these are unused now.
18587 (const rs6000_opt_masks) <"string">: Change mask to 0.
18588 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
18589 (MASK_STRING): Delete.
18590 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
18592 (load_multiple): Delete.
18599 (store_multiple): Delete.
18606 (movmemsi_8reg): Delete.
18607 (corresponding unnamed define_insn): Delete.
18608 (movmemsi_6reg): Delete.
18609 (corresponding unnamed define_insn): Delete.
18610 (movmemsi_4reg): Delete.
18611 (corresponding unnamed define_insn): Delete.
18612 (movmemsi_2reg): Delete.
18613 (corresponding unnamed define_insn): Delete.
18614 (movmemsi_1reg): Delete.
18615 (corresponding unnamed define_insn): Delete.
18616 * config/rs6000/rs6000.opt (mno-string): New.
18617 (mstring): Replace by deprecation warning stub.
18618 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
18620 2018-01-12 Jakub Jelinek <jakub@redhat.com>
18622 * regrename.c (regrename_do_replace): If replacing the same
18623 reg multiple times, try to reuse last created gen_raw_REG.
18626 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
18627 main to workaround a bug in GDB.
18629 2018-01-12 Tom de Vries <tom@codesourcery.com>
18632 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
18634 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
18636 PR rtl-optimization/80481
18637 * ira-color.c (get_cap_member): New function.
18638 (allocnos_conflict_by_live_ranges_p): Use it.
18639 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
18640 (setup_slot_coalesced_allocno_live_ranges): Ditto.
18642 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
18645 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
18646 (*saddl_se_1): Ditto.
18647 (*ssubsi_1): Ditto.
18648 (*ssubl_se_1): Ditto.
18650 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
18652 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
18653 rather than wi::to_widest for DR_INITs.
18654 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
18655 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
18656 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
18658 (vect_analyze_group_access_1): Note that here.
18660 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
18662 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
18663 polynomial type sizes.
18665 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
18667 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
18668 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
18669 (gimple_add_tmp_var): Likewise.
18671 2018-01-12 Martin Liska <mliska@suse.cz>
18673 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
18674 (gimple_alloc_sizes): Likewise.
18675 (dump_gimple_statistics): Use PRIu64 in printf format.
18676 * gimple.h: Change uint64_t to int.
18678 2018-01-12 Martin Liska <mliska@suse.cz>
18680 * tree-core.h: Use uint64_t instead of int.
18681 * tree.c (tree_node_counts): Likewise.
18682 (tree_node_sizes): Likewise.
18683 (dump_tree_statistics): Use PRIu64 in printf format.
18685 2018-01-12 Martin Liska <mliska@suse.cz>
18687 * Makefile.in: As qsort_chk is implemented in vec.c, add
18688 vec.o to linkage of gencfn-macros.
18689 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
18690 passing the info to record_node_allocation_statistics.
18691 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
18693 * ggc-common.c (struct ggc_usage): Add operator== and use
18694 it in operator< and compare function.
18695 * mem-stats.h (struct mem_usage): Likewise.
18696 * vec.c (struct vec_usage): Remove operator< and compare
18697 function. Can be simply inherited.
18699 2018-01-12 Martin Jambor <mjambor@suse.cz>
18702 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
18703 * tree-ssa-math-opts.c: Include domwalk.h.
18704 (convert_mult_to_fma_1): New function.
18705 (fma_transformation_info): New type.
18706 (fma_deferring_state): Likewise.
18707 (cancel_fma_deferring): New function.
18708 (result_of_phi): Likewise.
18709 (last_fma_candidate_feeds_initial_phi): Likewise.
18710 (convert_mult_to_fma): Added deferring logic, split actual
18711 transformation to convert_mult_to_fma_1.
18712 (math_opts_dom_walker): New type.
18713 (math_opts_dom_walker::after_dom_children): New method, body moved
18714 here from pass_optimize_widening_mul::execute, added deferring logic
18716 (pass_optimize_widening_mul::execute): Moved most of code to
18717 math_opts_dom_walker::after_dom_children.
18718 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
18719 * config/i386/i386.c (ix86_option_override_internal): Added
18720 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
18722 2018-01-12 Richard Biener <rguenther@suse.de>
18725 * dwarf2out.c (gen_variable_die): Do not reset old_die for
18726 inline instance vars.
18728 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
18731 * config/rx/rx.c (rx_is_restricted_memory_address):
18732 Handle SUBREG case.
18734 2018-01-12 Richard Biener <rguenther@suse.de>
18736 PR tree-optimization/80846
18737 * target.def (split_reduction): New target hook.
18738 * targhooks.c (default_split_reduction): New function.
18739 * targhooks.h (default_split_reduction): Declare.
18740 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
18741 target requests first reduce vectors by combining low and high
18743 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
18744 (get_vectype_for_scalar_type_and_size): Export.
18745 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
18746 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
18747 * doc/tm.texi: Regenerate.
18748 * config/i386/i386.c (ix86_split_reduction): Implement
18749 TARGET_VECTORIZE_SPLIT_REDUCTION.
18751 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
18754 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
18755 in PIC mode except for TARGET_VXWORKS_RTP.
18756 * config/sparc/sparc.c: Include cfgrtl.h.
18757 (TARGET_INIT_PIC_REG): Define.
18758 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
18759 (sparc_pic_register_p): New predicate.
18760 (sparc_legitimate_address_p): Use it.
18761 (sparc_legitimize_pic_address): Likewise.
18762 (sparc_delegitimize_address): Likewise.
18763 (sparc_mode_dependent_address_p): Likewise.
18764 (gen_load_pcrel_sym): Remove 4th parameter.
18765 (load_got_register): Adjust call to above. Remove obsolete stuff.
18766 (sparc_expand_prologue): Do not call load_got_register here.
18767 (sparc_flat_expand_prologue): Likewise.
18768 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
18769 (sparc_use_pseudo_pic_reg): New function.
18770 (sparc_init_pic_reg): Likewise.
18771 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
18772 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
18774 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
18776 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
18777 Add item for branch_cost.
18779 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
18781 PR rtl-optimization/83565
18782 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
18783 not extend the result to a larger mode for rotate operations.
18784 (num_sign_bit_copies1): Likewise.
18786 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
18789 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
18791 Use values-Xc.o for -pedantic.
18792 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
18794 2018-01-12 Martin Liska <mliska@suse.cz>
18797 * ipa-devirt.c (final_warning_record::grow_type_warnings):
18799 (possible_polymorphic_call_targets): Use it.
18800 (ipa_devirt): Likewise.
18802 2018-01-12 Martin Liska <mliska@suse.cz>
18804 * profile-count.h (enum profile_quality): Use 0 as invalid
18805 enum value of profile_quality.
18807 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
18809 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
18810 -mext-string options.
18812 2018-01-12 Richard Biener <rguenther@suse.de>
18814 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
18815 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
18816 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
18818 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
18820 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
18822 * configure.ac (--with-long-double-format): Add support for the
18823 configuration option to change the default long double format on
18825 * config.gcc (powerpc*-linux*-*): Likewise.
18826 * configure: Regenerate.
18827 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
18828 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
18829 used without modification.
18831 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18833 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
18834 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
18835 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
18836 MISC_BUILTIN_SPEC_BARRIER.
18837 (rs6000_init_builtins): Likewise.
18838 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
18840 (speculation_barrier): New define_insn.
18841 * doc/extend.texi: Document __builtin_speculation_barrier.
18843 2018-01-11 Jakub Jelinek <jakub@redhat.com>
18846 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
18847 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
18848 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
18850 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
18851 integral modes instead of "ss" and "sd".
18852 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
18853 vectors with 32-bit and 64-bit elements.
18854 (vecdupssescalarmodesuffix): New mode attribute.
18855 (vec_dup<mode>): Use it.
18857 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
18860 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
18861 frame if argument is passed on stack.
18863 2018-01-11 Jakub Jelinek <jakub@redhat.com>
18866 * ree.c (combine_reaching_defs): Optimize also
18867 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
18868 reg2=any_extend(exp); reg1=reg2;, formatting fix.
18870 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
18872 PR middle-end/83189
18873 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
18875 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
18877 PR middle-end/83718
18878 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
18879 after they are computed.
18881 2018-01-11 Bin Cheng <bin.cheng@arm.com>
18883 PR tree-optimization/83695
18884 * gimple-loop-linterchange.cc
18885 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
18886 reset cached scev information after interchange.
18887 (pass_linterchange::execute): Remove call to scev_reset_htab.
18889 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18891 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
18892 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
18893 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
18894 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
18895 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
18896 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
18897 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
18898 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
18899 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
18900 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
18901 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
18902 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
18903 (V_lane_reg): Likewise.
18904 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
18906 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
18907 (vfmal_lane_low<mode>_intrinsic,
18908 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
18909 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
18910 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
18911 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
18912 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
18913 vfmsl_lane_high<mode>_intrinsic): New define_insns.
18915 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18917 * config/arm/arm-cpus.in (fp16fml): New feature.
18918 (ALL_SIMD): Add fp16fml.
18919 (armv8.2-a): Add fp16fml as an option.
18920 (armv8.3-a): Likewise.
18921 (armv8.4-a): Add fp16fml as part of fp16.
18922 * config/arm/arm.h (TARGET_FP16FML): Define.
18923 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
18925 * config/arm/arm-modes.def (V2HF): Define.
18926 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
18927 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
18928 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
18929 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
18930 vfmsl_low, vfmsl_high): New set of builtins.
18931 * config/arm/iterators.md (PLUSMINUS): New code iterator.
18932 (vfml_op): New code attribute.
18933 (VFMLHALVES): New int iterator.
18934 (VFML, VFMLSEL): New mode attributes.
18935 (V_reg): Define mapping for V2HF.
18936 (V_hi, V_lo): New mode attributes.
18937 (VF_constraint): Likewise.
18938 (vfml_half, vfml_half_selector): New int attributes.
18939 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
18941 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
18942 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
18944 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
18945 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
18946 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
18947 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
18949 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
18950 Document new effective target and option set.
18952 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18954 * config/arm/arm-cpus.in (armv8_4): New feature.
18955 (ARMv8_4a): New fgroup.
18956 (armv8.4-a): New arch.
18957 * config/arm/arm-tables.opt: Regenerate.
18958 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
18959 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
18960 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
18961 Add matching rules for -march=armv8.4-a and extensions.
18962 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
18964 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
18967 * config/rx/rx.md (BW): New mode attribute.
18968 (sync_lock_test_and_setsi): Add mode suffix to insn output.
18970 2018-01-11 Richard Biener <rguenther@suse.de>
18972 PR tree-optimization/83435
18973 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
18974 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
18975 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
18977 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
18978 Alan Hayward <alan.hayward@arm.com>
18979 David Sherwood <david.sherwood@arm.com>
18981 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
18983 (aarch64_classify_address): Initialize it. Track polynomial offsets.
18984 (aarch64_print_address_internal): Use it to check for a zero offset.
18986 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
18987 Alan Hayward <alan.hayward@arm.com>
18988 David Sherwood <david.sherwood@arm.com>
18990 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
18991 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
18992 Return a poly_int64 rather than a HOST_WIDE_INT.
18993 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
18994 rather than a HOST_WIDE_INT.
18995 * config/aarch64/aarch64.h (aarch64_frame): Protect with
18996 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
18997 hard_fp_offset, frame_size, initial_adjust, callee_offset and
18998 final_offset from HOST_WIDE_INT to poly_int64.
18999 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
19000 to_constant when getting the number of units in an Advanced SIMD
19002 (aarch64_builtin_vectorized_function): Check for a constant number
19004 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
19006 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
19007 attribute instead of GET_MODE_NUNITS.
19008 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
19009 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
19010 GET_MODE_SIZE for fixed-size registers.
19011 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
19012 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
19013 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
19014 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
19015 (aarch64_print_operand, aarch64_print_address_internal)
19016 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
19017 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
19018 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
19019 Handle polynomial GET_MODE_SIZE.
19020 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
19021 wider than SImode without modification.
19022 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
19023 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
19024 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
19025 passing and returning SVE modes.
19026 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
19027 rather than GEN_INT.
19028 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
19029 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
19030 (aarch64_allocate_and_probe_stack_space): Likewise.
19031 (aarch64_layout_frame): Cope with polynomial offsets.
19032 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
19033 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
19034 polynomial offsets.
19035 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
19036 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
19037 poly_int64 rather than a HOST_WIDE_INT.
