1 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
3 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
4 and gimple stmt arguments with a stmt_vec_info.
5 (vect_record_base_alignments): Update calls accordingly.
6 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
7 and gimple stmt arguments with a stmt_vec_info.
8 (vect_build_slp_tree_1): Remove vinfo argument and update call
9 to vect_record_max_nunits.
10 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
11 and vect_record_max_nunits.
13 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
15 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
16 file and take a stmt_vec_info instead of a gimple stmt.
17 (supportable_widening_operation, vect_finish_replace_stmt)
18 (vect_finish_stmt_generation, vect_get_store_rhs)
19 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
20 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
21 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
22 (vect_get_smallest_scalar_type, vect_check_gather_scatter)
23 (vect_create_data_ref_ptr, bump_vector_ptr)
24 (vect_permute_store_chain, vect_setup_realignment)
25 (vect_transform_grouped_load, vect_record_grouped_load_vectors)
26 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
27 (vectorizable_reduction, vectorizable_induction)
28 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
29 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
30 than gimple stmts as arguments.
31 * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
32 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
33 (can_group_stmts_p, vect_check_gather_scatter)
34 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
35 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
36 (vect_permute_load_chain, vect_shift_permute_load_chain)
37 (vect_transform_grouped_load)
38 (vect_record_grouped_load_vectors): Likewise.
39 * tree-vect-loop.c (vect_fixup_reduc_chain)
40 (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
41 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
42 (vectorizable_reduction, vectorizable_induction)
43 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
44 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
45 (vect_get_load_store_mask): Likewise.
46 * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
47 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
48 * tree-vect-stmts.c (vect_mark_relevant)
49 (is_simple_and_all_uses_invariant)
50 (exist_non_indexing_operands_for_use_p, process_use)
51 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
52 (vect_get_vec_def_for_operand, vect_get_vec_defs)
53 (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
54 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
55 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
56 (get_negative_load_store_type, get_load_store_type)
57 (vect_check_load_store_mask, vect_check_store_rhs)
58 (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
59 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
60 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
61 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
62 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
63 (vectorizable_load, vectorizable_condition, vectorizable_comparison)
64 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
65 (supportable_widening_operation): Likewise.
67 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
69 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
70 a stmt_vec_info instead of a gcall.
71 (vect_check_gather_scatter): Update call accordingly.
72 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
74 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
75 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
76 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
77 instead of a gimple stmt.
78 (vect_transform_loop): Update calls accordingly.
79 * tree-vect-slp.c (vect_split_slp_store_group): Take and return
80 stmt_vec_infos instead of gimple stmts.
81 (vect_analyze_slp_instance): Update use accordingly.
82 * tree-vect-stmts.c (read_vector_array, write_vector_array)
83 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
84 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
85 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
86 (vect_gen_widened_results_half, vect_get_loop_based_defs)
87 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
88 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
91 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
93 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
94 of the worklist from a vector of gimple stmts to a vector of
96 * tree-vect-stmts.c (vect_mark_relevant, process_use)
97 (vect_mark_stmts_to_be_vectorized): Likewise
99 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
101 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
102 statement before passing it to vect_analyze_stmt.
103 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
104 the chain of phi vector definitions. Track the exit phi via its
106 (vectorizable_reduction): Set cond_stmt_vinfo directly from the
107 STMT_VINFO_REDUC_DEF.
108 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
109 stmt_vec_infos to handle the statement chains.
110 (vect_get_slp_defs): Record the first statement in the node
111 using a stmt_vec_info.
112 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
113 statements here and pass their stmt_vec_info down to subroutines.
114 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
115 down to vect_finish_stmt_generation.
116 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
117 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
118 stmt_vec_infos to subroutines.
119 (vect_remove_stores): Use stmt_vec_infos to handle the statement
122 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
124 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
125 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
126 (vect_permute_store_chain, vect_setup_realignment)
127 (vect_permute_load_chain, vect_shift_permute_load_chain)
128 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
129 stmts internally, and when passing values to other vectorizer routines.
130 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
131 * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
132 (vect_analyze_loop_operations, get_initial_def_for_reduction)
133 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
134 (vectorizable_reduction, vectorizable_induction)
135 (vectorizable_live_operation, vect_transform_loop_stmt)
136 (vect_transform_loop): Likewise.
137 * tree-vect-patterns.c (vect_reassociating_reduction_p)
138 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
139 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
140 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
141 (vect_slp_analyze_node_operations_1): Likewise.
142 * tree-vect-stmts.c (vect_mark_relevant, process_use)
143 (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
144 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
145 (vect_finish_stmt_generation_1, get_group_load_store_type)
146 (get_load_store_type, vect_build_gather_load_calls)
147 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
148 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
149 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
150 (vectorizable_store, vectorizable_load, vectorizable_condition)
151 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
152 (supportable_widening_operation): Likewise.
153 (vect_get_vector_types_for_stmt): Likewise.
154 * tree-vectorizer.h (vect_dr_behavior): Likewise.
156 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
158 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
159 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
160 (vect_permute_store_chain, vect_permute_load_chain)
161 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
162 repeated stmt_vec_info lookups.
163 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
164 (vect_update_ivs_after_vectorizer): Likewise.
165 * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
166 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
167 * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
168 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
169 (vect_bb_slp_scalar_cost): Likewise.
170 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
172 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
174 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
175 gcall rather than the generic gimple stmt to gimple_call_internal_fn.
176 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
177 to get gassigns and gcalls, rather than operating on generc gimple
179 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
180 (vect_mark_stmts_to_be_vectorized, vectorizable_store)
181 (vectorizable_load, vect_analyze_stmt): Likewise.
182 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
184 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
186 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
187 return stmt_vec_infos rather than gimple stmts. Do not accept
189 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
191 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
192 Update use of get_later_stmt.
193 (vect_get_constant_vectors): Update call accordingly.
194 (vect_schedule_slp_instance): Likewise
195 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
196 (vect_slp_analyze_instance_dependence): Likewise.
197 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
199 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
201 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
202 (stmt_info_for_cost::stmt_info): ...this new field.
203 (add_stmt_costs): Update accordingly.
204 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
205 (vect_get_known_peeling_cost): Likewise.
206 (vect_estimate_min_profitable_iters): Likewise.
207 * tree-vect-stmts.c (record_stmt_cost): Likewise.
209 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
211 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
212 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
213 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
215 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
217 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
219 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
220 a gimple stmt to a stmt_vec_info.
221 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
223 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
225 * tree-vectorizer.h (vec_info::grouped_stores): Change from
226 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
227 (_loop_vec_info::reduction_chains): Likewise.
228 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
230 * tree-vect-slp.c (vect_analyze_slp): Likewise.
232 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
234 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
235 a gimple stmt to a stmt_vec_info.
236 (_stmt_vec_info::next_element): Likewise.
237 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
238 (vect_slp_analyze_and_verify_node_alignment)
239 (vect_analyze_group_access_1, vect_analyze_group_access)
240 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
241 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
242 (vect_supportable_dr_alignment): Update accordingly.
243 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
244 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
245 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
246 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
247 * tree-vect-slp.c (vect_build_slp_tree_1)
248 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
249 (vect_split_slp_store_group, vect_analyze_slp_instance)
250 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
251 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
252 (get_group_load_store_type, get_load_store_type)
253 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
254 (vect_transform_stmt, vect_remove_stores): Likewise.
256 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
258 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
260 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
261 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
262 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
263 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
264 (vector_alignment_reachable_p, vect_get_data_access_cost)
265 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
266 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
267 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
268 (vect_analyze_group_access_1, vect_analyze_group_access)
269 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
270 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
271 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
272 result of vect_dr_stmt and use the stmt_vec_info instead of
273 the associated gimple stmt.
274 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
275 (vect_gen_prolog_loop_niters): Likewise.
276 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
278 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
280 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
281 to a vec<stmt_vec_info>.
282 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
283 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
285 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
286 to a vec<stmt_vec_info>.
287 (bst_traits::value_type, bst_traits::value_type): Likewise.
288 (bst_traits::hash): Update accordingly.
289 (vect_get_and_check_slp_defs): Change the stmts parameter from
290 a vec<gimple *> to a vec<stmt_vec_info>.
291 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
292 (vect_build_slp_tree): Likewise.
293 (vect_build_slp_tree_2): Likewise. Update uses of
294 SLP_TREE_SCALAR_STMTS.
295 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
296 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
297 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
298 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
299 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
300 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
301 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
302 (vect_get_constant_vectors, vect_get_slp_defs)
303 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
304 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
305 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
306 instead of gimple stmts.
307 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
308 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
309 (vect_slp_analyze_instance_dependence): Update uses of
310 SLP_TREE_SCALAR_STMTS.
311 (vect_slp_analyze_and_verify_node_alignment): Likewise.
312 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
313 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
314 (get_initial_defs_for_reduction): Likewise.
315 (vect_create_epilog_for_reduction): Likewise.
316 (vectorize_fold_left_reduction): Likewise.
317 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
318 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
319 (can_vectorize_live_stmts): Likewise.
321 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
323 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
324 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
325 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
327 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
329 (gather_scalar_reductions): Update after above interface changes.
330 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
331 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
333 (vect_force_simple_reduction): Likewise.
334 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
335 LOOP_VINFO_REDUCTIONS.
336 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
338 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
340 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
341 a gimple stmt to a stmt_vec_info.
342 * tree-vect-loop.c (vect_active_double_reduction_p)
343 (vect_force_simple_reduction, vectorizable_reduction): Update
346 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
348 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
349 vec<gimple *> to a vec<stmt_vec_info>.
350 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
351 the reduction_phis argument from a vec<gimple *> to a
353 (vectorizable_reduction): Likewise the phis local variable that
354 is passed to vect_create_epilog_for_reduction. Update for new type
355 of SLP_TREE_VEC_STMTS.
356 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
357 (vectorizable_live_operation): Likewise.
358 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
359 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
361 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
363 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
364 a gimple stmt to a stmt_vec_info.
365 (vectorizable_condition, vectorizable_live_operation)
366 (vectorizable_reduction, vectorizable_induction): Pass back the
367 vectorized statement as a stmt_vec_info.
368 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
369 use of STMT_VINFO_VEC_STMT.
370 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
371 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
372 as stmt_vec_infos rather than gimple stmts.
373 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
375 (vectorizable_live_operation): Likewise.
376 (vectorizable_reduction, vectorizable_induction): Likewise,
377 updating use of STMT_VINFO_VEC_STMT.
378 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
379 of STMT_VINFO_VEC_STMT.
380 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
381 (vectorizable_simd_clone_call, vectorizable_conversion)
382 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
383 (vectorizable_store, vectorizable_load, vectorizable_condition)
384 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
385 from a gimple stmt to a stmt_vec_info.
386 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
387 pointer to a stmt_vec_info to the vectorizable_* routines.
389 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
391 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
392 a gimple stmt to a stmt_vec_info.
393 (is_pattern_stmt_p): Update accordingly.
394 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
395 (vect_record_grouped_load_vectors): Likewise.
396 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
397 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
398 (vect_model_reduction_cost): Likewise.
399 (vect_create_epilog_for_reduction): Likewise.
400 (vectorizable_reduction, vectorizable_induction): Likewise.
401 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
402 Return the stmt_vec_info for the pattern statement.
403 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
404 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
405 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
406 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
407 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
408 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
409 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
410 (free_stmt_vec_info, vect_is_simple_use): Likewise.
412 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
414 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
415 (vect_finish_stmt_generation): Likewise.