19038 (aarch64_get_separate_components, aarch64_process_components)
19039 (aarch64_expand_prologue, aarch64_expand_epilogue)
19040 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
19041 (aarch64_anchor_offset): New function, split out from...
19042 (aarch64_legitimize_address): ...here.
19043 (aarch64_builtin_vectorization_cost): Handle polynomial
19044 TYPE_VECTOR_SUBPARTS.
19045 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
19047 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
19048 number of elements from the PARALLEL rather than the mode.
19049 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
19050 rather than GET_MODE_BITSIZE.
19051 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
19052 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
19053 (aarch64_expand_vec_perm_const_1): Handle polynomial
19054 d->perm.length () and d->perm elements.
19055 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
19056 Apply to_constant to d->perm elements.
19057 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
19058 polynomial CONST_VECTOR_NUNITS.
19059 (aarch64_move_pointer): Take amount as a poly_int64 rather
19061 (aarch64_progress_pointer): Avoid temporary variable.
19062 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
19063 the mode attribute instead of GET_MODE.
19065 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
19066 Alan Hayward <alan.hayward@arm.com>
19067 David Sherwood <david.sherwood@arm.com>
19069 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
19070 x exists before using it.
19071 (aarch64_add_constant_internal): Rename to...
19072 (aarch64_add_offset_1): ...this. Replace regnum with separate
19073 src and dest rtxes. Handle the case in which they're different,
19074 including when the offset is zero. Replace scratchreg with an rtx.
19075 Use 2 additions if there is no spare register into which we can
19076 move a 16-bit constant.
19077 (aarch64_add_constant): Delete.
19078 (aarch64_add_offset): Replace reg with separate src and dest
19079 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
19080 Use aarch64_add_offset_1.
19081 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
19082 an rtx rather than an int. Take the delta as a poly_int64
19083 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
19084 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
19085 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
19086 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
19087 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
19088 and aarch64_add_sp.
19089 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
19090 aarch64_add_constant.
19092 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
19094 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
19095 Use scalar_float_mode.
19097 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
19099 * config/aarch64/aarch64-simd.md
19100 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
19101 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
19102 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
19103 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
19104 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
19105 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
19106 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
19107 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
19108 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
19109 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
19111 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
19114 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
19115 targ_options->x_arm_arch_string is non NULL.
19117 2018-01-11 Tamar Christina <tamar.christina@arm.com>
19119 * config/aarch64/aarch64.h
19120 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
19122 2018-01-11 Sudakshina Das <sudi.das@arm.com>
19125 * expmed.c (emit_store_flag_force): Swap if const op0
19126 and change VOIDmode to mode of op0.
19128 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
19130 PR rtl-optimization/83761
19131 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
19132 than bytes to mode_for_size.
19134 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
19136 PR middle-end/83189
19137 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
19138 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
19141 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
19143 PR middle-end/83575
19144 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
19145 when in layout mode.
19146 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
19147 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
19150 2018-01-10 Michael Collison <michael.collison@arm.com>
19152 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
19153 * config/aarch64/aarch64-option-extension.def: Add
19154 AARCH64_OPT_EXTENSION of 'fp16fml'.
19155 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
19156 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
19157 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
19158 * config/aarch64/constraints.md (Ui7): New constraint.
19159 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
19160 (VFMLA_SEL_W): Ditto.
19163 (VFMLA16_LOW): New int iterator.
19164 (VFMLA16_HIGH): Ditto.
19165 (UNSPEC_FMLAL): New unspec.
19166 (UNSPEC_FMLSL): Ditto.
19167 (UNSPEC_FMLAL2): Ditto.
19168 (UNSPEC_FMLSL2): Ditto.
19169 (f16mac): New code attribute.
19170 * config/aarch64/aarch64-simd-builtins.def
19171 (aarch64_fmlal_lowv2sf): Ditto.
19172 (aarch64_fmlsl_lowv2sf): Ditto.
19173 (aarch64_fmlalq_lowv4sf): Ditto.
19174 (aarch64_fmlslq_lowv4sf): Ditto.
19175 (aarch64_fmlal_highv2sf): Ditto.
19176 (aarch64_fmlsl_highv2sf): Ditto.
19177 (aarch64_fmlalq_highv4sf): Ditto.
19178 (aarch64_fmlslq_highv4sf): Ditto.
19179 (aarch64_fmlal_lane_lowv2sf): Ditto.
19180 (aarch64_fmlsl_lane_lowv2sf): Ditto.
19181 (aarch64_fmlal_laneq_lowv2sf): Ditto.
19182 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
19183 (aarch64_fmlalq_lane_lowv4sf): Ditto.
19184 (aarch64_fmlsl_lane_lowv4sf): Ditto.
19185 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
19186 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
19187 (aarch64_fmlal_lane_highv2sf): Ditto.
19188 (aarch64_fmlsl_lane_highv2sf): Ditto.
19189 (aarch64_fmlal_laneq_highv2sf): Ditto.
19190 (aarch64_fmlsl_laneq_highv2sf): Ditto.
19191 (aarch64_fmlalq_lane_highv4sf): Ditto.
19192 (aarch64_fmlsl_lane_highv4sf): Ditto.
19193 (aarch64_fmlalq_laneq_highv4sf): Ditto.
19194 (aarch64_fmlsl_laneq_highv4sf): Ditto.
19195 * config/aarch64/aarch64-simd.md:
19196 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
19197 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
19198 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
19199 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
19200 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
19201 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
19202 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
19203 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
19204 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
19205 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
19206 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
19207 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
19208 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
19209 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
19210 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
19211 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
19212 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
19213 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
19214 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
19215 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
19216 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
19217 (vfmlsl_low_u32): Ditto.
19218 (vfmlalq_low_u32): Ditto.
19219 (vfmlslq_low_u32): Ditto.
19220 (vfmlal_high_u32): Ditto.
19221 (vfmlsl_high_u32): Ditto.
19222 (vfmlalq_high_u32): Ditto.
19223 (vfmlslq_high_u32): Ditto.
19224 (vfmlal_lane_low_u32): Ditto.
19225 (vfmlsl_lane_low_u32): Ditto.
19226 (vfmlal_laneq_low_u32): Ditto.
19227 (vfmlsl_laneq_low_u32): Ditto.
19228 (vfmlalq_lane_low_u32): Ditto.
19229 (vfmlslq_lane_low_u32): Ditto.
19230 (vfmlalq_laneq_low_u32): Ditto.
19231 (vfmlslq_laneq_low_u32): Ditto.
19232 (vfmlal_lane_high_u32): Ditto.
19233 (vfmlsl_lane_high_u32): Ditto.
19234 (vfmlal_laneq_high_u32): Ditto.
19235 (vfmlsl_laneq_high_u32): Ditto.
19236 (vfmlalq_lane_high_u32): Ditto.
19237 (vfmlslq_lane_high_u32): Ditto.
19238 (vfmlalq_laneq_high_u32): Ditto.
19239 (vfmlslq_laneq_high_u32): Ditto.
19240 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
19241 (AARCH64_FL_FOR_ARCH8_4): New.
19242 (AARCH64_ISA_F16FML): New ISA flag.
19243 (TARGET_F16FML): New feature flag for fp16fml.
19244 (doc/invoke.texi): Document new fp16fml option.
19246 2018-01-10 Michael Collison <michael.collison@arm.com>
19248 * config/aarch64/aarch64-builtins.c:
19249 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
19250 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
19251 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
19252 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
19253 (AARCH64_ISA_SHA3): New ISA flag.
19254 (TARGET_SHA3): New feature flag for sha3.
19255 * config/aarch64/iterators.md (sha512_op): New int attribute.
19256 (CRYPTO_SHA512): New int iterator.
19257 (UNSPEC_SHA512H): New unspec.
19258 (UNSPEC_SHA512H2): Ditto.
19259 (UNSPEC_SHA512SU0): Ditto.
19260 (UNSPEC_SHA512SU1): Ditto.
19261 * config/aarch64/aarch64-simd-builtins.def
19262 (aarch64_crypto_sha512hqv2di): New builtin.
19263 (aarch64_crypto_sha512h2qv2di): Ditto.
19264 (aarch64_crypto_sha512su0qv2di): Ditto.
19265 (aarch64_crypto_sha512su1qv2di): Ditto.
19266 (aarch64_eor3qv8hi): Ditto.
19267 (aarch64_rax1qv2di): Ditto.
19268 (aarch64_xarqv2di): Ditto.
19269 (aarch64_bcaxqv8hi): Ditto.
19270 * config/aarch64/aarch64-simd.md:
19271 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
19272 (aarch64_crypto_sha512su0qv2di): Ditto.
19273 (aarch64_crypto_sha512su1qv2di): Ditto.
19274 (aarch64_eor3qv8hi): Ditto.
19275 (aarch64_rax1qv2di): Ditto.
19276 (aarch64_xarqv2di): Ditto.
19277 (aarch64_bcaxqv8hi): Ditto.
19278 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
19279 (vsha512h2q_u64): Ditto.
19280 (vsha512su0q_u64): Ditto.
19281 (vsha512su1q_u64): Ditto.
19282 (veor3q_u16): Ditto.
19283 (vrax1q_u64): Ditto.
19284 (vxarq_u64): Ditto.
19285 (vbcaxq_u16): Ditto.
19286 * config/arm/types.md (crypto_sha512): New type attribute.
19287 (crypto_sha3): Ditto.
19288 (doc/invoke.texi): Document new sha3 option.
19290 2018-01-10 Michael Collison <michael.collison@arm.com>
19292 * config/aarch64/aarch64-builtins.c:
19293 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
19294 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
19295 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
19296 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
19297 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
19298 (AARCH64_ISA_SM4): New ISA flag.
19299 (TARGET_SM4): New feature flag for sm4.
19300 * config/aarch64/aarch64-simd-builtins.def
19301 (aarch64_sm3ss1qv4si): Ditto.
19302 (aarch64_sm3tt1aq4si): Ditto.
19303 (aarch64_sm3tt1bq4si): Ditto.
19304 (aarch64_sm3tt2aq4si): Ditto.
19305 (aarch64_sm3tt2bq4si): Ditto.
19306 (aarch64_sm3partw1qv4si): Ditto.
19307 (aarch64_sm3partw2qv4si): Ditto.
19308 (aarch64_sm4eqv4si): Ditto.
19309 (aarch64_sm4ekeyqv4si): Ditto.
19310 * config/aarch64/aarch64-simd.md:
19311 (aarch64_sm3ss1qv4si): Ditto.
19312 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
19313 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
19314 (aarch64_sm4eqv4si): Ditto.
19315 (aarch64_sm4ekeyqv4si): Ditto.
19316 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
19317 (sm3part_op): Ditto.
19318 (CRYPTO_SM3TT): Ditto.
19319 (CRYPTO_SM3PART): Ditto.
19320 (UNSPEC_SM3SS1): New unspec.
19321 (UNSPEC_SM3TT1A): Ditto.
19322 (UNSPEC_SM3TT1B): Ditto.
19323 (UNSPEC_SM3TT2A): Ditto.
19324 (UNSPEC_SM3TT2B): Ditto.
19325 (UNSPEC_SM3PARTW1): Ditto.
19326 (UNSPEC_SM3PARTW2): Ditto.
19327 (UNSPEC_SM4E): Ditto.
19328 (UNSPEC_SM4EKEY): Ditto.
19329 * config/aarch64/constraints.md (Ui2): New constraint.
19330 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
19331 * config/arm/types.md (crypto_sm3): New type attribute.
19332 (crypto_sm4): Ditto.
19333 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
19334 (vsm3tt1aq_u32): Ditto.
19335 (vsm3tt1bq_u32): Ditto.
19336 (vsm3tt2aq_u32): Ditto.
19337 (vsm3tt2bq_u32): Ditto.
19338 (vsm3partw1q_u32): Ditto.
19339 (vsm3partw2q_u32): Ditto.
19340 (vsm4eq_u32): Ditto.
19341 (vsm4ekeyq_u32): Ditto.
19342 (doc/invoke.texi): Document new sm4 option.
19344 2018-01-10 Michael Collison <michael.collison@arm.com>
19346 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
19347 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
19348 (AARCH64_FL_FOR_ARCH8_4): New.
19349 (AARCH64_FL_V8_4): New flag.
19350 (doc/invoke.texi): Document new armv8.4-a option.