416 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
417 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
418 (vect_build_gather_load_calls): Use the return value of the above
419 functions instead of a separate call to vinfo_for_stmt. Use narrow
420 scopes for the input gimple stmt and wider scopes for the associated
421 stmt_vec_info. Use vec_info::lookup_def when setting these
422 stmt_vec_infos from an SSA_NAME definition.
423 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
424 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
425 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
426 (vectorizable_store, vectorizable_load, vectorizable_condition)
427 (vectorizable_comparison): Likewise.
428 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
429 (vectorizable_reduction): Likewise.
431 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
433 * tree-vectorizer.h (vect_is_simple_use): Add an optional
434 stmt_vec_info * parameter before the optional gimple **.
435 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
436 (process_use, vect_get_vec_def_for_operand_1): Update callers.
437 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
438 * tree-vect-loop.c (vectorizable_reduction): Likewise.
439 (vectorizable_live_operation): Likewise.
440 * tree-vect-patterns.c (type_conversion_p): Likewise.
441 (vect_look_through_possible_promotion): Likewise.
442 (vect_recog_rotate_pattern): Likewise.
443 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
445 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
447 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
448 a typedef to a wrapper class.
449 (NULL_STMT_VEC_INFO): New macro.
450 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
451 (stmt_vec_info::operator*): New function.
452 (stmt_vec_info::operator gimple *): Likewise.
453 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
454 (add_stmt_costs): Likewise.
455 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
456 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
457 (vect_get_known_peeling_cost): Likewise.
458 (vect_estimate_min_profitable_iters): Likewise.
459 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
460 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
461 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
462 (vectorizable_store, free_stmt_vec_infos): Likewise.
463 (new_stmt_vec_info): Change return type of xcalloc to
466 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
468 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
469 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
470 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
471 a single_imm_use-based sequence.
472 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
474 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
476 * tree-vectorizer.h (vec_info::lookup_def): Declare.
477 * tree-vectorizer.c (vec_info::lookup_def): New function.
478 * tree-vect-patterns.c (vect_get_internal_def): Use it.
479 (vect_widened_op_tree): Likewise.
480 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
481 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
482 (vectorizable_reduction): Likewise.
483 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
485 (vect_is_slp_reduction): Update calls accordingly. Use
486 vec_info::lookup_def.
487 (vect_is_simple_reduction): Likewise
488 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
490 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
492 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
493 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
494 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
496 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
497 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
498 (vect_update_vf_for_slp, vect_analyze_loop_operations)
499 (vect_is_slp_reduction, vectorizable_induction)
500 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
501 * tree-vect-patterns.c (vect_init_pattern_stmt):
502 (vect_determine_min_output_precision_1, vect_determine_precisions)
503 (vect_pattern_recog): Likewise.
504 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
505 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
506 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
507 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
508 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
509 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
510 info field from a loop to a loop_vec_info.
512 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
514 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
515 (vec_info::add_stmt): Declare.
516 * tree-vectorizer.c (vec_info::add_stmt): New function.
517 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
518 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
519 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
520 (vectorizable_induction): Likewise.
521 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
522 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
523 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
524 (vectorizable_load): Likewise.
525 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
526 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
527 (vect_recog_gather_scatter_pattern): Likewise.
528 (append_pattern_def_seq): Likewise. Remove a check that is
529 performed by add_stmt itself.
531 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
533 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
534 which make_ssa_name was called with new_stmt before new_stmt
537 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
539 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
541 (vect_is_slp_reduction): ...here...
542 (vect_is_simple_reduction): ...and here. Remove repetition of tests
543 that are already known to be false.
545 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
547 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
548 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
549 STMT_VINFO_NUM_SLP_USES when it's true.
550 (vect_free_slp_instance): Add a final_p parameter and pass it to
552 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
553 (vect_analyze_slp_instance): Likewise.
554 (vect_slp_analyze_operations): Likewise.
555 (vect_slp_analyze_bb_1): Likewise.
556 * tree-vectorizer.c (vec_info): Likewise.
557 * tree-vect-loop.c (vect_transform_loop): Likewise.
559 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
561 * tree-vect-loop.c (vectorizable_reduction): Assert that the
562 function is not called for second and subsequent members of
565 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
567 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
568 cases for nested loops from here to ...
569 (vect_create_epilog_for_reduction): ...here. Only call
570 vect_is_simple_use for inner-loop reductions.
572 2018-07-31 Martin Liska <mliska@suse.cz>
574 PR gcov-profile/85338
575 PR gcov-profile/85350
576 PR gcov-profile/85372
577 * profile.c (struct location_triplet): New.
578 (struct location_triplet_hash): Likewise.
579 (output_location): Do not output a BB that
580 is already recorded for a line.
581 (branch_prob): Use streamed_locations.
583 2018-07-31 Martin Liska <mliska@suse.cz>
585 PR gcov-profile/85370
586 * coverage.c (coverage_begin_function): Do not mark target
587 clones as artificial functions.
589 2018-07-31 Martin Liska <mliska@suse.cz>
591 PR gcov-profile/83813
592 PR gcov-profile/84758
593 PR gcov-profile/85217
594 PR gcov-profile/85332
595 * profile.c (branch_prob): Do not record GOTO expressions
596 for GIMPLE statements which locations are already streamed.
598 2018-07-31 Olivier Hainque <hainque@adacore.com>
600 * gcc.c (handle_spec_function): Accept a soft_matched_part
601 argument, as do_spec_1. Pass it down to ...
602 (eval_spec_function): Accept a soft_matched_part argument,
603 and pass it down to ...
604 (do_spec_2): Accept a soft_matched_part argument, and pass
605 it down to do_spec_1.
606 (do_spec_1): Pass soft_matched_part to handle_spec_function.
607 (handle_braces): Update call to handle_spec_function.
608 (driver::set_up_specs): Update calls to do_spec_2.
609 (compare_debug_dump_opt_spec_function): Likewise.
610 (compare_debug_self_opt_spec_function): Likewise.
612 2018-07-31 Olivier Hainque <hainque@adacore.com>
614 * common.opt (nolibc): New option.
615 * doc/invoke.texi (Link Options): Document it.
616 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
617 * config/alpha/linux.h: Likewise.
618 * config/arc/elf.h: Likewise.
619 * config/arm/uclinux-elf.h: Likewise.
620 * config/arm/unknown-elf.h: Likewise.
621 * config/avr/avrlibc.h: Likewise.
622 * config/bfin/bfin.h: Likewise.
623 * config/bfin/linux.h: Likewise.
624 * config/bfin/uclinux.h: Likewise.
625 * config/darwin.h: Likewise.
626 * config/darwin10.h: Likewise.
627 * config/darwin12.h: Likewise.
628 * config/gnu-user.h: Likewise.
629 * config/lm32/uclinux-elf.h: Likewise.
630 * config/pa/pa-hpux11.h: Likewise.
631 * config/pa/pa64-hpux.h: Likewise.
632 * config/sparc/sparc.h: Likewise.
634 2018-07-31 Olivier Hainque <hainque@adacore.com>
636 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
639 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
642 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
645 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
647 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
648 terminated string literal.
650 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
652 PR rtl-optimization/85160
653 * combine.c (is_just_move): New function.
654 (try_combine): Allow combining two instructions into two if neither of
655 the original instructions was a move.
657 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
660 * doc/extend.texi (Global Register Variables): Discourage use of type
662 (Local Register Variables): Likewise.
664 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
666 PR tree-optimization/86506
667 * hwint.h (ceil_log2): Resync with hwint.c implementation.
669 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
672 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
673 hard_regno, make sure no insn between `from` and `to` clobbers it.
675 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
676 Tom de Vries <tdevries@suse.de>
678 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
679 (PTX_DEFAULT_RUNTIME_DIM): ... this.
680 (nvptx_goacc_validate_dims): Set default worker and gang dims to
681 PTX_DEFAULT_RUNTIME_DIM.
682 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
684 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
686 * config/pa/pa.c (pa_output_addr_vec): Align address table.
687 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
688 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
690 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
692 * config/rs6000/constraints.md (wG constraint): Delete, no longer
694 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
695 predicate to reflect toc fusion has been deleted.
696 (toc_fusion_mem_raw): Delete, no longer used.
697 (toc_fusion_mem_wrapped): Likewise.
698 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
700 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
701 Delete, no longer used.
702 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
703 meant to be used for toc fusion.
704 (rs6000_debug_print_mode): Delete toc fusion debugging.
705 (rs6000_debug_reg_global): Likewise.
706 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
707 fusion and secondary reload support that were never used.
708 (rs6000_option_override_internal): Delete TOC fusion, that was only
709 partially defined, and it did not work unless you also used the
711 (rs6000_legitimate_address_p): Delete TOC fusion support.
712 (rs6000_opt_masks): Likewise.
713 (fusion_wrap_memory_address): Delete function, no longer used.
714 (fusion_split_address); Delete TOC fusion support.
715 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
716 longer used with toc fusion being deleted.
717 (TARGET_TOC_FUSION_FP): Likewise.
718 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
720 (toc fusion spliter): Delete TOC fusion support.
721 (toc_fusionload_<mode>): Likewise.
722 (toc_fusionload_di): Likewise.
723 (fusion_gpr_load_<mode>): Delete generator function, this insn no
724 longer needs to be named. Rename predicate to delete TOC fusion.
725 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
726 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
727 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
728 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
729 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
731 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
733 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
734 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
735 __int128 in built-in function prototypes.
736 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
737 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
739 2018-07-27 Martin Sebor <msebor@redhat.com>
741 PR tree-optimization/86696
742 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
743 types, including enums.
744 (handle_char_store): Be prepared for the above function to fail.
746 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
748 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
749 when optimization level is lower than 2 or optimize for size.
751 2018-07-26 Martin Sebor <msebor@redhat.com>
753 PR tree-optimization/86043
754 PR tree-optimization/86042
755 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
756 (get_string_cst_length): Rename...
757 (get_min_string_length): ...to this. Add argument.
758 (handle_char_store): Extend to handle multi-character stores by
760 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
761 * tree.h (initializer_zerop): Add argument.
763 2018-07-26 Jakub Jelinek <jakub@redhat.com>
766 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
767 declare target to variables if they have always,{to,from,tofrom} map
770 2018-07-26 Martin Liska <mliska@suse.cz>
773 * lto-wrapper.c: Add linker_output as prefix
774 for ltrans_output_file.
776 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
778 PR rtl-optimization/85805
779 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
780 value for hard registers if that was written in the same mode.
782 2018-07-26 Martin Liska <mliska@suse.cz>
784 PR gcov-profile/86536
785 * gcov.c (format_gcov): Use printf format %.*f directly
786 and do not handle special values.
788 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
790 * common/config/arc/arc-common.c (arc_option_optimization_table):
791 Update default optimizations for size.
793 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
795 * config/arc/arc.md (movsf_insn): Add short instruction selection.
796 * config/arc/constraints.md (CfZ): New constraint.
797 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
798 (subsf3_fpu): Likewise.
799 (cmpsf_fpu): Likewise.
800 (cmpsf_fpu_uneq): Likewise.
802 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
804 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
806 (arc_is_uncached_mem_p): uncached applies to both the variable and
809 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
811 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
814 2018-07-25 David Malcolm <dmalcolm@redhat.com>
816 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
817 field "m_scopes" from vec to auto_vec.
819 2018-07-25 Martin Liska <mliska@suse.cz>
821 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
824 2018-07-25 Richard Biener <rguenther@suse.de>
827 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
828 special wrt context_die late.
829 (gen_subprogram_die): Re-use DIEs in local scope.
831 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
833 PR tree-optimization/86644
834 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
836 2018-07-25 Martin Liska <mliska@suse.cz>
839 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
840 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
842 2018-07-25 Martin Liska <mliska@suse.cz>
845 * params.def: Explain ASan abbreviation and provide
846 a documentation link.