19352 2018-01-10 Michael Collison <michael.collison@arm.com>
19354 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
19355 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
19356 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
19357 * config/aarch64/aarch64-option-extension.def: Add
19358 AARCH64_OPT_EXTENSION of 'sha2'.
19359 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
19360 (crypto): Disable sha2 and aes if crypto disabled.
19361 (crypto): Enable aes and sha2 if enabled.
19362 (simd): Disable sha2 and aes if simd disabled.
19363 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
19365 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
19366 (TARGET_SHA2): New feature flag for sha2.
19367 (TARGET_AES): New feature flag for aes.
19368 * config/aarch64/aarch64-simd.md:
19369 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
19370 conditional on TARGET_AES.
19371 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
19372 (aarch64_crypto_sha1hsi): Make pattern conditional
19374 (aarch64_crypto_sha1hv4si): Ditto.
19375 (aarch64_be_crypto_sha1hv4si): Ditto.
19376 (aarch64_crypto_sha1su1v4si): Ditto.
19377 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
19378 (aarch64_crypto_sha1su0v4si): Ditto.
19379 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
19380 (aarch64_crypto_sha256su0v4si): Ditto.
19381 (aarch64_crypto_sha256su1v4si): Ditto.
19382 (doc/invoke.texi): Document new aes and sha2 options.
19384 2018-01-10 Martin Sebor <msebor@redhat.com>
19386 PR tree-optimization/83781
19387 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
19390 2018-01-11 Martin Sebor <msebor@gmail.com>
19391 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
19393 PR tree-optimization/83501
19394 PR tree-optimization/81703
19396 * tree-ssa-strlen.c (get_string_cst): Rename...
19397 (get_string_len): ...to this. Handle global constants.
19398 (handle_char_store): Adjust.
19400 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
19401 Jim Wilson <jimw@sifive.com>
19403 * config/riscv/riscv-protos.h (riscv_output_return): New.
19404 * config/riscv/riscv.c (struct machine_function): New naked_p field.
19405 (riscv_attribute_table, riscv_output_return),
19406 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
19407 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
19408 (riscv_compute_frame_info): Only compute frame->mask if not a naked
19410 (riscv_expand_prologue): Add early return for naked function.
19411 (riscv_expand_epilogue): Likewise.
19412 (riscv_function_ok_for_sibcall): Return false for naked function.
19413 (riscv_set_current_function): New.
19414 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
19415 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
19416 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
19417 * doc/extend.texi (RISC-V Function Attributes): New.
19419 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
19421 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
19422 check for 128-bit long double before checking TCmode.
19423 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
19424 128-bit long doubles before checking TFmode or TCmode.
19425 (FLOAT128_IBM_P): Likewise.
19427 2018-01-10 Martin Sebor <msebor@redhat.com>
19429 PR tree-optimization/83671
19430 * builtins.c (c_strlen): Unconditionally return zero for the empty
19432 Use -Warray-bounds for warnings.
19433 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
19434 for non-constant array indices with COMPONENT_REF, arrays of
19435 arrays, and pointers to arrays.
19436 (gimple_fold_builtin_strlen): Determine and set length range for
19437 non-constant character arrays.
19439 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
19441 PR middle-end/81897
19442 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
19445 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
19447 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
19449 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
19452 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
19453 VECTOR_MEM_ALTIVEC_OR_VSX_P.
19454 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
19455 indexed_or_indirect_operand predicate.
19456 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
19457 (*vsx_le_perm_load_v8hi): Likewise.
19458 (*vsx_le_perm_load_v16qi): Likewise.
19459 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
19460 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
19461 (*vsx_le_perm_store_v8hi): Likewise.
19462 (*vsx_le_perm_store_v16qi): Likewise.
19463 (eight unnamed splitters): Likewise.
19465 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
19467 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
19468 * config/rs6000/emmintrin.h: Likewise.
19469 * config/rs6000/mmintrin.h: Likewise.
19470 * config/rs6000/xmmintrin.h: Likewise.
19472 2018-01-10 David Malcolm <dmalcolm@redhat.com>
19475 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
19477 * tree.c (tree_nop_conversion): Return true for location wrapper
19479 (maybe_wrap_with_location): New function.
19480 (selftest::check_strip_nops): New function.
19481 (selftest::test_location_wrappers): New function.
19482 (selftest::tree_c_tests): Call it.
19483 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
19484 (maybe_wrap_with_location): New decl.
19485 (EXPR_LOCATION_WRAPPER_P): New macro.
19486 (location_wrapper_p): New inline function.
19487 (tree_strip_any_location_wrapper): New inline function.
19489 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
19492 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
19493 stack_realign_offset for the largest alignment of stack slot
19495 (ix86_find_max_used_stack_alignment): New function.
19496 (ix86_finalize_stack_frame_flags): Use it. Set
19497 max_used_stack_alignment if we don't realign stack.
19498 * config/i386/i386.h (machine_function): Add
19499 max_used_stack_alignment.
19501 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
19503 * config/arm/arm.opt (-mbranch-cost): New option.
19504 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
19507 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
19510 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
19511 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
19513 2018-01-10 Richard Biener <rguenther@suse.de>
19516 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
19517 early out so it also covers the case where we have a non-NULL
19520 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
19522 PR tree-optimization/83753
19523 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
19524 for non-strided grouped accesses if the number of elements is 1.
19526 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
19529 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
19530 * i386.h (TARGET_USE_GATHER): Define.
19531 * x86-tune.def (X86_TUNE_USE_GATHER): New.
19533 2018-01-10 Martin Liska <mliska@suse.cz>
19536 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
19537 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
19539 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
19540 CLEANUP_NO_PARTITIONING is not set.
19542 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
19544 * doc/rtl.texi: Remove documentation of (const ...) wrappers
19545 for vectors, as a partial revert of r254296.
19546 * rtl.h (const_vec_p): Delete.
19547 (const_vec_duplicate_p): Don't test for vector CONSTs.
19548 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
19549 * expmed.c (make_tree): Likewise.
19552 * common.md (E, F): Use CONSTANT_P instead of checking for
19554 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
19555 checking for CONST_VECTOR.
19557 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
19559 PR middle-end/83575
19560 * predict.c (force_edge_cold): Handle in more sane way edges
19561 with no prediction.
19563 2018-01-09 Carl Love <cel@us.ibm.com>
19565 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
19567 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
19568 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
19569 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
19570 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
19571 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
19572 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
19573 * config/rs6000/rs6000-protos.h: Add extern defition for
19574 rs6000_generate_float2_double_code.
19575 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
19577 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
19578 (float2_v2df): Add define_expand.
19580 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
19583 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
19584 op_mode in the force_to_mode call.
19586 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
19588 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
19589 instead of checking each element individually.
19590 (aarch64_evpc_uzp): Likewise.
19591 (aarch64_evpc_zip): Likewise.
19592 (aarch64_evpc_ext): Likewise.
19593 (aarch64_evpc_rev): Likewise.
19594 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
19595 instead of checking each element individually. Return true without
19597 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
19598 whether all selected elements come from the same input, instead of
19599 checking each element individually. Remove calls to gen_rtx_REG,
19600 start_sequence and end_sequence and instead assert that no rtl is
19603 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
19605 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
19606 order of HIGH and CONST checks.
19608 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
19610 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
19611 if the destination isn't an SSA_NAME.
19613 2018-01-09 Richard Biener <rguenther@suse.de>
19615 PR tree-optimization/83668
19616 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
19618 (canonicalize_loop_form): ... here, renamed from ...
19619 (canonicalize_loop_closed_ssa_form): ... this and amended to
19620 swap successor edges for loop exit blocks to make us use
19621 the RPO order we need for initial schedule generation.
19623 2018-01-09 Joseph Myers <joseph@codesourcery.com>
19625 PR tree-optimization/64811
19626 * match.pd: When optimizing comparisons with Inf, avoid
19627 introducing or losing exceptions from comparisons with NaN.
19629 2018-01-09 Martin Liska <mliska@suse.cz>
19632 * asan.c (shadow_mem_size): Add gcc_assert.
19634 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
19636 Don't save registers in main().
19639 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
19640 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
19641 * config/avr/avr.c (avr_set_current_function): Don't error if
19642 naked, OS_task or OS_main are specified at the same time.
19643 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
19645 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
19647 * common/config/avr/avr-common.c (avr_option_optimization_table):
19648 Switch on -mmain-is-OS_task for optimizing compilations.
19650 2018-01-09 Richard Biener <rguenther@suse.de>
19652 PR tree-optimization/83572
19653 * graphite.c: Include cfganal.h.
19654 (graphite_transform_loops): Connect infinite loops to exit
19655 and remove fake edges at the end.
19657 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
19659 * ipa-inline.c (edge_badness): Revert accidental checkin.
19661 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
19664 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
19665 symbols; not inline clones.
19667 2018-01-09 Jakub Jelinek <jakub@redhat.com>
19670 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
19671 hard registers. Formatting fixes.
19673 PR preprocessor/83722
19674 * gcc.c (try_generate_repro): Pass
19675 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
19676 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
19679 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
19680 Kito Cheng <kito.cheng@gmail.com>
19682 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
19683 (riscv_leaf_function_p): Delete.
19684 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
19686 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19688 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
19690 (do_ifelse): New function.
19691 (do_isel): New function.
19692 (do_sub3): New function.
19693 (do_add3): New function.
19694 (do_load_mask_compare): New function.
19695 (do_overlap_load_compare): New function.
19696 (expand_compare_loop): New function.
19697 (expand_block_compare): Call expand_compare_loop() when appropriate.
19698 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
19699 option description.
19700 (-mblock-compare-inline-loop-limit): New option.
19702 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19705 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
19706 Reverse order of second and third operands in first alternative.
19707 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
19708 of first and second elements in UNSPEC_VPERMR vector.
19709 (altivec_expand_vec_perm_le): Likewise.
19711 2018-01-08 Jeff Law <law@redhat.com>
19713 PR rtl-optimizatin/81308
19714 * tree-switch-conversion.c (cfg_altered): New file scoped static.
19715 (process_switch): If group_case_labels makes a change, then set
19717 (pass_convert_switch::execute): If a switch is converted, then
19718 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
19720 PR rtl-optimization/81308
19721 * recog.c (split_all_insns): Conditionally cleanup the CFG after
19724 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
19726 PR target/83663 - Revert r255946
19727 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
19728 generation for cases where splatting a value is not useful.
19729 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
19730 across a vec_duplicate and a paradoxical subreg forming a vector
19731 mode to a vec_concat.
19733 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19735 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
19736 -march=armv8.3-a variants.
19737 * config/arm/t-multilib: Likewise.
19738 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
19740 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19742 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
19744 (cceq_ior_compare_complement): Give it a name so I can use it, and
19745 change boolean_or_operator predicate to boolean_operator so it can
19746 be used to generate a crand.
19747 (eqne): New code iterator.
19748 (bd/bd_neg): New code_attrs.
19749 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
19750 a single define_insn.
19751 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
19752 decrement (bdnzt/bdnzf/bdzt/bdzf).
19753 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
19754 with the new names of the branch decrement patterns, and added the
19755 names of the branch decrement conditional patterns.
19757 2018-01-08 Richard Biener <rguenther@suse.de>
19759 PR tree-optimization/83563
19760 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
19763 2018-01-08 Richard Biener <rguenther@suse.de>
19765 PR middle-end/83713
19766 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
19768 2018-01-08 Richard Biener <rguenther@suse.de>
19770 PR tree-optimization/83685
19771 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
19772 references to abnormals.
19774 2018-01-08 Richard Biener <rguenther@suse.de>
19777 * dwarf2out.c (output_indirect_strings): Handle empty
19778 skeleton_debug_str_hash.
19779 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
19781 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
19783 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
19784 (emit_store_direct): Likewise.
19785 (arc_trampoline_adjust_address): Likewise.
19786 (arc_asm_trampoline_template): New function.
19787 (arc_initialize_trampoline): Use asm_trampoline_template.
19788 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
19789 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
19790 * config/arc/arc.md (flush_icache): Delete pattern.
19792 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
19794 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
19795 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
19798 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
19801 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
19802 by not USED_FOR_TARGET.
19803 (make_pass_resolve_sw_modes): Likewise.
19805 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
19807 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
19810 2018-01-08 Richard Biener <rguenther@suse.de>
19812 PR middle-end/83580
19813 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
19815 2018-01-08 Richard Biener <rguenther@suse.de>
19817 PR middle-end/83517
19818 * match.pd ((t * 2) / 2) -> t): Add missing :c.
19820 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
19822 PR middle-end/81897
19823 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
19824 basic blocks with a small number of successors.
19825 (convert_control_dep_chain_into_preds): Improve handling of
19827 (dump_predicates): Split apart into...
19828 (dump_pred_chain): ...here...
19829 (dump_pred_info): ...and here.
19830 (can_one_predicate_be_invalidated_p): Add debugging printfs.
19831 (can_chain_union_be_invalidated_p): Improve check for invalidation
19833 (uninit_uses_cannot_happen): Avoid unnecessary if
19834 convert_control_dep_chain_into_preds yielded nothing.
19836 2018-01-06 Martin Sebor <msebor@redhat.com>
19838 PR tree-optimization/83640
19839 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
19840 subtracting negative offset from size.
19841 (builtin_access::overlap): Adjust offset bounds of the access to fall
19842 within the size of the object if possible.
19844 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
19846 PR rtl-optimization/83699
19847 * expmed.c (extract_bit_field_1): Restrict the vector usage of
19848 extract_bit_field_as_subreg to cases in which the extracted
19849 value is also a vector.
19851 * lra-constraints.c (process_alt_operands): Test for the equivalence
19852 substitutions when detecting a possible reload cycle.
19854 2018-01-06 Jakub Jelinek <jakub@redhat.com>
19857 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
19858 by default if flag_selective_schedling{,2}. Formatting fixes.
19860 PR rtl-optimization/83682
19861 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
19862 if it has non-VECTOR_MODE element mode.
19863 (vec_duplicate_p): Likewise.
19865 PR middle-end/83694
19866 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
19867 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
19869 2018-01-05 Jakub Jelinek <jakub@redhat.com>
19872 * config/i386/i386-builtin.def
19873 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
19874 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
19875 Require also OPTION_MASK_ISA_AVX512F in addition to
19876 OPTION_MASK_ISA_GFNI.
19877 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
19878 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
19879 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
19880 to OPTION_MASK_ISA_GFNI.
19881 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
19882 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
19883 OPTION_MASK_ISA_AVX512BW.
19884 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
19885 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
19886 addition to OPTION_MASK_ISA_GFNI.
19887 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
19888 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
19889 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
19890 to OPTION_MASK_ISA_GFNI.
19891 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
19892 a requirement for all ISAs rather than any of them with a few
19894 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
19896 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
19897 bitmasks to be enabled with 3 exceptions, instead of requiring any
19898 enabled ISA with lots of exceptions.
19899 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
19900 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
19901 Change avx512bw in isa attribute to avx512f.
19902 * config/i386/sgxintrin.h: Add license boilerplate.
19903 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
19904 to __AVX512F__ and __AVX512VL to __AVX512VL__.
19905 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
19906 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
19908 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
19909 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
19910 temporarily sse2 rather than sse if not enabled already.
19913 * config/i386/sse.md (VI248_VLBW): Rename to ...
19914 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
19915 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
19916 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
19917 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
19918 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
19919 mode iterator instead of VI248_VLBW.
19921 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
19923 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
19924 (record_modified): Skip clobbers; add debug output.
19925 (param_change_prob): Use sreal frequencies.
19927 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
19929 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
19930 punt for user-aligned variables.
19932 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
19934 * tree-chrec.c (chrec_contains_symbols): Return true for
19937 2018-01-05 Sudakshina Das <sudi.das@arm.com>
19940 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
19941 of (x|y) == x for BICS pattern.
19943 2018-01-05 Jakub Jelinek <jakub@redhat.com>
19945 PR tree-optimization/83605
19946 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
19947 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
19950 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
19952 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
19953 * config/epiphany/rtems.h: New file.
19955 2018-01-04 Jakub Jelinek <jakub@redhat.com>
19956 Uros Bizjak <ubizjak@gmail.com>
19959 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
19960 QIreg_operand instead of register_operand predicate.
19961 * config/i386/i386.c (ix86_rop_should_change_byte_p,
19962 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
19963 comments instead of -fmitigate[-_]rop.
19965 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
19968 * cgraphunit.c (symbol_table::compile): Switch to text_section
19969 before calling assembly_start debug hook.
19970 * run-rtl-passes.c (run_rtl_passes): Likewise.
19973 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
19975 * tree-vrp.c (extract_range_from_binary_expr_1): Check
19976 range_int_cst_p rather than !symbolic_range_p before calling
19977 extract_range_from_multiplicative_op_1.
19979 2018-01-04 Jeff Law <law@redhat.com>
19981 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
19982 redundant test in assertion.
19984 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
19986 * doc/rtl.texi: Document machine_mode wrapper classes.
19988 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
19990 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
19991 using tree_to_uhwi.
19993 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
19995 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
19996 the VEC_PERM_EXPR fold to fail.
19998 2018-01-04 Jakub Jelinek <jakub@redhat.com>
20001 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
20002 to switched_sections.
20004 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
20007 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
20008 test for d.testing.
20010 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
20013 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
20014 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
20016 2018-01-04 Jakub Jelinek <jakub@redhat.com>
20019 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
20020 is BLKmode and bitpos not zero or mode change is needed.
20022 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
20025 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
20028 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
20031 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
20032 instead of MULT rtx. Update all corresponding splitters.
20033 (*saddl_se): Ditto.
20034 (*ssub<modesuffix>): Ditto.
20035 (*ssubl_se): Ditto.
20036 (*cmp_sadd_di): Update split patterns.
20037 (*cmp_sadd_si): Ditto.
20038 (*cmp_sadd_sidi): Ditto.
20039 (*cmp_ssub_di): Ditto.
20040 (*cmp_ssub_si): Ditto.
20041 (*cmp_ssub_sidi): Ditto.
20042 * config/alpha/predicates.md (const23_operand): New predicate.
20043 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
20044 Look for ASHIFT, not MULT inner operand.
20045 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
20047 2018-01-04 Martin Liska <mliska@suse.cz>
20049 PR gcov-profile/83669
20050 * gcov.c (output_intermediate_file): Add version to intermediate
20052 * doc/gcov.texi: Document new field 'version' in intermediate
20053 file format. Fix location of '-k' option of gcov command.
20055 2018-01-04 Martin Liska <mliska@suse.cz>
20058 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
20060 2018-01-04 Jakub Jelinek <jakub@redhat.com>
20062 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
20064 2018-01-03 Martin Sebor <msebor@redhat.com>
20066 PR tree-optimization/83655
20067 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
20068 checking calls with invalid arguments.
20070 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20072 * tree-vect-stmts.c (vect_get_store_rhs): New function.
20073 (vectorizable_mask_load_store): Delete.
20074 (vectorizable_call): Return false for masked loads and stores.
20075 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
20076 instead of gimple_assign_rhs1.
20077 (vectorizable_load): Handle IFN_MASK_LOAD.
20078 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
20080 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20082 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
20084 (vectorizable_mask_load_store): ...here.
20085 (vectorizable_load): ...and here.
20087 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20089 * tree-vect-stmts.c (vect_build_all_ones_mask)
20090 (vect_build_zero_merge_argument): New functions, split out from...
20091 (vectorizable_load): ...here.
20093 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20095 * tree-vect-stmts.c (vect_check_store_rhs): New function,
20097 (vectorizable_mask_load_store): ...here.
20098 (vectorizable_store): ...and here.
20100 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20102 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
20104 (vectorizable_mask_load_store): ...here.
20106 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20108 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
20109 (vect_model_store_cost): Take a vec_load_store_type instead of a
20111 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
20112 (vect_model_store_cost): Take a vec_load_store_type instead of a
20114 (vectorizable_mask_load_store): Update accordingly.
20115 (vectorizable_store): Likewise.
20116 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
20118 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20120 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
20121 IFN_MASK_LOAD calls here rather than...
20122 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
20124 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20125 Alan Hayward <alan.hayward@arm.com>
20126 David Sherwood <david.sherwood@arm.com>
20128 * expmed.c (extract_bit_field_1): For vector extracts,
20129 fall back to extract_bit_field_as_subreg if vec_extract
20132 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20133 Alan Hayward <alan.hayward@arm.com>
20134 David Sherwood <david.sherwood@arm.com>
20136 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
20137 they are variable or constant sized.
20138 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
20139 slots for constant-sized data.
20141 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20142 Alan Hayward <alan.hayward@arm.com>
20143 David Sherwood <david.sherwood@arm.com>
20145 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
20146 handling COND_EXPRs with boolean comparisons, try to find a better
20147 basis for the mask type than the boolean itself.
20149 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20151 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
20152 is calculated and how it can be overridden.
20153 * genmodes.c (max_bitsize_mode_any_mode): New variable.
20154 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
20156 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
20159 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20160 Alan Hayward <alan.hayward@arm.com>
20161 David Sherwood <david.sherwood@arm.com>
20163 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
20164 Remove the mode argument.
20165 (aarch64_simd_valid_immediate): Remove the mode and inverse
20167 * config/aarch64/iterators.md (bitsize): New iterator.
20168 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
20169 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
20170 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
20171 aarch64_simd_valid_immediate.
20172 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
20173 (aarch64_reg_or_bic_imm): Likewise.
20174 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
20175 with an insn_type enum and msl with a modifier_type enum.
20176 Replace element_width with a scalar_mode. Change the shift
20177 to unsigned int. Add constructors for scalar_float_mode and
20178 scalar_int_mode elements.
20179 (aarch64_vect_float_const_representable_p): Delete.
20180 (aarch64_can_const_movi_rtx_p)
20181 (aarch64_simd_scalar_immediate_valid_for_move)
20182 (aarch64_simd_make_constant): Update call to
20183 aarch64_simd_valid_immediate.
20184 (aarch64_advsimd_valid_immediate_hs): New function.
20185 (aarch64_advsimd_valid_immediate): Likewise.
20186 (aarch64_simd_valid_immediate): Remove mode and inverse
20187 arguments. Rewrite to use the above. Use const_vec_duplicate_p
20188 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
20189 and aarch64_float_const_representable_p on the result.
20190 (aarch64_output_simd_mov_immediate): Remove mode argument.
20191 Update call to aarch64_simd_valid_immediate and use of
20192 simd_immediate_info.
20193 (aarch64_output_scalar_simd_mov_immediate): Update call
20196 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20197 Alan Hayward <alan.hayward@arm.com>
20198 David Sherwood <david.sherwood@arm.com>
20200 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
20201 (mode_nunits): Likewise CONST_MODE_NUNITS.
20202 * machmode.def (ADJUST_NUNITS): Document.
20203 * genmodes.c (mode_data::need_nunits_adj): New field.
20204 (blank_mode): Update accordingly.
20205 (adj_nunits): New variable.
20206 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
20208 (emit_mode_size_inline): Set need_bytesize_adj for all modes
20209 listed in adj_nunits.
20210 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
20211 listed in adj_nunits. Don't emit case statements for such modes.
20212 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
20213 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
20214 nothing if adj_nunits is nonnull.
20215 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
20216 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
20217 (emit_mode_fbit): Update use of print_maybe_const_decl.
20218 (emit_move_size): Likewise. Treat the array as non-const
20220 (emit_mode_adjustments): Handle adj_nunits.
20222 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20224 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
20225 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
20226 (VECTOR_MODES): Use it.
20227 (make_vector_modes): Take the prefix as an argument.
20229 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20230 Alan Hayward <alan.hayward@arm.com>
20231 David Sherwood <david.sherwood@arm.com>
20233 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
20234 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
20235 for MODE_VECTOR_BOOL.
20236 * machmode.def (VECTOR_BOOL_MODE): Document.
20237 * genmodes.c (VECTOR_BOOL_MODE): New macro.
20238 (make_vector_bool_mode): New function.
20239 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
20241 * lto-streamer-in.c (lto_input_mode_table): Likewise.
20242 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
20244 * stor-layout.c (int_mode_for_mode): Likewise.
20245 * tree.c (build_vector_type_for_mode): Likewise.
20246 * varasm.c (output_constant_pool_2): Likewise.
20247 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
20248 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
20249 for MODE_VECTOR_BOOL.
20250 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
20251 of mode class checks.
20252 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
20253 instead of a list of mode class checks.
20254 (expand_vector_scalar_condition): Likewise.
20255 (type_for_widest_vector_mode): Handle BImode as an inner mode.