848 2018-07-24 Martin Sebor <msebor@redhat.com>
850 PR tree-optimization/86622
851 PR tree-optimization/86532
852 * builtins.h (string_length): Declare.
853 * builtins.c (c_strlen): Correct handling of non-constant offsets.
854 (check_access): Be prepared for non-constant length ranges.
855 (string_length): Make extern.
856 * expr.c (string_constant): Only handle the minor non-constant
857 array index. Use string_constant to compute the length of
858 a generic string constant.
860 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
862 PR tree-optimization/86618
863 * tree-vect-stmts.c (vectorizable_call): Don't take the address
864 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
866 2018-07-24 David Malcolm <dmalcolm@redhat.com>
868 PR tree-optimization/86636
869 * json.cc (json::object::set): Fix comment. Add assertions.
870 (json::array::append): Move here from json.h. Add comment and an
872 (json::string::string): Likewise.
873 * json.h (json::array::append): Move to json.cc.
874 (json::string::string): Likewise.
875 * optinfo-emit-json.cc
876 (optrecord_json_writer::impl_location_to_json): Assert that we
877 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
878 wrapper around it. Expand the location once, rather than three
880 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
881 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
883 (optrecord_json_writer::optinfo_to_json): Likewise, in four
884 places. Fix some overlong lines.
886 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
888 * config/aarch64/aarch64-simd.md
889 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
890 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
891 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
892 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
893 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
894 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
895 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
896 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
897 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
899 2018-07-24 Jakub Jelinek <jakub@redhat.com>
902 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
903 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
904 and abs_d == d, do the power of two handling if profitable.
906 2018-07-24 Richard Biener <rguenther@suse.de>
908 * match.pd: Add BIT_FIELD_REF canonicalizations.
910 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
913 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
915 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
917 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
918 terminated STRING_CST object.
920 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
922 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
924 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
926 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
927 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
928 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
929 the elements into a register.
930 (rs6000_split_v4si_init_di_reg): Delete.
931 (rs6000_split_v4si_init): Delete.
932 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
933 (vsx_init_v4si): Rewrite as a define_expand.
935 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
937 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
938 zero_extend argument from memory): New.
940 2018-07-22 Martin Sebor <msebor@redhat.com>
943 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
944 diagnosing calls with unknown arguments unless -Walloca-larger-than
945 is restricted to less than PTRDIFF_MAX bytes.
947 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
949 * doc/gcov.texi (Invoking Gcov): Editorial changes.
951 2018-07-20 David Malcolm <dmalcolm@redhat.com>
953 * pretty-print.c (text_info::set_location): Remove redundant
954 "line_table" parameter from call to rich_location::set_range.
956 2018-07-20 Martin Sebor <msebor@redhat.com>
959 * builtins.c (expand_builtin_alloca): Adjust.
960 * calls.c (alloc_max_size): Simplify.
961 * cgraphunit.c (cgraph_node::expand): Adjust.
962 * common.opt (larger_than_size, warn_frame_larger_than): Remove
964 (frame_larger_than_size): Same.
965 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
966 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
968 * doc/invoke.texi (GCC Command Options): Document option arguments.
969 Explain byte-size arguments and suffixes.
970 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
971 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
972 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
973 * doc/options.texi (UInteger): Expand.
974 (Host_Wide_Int, ByteSize): Document new properties.
975 * final.c (final_start_function_1): Include sizes in an error message.
976 * function.c (frame_offset_overflow): Same.
977 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
978 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
979 Diagnose unbounded alloca calls only for limits of less than
981 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
982 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
984 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
985 only for limits of less than PTRDIFF_MAX.
986 * langhooks-def.h (lhd_handle_option): Change function argument
988 * langhooks.c (lhd_handle_option): Same.
989 * langhooks.h (handle_option): Same.
990 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
992 (var_type, var_type_struct): Same.
993 (var_set): Handle ByteSize flag.
994 * optc-gen.awk: Add comments to output to ease debugging. Make
995 use of HOST_WIDE_INT where appropriate.
996 * opts-gen-save.awk: Use %lx to format unsigned long.
997 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
998 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
999 arguments. Parse bytes-size suffixes.
1000 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
1001 (enum_value_to_arg): Same.
1002 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
1003 (handle_option): Adjust.
1004 (generate_option): Change function argument to HOST_WIDE_INT.
1005 (cmdline_handle_error): Adjust.
1006 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
1007 (set_option): Change function argument to HOST_WIDE_INT.
1008 (option_enabled): Handle cl_host_wide_int.
1009 (get_option_state): Handle CLVC_SIZE.
1010 (control_warning_option): Same.
1011 * opts.c (common_handle_option): Change function argument to
1012 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
1013 OPT_Wvla_larger_than_.
1014 * opts.h (enum cl_var_type): Add an enumerator.
1015 * stor-layout.c (layout_decl): Print a more meaningful warning.
1016 * toplev.c (output_stack_usage): Adjust.
1018 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
1020 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
1021 call to inline_expand_builtin_string_cmp.
1022 (expand_builtin_strcmp): Likewise.
1023 (expand_builtin_strncmp): Likewise.
1024 (inline_string_cmp): Delete the last parameter, change char_type_node
1025 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
1027 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
1028 the inlining expansion on target where the type of the call has same or
1029 narrower precision than unsigned char.
1031 2018-07-20 David Malcolm <dmalcolm@redhat.com>
1033 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
1034 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
1035 * common.opt (fsave-optimization-record): New option.
1036 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
1037 * doc/invoke.texi (-fsave-optimization-record): New option.
1038 * dumpfile.c: Include "optinfo-emit-json.h".
1039 (struct kv_pair): Move to coretypes.h.
1040 (optgroup_options): Make non-static.
1041 (dump_context::end_scope): Call
1042 optimization_records_maybe_pop_dump_scope.
1043 * dumpfile.h (optgroup_options): New decl.
1044 * json.cc: New file.
1046 * optinfo-emit-json.cc: New file.
1047 * optinfo-emit-json.h: New file.
1048 * optinfo.cc: Include "optinfo-emit-json.h".
1049 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
1050 (optinfo_enabled_p): Check optimization_records_enabled_p.
1051 (optinfo_wants_inlining_info_p): Likewise.
1052 * optinfo.h: Update comment.
1053 * profile-count.c (profile_quality_as_string): New function.
1054 * profile-count.h (profile_quality_as_string): New decl.
1055 (profile_count::quality): New accessor.
1056 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
1057 and optinfo_emit_json_cc_tests.
1058 * selftest.h (selftest::json_cc_tests): New decl.
1059 (selftest::optinfo_emit_json_cc_tests): New decl.
1060 * toplev.c: Include "optinfo-emit-json.h".
1061 (compile_file): Call optimization_records_finish.
1062 (do_compile): Call optimization_records_start.
1063 * tree-ssa-live.c: Include optinfo.h.
1064 (remove_unused_scope_block_p): Retain inlining information if
1065 optinfo_wants_inlining_info_p returns true.
1067 2018-07-20 Richard Biener <rguenther@suse.de>
1070 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
1071 to cover -flto-partition=none.
1073 2018-07-20 Martin Liska <mliska@suse.cz>
1075 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
1076 (get_decl_source_range): Remove unused function.
1078 2018-07-20 Richard Biener <rguenther@suse.de>
1080 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
1081 (struct vn_phi_s): Likewise.
1082 (struct vn_reference_s): Likewise.
1083 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
1084 for searching the slot of an entry known to be in the hash itself.
1085 (vn_phi_hasher::equal): Likewise.
1086 (vn_reference_hasher::equal): Likewise.
1087 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
1089 (optimistic_info, current_info): Remove, keeping only valid_info.
1090 (vn_reference_lookup_1): Remove fallback lookup.
1091 (vn_reference_lookup_2): Likewise.
1092 (vn_nary_op_lookup_1): Likewise.
1093 (vn_phi_lookup): Likewise.
1094 (vn_nary_build_or_lookup_1): Make sure to not chain the built
1096 (vn_reference_insert): Adjust, chain the inserted hash element
1097 at last_inserted_ref.
1098 (vn_reference_insert_pieces): Likewise.
1099 (visit_reference_op_call): Likewise.
1100 (vn_nary_op_insert_into): Chain the inserted hash element at
1102 (vn_nary_op_insert_pieces): Adjust.
1103 (vn_nary_op_insert): Likewise.
1104 (vn_nary_op_insert_stmt): Likewise.
1105 (vn_phi_insert): Adjust, chain the inserted hash element at
1107 (process_scc): Remove clearing and copying the optimistic
1108 table. Instead remove elements inserted during an optimistic
1109 iteration from the single table we maintain.
1110 (init_scc_vn): Adjust.
1111 (free_scc_vn): Likewise.
1112 (sccvn_dom_walker::record_cond): Likewise.
1113 (sccvn_dom_walker::after_dom_children): Likewise.
1115 2018-07-19 Martin Sebor <msebor@redhat.com>
1117 PR tree-optimization/84047
1118 PR tree-optimization/83776
1119 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
1120 (check_array_bounds): Call it.
1122 2018-07-19 Martin Sebor <msebor@redhat.com>
1124 * align.h (align_flags): Use member initialization.
1126 2018-07-19 David Malcolm <dmalcolm@redhat.com>
1128 * Makefile.in (OBJS): Add optinfo.o.
1129 * coretypes.h (class symtab_node): New forward decl.
1130 (struct cgraph_node): New forward decl.
1131 (class varpool_node): New forward decl.
1132 * dump-context.h: New file.
1133 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
1135 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
1136 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
1137 (set_alt_dump_file): Likewise.
1138 (dump_context::~dump_context): New dtor.
1139 (dump_gimple_stmt): Move implementation to...
1140 (dump_context::dump_gimple_stmt): ...this new member function.
1141 Add the stmt to any pending optinfo, creating one if need be.
1142 (dump_gimple_stmt_loc): Move implementation to...
1143 (dump_context::dump_gimple_stmt_loc): ...this new member function.
1144 Start a new optinfo and add the stmt to it.
1145 (dump_gimple_expr): Move implementation to...
1146 (dump_context::dump_gimple_expr): ...this new member function.
1147 Add the stmt to any pending optinfo, creating one if need be.
1148 (dump_gimple_expr_loc): Move implementation to...
1149 (dump_context::dump_gimple_expr_loc): ...this new member function.
1150 Start a new optinfo and add the stmt to it.
1151 (dump_generic_expr): Move implementation to...
1152 (dump_context::dump_generic_expr): ...this new member function.
1153 Add the tree to any pending optinfo, creating one if need be.
1154 (dump_generic_expr_loc): Move implementation to...
1155 (dump_context::dump_generic_expr_loc): ...this new member
1156 function. Add the tree to any pending optinfo, creating one if
1158 (dump_printf): Move implementation to...
1159 (dump_context::dump_printf_va): ...this new member function. Add
1160 the text to any pending optinfo, creating one if need be.
1161 (dump_printf_loc): Move implementation to...
1162 (dump_context::dump_printf_loc_va): ...this new member function.
1163 Start a new optinfo and add the stmt to it.
1164 (dump_dec): Move implementation to...
1165 (dump_context::dump_dec): ...this new member function. Add the
1166 value to any pending optinfo, creating one if need be.
1167 (dump_context::dump_symtab_node): New member function.
1168 (dump_context::get_scope_depth): New member function.
1169 (dump_context::begin_scope): New member function.
1170 (dump_context::end_scope): New member function.
1171 (dump_context::ensure_pending_optinfo): New member function.
1172 (dump_context::begin_next_optinfo): New member function.