20257 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20258 Alan Hayward <alan.hayward@arm.com>
20259 David Sherwood <david.sherwood@arm.com>
20261 * machmode.h (mode_size): Change from unsigned short to
20263 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
20264 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
20265 or if measurement_type is not polynomial.
20266 (fixed_size_mode::includes_p): Check for constant-sized modes.
20267 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
20268 return a poly_uint16 rather than an unsigned short.
20269 (emit_mode_size): Change the type of mode_size from unsigned short
20270 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
20271 (emit_mode_adjustments): Cope with polynomial vector sizes.
20272 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
20274 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
20276 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
20277 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
20278 * caller-save.c (setup_save_areas): Likewise.
20279 (replace_reg_with_saved_mem): Likewise.
20280 * calls.c (emit_library_call_value_1): Likewise.
20281 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
20282 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
20283 (gen_lowpart_for_combine): Likewise.
20284 * convert.c (convert_to_integer_1): Likewise.
20285 * cse.c (equiv_constant, cse_insn): Likewise.
20286 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
20287 (cselib_subst_to_values): Likewise.
20288 * dce.c (word_dce_process_block): Likewise.
20289 * df-problems.c (df_word_lr_mark_ref): Likewise.
20290 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
20291 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
20292 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
20293 (rtl_for_decl_location): Likewise.
20294 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
20295 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
20296 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
20297 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
20298 (expand_expr_real_1): Likewise.
20299 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
20300 (pad_below): Likewise.
20301 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
20302 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
20303 * ira.c (get_subreg_tracking_sizes): Likewise.
20304 * ira-build.c (ira_create_allocno_objects): Likewise.
20305 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
20306 (ira_sort_regnos_for_alter_reg): Likewise.
20307 * ira-costs.c (record_operand_costs): Likewise.
20308 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
20309 (resolve_simple_move): Likewise.
20310 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
20311 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
20312 (lra_constraints): Likewise.
20313 (CONST_POOL_OK_P): Reject variable-sized modes.
20314 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
20315 (add_pseudo_to_slot, lra_spill): Likewise.
20316 * omp-low.c (omp_clause_aligned_alignment): Likewise.
20317 * optabs-query.c (get_best_extraction_insn): Likewise.
20318 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
20319 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
20320 (expand_mult_highpart, valid_multiword_target_p): Likewise.
20321 * recog.c (offsettable_address_addr_space_p): Likewise.
20322 * regcprop.c (maybe_mode_change): Likewise.
20323 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
20324 * regrename.c (build_def_use): Likewise.
20325 * regstat.c (dump_reg_info): Likewise.
20326 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
20327 (find_reloads, find_reloads_subreg_address): Likewise.
20328 * reload1.c (eliminate_regs_1): Likewise.
20329 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
20330 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
20331 (simplify_binary_operation_1, simplify_subreg): Likewise.
20332 * targhooks.c (default_function_arg_padding): Likewise.
20333 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
20334 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
20335 (verify_gimple_assign_ternary): Likewise.
20336 * tree-inline.c (estimate_move_cost): Likewise.
20337 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
20338 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
20339 (get_address_cost_ainc): Likewise.
20340 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
20341 (vect_supportable_dr_alignment): Likewise.
20342 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
20343 (vectorizable_reduction): Likewise.
20344 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
20345 (vectorizable_operation, vectorizable_load): Likewise.
20346 * tree.c (build_same_sized_truth_vector_type): Likewise.
20347 * valtrack.c (cleanup_auto_inc_dec): Likewise.
20348 * var-tracking.c (emit_note_insn_var_location): Likewise.
20349 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
20350 (ADDR_VEC_ALIGN): Likewise.
20352 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20353 Alan Hayward <alan.hayward@arm.com>
20354 David Sherwood <david.sherwood@arm.com>
20356 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
20358 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
20359 or if measurement_type is polynomial.
20360 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
20361 * combine.c (make_extraction): Likewise.
20362 * dse.c (find_shift_sequence): Likewise.
20363 * dwarf2out.c (mem_loc_descriptor): Likewise.
20364 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
20365 (extract_bit_field, extract_low_bits): Likewise.
20366 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
20367 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
20368 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
20369 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
20370 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
20371 * reload.c (find_reloads): Likewise.
20372 * reload1.c (alter_reg): Likewise.
20373 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
20374 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
20375 * tree-if-conv.c (predicate_mem_writes): Likewise.
20376 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
20377 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
20378 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
20379 * valtrack.c (dead_debug_insert_temp): Likewise.
20380 * varasm.c (mergeable_constant_section): Likewise.
20381 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
20383 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20384 Alan Hayward <alan.hayward@arm.com>
20385 David Sherwood <david.sherwood@arm.com>
20387 * expr.c (expand_assignment): Cope with polynomial mode sizes
20388 when assigning to a CONCAT.
20390 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20391 Alan Hayward <alan.hayward@arm.com>
20392 David Sherwood <david.sherwood@arm.com>
20394 * machmode.h (mode_precision): Change from unsigned short to
20396 (mode_to_precision): Return a poly_uint16 rather than an unsigned
20398 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
20399 or if measurement_type is not polynomial.
20400 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
20401 in which the mode is already known to be a scalar_int_mode.
20402 * genmodes.c (emit_mode_precision): Change the type of mode_precision
20403 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
20405 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
20406 for GET_MODE_PRECISION.
20407 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
20408 for GET_MODE_PRECISION.
20409 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
20411 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
20412 (expand_field_assignment, make_extraction): Likewise.
20413 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
20414 (get_last_value): Likewise.
20415 * convert.c (convert_to_integer_1): Likewise.
20416 * cse.c (cse_insn): Likewise.
20417 * expr.c (expand_expr_real_1): Likewise.
20418 * lra-constraints.c (simplify_operand_subreg): Likewise.
20419 * optabs-query.c (can_atomic_load_p): Likewise.
20420 * optabs.c (expand_atomic_load): Likewise.
20421 (expand_atomic_store): Likewise.
20422 * ree.c (combine_reaching_defs): Likewise.
20423 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
20424 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
20425 * tree.h (type_has_mode_precision_p): Likewise.
20426 * ubsan.c (instrument_si_overflow): Likewise.
20428 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20429 Alan Hayward <alan.hayward@arm.com>
20430 David Sherwood <david.sherwood@arm.com>
20432 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
20433 polynomial numbers of units.
20434 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
20435 (valid_vector_subparts_p): New function.
20436 (build_vector_type): Remove temporary shim and take the number
20437 of units as a poly_uint64 rather than an int.
20438 (build_opaque_vector_type): Take the number of units as a
20439 poly_uint64 rather than an int.
20440 * tree.c (build_vector_from_ctor): Handle polynomial
20441 TYPE_VECTOR_SUBPARTS.
20442 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
20443 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
20444 (build_vector_from_val): If the number of units is variable,
20445 use build_vec_duplicate_cst for constant operands and
20446 VEC_DUPLICATE_EXPR otherwise.
20447 (make_vector_type): Remove temporary is_constant ().
20448 (build_vector_type, build_opaque_vector_type): Take the number of
20449 units as a poly_uint64 rather than an int.
20450 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
20452 * cfgexpand.c (expand_debug_expr): Likewise.
20453 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
20454 (store_constructor, expand_expr_real_1): Likewise.
20455 (const_scalar_mask_from_tree): Likewise.
20456 * fold-const-call.c (fold_const_reduction): Likewise.
20457 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
20458 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
20459 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
20460 (fold_relational_const): Likewise.
20461 (native_interpret_vector): Likewise. Change the size from an
20462 int to an unsigned int.
20463 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
20464 TYPE_VECTOR_SUBPARTS.
20465 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
20466 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
20467 duplicating a non-constant operand into a variable-length vector.
20468 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
20469 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
20470 * ipa-icf.c (sem_variable::equals): Likewise.
20471 * match.pd: Likewise.
20472 * omp-simd-clone.c (simd_clone_subparts): Likewise.
20473 * print-tree.c (print_node): Likewise.
20474 * stor-layout.c (layout_type): Likewise.
20475 * targhooks.c (default_builtin_vectorization_cost): Likewise.
20476 * tree-cfg.c (verify_gimple_comparison): Likewise.
20477 (verify_gimple_assign_binary): Likewise.
20478 (verify_gimple_assign_ternary): Likewise.
20479 (verify_gimple_assign_single): Likewise.
20480 * tree-pretty-print.c (dump_generic_node): Likewise.
20481 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
20482 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
20483 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
20484 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
20485 (vect_shift_permute_load_chain): Likewise.
20486 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
20487 (expand_vector_condition, optimize_vector_constructor): Likewise.
20488 (lower_vec_perm, get_compute_type): Likewise.
20489 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
20490 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
20491 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
20492 (vect_recog_mask_conversion_pattern): Likewise.
20493 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
20494 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
20495 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
20496 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
20497 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
20498 (vectorizable_shift, vectorizable_operation, vectorizable_store)
20499 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
20500 (supportable_widening_operation): Likewise.
20501 (supportable_narrowing_operation): Likewise.
20502 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
20504 * varasm.c (output_constant): Likewise.
20506 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20507 Alan Hayward <alan.hayward@arm.com>
20508 David Sherwood <david.sherwood@arm.com>
20510 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
20511 so that both the length == 3 and length != 3 cases set up their
20512 own permute vectors. Add comments explaining why we know the
20513 number of elements is constant.
20514 (vect_permute_load_chain): Likewise.
20516 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20517 Alan Hayward <alan.hayward@arm.com>
20518 David Sherwood <david.sherwood@arm.com>
20520 * machmode.h (mode_nunits): Change from unsigned char to
20522 (ONLY_FIXED_SIZE_MODES): New macro.
20523 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
20524 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
20525 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
20527 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
20528 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
20529 or if measurement_type is not polynomial.
20530 * genmodes.c (ZERO_COEFFS): New macro.
20531 (emit_mode_nunits_inline): Make mode_nunits_inline return a
20533 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
20534 Use ZERO_COEFFS when emitting initializers.
20535 * data-streamer.h (bp_pack_poly_value): New function.
20536 (bp_unpack_poly_value): Likewise.
20537 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
20538 for GET_MODE_NUNITS.
20539 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
20540 for GET_MODE_NUNITS.
20541 * tree.c (make_vector_type): Remove temporary shim and make
20542 the real function take the number of units as a poly_uint64
20543 rather than an int.
20544 (build_vector_type_for_mode): Handle polynomial nunits.
20545 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
20546 * emit-rtl.c (const_vec_series_p_1): Likewise.
20547 (gen_rtx_CONST_VECTOR): Likewise.
20548 * fold-const.c (test_vec_duplicate_folding): Likewise.
20549 * genrecog.c (validate_pattern): Likewise.
20550 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
20551 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
20552 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
20553 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
20554 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
20555 * rtlanal.c (subreg_get_info): Likewise.
20556 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
20557 (vect_grouped_load_supported): Likewise.
20558 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
20559 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
20560 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
20561 (simplify_const_unary_operation, simplify_binary_operation_1)
20562 (simplify_const_binary_operation, simplify_ternary_operation)
20563 (test_vector_ops_duplicate, test_vector_ops): Likewise.
20564 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
20565 instead of CONST_VECTOR_NUNITS.
20566 * varasm.c (output_constant_pool_2): Likewise.
20567 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
20568 explicit-encoded elements in the XVEC for variable-length vectors.
20570 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20572 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
20574 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20575 Alan Hayward <alan.hayward@arm.com>
20576 David Sherwood <david.sherwood@arm.com>
20578 * coretypes.h (fixed_size_mode): Declare.
20579 (fixed_size_mode_pod): New typedef.
20580 * builtins.h (target_builtins::x_apply_args_mode)
20581 (target_builtins::x_apply_result_mode): Change type to
20582 fixed_size_mode_pod.
20583 * builtins.c (apply_args_size, apply_result_size, result_vector)
20584 (expand_builtin_apply_args_1, expand_builtin_apply)
20585 (expand_builtin_return): Update accordingly.
20587 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20589 * cse.c (hash_rtx_cb): Hash only the encoded elements.
20590 * cselib.c (cselib_hash_rtx): Likewise.
20591 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
20592 CONST_VECTOR encoding.
20594 2018-01-03 Jakub Jelinek <jakub@redhat.com>
20595 Jeff Law <law@redhat.com>
20598 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
20599 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
20600 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
20601 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
20604 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
20605 explicitly probe *sp in a noreturn function if there were any callee
20606 register saves or frame pointer is needed.
20608 2018-01-03 Jakub Jelinek <jakub@redhat.com>
20611 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
20612 BLKmode for ternary, binary or unary expressions.
20615 * var-tracking.c (delete_vta_debug_insn): New inline function.
20616 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
20617 insns from get_insns () to NULL instead of each bb separately.
20618 Use delete_vta_debug_insn. No longer static.
20619 (vt_debug_insns_local, variable_tracking_main_1): Adjust
20620 delete_vta_debug_insns callers.
20621 * rtl.h (delete_vta_debug_insns): Declare.
20622 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
20623 instead of variable_tracking_main.
20625 2018-01-03 Martin Sebor <msebor@redhat.com>
20627 PR tree-optimization/83603
20628 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
20629 arguments past the endof the argument list in functions declared
20630 without a prototype.
20631 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
20632 Avoid checking when arguments are null.
20634 2018-01-03 Martin Sebor <msebor@redhat.com>
20637 * doc/extend.texi (attribute const): Fix a typo.
20638 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
20639 issuing -Wsuggest-attribute for void functions.
20641 2018-01-03 Martin Sebor <msebor@redhat.com>
20643 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
20644 offset_int::from instead of wide_int::to_shwi.
20645 (maybe_diag_overlap): Remove assertion.
20646 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
20647 * gimple-ssa-sprintf.c (format_directive): Same.
20648 (parse_directive): Same.
20649 (sprintf_dom_walker::compute_format_length): Same.
20650 (try_substitute_return_value): Same.
20652 2018-01-03 Jeff Law <law@redhat.com>
20654 PR middle-end/83654
20655 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
20656 non-constant residual for zero at runtime and avoid probing in
20657 that case. Reorganize code for trailing problem to mirror handling
20660 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
20662 PR tree-optimization/83501
20663 * tree-ssa-strlen.c (get_string_cst): New.
20664 (handle_char_store): Call get_string_cst.
20666 2018-01-03 Martin Liska <mliska@suse.cz>
20668 PR tree-optimization/83593
20669 * tree-ssa-strlen.c: Include tree-cfg.h.
20670 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
20671 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
20672 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
20674 (strlen_dom_walker::before_dom_children): Call
20675 gimple_purge_dead_eh_edges. Dump tranformation with details
20677 (strlen_dom_walker::before_dom_children): Update call by adding
20678 new argument cleanup_eh.
20679 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
20681 2018-01-03 Martin Liska <mliska@suse.cz>
20684 * cif-code.def (VARIADIC_THUNK): New enum value.
20685 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
20688 2018-01-03 Jan Beulich <jbeulich@suse.com>
20690 * sse.md (mov<mode>_internal): Tighten condition for when to use
20691 vmovdqu<ssescalarsize> for TI and OI modes.
20693 2018-01-03 Jakub Jelinek <jakub@redhat.com>
20695 Update copyright years.
20697 2018-01-03 Martin Liska <mliska@suse.cz>
20700 * ipa-visibility.c (function_and_variable_visibility): Skip
20701 functions with noipa attribure.
20703 2018-01-03 Jakub Jelinek <jakub@redhat.com>
20705 * gcc.c (process_command): Update copyright notice dates.
20706 * gcov-dump.c (print_version): Ditto.
20707 * gcov.c (print_version): Ditto.
20708 * gcov-tool.c (print_version): Ditto.
20709 * gengtype.c (create_file): Ditto.
20710 * doc/cpp.texi: Bump @copying's copyright year.
20711 * doc/cppinternals.texi: Ditto.
20712 * doc/gcc.texi: Ditto.
20713 * doc/gccint.texi: Ditto.
20714 * doc/gcov.texi: Ditto.
20715 * doc/install.texi: Ditto.
20716 * doc/invoke.texi: Ditto.
20718 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20720 * vector-builder.h (vector_builder::m_full_nelts): Change from
20721 unsigned int to poly_uint64.
20722 (vector_builder::full_nelts): Update prototype accordingly.
20723 (vector_builder::new_vector): Likewise.
20724 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
20725 (vector_builder::operator ==): Likewise.
20726 (vector_builder::finalize): Likewise.
20727 * int-vector-builder.h (int_vector_builder::int_vector_builder):
20728 Take the number of elements as a poly_uint64 rather than an
20730 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
20731 from unsigned int to poly_uint64.
20732 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
20733 (vec_perm_indices::new_vector): Likewise.
20734 (vec_perm_indices::length): Likewise.
20735 (vec_perm_indices::nelts_per_input): Likewise.
20736 (vec_perm_indices::input_nelts): Likewise.
20737 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
20738 number of elements per input as a poly_uint64 rather than an
20739 unsigned int. Use the original encoding for variable-length
20740 vectors, rather than clamping each individual element.
20741 For the second and subsequent elements in each pattern,
20742 clamp the step and base before clamping their sum.
20743 (vec_perm_indices::series_p): Handle polynomial element counts.
20744 (vec_perm_indices::all_in_range_p): Likewise.
20745 (vec_perm_indices_to_tree): Likewise.
20746 (vec_perm_indices_to_rtx): Likewise.
20747 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
20748 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
20749 (tree_vector_builder::new_binary_operation): Handle polynomial
20750 element counts. Return false if we need to know the number
20751 of elements at compile time.
20752 * fold-const.c (fold_vec_perm): Punt if the number of elements
20753 isn't known at compile time.
20755 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20757 * vec-perm-indices.h (vec_perm_builder): Change element type
20758 from HOST_WIDE_INT to poly_int64.
20759 (vec_perm_indices::element_type): Update accordingly.
20760 (vec_perm_indices::clamp): Handle polynomial element_types.
20761 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
20762 (vec_perm_indices::all_in_range_p): Likewise.
20763 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
20765 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
20766 polynomial vec_perm_indices element types.
20767 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
20768 * fold-const.c (fold_vec_perm): Likewise.
20769 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
20770 * tree-vect-generic.c (lower_vec_perm): Likewise.
20771 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
20772 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
20773 element type to HOST_WIDE_INT.
20775 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20776 Alan Hayward <alan.hayward@arm.com>
20777 David Sherwood <david.sherwood@arm.com>
20779 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
20780 rather than an int. Use plus_constant.
20781 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
20782 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
20784 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20785 Alan Hayward <alan.hayward@arm.com>
20786 David Sherwood <david.sherwood@arm.com>
20788 * calls.c (emit_call_1, expand_call): Change struct_value_size from
20789 a HOST_WIDE_INT to a poly_int64.
20791 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20792 Alan Hayward <alan.hayward@arm.com>
20793 David Sherwood <david.sherwood@arm.com>
20795 * calls.c (load_register_parameters): Cope with polynomial
20796 mode sizes. Require a constant size for BLKmode parameters
20797 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
20798 forces a parameter to be padded at the lsb end in order to
20799 fill a complete number of words, require the parameter size
20800 to be ordered wrt UNITS_PER_WORD.
20802 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20803 Alan Hayward <alan.hayward@arm.com>
20804 David Sherwood <david.sherwood@arm.com>
20806 * reload1.c (spill_stack_slot_width): Change element type
20807 from unsigned int to poly_uint64_pod.
20808 (alter_reg): Treat mode sizes as polynomial.
20810 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20811 Alan Hayward <alan.hayward@arm.com>
20812 David Sherwood <david.sherwood@arm.com>
20814 * reload.c (complex_word_subreg_p): New function.
20815 (reload_inner_reg_of_subreg, push_reload): Use it.
20817 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20818 Alan Hayward <alan.hayward@arm.com>
20819 David Sherwood <david.sherwood@arm.com>
20821 * lra-constraints.c (process_alt_operands): Reject matched
20822 operands whose sizes aren't ordered.
20823 (match_reload): Refer to this check here.
20825 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20826 Alan Hayward <alan.hayward@arm.com>
20827 David Sherwood <david.sherwood@arm.com>
20829 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
20830 that the mode size is in the set {1, 2, 4, 8, 16}.
20832 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20833 Alan Hayward <alan.hayward@arm.com>
20834 David Sherwood <david.sherwood@arm.com>
20836 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
20837 Use plus_constant instead of gen_rtx_PLUS.
20839 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20840 Alan Hayward <alan.hayward@arm.com>
20841 David Sherwood <david.sherwood@arm.com>
20843 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
20844 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
20845 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
20846 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
20847 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
20848 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
20849 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
20850 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
20851 * config/i386/i386.c (ix86_push_rounding): ...this new function.
20852 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
20854 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
20855 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
20856 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
20857 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
20858 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
20859 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
20860 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
20861 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
20862 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
20863 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
20865 * expr.c (emit_move_resolve_push): Treat the input and result
20866 of PUSH_ROUNDING as a poly_int64.
20867 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
20868 (emit_push_insn): Likewise.
20869 * lra-eliminations.c (mark_not_eliminable): Likewise.
20870 * recog.c (push_operand): Likewise.
20871 * reload1.c (elimination_effects): Likewise.
20872 * rtlanal.c (nonzero_bits1): Likewise.
20873 * calls.c (store_one_arg): Likewise. Require the padding to be
20874 known at compile time.
20876 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20877 Alan Hayward <alan.hayward@arm.com>
20878 David Sherwood <david.sherwood@arm.com>
20880 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
20881 Use plus_constant instead of gen_rtx_PLUS.
20883 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20884 Alan Hayward <alan.hayward@arm.com>
20885 David Sherwood <david.sherwood@arm.com>
20887 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
20888 rather than an int.
20890 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20891 Alan Hayward <alan.hayward@arm.com>
20892 David Sherwood <david.sherwood@arm.com>
20894 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
20895 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
20896 via stack temporaries. Treat the mode size as polynomial too.
20898 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20899 Alan Hayward <alan.hayward@arm.com>
20900 David Sherwood <david.sherwood@arm.com>
20902 * expr.c (expand_expr_real_2): When handling conversions involving
20903 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
20904 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
20905 as a poly_uint64 too.
20907 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20908 Alan Hayward <alan.hayward@arm.com>
20909 David Sherwood <david.sherwood@arm.com>
20911 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
20913 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20914 Alan Hayward <alan.hayward@arm.com>
20915 David Sherwood <david.sherwood@arm.com>
20917 * combine.c (can_change_dest_mode): Handle polynomial
20918 REGMODE_NATURAL_SIZE.
20919 * expmed.c (store_bit_field_1): Likewise.
20920 * expr.c (store_constructor): Likewise.
20921 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
20922 and polynomial REGMODE_NATURAL_SIZE.
20923 (gen_lowpart_common): Likewise.
20924 * reginfo.c (record_subregs_of_mode): Likewise.
20925 * rtlanal.c (read_modify_subreg_p): Likewise.
20927 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20928 Alan Hayward <alan.hayward@arm.com>
20929 David Sherwood <david.sherwood@arm.com>
20931 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
20932 numbers of elements.
20934 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20935 Alan Hayward <alan.hayward@arm.com>
20936 David Sherwood <david.sherwood@arm.com>
20938 * match.pd: Cope with polynomial numbers of vector elements.
20940 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20941 Alan Hayward <alan.hayward@arm.com>
20942 David Sherwood <david.sherwood@arm.com>
20944 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
20945 in a POINTER_PLUS_EXPR.
20947 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20948 Alan Hayward <alan.hayward@arm.com>
20949 David Sherwood <david.sherwood@arm.com>
20951 * omp-simd-clone.c (simd_clone_subparts): New function.
20952 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
20953 (ipa_simd_modify_function_body): Likewise.
20955 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20956 Alan Hayward <alan.hayward@arm.com>
20957 David Sherwood <david.sherwood@arm.com>
20959 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
20960 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
20961 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
20962 (expand_vector_condition, vector_element): Likewise.
20963 (subparts_gt): New function.
20964 (get_compute_type): Use subparts_gt.
20965 (count_type_subparts): Delete.
20966 (expand_vector_operations_1): Use subparts_gt instead of
20967 count_type_subparts.
20969 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20970 Alan Hayward <alan.hayward@arm.com>
20971 David Sherwood <david.sherwood@arm.com>
20973 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
20974 (vect_compile_time_alias): ...this new function. Do the calculation
20975 on poly_ints rather than trees.
20976 (vect_prune_runtime_alias_test_list): Update call accordingly.
20978 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20979 Alan Hayward <alan.hayward@arm.com>
20980 David Sherwood <david.sherwood@arm.com>
20982 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
20984 (vect_schedule_slp_instance): Likewise.