1173 (dump_context::end_any_optinfo): New member function.
1174 (dump_context::s_current): New global.
1175 (dump_context::s_default): New global.
1176 (dump_scope_depth): Delete global.
1177 (dumpfile_ensure_any_optinfo_are_flushed): New function.
1178 (dump_symtab_node): New function.
1179 (get_dump_scope_depth): Reimplement in terms of dump_context.
1180 (dump_begin_scope): Likewise.
1181 (dump_end_scope): Likewise.
1182 (selftest::temp_dump_context::temp_dump_context): New ctor.
1183 (selftest::temp_dump_context::~temp_dump_context): New dtor.
1184 (selftest::verify_item): New function.
1185 (ASSERT_IS_TEXT): New macro.
1186 (ASSERT_IS_TREE): New macro.
1187 (ASSERT_IS_GIMPLE): New macro.
1188 (selftest::test_capture_of_dump_calls): New test.
1189 (selftest::dumpfile_c_tests): Call it.
1190 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
1191 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
1192 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
1193 descriptive comment.
1194 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
1195 (dump_node, dump_bb): Move these unrelated decls.
1196 (class dump_manager): Add leading comment.
1197 * optinfo.cc: New file.
1198 * optinfo.h: New file.
1200 2018-07-19 Michael Collison <michael.collison@arm.com>
1201 Richard Henderson <rth@redhat.com>
1203 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
1204 (subti): Handle op1 zero.
1205 (subvti4, usub4ti4): New.
1206 (*sub<GPI>3_compare1_imm): New.
1207 (sub<GPI>3_carryinCV): New.
1208 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
1209 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
1211 2018-07-19 Michael Collison <michael.collison@arm.com>
1212 Richard Henderson <rth@redhat.com>
1214 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
1215 (addti3): Create simpler code if low part is already known to be 0.
1216 (addvti4, uaddvti4): New.
1217 (*add<GPI>3_compareC_cconly_imm): New.
1218 (*add<GPI>3_compareC_cconly): New.
1219 (*add<GPI>3_compareC_imm): New.
1220 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
1221 handle constants within this pattern..
1222 (*add<GPI>3_compareV_cconly_imm): New.
1223 (*add<GPI>3_compareV_cconly): New.
1224 (*add<GPI>3_compareV_imm): New.
1225 (add<GPI>3_compareV): New.
1226 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
1227 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
1228 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
1229 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
1230 with 'comparison' operator.
1231 (*add<GPI>3_compareV_cconly_imm): Ditto.
1232 (*add<GPI>3_compareV_cconly): Ditto.
1233 (*add<GPI>3_compareV_imm): Ditto.
1234 (add<GPI>3_compareV): Ditto.
1235 (add<mode>3_carryinC): Ditto.
1236 (*add<mode>3_carryinC_zero): Ditto.
1237 (*add<mode>3_carryinC): Ditto.
1238 (add<mode>3_carryinV): Ditto.
1239 (*add<mode>3_carryinV_zero): Ditto.
1240 (*add<mode>3_carryinV): Ditto.
1242 2018-07-19 Michael Collison <michael.collison@arm.com>
1243 Richard Henderson <rth@redhat.com>
1245 * config/aarch64/aarch64-modes.def (CC_V): New.
1246 * config/aarch64/aarch64-protos.h
1247 (aarch64_addti_scratch_regs): Declare
1248 (aarch64_subvti_scratch_regs): Declare.
1249 (aarch64_expand_subvti): Declare.
1250 (aarch64_gen_unlikely_cbranch): Declare
1251 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
1252 for signed overflow using CC_Vmode.
1253 (aarch64_get_condition_code_1): Handle CC_Vmode.
1254 (aarch64_gen_unlikely_cbranch): New function.
1255 (aarch64_addti_scratch_regs): New function.
1256 (aarch64_subvti_scratch_regs): New function.
1257 (aarch64_expand_subvti): New function.
1259 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
1261 * config/aarch64/aarch64-option-extensions.def: New entry for profile
1263 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
1264 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
1267 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
1270 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
1271 address check not strict.
1273 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
1275 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
1277 (store_pair_lanes<mode>): Likewise.
1278 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
1279 enum value 'ADDR_QUERY_LDP_STP_N'.
1280 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
1281 (aarch64_print_address_internal): Add declaration.
1282 (aarch64_print_ldpstp_address): Remove.
1283 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
1284 (aarch64_print_operand): Change printing of 'y'.
1285 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
1286 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
1287 'true' rather than '1'.
1288 * config/aarch64/constraints.md (Uml): Likewise.
1289 (Uml): Rename to Umn.
1292 2018-07-19 Richard Biener <rguenther@suse.de>
1294 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
1296 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
1297 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
1298 (vn_reference_hasher): Likewise.
1299 (struct vn_tables_s): Remove obstack and alloc-pool members.
1300 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
1301 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
1302 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
1303 (vn_reference_insert_pieces): Likewise.
1304 (alloc_vn_nary_op_noinit): Adjust.
1305 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
1306 (vn_phi_eq): Adjust.
1307 (shared_lookup_phiargs): Remove.
1308 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
1309 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
1310 (visit_reference_op_call): Likewise.
1311 (copy_nary, copy_phi, copy_reference): Remove.
1312 (process_scc): Rewind the obstack when iterating. Do not
1313 copy the elements to valid_info but just move them from one
1314 hashtable to the other.
1315 (allocate_vn_table): Adjust.
1316 (free_vn_table): Likewise.
1317 (init_scc_vn): Likewise.
1318 (free_scc_vn): Likewise.
1320 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
1323 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
1324 indirect_return as function type attribute.
1325 (ix86_attribute_table): Change indirect_return to function
1327 * doc/extend.texi: Update indirect_return attribute.
1329 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
1331 * wide-int.h (widest2_int): New.
1332 * gimple-fold.c (arith_overflowed_p): Use it.
1333 * tree.h (widest2_int_cst): New.
1334 * tree-vrp.c (wide_int_binop_overflow): Rename from
1335 vrp_int_const_binop.
1336 Rewrite to work on trees.
1337 (extract_range_from_multiplicative_op_1): Abstract code to...
1338 (wide_int_range_min_max): ...here.
1339 (wide_int_range_cross_product): ...and here.
1340 (extract_range_from_binary_expr_1): Abstract overflow code to...
1341 (wide_int_range_mult_wrapping): ...here.
1342 * tree-vrp.h (wide_int_range_cross_product): New.
1343 (wide_int_range_mult_wrapping): New.
1345 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
1346 Julia Koval <julia.koval@intel.com>
1348 * config/i386/x86-tune-costs.h (skylake_memcpy,
1349 skylake_memset): Replace rep_prefix with unrolling for size 512.
1351 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
1354 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle comparision with EQ_EXPR
1357 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
1359 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
1360 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
1361 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
1362 previously known as "PowerPC AltiVec Built-in Functions". Move
1363 some material to new subsubsections "PowerPC AltiVec Built-in
1364 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
1366 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
1367 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
1368 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
1369 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
1371 2018-07-18 Richard Biener <rguenther@suse.de>
1373 PR tree-optimization/86557
1374 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
1377 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
1379 * config/s390/s390.c (s390_function_profiler): Generate CFI.
1381 2018-07-17 Jeff Law <law@redhat.com>
1383 * config/arm/arm.c (get_label_padding): Update for recent
1384 changes to label_to_alignment.
1386 PR tree-optimization/86010
1387 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
1389 * config/mips/mips.c (vr4130_align_insns): Update for recent
1390 changes to label_to_alignment.
1392 * config/frv/frv.c (frv_label_align): Update for recent changes
1393 to label_to_alignment.
1395 * config/nios2/nios2.c (nios2_label_align): Update for recent
1396 changes which dropped ALIGN_LABELS_LOG.
1398 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
1400 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
1401 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
1403 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
1405 * config/arc/arc.c (arc_label_align): Use align_labels instead of
1406 deprecated align_labels_log.
1408 2018-07-17 Richard Biener <rguenther@suse.de>
1411 * dwarf2out.c (init_sections_and_labels): Always generate
1412 a debug_line_str_section for early LTO debug.
1413 (dwarf2out_finish): Reset debug_line_str_hash output early.
1414 Bump counter for extra dwarf5 .debug_loc labels to not conflict
1416 (dwarf2out_early_finish): Output debug_line_str.
1418 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
1420 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
1421 index register on z196 or later.
1423 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
1425 * config/s390/s390.c (s390_default_align): Set default function
1427 (s390_override_options_after_change): Call s390_default align.
1428 (s390_option_override_internal): Call s390_default_align.
1429 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
1431 2018-07-17 Jakub Jelinek <jakub@redhat.com>
1434 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
1435 to _looptemp_ clauses, other than the first two.
1437 2018-07-17 Martin Liska <mliska@suse.cz>
1439 * opts.c: Do not enable OPT_falign_* for -Os.
1441 2018-07-17 Martin Liska <mliska@suse.cz>
1443 * align.h (MAX_CODE_ALIGN): New.
1444 (MAX_CODE_ALIGN_VALUE): New.
1445 * common/config/i386/i386-common.c (ix86_handle_option):
1446 (MAX_CODE_ALIGN): Moved to align.h.
1447 * final.c (MAX_CODE_ALIGN): Likewise.
1448 * opts.c (parse_and_check_align_values):
1449 (MAX_CODE_ALIGN): Likewise.
1450 (MAX_CODE_ALIGN_VALUE): Likewise.
1452 2018-07-17 Martin Liska <mliska@suse.cz>
1454 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
1455 in order to fulfil coding style.
1456 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
1457 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1458 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
1459 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
1460 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
1461 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
1462 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
1463 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
1464 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1466 2018-07-17 Martin Liska <mliska@suse.cz>
1468 * align.h: New file.
1469 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions directly.
1470 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
1471 align_flags of label_to_alignment.
1472 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into align_flags
1474 * config/m68k/m68k.c: Do not use removed align_labels_value and
1476 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
1477 (LOOP_ALIGN): Likewise.
1478 (LABEL_ALIGN): Likewise.
1479 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
1480 Remove not used macro.
1481 (rs6000_loop_align): Change return type to align_flags.
1482 (rs6000_loop_align_max_skip): Remove.
1483 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
1484 Change return type to align_flags.
1485 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
1486 Remove not used macro.
1487 (rs6000_loop_align): Change return type to align_flags.
1488 (rs6000_loop_align_max_skip): Remove.
1489 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
1490 * config/rx/rx-protos.h (rx_align_for_label): Make it
1492 * config/rx/rx.c (rx_align_for_label): Change return type
1494 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
1496 into align_flags class.
1497 (LABEL_ALIGN): Likewise.
1498 (LOOP_ALIGN): Likewise.
1499 * config/s390/s390.c (s390_label_align): Use align_flags
1501 (s390_asm_output_function_label): Likewise.
1502 * config/sh/sh.c (sh_override_options_after_change):
1503 Use align_flags class directly without macros.
1504 (find_barrier): Likewise.
1505 (barrier_align): Likewise.
1506 (sh_loop_align): Likewise.
1507 * config/spu/spu.c (spu_option_override):
1508 Use align_flags_tuple::get_value instead of removed macros.
1509 (spu_sched_init): Likewise.
1510 * config/spu/spu.h (GTY): Likewise.
1511 * config/visium/visium.c (visium_option_override):
1512 Set "8" as default secondary alignment.
1513 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
1514 in order to guarantee secondary alignment of 8.
1515 * coretypes.h: Include align.h header file.
1516 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
1517 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
1518 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
1519 * doc/tm.texi.in: Likewise.
1520 * final.c (struct label_alignment): Remove not used structure.