20986 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20987 Alan Hayward <alan.hayward@arm.com>
20988 David Sherwood <david.sherwood@arm.com>
20990 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
20991 constant and extern definitions for variable-length vectors.
20992 (vect_get_constant_vectors): Note that the number of units
20993 is known to be constant.
20995 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20996 Alan Hayward <alan.hayward@arm.com>
20997 David Sherwood <david.sherwood@arm.com>
20999 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
21000 of units as polynomial. Choose between WIDE and NARROW based
21003 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21004 Alan Hayward <alan.hayward@arm.com>
21005 David Sherwood <david.sherwood@arm.com>
21007 * tree-vect-stmts.c (simd_clone_subparts): New function.
21008 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
21010 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21011 Alan Hayward <alan.hayward@arm.com>
21012 David Sherwood <david.sherwood@arm.com>
21014 * tree-vect-stmts.c (vectorizable_call): Treat the number of
21015 vectors as polynomial. Use build_index_vector for
21016 IFN_GOMP_SIMD_LANE.
21018 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21019 Alan Hayward <alan.hayward@arm.com>
21020 David Sherwood <david.sherwood@arm.com>
21022 * tree-vect-stmts.c (get_load_store_type): Treat the number of
21023 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
21024 for variable-length vectors.
21025 (vectorizable_mask_load_store): Treat the number of units as
21026 polynomial, asserting that it is constant if the condition has
21027 already been enforced.
21028 (vectorizable_store, vectorizable_load): Likewise.
21030 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21031 Alan Hayward <alan.hayward@arm.com>
21032 David Sherwood <david.sherwood@arm.com>
21034 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
21035 of units as polynomial. Punt if we can't tell at compile time
21036 which vector contains the final result.
21038 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21039 Alan Hayward <alan.hayward@arm.com>
21040 David Sherwood <david.sherwood@arm.com>
21042 * tree-vect-loop.c (vectorizable_induction): Treat the number
21043 of units as polynomial. Punt on SLP inductions. Use an integer
21044 VEC_SERIES_EXPR for variable-length integer reductions. Use a
21045 cast of such a series for variable-length floating-point
21048 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21049 Alan Hayward <alan.hayward@arm.com>
21050 David Sherwood <david.sherwood@arm.com>
21052 * tree.h (build_index_vector): Declare.
21053 * tree.c (build_index_vector): New function.
21054 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
21055 of units as polynomial, forcibly converting it to a constant if
21056 vectorizable_reduction has already enforced the condition.
21057 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
21058 to create a {1,2,3,...} vector.
21059 (vectorizable_reduction): Treat the number of units as polynomial.
21060 Choose vectype_in based on the largest scalar element size rather
21061 than the smallest number of units. Enforce the restrictions
21064 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21065 Alan Hayward <alan.hayward@arm.com>
21066 David Sherwood <david.sherwood@arm.com>
21068 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
21069 number of units as polynomial.
21071 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21072 Alan Hayward <alan.hayward@arm.com>
21073 David Sherwood <david.sherwood@arm.com>
21075 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
21076 * target.def (autovectorize_vector_sizes): Return the vector sizes
21077 by pointer, using vector_sizes rather than a bitmask.
21078 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
21079 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
21080 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
21082 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
21083 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
21084 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
21085 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
21086 * omp-general.c (omp_max_vf): Likewise.
21087 * omp-low.c (omp_clause_aligned_alignment): Likewise.
21088 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
21089 * tree-vect-loop.c (vect_analyze_loop): Likewise.
21090 * tree-vect-slp.c (vect_slp_bb): Likewise.
21091 * doc/tm.texi: Regenerate.
21092 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
21094 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
21095 the vector size as a poly_uint64 rather than an unsigned int.
21096 (current_vector_size): Change from an unsigned int to a poly_uint64.
21097 (get_vectype_for_scalar_type): Update accordingly.
21098 * tree.h (build_truth_vector_type): Take the size and number of
21099 units as a poly_uint64 rather than an unsigned int.
21100 (build_vector_type): Add a temporary overload that takes
21101 the number of units as a poly_uint64 rather than an unsigned int.
21102 * tree.c (make_vector_type): Likewise.
21103 (build_truth_vector_type): Take the number of units as a poly_uint64
21104 rather than an unsigned int.
21106 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21107 Alan Hayward <alan.hayward@arm.com>
21108 David Sherwood <david.sherwood@arm.com>
21110 * target.def (get_mask_mode): Take the number of units and length
21111 as poly_uint64s rather than unsigned ints.
21112 * targhooks.h (default_get_mask_mode): Update accordingly.
21113 * targhooks.c (default_get_mask_mode): Likewise.
21114 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
21115 * doc/tm.texi: Regenerate.
21117 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21118 Alan Hayward <alan.hayward@arm.com>
21119 David Sherwood <david.sherwood@arm.com>
21121 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
21122 * omp-general.c (omp_max_vf): Likewise.
21123 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
21124 (expand_omp_simd): Handle polynomial safelen.
21125 * omp-low.c (omplow_simd_context): Add a default constructor.
21126 (omplow_simd_context::max_vf): Change from int to poly_uint64.
21127 (lower_rec_simd_input_clauses): Update accordingly.
21128 (lower_rec_input_clauses): Likewise.
21130 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21131 Alan Hayward <alan.hayward@arm.com>
21132 David Sherwood <david.sherwood@arm.com>
21134 * tree-vectorizer.h (vect_nunits_for_cost): New function.
21135 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
21136 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
21137 (vect_analyze_slp_cost): Likewise.
21138 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
21139 (vect_model_load_cost): Likewise.
21141 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21142 Alan Hayward <alan.hayward@arm.com>
21143 David Sherwood <david.sherwood@arm.com>
21145 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
21146 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
21147 from an unsigned int * to a poly_uint64_pod *.
21148 (calculate_unrolling_factor): New function.
21149 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
21151 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21152 Alan Hayward <alan.hayward@arm.com>
21153 David Sherwood <david.sherwood@arm.com>
21155 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
21156 from an unsigned int to a poly_uint64.
21157 (_loop_vec_info::slp_unrolling_factor): Likewise.
21158 (_loop_vec_info::vectorization_factor): Change from an int
21160 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
21161 (vect_get_num_vectors): New function.
21162 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
21163 (vect_get_num_copies): Use vect_get_num_vectors.
21164 (vect_analyze_data_ref_dependences): Change max_vf from an int *
21165 to an unsigned int *.
21166 (vect_analyze_data_refs): Change min_vf from an int * to a
21168 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
21169 than an unsigned HOST_WIDE_INT.
21170 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
21171 (vect_analyze_data_ref_dependence): Change max_vf from an int *
21172 to an unsigned int *.
21173 (vect_analyze_data_ref_dependences): Likewise.
21174 (vect_compute_data_ref_alignment): Handle polynomial vf.
21175 (vect_enhance_data_refs_alignment): Likewise.
21176 (vect_prune_runtime_alias_test_list): Likewise.
21177 (vect_shift_permute_load_chain): Likewise.
21178 (vect_supportable_dr_alignment): Likewise.
21179 (dependence_distance_ge_vf): Take the vectorization factor as a
21180 poly_uint64 rather than an unsigned HOST_WIDE_INT.
21181 (vect_analyze_data_refs): Change min_vf from an int * to a
21183 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
21184 vfm1 as a poly_uint64 rather than an int. Make the same change
21185 for the returned bound_scalar.
21186 (vect_gen_vector_loop_niters): Handle polynomial vf.
21187 (vect_do_peeling): Likewise. Update call to
21188 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
21189 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
21191 * tree-vect-loop.c (vect_determine_vectorization_factor)
21192 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
21193 (vect_get_known_peeling_cost): Likewise.
21194 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
21195 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
21196 (vect_transform_loop): Likewise. Use the lowest possible VF when
21197 updating the upper bounds of the loop.
21198 (vect_min_worthwhile_factor): Make static. Return an unsigned int
21199 rather than an int.
21200 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
21201 polynomial unroll factors.
21202 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
21203 (vect_make_slp_decision): Likewise.
21204 (vect_supported_load_permutation_p): Likewise, and polynomial
21206 (vect_analyze_slp_cost): Handle polynomial vf.
21207 (vect_slp_analyze_node_operations): Likewise.
21208 (vect_slp_analyze_bb_1): Likewise.
21209 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
21210 than an unsigned HOST_WIDE_INT.
21211 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
21212 (vectorizable_load): Handle polynomial vf.
21213 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
21215 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
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 * match.pd: Handle bit operations involving three constants
21222 and try to fold one pair.
21224 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21226 * tree-vect-loop-manip.c: Include gimple-fold.h.
21227 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
21228 niters_maybe_zero parameters. Handle other cases besides a step of 1.
21229 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
21230 Add a path that uses a step of VF instead of 1, but disable it
21232 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
21233 and niters_no_overflow parameters. Update calls to
21234 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
21235 Create a new SSA name if the latter choses to use a ste other
21236 than zero, and return it via niters_vector_mult_vf_var.
21237 * tree-vect-loop.c (vect_transform_loop): Update calls to
21238 vect_do_peeling, vect_gen_vector_loop_niters and
21239 slpeel_make_loop_iterate_ntimes.
21240 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
21241 (vect_gen_vector_loop_niters): Update declarations after above changes.
21243 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
21245 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
21246 128-bit round to integer instructions.
21247 (ceil<mode>2): Likewise.
21248 (btrunc<mode>2): Likewise.
21249 (round<mode>2): Likewise.
21251 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
21253 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
21254 unaligned VSX load/store on P8/P9.
21255 (expand_block_clear): Allow the use of unaligned VSX
21256 load/store on P8/P9.
21258 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21260 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
21262 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
21263 swap associated with both a load and a store.
21265 2018-01-02 Andrew Waterman <andrew@sifive.com>
21267 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
21268 * config/riscv/riscv.md (clear_cache): Use it.
21270 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
21272 * web.c: Remove out-of-date comment.
21274 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21276 * expr.c (fixup_args_size_notes): Check that any existing
21277 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
21278 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
21279 (emit_single_push_insn): ...here.
21281 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21283 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
21284 (const_vector_encoded_nelts): New function.
21285 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
21286 (const_vector_int_elt, const_vector_elt): Declare.
21287 * emit-rtl.c (const_vector_int_elt_1): New function.
21288 (const_vector_elt): Likewise.
21289 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
21290 of CONST_VECTOR_ELT.
21292 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21294 * expr.c: Include rtx-vector-builder.h.
21295 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
21296 directly on the tree encoding.
21297 (const_vector_from_tree): Likewise.
21298 * optabs.c: Include rtx-vector-builder.h.
21299 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
21300 sequence of "u" values.
21301 * vec-perm-indices.c: Include rtx-vector-builder.h.
21302 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
21303 directly on the vec_perm_indices encoding.
21305 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21307 * doc/rtl.texi (const_vector): Describe new encoding scheme.
21308 * Makefile.in (OBJS): Add rtx-vector-builder.o.
21309 * rtx-vector-builder.h: New file.
21310 * rtx-vector-builder.c: Likewise.
21311 * rtl.h (rtx_def::u2): Add a const_vector field.
21312 (CONST_VECTOR_NPATTERNS): New macro.
21313 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
21314 (CONST_VECTOR_DUPLICATE_P): Likewise.
21315 (CONST_VECTOR_STEPPED_P): Likewise.
21316 (CONST_VECTOR_ENCODED_ELT): Likewise.
21317 (const_vec_duplicate_p): Check for a duplicated vector encoding.
21318 (unwrap_const_vec_duplicate): Likewise.
21319 (const_vec_series_p): Check for a non-duplicated vector encoding.
21320 Say that the function only returns true for integer vectors.
21321 * emit-rtl.c: Include rtx-vector-builder.h.
21322 (gen_const_vec_duplicate_1): Delete.
21323 (gen_const_vector): Call gen_const_vec_duplicate instead of
21324 gen_const_vec_duplicate_1.
21325 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
21326 (gen_const_vec_duplicate): Use rtx_vector_builder.
21327 (gen_const_vec_series): Likewise.
21328 (gen_rtx_CONST_VECTOR): Likewise.
21329 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
21330 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
21331 Build a new vector rather than modifying a CONST_VECTOR in-place.
21332 (handle_special_swappables): Update call accordingly.
21333 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
21334 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
21335 Build a new vector rather than modifying a CONST_VECTOR in-place.
21336 (handle_special_swappables): Update call accordingly.