1521 (LABEL_ALIGN): Change type to align_flags.
1522 (LOOP_ALIGN): Likewise.
1523 (JUMP_ALIGN): Likewise.
1524 (default_loop_align_max_skip): Remove.
1525 (default_label_align_max_skip): Likewise.
1526 (default_jump_align_max_skip): Likewise.
1527 (default_label_align_after_barrier_max_skip):
1528 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
1529 (LABEL_TO_MAX_SKIP): Remove.
1530 (label_to_alignment): Return align_flags type instead of integer.
1531 (label_to_max_skip): Remove.
1532 (align_fuzz): Use align_flags type.
1533 (compute_alignments): Use align_flags type and use align_flags::max
1534 to combine multiple alignments.
1535 (grow_label_align): Grow vec instead of C array.
1536 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
1537 (shorten_branches): Use align_flags type and use align_flags::max
1538 to combine multiple alignments.
1539 (final_scan_insn_1): Remove usage of secondary alignment that comes
1540 from label alignment, but instead use proper secondary alignment
1541 which is computed in grow_label_align.
1542 * flags.h (struct align_flags_tuple): Move to align.h.
1543 (struct align_flags): Likewise.
1544 (state_align_loops): Rename to align_loops.
1545 (state_align_jumps): Rename to align_jumps.
1546 (state_align_labels): Rename to align_labels.
1547 (state_align_functions): Rename to align_functions.
1548 (align_loops_log): Remove.
1549 (align_jumps_log): Remove.
1550 (align_labels_log): Remove.
1551 (align_functions_log): Remove.
1552 (align_loops_max_skip): Remove.
1553 (align_jumps_max_skip): Remove.
1554 (align_labels_max_skip): Remove.
1555 (align_functions_max_skip): Remove.
1556 (align_loops_value): Remove.
1557 (align_jumps_value): Remove.
1558 (align_labels_value): Remove.
1559 (align_functions_value): Remove.
1560 * output.h (label_to_alignment): Change return type to align_flags.
1561 (label_to_max_skip): Remove.
1562 * target.def: Remove loop_align_max_skip, label_align_max_skip,
1563 jump_align_max_skip macros.
1564 * targhooks.h (default_loop_align_max_skip): Remove.
1565 (default_label_align_max_skip): Likewise.
1566 (default_jump_align_max_skip): Likewise.
1567 (default_label_align_after_barrier_max_skip): Remove.
1568 * toplev.c (read_log_maxskip): Use ::normalize function.
1569 (parse_N_M): Remove not used argument and also call ::normalize.
1570 (parse_alignment_opts): Do not pass unused arguments.
1571 * varasm.c (assemble_start_function): Use directly align_functions
1572 instead of removed macros.
1573 * system.h: Do not poison removed macros.
1575 2018-07-17 Jakub Jelinek <jakub@redhat.com>
1578 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
1579 and cond temporaries don't have reference type if iterator has
1580 pointer type. For init use &for_pre_body instead of pre_p if
1581 for_pre_body is non-empty.
1583 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
1585 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
1586 double-double modes to SFmode directly directly.
1587 (trunc<mode>sf2_fprs): Delete.
1589 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
1591 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
1592 for conversions between IFmode and the decimal floating point modes.
1593 (init_float128_ieee): Use the correct names for conversions between
1594 KFmode and the decimal floating point modes.
1596 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
1598 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
1599 for the conversions between TDmode and IFmode.
1600 (init_float128_ieee): Use more correct names for the conversions
1601 between TDmode and KFmode.
1603 2018-07-16 Jakub Jelinek <jakub@redhat.com>
1605 PR tree-optimization/86526
1606 * builtins.c (expand_builtin_memcmp): Formatting fixes.
1607 (inline_expand_builtin_string_cmp): Likewise.
1608 (inline_string_cmp): Likewise. Use c_readstr instead of
1609 builtin_memcpy_read_str. Add unit_mode temporary.
1611 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
1614 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
1615 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
1617 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
1619 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
1620 Alphabetize prototypes of built-in functions, separating out
1621 built-in functions that are listed in this section but should be
1622 described elsewhere.
1624 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
1627 * expmed.c (emit_store_flag): Do not emit setcc followed by a
1628 conditional move when trapping comparison was split to a
1629 non-trapping one (and vice versa).
1631 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
1633 * config/s390/s390.c (s390_function_profiler): Generate nops
1634 instead of profiler call sequences.
1635 * config/s390/s390.opt: Add the new option.
1637 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
1639 * config/s390/s390.c (s390_function_profiler): Generate
1640 __mcount_loc section.
1641 * config/s390/s390.opt: Add the new option.
1643 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
1645 * common.opt: Add the new warning.
1646 * config/s390/s390.c (s390_function_profiler): Emit "brasl
1647 %r0,__fentry__" when -mfentry is specified.
1648 (s390_option_override_internal): Disallow -mfentry for 31-bit
1650 * config/s390/s390.opt: Add the new option.
1652 2018-07-16 Richard Biener <rguenther@suse.de>
1655 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
1656 for function-local FUNCTION_DECL and RESULT_DECL immediately.
1658 2018-07-16 Martin Liska <mliska@suse.cz>
1661 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
1664 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
1666 * config/arc/arcHS.md: Update ARCHS scheduling rules.
1668 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
1670 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
1672 * config/arc/arc-cpus.def (hs4x): New cpu.
1674 * config/arc/arc-tables.opt: Regenerate.
1675 * config/arc/arc.c (arc_sched_issue_rate): New function.
1676 (TARGET_SCHED_ISSUE_RATE): Define.
1677 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
1678 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
1680 (attr tune): Add ARCHS4x tune values.
1681 (attr tune_dspmpy): Define.
1682 (*tst): Correct instruction type.
1683 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
1684 * config/arc/arcHS4x.md: New file.
1685 * config/arc/fpu.md: Update instruction type attributes.
1686 * config/arc/t-multilib: Regenerate.
1688 2018-07-16 Tom de Vries <tdevries@suse.de>
1691 * var-tracking.c (vt_initialize): Fix pre_dec handling.
1693 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1695 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
1698 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
1700 PR tree-optimization/86514
1701 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
1702 conversion to a boolean type from a type with greater precision.
1704 2018-07-16 Tom de Vries <tdevries@suse.de>
1706 * var-tracking.c (vt_initialize): Print adjusted insn slim if
1707 dump_flags request TDF_SLIM.
1709 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
1711 * fold-const.c (int_const_binop_1): Abstract...
1712 (wide_int_binop): ...wide int code here.
1713 (poly_int_binop): ...poly int code here.
1714 Abstract the rest of int_const_binop_1 into int_const_binop.
1715 * fold-const.h (wide_int_binop): New.
1716 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
1717 Remove useless PLUS/MINUS_EXPR case.
1718 (zero_nonzero_bits_from_vr): Move wide int code...
1719 (zero_nonzero_bits_from_bounds): ...here.
1720 (extract_range_from_binary_expr_1): Move mask optimization code...
1721 (range_easy_mask_min_max): ...here.
1722 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
1723 (range_easy_mask_min_max): New.
1725 2018-07-15 Jeff Law <law@redhat.com>
1728 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
1731 2018-07-14 Jim Wilson <jimw@sifive.com>
1733 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
1735 2018-07-14 Paul Koning <ni1d@arrl.net>
1737 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
1739 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
1741 * lto-streamer-out.c (copy_function_or_variable): Dump info about
1744 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
1745 Steve Munroe <munroesj52@gmail.com>
1747 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
1748 (_mm_andnot_si128): Likewise.
1749 (_mm_or_si128): Likewise.
1750 (_mm_xor_si128): Likewise.
1752 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
1755 * builtins.c (expand_builtin_memcmp): Inline the calls first
1756 when result_eq is false.
1757 (expand_builtin_strcmp): Inline the calls first.
1758 (expand_builtin_strncmp): Likewise.
1759 (inline_string_cmp): New routine. Expand a string compare
1760 call by using a sequence of char comparison.
1761 (inline_expand_builtin_string_cmp): New routine. Inline expansion
1762 a call to str(n)cmp/memcmp.
1763 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
1765 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
1767 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
1769 * config/arm/driver-arm.c: Include arm-native.h.
1770 (host_detect_local_cpu): Use auto-generated data tables.
1771 (vendors, arm_cpu_table): Delete. Move part information to ...
1772 * config/arm/arm-cpus.in: ... here.
1773 * config/arm/parsecpu.awk (gen_native): New function.
1774 (vendor, part): New CPU fields.
1775 (END): Add support for building the native CPU detection tables.
1776 * config/arm/t-arm (arm-native.h): Add build rule.
1777 (driver-arm.o): Add dependency on arm-native.h.
1779 2018-07-13 Richard Biener <rguenther@suse.de>
1782 * match.pd (addr1 - addr2): Allow either of the operand to
1785 2018-07-13 Tom de Vries <tdevries@suse.de>
1787 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
1790 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
1792 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
1793 arrays instead of numbered variables.
1795 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
1797 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
1798 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
1800 2018-07-13 Richard Biener <rguenther@suse.de>
1803 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
1804 instead of get_context_die.
1806 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
1807 Richard Biener <rguenther@suse.de>
1810 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
1811 that the loop latch destination where phi is defined.
1813 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
1815 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
1816 (riscv_expand_epilogue): Add assertion to check interrupt mode.
1817 (riscv_set_current_function): Extract getting interrupt type to new
1819 (riscv_get_interrupt_type): New function.
1820 (riscv_merge_decl_attributes): New function, checking interrupt type is
1822 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
1824 2018-07-12 Paul Koning <ni1d@arrl.net>
1826 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
1829 2018-07-12 Paul Koning <ni1d@arrl.net>
1831 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
1832 zero reference, add doloop_end instead.
1833 * doc/md.texi (decrement_and_branch_until_zero): Remove.
1834 (Looping patterns): Remove decrement_and_branch_until_zero. Add
1835 detail for doloop_end.
1837 2018-07-12 Martin Sebor <msebor@redhat.com>
1840 * attribs.c (decl_attributes): Reject conflicting attributes before
1841 calling attribute handlers.
1843 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
1845 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
1847 (gcc::dump_manager::get_dump_file_name): likewise.
1848 (dump_begin): Likewise.
1849 * dumpfile.h (dump_begin): Update prototype.
1850 (gcc::dump_manager::get_dump_file_name,
1851 gcc::dump_manager::get_dump_file_name): Update prototype.
1853 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
1855 * internal-fn.h (vectorizable_internal_fn_p): New function.
1856 * tree-vect-slp.c (compatible_calls_p): Likewise.
1857 (vect_build_slp_tree_1): Remove nops argument. Handle calls
1858 to internal functions.
1859 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
1861 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
1863 * fold-const.h (inverse_conditions_p): Declare.
1864 * fold-const.c (inverse_conditions_p): New function.
1865 * match.pd: Use inverse_conditions_p. Add folds of view_converts
1866 that test the inverse condition of a conditional internal function.
1867 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
1868 * internal-fn.c (internal_fn_mask_index): Handle conditional
1870 (vectorized_internal_fn_supported_p): New function.
1871 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
1872 (any_pred_load_store): Replace with...
1873 (need_to_predicate): ...this new variable.
1874 (redundant_ssa_names): New variable.
1875 (ifcvt_can_use_mask_load_store): Move initial checks to...
1876 (ifcvt_can_predicate): ...this new function. Handle tree codes
1877 for which a conditional internal function exists.
1878 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
1879 instead of ifcvt_can_use_mask_load_store. Update after variable
1881 (predicate_load_or_store): New function, split out from
1882 predicate_mem_writes.
1883 (check_redundant_cond_expr): New function.
1884 (value_available_p): Likewise.
1885 (predicate_rhs_code): Likewise.
1886 (predicate_mem_writes): Rename to...
1887 (predicate_statements): ...this. Use predicate_load_or_store
1888 and predicate_rhs_code.
1889 (combine_blocks, tree_if_conversion): Update after above name changes.
1890 (ifcvt_local_dce): Handle redundant_ssa_names.
1891 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
1892 general conditional functions.
1893 * tree-vect-stmts.c (vectorizable_call): Likewise.
1895 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
1896 Alan Hayward <alan.hayward@arm.com>
1897 David Sherwood <david.sherwood@arm.com>
1899 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
1900 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
1901 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
1902 plus and minus and convert them into IFN_COND_FMA-based sequences.
1903 (convert_mult_to_fma): Handle conditional plus and minus.
1905 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
1907 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
1908 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
1909 (cond_fnms_optab): New optabs.
1910 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
1912 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
1913 * internal-fn.h (get_conditional_internal_fn): Declare.
1914 (get_unconditional_internal_fn): Likewise.
1915 * internal-fn.c (cond_ternary_direct): New macro.
1916 (expand_cond_ternary_optab_fn): Likewise.
1917 (direct_cond_ternary_optab_supported_p): Likewise.
1918 (FOR_EACH_COND_FN_PAIR): Likewise.
1919 (get_conditional_internal_fn): New function.
1920 (get_unconditional_internal_fn): Likewise.
1921 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
1922 (gimple_match_op::gimple_match_op): Add a new overload for 5
1924 (gimple_match_op::set_op): Likewise.
1925 (gimple_resimplify5): Declare.
1926 * genmatch.c (decision_tree::gen): Generate simplifications for
1928 * gimple-match-head.c (gimple_simplify): Define an overload for
1929 5 operands. Handle calls with 5 arguments in the top-level overload.
1930 (convert_conditional_op): Handle conversions from unconditional
1931 internal functions to conditional ones.
1932 (gimple_resimplify5): New function.
1933 (build_call_internal): Pass a fifth operand.
1934 (maybe_push_res_to_seq): Likewise.
1935 (try_conditional_simplification): Try converting conditional
1936 internal functions to unconditional internal functions.
1937 Handle 3-operand unconditional forms.
1938 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
1939 Define ternary equivalents of the current rules for binary conditional
1941 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
1943 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
1944 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
1945 (optab): Handle them.
1946 (SVE_COND_FP_TERNARY): New int iterator.
1947 (sve_fmla_op, sve_fmad_op): New int attributes.
1948 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
1949 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
1950 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
1952 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
1954 * target.def (preferred_else_value): New target hook.
1955 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
1956 * doc/tm.texi: Regenerate.
1957 * targhooks.h (default_preferred_else_value): Declare.
1958 * targhooks.c (default_preferred_else_value): New function.
1959 * internal-fn.h (conditional_internal_fn_code): Declare.
1960 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
1961 (get_conditional_internal_fn): Use it.
1962 (conditional_internal_fn_code): New function.
1963 * gimple-match.h (gimple_match_cond): New struct.
1964 (gimple_match_op): Add a cond member function.
1965 (gimple_match_op::gimple_match_op): Update all forms to take a
1967 * genmatch.c (expr::gen_transform): Use the same condition as res_op
1968 for the suboperation, but don't specify a particular else_value.
1969 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
1970 (visit_nary_op, visit_reference_op_load): Pass
1971 gimple_match_cond::UNCOND to the gimple_match_op constructor.
1972 * gimple-match-head.c: Include tree-eh.h
1973 (convert_conditional_op): New function.
1974 (maybe_resimplify_conditional_op): Likewise.
1975 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
1976 (gimple_resimplify2): Likewise.
1977 (gimple_resimplify3): Likewise.
1978 (gimple_resimplify4): Likewise.
1979 (maybe_push_res_to_seq): Return null for conditional operations.
1980 (try_conditional_simplification): New function.
1981 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
1983 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
1985 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
1987 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
1989 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
1991 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
1993 (hash_tree): Do not hash DECL_FCONTEXT
1994 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
1995 Do not stream DECL_FCONTEXT.
1996 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
1997 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
1999 2018-07-12 Richard Biener <rguenther@suse.de>
2002 * dwarf2out.c (gen_block_die): Only output blocks when they have
2003 at least one !DECL_IGNORED_P variable.
2005 2018-07-12 Richard Biener <rguenther@suse.de>
2008 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
2009 Remove -mieee-fp handling.
2011 2018-07-12 Richard Biener <rguenther@suse.de>
2013 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
2014 left-over from last patch.
2016 2018-07-12 Jakub Jelinek <jakub@redhat.com>
2018 PR tree-optimization/86492
2019 * gimple-ssa-store-merging.c
2020 (imm_store_chain_info::coalesce_immediate_stores): Call
2021 check_no_overlap even for the merge_overlapping case. Formatting fix.
2023 2018-07-12 Richard Biener <rguenther@suse.de>
2026 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
2027 move possibly trapping operations into the conditional.
2029 2018-07-12 Richard Biener <rguenther@suse.de>
2031 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
2032 (vn_lookup_simplify_result): Remove recursion limit applied
2034 (vn_nary_build_or_lookup_1): Adjust.
2035 (try_to_simplify): Likewise.
2036 * gimple-match-head.c (gimple_resimplify1): Instead apply one
2038 (gimple_resimplify2): Likewise.
2039 (gimple_resimplify3): Likewise.
2040 (gimple_resimplify4): Likewise.
2042 2018-07-11 Jakub Jelinek <jakub@redhat.com>
2044 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
2045 Use __mmask64 type instead of __mmask8 for __M argument.
2046 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
2047 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
2049 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
2050 __mmask16 for __M argument.
2051 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
2052 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
2053 to __mmask16 instead of __mmask8.
2054 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
2055 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
2056 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
2057 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
2058 instead of __mmask16 for __U argument.
2059 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
2060 __mmask16 instead of __mmask8 for __U argument.
2061 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
2063 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
2065 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
2067 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
2069 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
2071 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
2073 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
2075 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
2077 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
2079 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
2081 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
2082 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
2083 return type as well as __M argument type and all casts from __mmask8
2085 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
2086 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
2087 return type as well as __M argument type and all casts from __mmask8
2089 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
2090 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
2091 return type as well as __M argument type and all casts from __mmask8
2093 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
2094 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
2095 return type as well as __M argument type and all casts from __mmask8
2097 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
2098 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
2101 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
2103 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
2104 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
2107 2018-07-11 Paul Koning <ni1d@arrl.net>
2109 * doc/md.texi (define_subst): Document how multiple occurrences of
2110 the same argument in the replacement pattern are handled.
2112 2018-07-11 Paul Koning <ni1d@arrl.net>
2114 * doc/extend.texi (Common Variable Attributes): Move "mode" into
2116 (Common Type Attributes): Add "mode" attribute.
2118 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
2120 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
2121 stream DECL_ORIGINAL_TYPE.
2122 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
2123 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
2124 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
2125 Do not walk original type.
2126 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
2128 (write_ts_decl_non_common_tree_pointers): Do not stream
2130 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
2131 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
2133 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
2135 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
2136 thread twice from the same starting edge.
2138 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
2140 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
2141 * gimple.c (gimple_call_nonnull_result_p): ...here...
2142 (gimple_call_nonnull_arg): ...and here.
2143 * gimple.h (gimple_call_nonnull_result_p): New.
2144 (gimple_call_nonnull_arg): New.
2146 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
2148 * config/arm/arm-cpus.in: Move information from fpu field of each
2149 cpu definition to the isa field.
2150 * config/arm/parsecpu.awk (fpu): Delete match rule.
2151 (gen_comm_data): Don't add bits from the CPU's FPU entry.
2153 2018-07-11 Richard Biener <rguenther@suse.de>
2156 * dwarf2out.c (init_sections_and_labels): Use
2157 output_asm_line_debug_info consistently.
2158 (dwarf2out_early_finish): Likewise.
2159 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
2162 2018-07-11 Richard Biener <rguenther@suse.de>
2164 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
2165 Rework father_bb setting in a way to avoid propagating constants
2166 multiple times on a loop body.
2168 2018-07-10 Mark Wielaard <mark@klomp.org>
2171 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
2173 2018-07-10 Richard Biener <rguenther@suse.de>
2175 * hash-map.h (hash_map::iterator::operator*): Return
2176 references to key and value.
2178 2018-07-10 Jakub Jelinek <jakub@redhat.com>
2181 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
2182 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
2183 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
2184 (gimplify_omp_for): For composite loops, move outer
2185 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
2186 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
2187 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
2188 TREE_LIST for both the original class iterator and the "last" helper
2189 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
2190 loop, remember has_decl_expr from outer composite loops for the
2191 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
2193 2018-07-09 Martin Sebor <msebor@redhat.com>
2197 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
2198 accessing implicitly initialized array elements.
2199 * expr.c (string_constant): Handle string initializers of
2200 character arrays within aggregates.
2201 * gimple-fold.c (fold_array_ctor_reference): Add argument.
2202 Store element offset. As a special case, handle zero size.
2203 (fold_nonarray_ctor_reference): Same.
2204 (fold_ctor_reference): Add argument. Store subobject offset.
2205 * gimple-fold.h (fold_ctor_reference): Add argument.
2207 2018-07-09 Paul Koning <ni1d@arrl.net>
2209 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
2210 (pdp11_insn_cost): New function.
2211 (pdp11_md_asm_adjust): New function.
2212 (TARGET_INVALID_WITHIN_DOLOOP): Define.
2213 (pdp11_rtx_costs): Update to match machine better.
2214 (output_addr_const_pdp11): Correct format mismatch warnings.
2215 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
2216 * config/pdp11/pdp11.md: General change to add base_cost and/or
2217 length attributes for use by new pdp11_insn_cost function.
2218 (MIN_BRANCH): Correct definition.
2220 (doloop_end): Use standard pattern name for looping pattern.
2221 (doloop_end_nocc): New.
2222 (movsf): Add another constraint alternative.
2223 (zero_extendqihi2): Add constraint alternatives for not in place
2225 (zero_extendhisi2): Remove.
2226 (shift patterns): Add CC handling variants.
2230 (define_peephole2): New peephole to recognize mov that sets CC for
2233 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2235 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
2236 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
2239 2018-07-09 Jakub Jelinek <jakub@redhat.com>
2242 * real.c (real_nextafter): Return true if result is denormal.
2244 2018-07-09 Martin Liska <mliska@suse.cz>
2246 * common.opt: Add back wrongly removed attribute.
2248 2018-07-09 Richard Biener <rguenther@suse.de>
2251 * dwarf2out.c (gen_block_die): For an early generated DIE
2252 always output high/low PC attributes.
2254 2018-07-09 Tom de Vries <tdevries@suse.de>
2256 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
2257 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
2258 onto VAR_DECL with abstract origin.
2260 2018-07-07 Jim Wilson <jimw@sifive.com>
2262 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
2264 2018-07-07 Tom de Vries <tdevries@suse.de>
2266 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
2268 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
2270 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
2272 (combine_bound): Use wide-int overflow calculation instead of
2274 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
2276 * fold-const.c (int_const_binop_2): Same.
2277 (extract_muldiv_1): Same.
2278 (fold_div_compare): Same.
2279 (fold_abs_const): Same.
2281 * poly-int.h (add): Same.
2285 * predict.c (predict_iv_comparison): Same.
2286 * profile-count.c (slow_safe_scale_64bit): Same.
2287 * simplify-rtx.c (simplify_const_binary_operation): Same.