21338 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21340 * simplify-rtx.c (simplify_const_binary_operation): Use
21341 CONST_VECTOR_ELT instead of XVECEXP.
21343 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21345 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
21346 the selector elements to be different from the data elements
21347 if the selector is a VECTOR_CST.
21348 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
21349 ssizetype for the selector.
21351 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21353 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
21354 before testing each element individually.
21355 * tree-vect-generic.c (lower_vec_perm): Likewise.
21357 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21359 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
21360 * selftest-run-tests.c (selftest::run_tests): Call it.
21361 * vector-builder.h (vector_builder::operator ==): New function.
21362 (vector_builder::operator !=): Likewise.
21363 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
21364 (vec_perm_indices::all_from_input_p): New function.
21365 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
21366 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
21367 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
21368 instead of reading the VECTOR_CST directly. Detect whether both
21369 vector inputs are the same before constructing the vec_perm_indices,
21370 and update the number of inputs argument accordingly. Use the
21371 utility functions added above. Only construct sel2 if we need to.
21373 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21375 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
21376 the broadcast of the low byte.
21377 (expand_mult_highpart): Use an explicit encoding for the permutes.
21378 * optabs-query.c (can_mult_highpart_p): Likewise.
21379 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
21380 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
21381 (vectorizable_bswap): Likewise.
21382 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
21383 explicit encoding for the power-of-2 permutes.
21384 (vect_permute_store_chain): Likewise.
21385 (vect_grouped_load_supported): Likewise.
21386 (vect_permute_load_chain): Likewise.
21388 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21390 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
21391 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
21392 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
21393 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
21394 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
21395 (vect_gen_perm_mask_any): Likewise.
21397 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21399 * int-vector-builder.h: New file.
21400 * vec-perm-indices.h: Include int-vector-builder.h.
21401 (vec_perm_indices): Redefine as an int_vector_builder.
21402 (auto_vec_perm_indices): Delete.
21403 (vec_perm_builder): Redefine as a stand-alone class.
21404 (vec_perm_indices::vec_perm_indices): New function.
21405 (vec_perm_indices::clamp): Likewise.
21406 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
21407 (vec_perm_indices::new_vector): New function.
21408 (vec_perm_indices::new_expanded_vector): Update for new
21409 vec_perm_indices class.
21410 (vec_perm_indices::rotate_inputs): New function.
21411 (vec_perm_indices::all_in_range_p): Operate directly on the
21412 encoded form, without computing elided elements.
21413 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
21414 encoding. Update for new vec_perm_indices class.
21415 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
21416 the given vec_perm_builder.
21417 (expand_vec_perm_var): Update vec_perm_builder constructor.
21418 (expand_mult_highpart): Use vec_perm_builder instead of
21419 auto_vec_perm_indices.
21420 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
21421 vec_perm_indices instead of auto_vec_perm_indices. Use a single
21422 or double series encoding as appropriate.
21423 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
21424 vec_perm_indices instead of auto_vec_perm_indices.
21425 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
21426 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
21427 (vect_permute_store_chain): Likewise.
21428 (vect_grouped_load_supported): Likewise.
21429 (vect_permute_load_chain): Likewise.
21430 (vect_shift_permute_load_chain): Likewise.
21431 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
21432 (vect_transform_slp_perm_load): Likewise.
21433 (vect_schedule_slp_instance): Likewise.
21434 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
21435 (vectorizable_mask_load_store): Likewise.
21436 (vectorizable_bswap): Likewise.
21437 (vectorizable_store): Likewise.
21438 (vectorizable_load): Likewise.
21439 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
21440 vec_perm_indices instead of auto_vec_perm_indices. Use
21441 tree_to_vec_perm_builder to read the vector from a tree.
21442 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
21443 vec_perm_builder instead of a vec_perm_indices.
21444 (have_whole_vector_shift): Use vec_perm_builder and
21445 vec_perm_indices instead of auto_vec_perm_indices. Leave the
21446 truncation to calc_vec_perm_mask_for_shift.
21447 (vect_create_epilog_for_reduction): Likewise.
21448 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
21449 from auto_vec_perm_indices to vec_perm_indices.
21450 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
21451 instead of changing individual elements.
21452 (aarch64_vectorize_vec_perm_const): Use new_vector to install
21453 the vector in d.perm.
21454 * config/arm/arm.c (expand_vec_perm_d::perm): Change
21455 from auto_vec_perm_indices to vec_perm_indices.
21456 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
21457 instead of changing individual elements.
21458 (arm_vectorize_vec_perm_const): Use new_vector to install
21459 the vector in d.perm.
21460 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
21461 Update vec_perm_builder constructor.
21462 (rs6000_expand_interleave): Likewise.
21463 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
21464 (rs6000_expand_interleave): Likewise.
21466 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21468 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
21469 to qimode could truncate the indices.
21470 * optabs.c (expand_vec_perm_var): Likewise.
21472 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21474 * Makefile.in (OBJS): Add vec-perm-indices.o.
21475 * vec-perm-indices.h: New file.
21476 * vec-perm-indices.c: Likewise.
21477 * target.h (vec_perm_indices): Replace with a forward class
21479 (auto_vec_perm_indices): Move to vec-perm-indices.h.
21480 * optabs.h: Include vec-perm-indices.h.
21481 (expand_vec_perm): Delete.
21482 (selector_fits_mode_p, expand_vec_perm_var): Declare.
21483 (expand_vec_perm_const): Declare.
21484 * target.def (vec_perm_const_ok): Replace with...
21485 (vec_perm_const): ...this new hook.
21486 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
21487 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
21488 * doc/tm.texi: Regenerate.
21489 * optabs.def (vec_perm_const): Delete.
21490 * doc/md.texi (vec_perm_const): Likewise.
21491 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
21492 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
21493 expand_vec_perm for constant permutation vectors. Assert that
21494 the mode of variable permutation vectors is the integer equivalent
21495 of the mode that is being permuted.
21496 * optabs-query.h (selector_fits_mode_p): Declare.
21497 * optabs-query.c: Include vec-perm-indices.h.
21498 (selector_fits_mode_p): New function.
21499 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
21500 is defined, instead of checking whether the vec_perm_const_optab
21501 exists. Use targetm.vectorize.vec_perm_const instead of
21502 targetm.vectorize.vec_perm_const_ok. Check whether the indices
21503 fit in the vector mode before using a variable permute.
21504 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
21505 vec_perm_indices instead of an rtx.
21506 (expand_vec_perm): Replace with...
21507 (expand_vec_perm_const): ...this new function. Take the selector
21508 as a vec_perm_indices rather than an rtx. Also take the mode of
21509 the selector. Update call to shift_amt_for_vec_perm_mask.
21510 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
21511 Use vec_perm_indices::new_expanded_vector to expand the original
21512 selector into bytes. Check whether the indices fit in the vector
21513 mode before using a variable permute.
21514 (expand_vec_perm_var): Make global.
21515 (expand_mult_highpart): Use expand_vec_perm_const.
21516 * fold-const.c: Includes vec-perm-indices.h.
21517 * tree-ssa-forwprop.c: Likewise.
21518 * tree-vect-data-refs.c: Likewise.
21519 * tree-vect-generic.c: Likewise.
21520 * tree-vect-loop.c: Likewise.
21521 * tree-vect-slp.c: Likewise.
21522 * tree-vect-stmts.c: Likewise.
21523 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
21525 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
21526 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
21527 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
21528 (aarch64_vectorize_vec_perm_const): ...this new function.
21529 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
21530 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
21531 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
21532 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
21533 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
21534 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
21535 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
21537 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
21538 check for NEON modes.
21539 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
21540 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
21541 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
21542 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
21544 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
21545 the old VEC_PERM_CONST conditions.
21546 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
21547 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
21548 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
21549 (ia64_vectorize_vec_perm_const_ok): Merge into...
21550 (ia64_vectorize_vec_perm_const): ...this new function.
21551 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
21552 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
21553 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
21554 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
21555 * config/mips/mips.c (mips_expand_vec_perm_const)
21556 (mips_vectorize_vec_perm_const_ok): Merge into...
21557 (mips_vectorize_vec_perm_const): ...this new function.
21558 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
21559 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
21560 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
21561 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
21562 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
21563 (rs6000_expand_vec_perm_const): Delete.
21564 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
21566 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
21567 (altivec_expand_vec_perm_const_le): Take each operand individually.
21568 Operate on constant selectors rather than rtxes.
21569 (altivec_expand_vec_perm_const): Likewise. Update call to
21570 altivec_expand_vec_perm_const_le.
21571 (rs6000_expand_vec_perm_const): Delete.
21572 (rs6000_vectorize_vec_perm_const_ok): Delete.
21573 (rs6000_vectorize_vec_perm_const): New function.
21574 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
21575 an element count and rtx array.
21576 (rs6000_expand_extract_even): Update call accordingly.
21577 (rs6000_expand_interleave): Likewise.
21578 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
21579 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
21580 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
21581 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
21582 (rs6000_expand_vec_perm_const): Delete.
21583 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
21584 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
21585 (altivec_expand_vec_perm_const_le): Take each operand individually.
21586 Operate on constant selectors rather than rtxes.
21587 (altivec_expand_vec_perm_const): Likewise. Update call to
21588 altivec_expand_vec_perm_const_le.
21589 (rs6000_expand_vec_perm_const): Delete.
21590 (rs6000_vectorize_vec_perm_const_ok): Delete.
21591 (rs6000_vectorize_vec_perm_const): New function. Remove stray
21592 reference to the SPE evmerge intructions.
21593 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
21594 an element count and rtx array.
21595 (rs6000_expand_extract_even): Update call accordingly.
21596 (rs6000_expand_interleave): Likewise.
21597 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
21598 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
21600 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
21602 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21604 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
21605 vector mode and that that mode matches the mode of the data
21607 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
21608 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
21609 directly using expand_vec_perm_1 when forcing selectors into
21611 (expand_vec_perm_var): New function, split out from expand_vec_perm.
21613 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21615 * optabs-query.h (can_vec_perm_p): Delete.
21616 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
21617 * optabs-query.c (can_vec_perm_p): Split into...
21618 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
21619 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
21620 particular selector is valid.
21621 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
21622 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
21623 (vect_grouped_load_supported): Likewise.
21624 (vect_shift_permute_load_chain): Likewise.
21625 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
21626 (vect_transform_slp_perm_load): Likewise.
21627 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
21628 (vectorizable_bswap): Likewise.
21629 (vect_gen_perm_mask_checked): Likewise.
21630 * fold-const.c (fold_ternary_loc): Likewise. Don't take
21631 implementations of variable permutation vectors into account
21632 when deciding which selector to use.
21633 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
21634 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
21635 with a false third argument.
21636 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
21637 to test whether the constant selector is valid and can_vec_perm_var_p
21638 to test whether a variable selector is valid.
21640 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21642 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
21643 * optabs-query.c (can_vec_perm_p): Likewise.
21644 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
21645 instead of vec_perm_indices.
21646 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
21647 (vect_gen_perm_mask_checked): Likewise,
21648 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
21649 (vect_gen_perm_mask_checked): Likewise,
21651 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21653 * optabs-query.h (qimode_for_vec_perm): Declare.
21654 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
21655 (qimode_for_vec_perm): ...this new function.
21656 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
21658 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
21660 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
21661 does not have a conditional at the top.
21663 2018-01-02 Richard Biener <rguenther@suse.de>
21665 * ipa-inline.c (big_speedup_p): Fix expression.
21667 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
21670 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
21673 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
21677 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
21678 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
21679 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
21680 cond_taken_branch_cost 3->4.
21682 2018-01-01 Jakub Jelinek <jakub@redhat.com>
21684 PR tree-optimization/83581
21685 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
21686 TODO_cleanup_cfg if any changes have been made.
21688 PR middle-end/83608
21689 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
21690 convert_modes if target mode has the right side, but different mode
21693 PR middle-end/83609
21694 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
21695 last argument when extracting from CONCAT. If either from_real or
21696 from_imag is NULL, use expansion through memory. If result is not
21697 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
21698 the parts directly to inner mode, if even that fails, use expansion
21701 PR middle-end/83623
21702 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
21703 check for bswap in mode rather than HImode and use that in expand_unop
21706 Copyright (C) 2018 Free Software Foundation, Inc.
21708 Copying and distribution of this file, with or without modification,
21709 are permitted in any medium without royalty provided the copyright
21710 notice and this notice are preserved.