2288 * tree-chrec.c (tree_fold_binomial): Same.
2289 * tree-data-ref.c (split_constant_offset_1): Same.
2290 * tree-if-conv.c (idx_within_array_bound): Same.
2291 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
2292 * tree-ssa-phiopt.c (minmax_replacement): Same.
2293 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
2294 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
2295 * vr-values.c (vr_values::adjust_range_with_scev): Same.
2296 * wide-int.cc (wi::add_large): Same.
2297 (wi::mul_internal): Same.
2298 (wi::sub_large): Same.
2299 (wi::divmod_internal): Same.
2300 * wide-int.h: Change overflow type to overflow_type for neg, add,
2301 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
2302 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
2303 mul_internal, divmod_internal.
2304 (overflow_type): New enum.
2305 (accumulate_overflow): New.
2307 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
2309 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
2310 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
2312 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
2314 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
2315 argument is checked for zero before entering loop, avoid checking again.
2317 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
2319 * gimplify.h (generic_expr_could_trap_p): Set as global function.
2320 * gimplify.h (generic_expr_could_trap_p): Likwise.
2321 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
2323 2018-07-06 Jakub Jelinek <jakub@redhat.com>
2325 PR tree-optimization/86401
2326 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
2327 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
2328 (fold_bit_and_mask): ... here. New helper function for match.pd.
2329 * fold-const.h (fold_bit_and_mask): Declare.
2330 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
2332 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
2335 * target.def (translate_mode_attribute): New hook.
2336 * targhooks.h (default_translate_mode_attribute): Declare.
2337 * targhooks.c (default_translate_mode_attribute): New function.
2338 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
2339 * doc/tm.texi: Regenerate.
2340 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
2341 (rs6000_translate_mode_attribute): New function.
2343 2018-07-06 Paul Koning <ni1d@arrl.net>
2345 * doc/md.texi (define_split): Document DONE and FAIL.
2346 (define_peephole2): Ditto.
2348 2018-07-05 Jeff Law <law@redhat.com>
2350 PR tree-optimization/86010
2351 * tree-ssa-dse.c (compute_trims): More aggressively trim at
2352 both the head and tail of mem* and str* calls.
2354 2018-07-05 Jim Wilson <jimw@sifive.com>
2356 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
2358 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
2360 * config/aarch64/aarch64-simd.md: correct flags text for
2361 MIN_EXPR replacement.
2363 2018-07-05 James Clarke <jrtc27@jrtc27.com>
2365 * configure: Regenerated.
2367 2018-07-05 Carl Love <cel@us.ibm.com>
2369 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
2370 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
2371 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
2372 VSX_BUILTIN_DOUBLEL_V4SF.
2374 2018-07-05 Martin Sebor <msebor@redhat.com>
2377 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
2378 than its domain to compute its the upper bound of a char array.
2380 2018-07-05 Nathan Sidwell <nathan@acm.org>
2382 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
2383 * doc/cpp.texi: Update comment.
2384 * doc/tm.texi: Rebuilt.
2385 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
2386 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
2387 * doc/extend.texi (Backwards Compatibility): Clarify it is system
2388 headers affected by extern "C".
2389 * system.h: Poison NO_IMPLICIT_EXTERN_C.
2390 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
2391 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
2392 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
2393 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
2394 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
2395 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
2396 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
2397 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
2398 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
2399 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
2400 config/sparc/sp64-elf.h, config/spu/spu.h,
2401 config/stormy16/stormy16.h, config/v850/v850.h,
2402 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
2403 define NO_IMPLICIT_EXTERN_C.
2404 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
2406 2018-07-05 Tamar Christina <tamar.christina@arm.com>
2409 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
2410 instead of GET_MODE_SIZE when comparing Units.
2412 2018-07-05 Tamar Christina <tamar.christina@arm.com>
2415 * rtlanal.c (set_noop_p): Constrain on mode change,
2416 include hard-reg-set.h
2418 2018-07-05 Tamar Christina <tamar.christina@arm.com>
2420 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
2422 2018-07-05 Jakub Jelinek <jakub@redhat.com>
2425 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
2428 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
2430 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
2432 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
2435 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
2437 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
2439 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
2442 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
2444 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
2446 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
2447 neg_*_op* variables.
2449 2018-07-04 Martin Liska <mliska@suse.cz>
2451 * tree-switch-conversion.c: Define
2452 max_ratio_for_speed and max_ratio_for_size constants.
2454 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
2455 Martin Liska <mliska@suse.cz>
2460 * common.opt: Rename align options with 'str_' prefix.
2461 * common/config/i386/i386-common.c (set_malign_value): New
2463 (ix86_handle_option): Use it to set -falign-* options/
2464 * config/aarch64/aarch64-protos.h (struct tune_params): Change
2465 type from int to string.
2466 * config/aarch64/aarch64.c: Update default values from int
2468 * config/alpha/alpha.c (alpha_override_options_after_change):
2470 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
2471 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
2472 max skip conditionally.
2473 * config/i386/freebsd.h (SUBALIGN_LOG): New.
2474 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
2475 max skip conditionally.
2476 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
2477 max skip conditionally.
2478 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
2479 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
2480 max skip conditionally.
2481 * config/i386/i386.c (struct ptt): Change type from int to
2483 (ix86_default_align): Set default values.
2484 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
2485 max skip conditionally.
2486 * config/i386/iamcu.h (SUBALIGN_LOG): New.
2487 (ASM_OUTPUT_MAX_SKIP_ALIGN):
2488 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
2489 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
2490 max skip conditionally.
2491 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
2492 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
2493 * config/i386/x86-64.h (SUBALIGN_LOG): New.
2494 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
2495 max skip conditionally.
2496 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
2497 * config/ia64/ia64.c (ia64_option_override): Set default values
2498 for alignment options.
2499 * config/m68k/m68k.c: Handle new str_align_* options.
2500 * config/mips/mips.c (mips_set_compression_mode): Change
2502 (mips_option_override): Set default values for options.
2503 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
2505 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2507 * config/rx/rx.c (rx_option_override): Likewise.
2508 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
2509 (LABEL_ALIGN): Use align_labels_log.
2510 (LOOP_ALIGN): Use align_loops_align.
2511 * config/s390/s390.c (s390_asm_output_function_label): Use new
2513 * config/sh/sh.c (sh_override_options_after_change):
2514 Change type of constants.
2515 * config/spu/spu.c (spu_sched_init): Likewise.
2516 * config/sparc/sparc.c (sparc_option_override): Set default
2518 * config/visium/visium.c (visium_option_override): Likewise.
2519 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
2520 emit p2align format with last argument if it's not needed.
2521 * doc/invoke.texi: Document extended format of -falign-*.
2522 * final.c: Use align_labels alignment.
2523 * flags.h (struct target_flag_state): Change type to use
2525 (struct align_flags_tuple): New.
2526 (struct align_flags): Likewise.
2527 (align_loops_log): Redefine macro to use new types.
2528 (align_loops_max_skip): Redefine macro to use new types.
2529 (align_jumps_log): Redefine macro to use new types.
2530 (align_jumps_max_skip): Redefine macro to use new types.
2531 (align_labels_log): Redefine macro to use new types.
2532 (align_labels_max_skip): Redefine macro to use new types.
2533 (align_functions_log): Redefine macro to use new types.
2534 (align_loops): Redefine macro to use new types.
2535 (align_jumps): Redefine macro to use new types.
2536 (align_labels): Redefine macro to use new types.
2537 (align_functions): Redefine macro to use new types.
2538 (align_functions_max_skip): Redefine macro to use new types.
2539 (align_loops_value): New macro.
2540 (align_jumps_value): New macro.
2541 (align_labels_value): New macro.
2542 (align_functions_value): New macro.
2543 * function.c (invoke_set_current_function_hook): Propagate
2544 alignment values from flags to global variables default in
2546 * ipa-icf.c (sem_function::equals_wpa): Use
2547 cl_optimization_option_eq instead of memcmp.
2548 * lto-streamer.h (cl_optimization_stream_out): Support streaming
2550 (cl_optimization_stream_in): Likewise.
2551 * optc-save-gen.awk: Support strings in cl_optimization.
2552 * opth-gen.awk: Likewise.
2553 * opts.c (finish_options): Remove error checking of invalid
2555 (MAX_CODE_ALIGN): Remove.
2556 (MAX_CODE_ALIGN_VALUE): Likewise.
2557 (parse_and_check_align_values): New function.
2558 (check_alignment_argument): Likewise.
2559 (common_handle_option): Use check_alignment_argument.
2560 * opts.h (parse_and_check_align_values): Declare.
2561 * toplev.c (init_alignments): Remove.
2562 (read_log_maxskip): New.
2563 (parse_N_M): Likewise.
2564 (parse_alignment_opts): Likewise.
2565 (backend_init_target): Remove usage of init_alignments.
2566 * toplev.h (parse_alignment_opts): Declare.
2567 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
2569 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
2570 * tree.c (cl_option_hasher::equal): New.
2571 * varasm.c: Use new global macros.
2573 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
2575 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
2576 Use a simpler align directive also if MAXSKIP = ALIGN-1.
2577 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2578 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2579 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2580 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
2581 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
2582 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
2583 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
2584 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2585 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2586 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2587 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
2589 2018-07-04 Martin Liska <mliska@suse.cz>
2590 Jonathan Wakely <jwakely@redhat.com>
2592 * coverage.c: Use correct type.
2593 * doc/invoke.texi: Language correction.
2595 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
2598 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
2599 ENDBRANCH for non-tail call which may return via indirect branch.
2600 * doc/extend.texi: Document indirect_return attribute.
2602 2018-07-03 Martin Sebor <msebor@redhat.com>
2604 PR tree-optimization/86274
2605 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
2607 (format_floating): Correct handling of infinities and NaNs.
2609 2018-07-03 Martin Sebor <msebor@redhat.com>
2611 * print-tree.c (print_real_cst): New function.
2612 (print_node_brief): Call it.
2613 (print_node): Ditto.
2615 2018-07-03 Jeff Law <law@redhat.com>
2617 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
2618 into a single pattern.
2620 * config/h8300/h8300.md (ors code_iterator): New.
2621 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
2622 a single pattern and single splitter.
2623 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
2624 (iorqi3_1, xorqi3_1): Likewise.
2625 (iorqi3, xorqi3 expanders): Similarly.
2627 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
2628 (movmd_internal) into a single pattern using the P mode iterator.
2629 (movmd splitters): Similarly.
2630 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
2631 (movsd splitters): Similarly.
2633 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
2634 ADDB, ADDW and ADDL into a single ADD attribute which selects the
2635 right table based on the size of the operand.
2636 * config/h8300/h8300.md (length_table): Corresponding changes. All
2637 references to "addb", "addw" and "addl" changed to "add".
2638 (btst patterns): Merge two variants into a single pattern.
2639 (tstqi, tsthi): Likewise.
2640 (addhi3_incdec, addsi3_incdec): Likewise.
2641 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
2642 (mulhi3, mulsi3): Likewise.
2643 (udivhi3, udivsi3): Likewise.
2644 (divhi3, divsi3): Likewise.
2645 (andorqi3, andorhi3, andorsi3): Likewise.
2647 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
2650 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
2651 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
2653 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2655 PR tree-optimization/85694
2656 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
2657 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
2659 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
2660 (<u>avg<mode>3_ceil): New patterns.
2662 2018-07-03 David Malcolm <dmalcolm@redhat.com>
2664 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
2665 scan-tree-dump directive.
2666 * gcc.dg/vect/slp-perm-2.c: Likewise.
2667 * gcc.dg/vect/slp-perm-3.c: Likewise.
2668 * gcc.dg/vect/slp-perm-5.c: Likewise.
2669 * gcc.dg/vect/slp-perm-6.c: Likewise.
2670 * gcc.dg/vect/slp-perm-7.c: Likewise.
2671 * gcc.dg/vect/slp-perm-8.c: Likewise.
2673 2018-07-03 Marek Polacek <polacek@redhat.com>
2676 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
2678 2018-07-03 Richard Biener <rguenther@suse.de>
2681 * tree-ssa-structalias.c (find_func_clobbers): Properly
2682 handle indirect calls.
2684 2018-07-03 Jeff Law <law@redhat.com>
2686 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
2687 (shifts): New code iterator.
2688 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
2689 expander. Fix HImode handling on H8/SX.
2690 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
2691 (subqi3, subhi3, subsi3 expanders): Likewise.
2692 (andqi3, andhi3, andsi3 expanders): Likewise.
2693 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
2694 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
2695 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
2696 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
2697 (zero_extendqihi2, zero_extendqisi2): Likewise.
2698 (extendqihi2, extendqisi2): Likewise.
2699 (rotlqi3, rotlhi3, rotlsi3): Likewise.
2700 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
2701 (rotlqi3_1, rotlhi3_1): Likewise.
2702 (logicalhi3_sn, logicalsi3_sn): Likewise.
2703 (logicalhi3, logicalsi3): Likewise.
2705 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2707 * tree-vect-patterns.c (vect_recog_rotate_pattern)
2708 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
2709 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
2710 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
2711 type to append_pattern_def_seq instead of creating a stmt_vec_info
2713 (build_mask_conversion): Likewise. Remove vinfo argument.
2714 (vect_add_conversion_to_patterm): Likewise, renaming to...
2715 (vect_add_conversion_to_pattern): ...this.
2716 (vect_recog_mask_conversion_pattern): Update call to
2717 build_mask_conversion. Pass the vector type to
2718 append_pattern_def_seq here too.
2719 (vect_recog_gather_scatter_pattern): Update call to
2720 vect_add_conversion_to_pattern.
2722 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2724 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
2725 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
2726 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
2727 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
2728 STMT_VINFO_PATTERN_DEF_SEQ to null here.
2729 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
2730 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
2731 append_pattern_def_seq instead of new_pattern_def_seq.
2732 (vect_recog_divmod_pattern): Do both of the above.
2733 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
2736 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2738 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
2739 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
2740 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
2741 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
2742 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
2743 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
2744 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
2745 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
2746 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
2747 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
2748 parameter with a single stmt_vec_info.
2749 (vect_recog_func_ptr): Likewise.
2750 (vect_recog_gather_scatter_pattern): Likewise, folding in...
2751 (vect_try_gather_scatter_pattern): ...this.
2752 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
2753 the stmt_vec_info of the statement to be matched. Don't clear
2754 STMT_VINFO_RELATED_STMT.
2755 (vect_pattern_recog): Update call accordingly.
2757 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2759 PR tree-optimization/85694
2760 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
2761 (uavgM3_ceil): Document new optabs.
2762 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
2763 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
2765 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
2766 (savg_ceil_optab): New optabs.
2767 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
2768 (vect_vect_recog_func_ptrs): Add it.
2769 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
2770 constant directly from the associated lhs.
2772 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2774 * tree-vect-patterns.c (vect_split_statement): New function.
2775 (vect_convert_input): Use it to try to split an existing cast.
2777 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2779 * poly-int.h (print_hex): New function.
2780 * dumpfile.h (dump_dec, dump_hex): Declare.
2781 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
2782 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
2783 min_input_precision, operation_precision and operation_sign.
2784 * tree-vect-patterns.c (vect_get_range_info): New function.
2785 (vect_same_loop_or_bb_p, vect_single_imm_use)
2786 (vect_operation_fits_smaller_type): Delete.
2787 (vect_look_through_possible_promotion): Add an optional
2788 single_use_p parameter.
2789 (vect_recog_over_widening_pattern): Rewrite to use new
2790 stmt_vec_info infomration. Handle one operation at a time.
2791 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
2792 (vect_truncatable_operation_p, vect_set_operation_type)
2793 (vect_set_min_input_precision): New functions.
2794 (vect_determine_min_output_precision_1): Likewise.
2795 (vect_determine_min_output_precision): Likewise.
2796 (vect_determine_precisions_from_range): Likewise.
2797 (vect_determine_precisions_from_users): Likewise.
2798 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
2799 (vect_vect_recog_func_ptrs): Put over_widening first.
2801 (vect_pattern_recog): Call vect_determine_precisions.
2803 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
2805 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
2806 statements that have been replaced by further pattern statements.
2807 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
2809 2018-07-03 Richard Biener <rguenther@suse.de>
2811 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
2812 always set *dt. Dump vectype in vectype overload.
2813 * dumpfile.h (dump_gimple_expr): New function.
2814 (dump_gimple_expr_loc): Likewise.
2815 * dumpfile.c (dump_gimple_expr): New function.
2816 (dump_gimple_expr_loc): Likewise.
2818 2018-07-02 Jeff Law <law@redhat.com>
2820 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
2821 the H8/300, H8/300H and H8/S variants into a single pattern.
2822 (movhi_h8300, movqi_h8300hs): Similarly.
2823 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
2824 (QHI mode iterator): New.
2826 * config/h8300/h8300.md: Remove trailing whitespace.
2828 2018-07-02 Jim Wilson <jimw@sifive.com>
2830 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
2831 instead of emit_insn for interrupt returns.
2832 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
2833 (riscv_sret, riscv_uret): Likewise.
2835 2018-07-02 David Malcolm <dmalcolm@redhat.com>
2837 * pretty-print.c (selftest::test_pp_format): Move save and restore
2838 of quotes to class auto_fix_quotes, and add an instance.
2839 * selftest.c: Include "intl.h".
2840 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
2841 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
2842 * selftest.h (selftest::auto_fix_quotes): New class.
2844 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
2846 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
2847 (aarch64_sve_prepare_conditional_op): Remove.
2848 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
2849 Allow aarch64_simd_reg_or_zero as select operand; remove
2850 the aarch64_sve_prepare_conditional_op call.
2851 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
2852 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
2853 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
2854 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
2855 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
2856 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
2857 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
2858 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
2859 and a splitters to match all of the *_any patterns.
2860 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
2862 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
2863 (SVE_COND_FP_BINARY_REV): Remove.
2864 (sve_int_op_rev, sve_fp_op_rev): New.
2865 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
2866 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
2867 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
2868 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
2869 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
2870 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
2871 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
2872 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
2873 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
2875 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
2876 Remove match_dup 1 from the inner unspec.
2877 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
2879 * config/aarch64/aarch64.md (movprfx): New attr.
2880 (length): Default movprfx to 8.
2881 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
2882 (*madd<SVE_I>, *msub<SVE_I): Likewise.
2883 (*<su>mul<SVE_I>3_highpart): Likewise.
2884 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
2885 (*v<ASHIFT><SVE_I>3): Likewise.
2886 (*<su><MAXMIN><SVE_I>3): Likewise.
2887 (*<su><MAXMIN><SVE_F>3): Likewise.
2888 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
2889 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
2890 (*div<SVE_F>4): Likewise.
2892 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
2894 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
2897 2018-07-02 Richard Biener <rguenther@suse.de>
2899 PR tree-optimization/86363
2900 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
2901 memset argument refers to a non-variable address.
2903 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
2905 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
2906 {PLUS,MINUS}_EXPR code to...
2907 (adjust_symbolic_bound): ...here,
2908 (combine_bound): ...here,
2909 (set_value_range_with_overflow): ...and here.
2911 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
2913 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
2915 (extract_range_from_abs_expr): ...here.
2917 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
2919 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
2920 -fno-omit-frame-pointer when not optimizing.
2922 2018-07-02 Martin Liska <mliska@suse.cz>
2925 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
2926 (propagate_nothrow): Likewise.
2928 2018-07-02 Martin Liska <mliska@suse.cz>
2931 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
2933 2018-07-02 David Malcolm <dmalcolm@redhat.com>
2935 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
2936 function in r262149, changing "loc" param from source_location to
2937 const dump_location_t &.
2938 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
2939 declaration, as above.
2941 2018-07-01 Paul Koning <ni1d@arrl.net>
2943 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
2944 -munit-asm, -mgnu-asm, -mdec-asm.
2945 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
2946 (pdp11_output_labelref): New.
2947 (pdp11_output_def): New.
2948 (pdp11_output_addr_vec_elt): New.
2949 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
2950 %# and %@ format codes.
2951 (pdp11_option_override): New.
2952 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
2953 (pdp11_output_ident): New.
2954 (pdp11_asm_named_section): New.
2955 (pdp11_asm_init_sections): New.
2956 (pdp11_file_start): New.
2957 (pdp11_file_end): New.
2958 (output_ascii): Use .ascii/.asciz for -mdec-asm.
2959 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
2960 %o, like %c but octal.
2961 (pdp11_option_override): New.
2962 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
2964 (DATA_SECTION_ASM_OP): Ditto.
2965 (READONLY_DATA_SECTION_ASM_OP): New.
2966 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
2967 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
2968 (ASM_OUTPUT_LABELREF): Ditto.
2969 (ASM_OUTPUT_DEF): Ditto.
2970 (ASM_OUTPUT_EXTERNAL): New.
2971 (ASM_OUTPUT_SOURCE_FILENAME): New.
2972 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
2973 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
2974 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
2975 %# and %@ format codes.
2976 * config/pdp11/pdp11.opt (mgnu-asm): New.
2977 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
2978 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
2979 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
2981 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
2983 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
2984 dereferencing path[] beyond its length.
2986 (debug_all_paths): New.
2987 (rewire_first_differing_edge): New.
2988 (adjust_paths_after_duplication): New.
2989 (duplicate_thread_path): Call adjust_paths_after_duplication.
2991 (thread_through_all_blocks): Add new argument to
2992 duplicate_thread_path.
2994 2018-06-30 Jim Wilson <jimw@sifive.com>
2996 * config/riscv/predicates.md (p2m1_shift_operand): New.
2997 (high_mask_shift_operand): New.
2998 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
2999 pattern using p2m1_shift_operand.
3000 (lshsi3_zero_extend_3+2): New combiner pattern using
3001 high_mask_shift_operand.
3003 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
3005 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
3007 (vect_recog_rotate_pattern): ...here.
3008 (vect_convert_input): Try to insert casts of invariants in the
3010 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
3011 preheader to be empty.
3013 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
3015 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
3016 vector type. If given, install it in the new statement's
3018 (vect_element_precision): New function.
3019 (vect_unpromoted_value): New struct.
3020 (vect_unpromoted_value::vect_unpromoted_value): New function.
3021 (vect_unpromoted_value::set_op): Likewise.
3022 (vect_look_through_possible_promotion): Likewise.
3023 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
3024 (vect_widened_op_tree, vect_convert_input): Likewise.
3025 (vect_convert_inputs, vect_convert_output): Likewise.
3026 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
3027 to handle the optional cast of the multiplication result and
3028 vect_widened_op_tree to detect the widened multiplication itself.
3029 Do not require the input and output of promotion casts to have
3030 the same sign, but base the signedness of the operation on the
3031 input rather than the result. If the pattern includes two
3032 promotions, check that those promotions have the same sign.
3033 Do not restrict the MULT_EXPR handling to a double-width result;
3034 handle quadruple-width results and wider. Use vect_convert_inputs
3035 to convert the inputs to the common type.
3036 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
3037 to handle the optional cast of the ABS result. Also allow a sign
3038 change or a sign extension between the ABS and MINUS.
3039 Use vect_widened_op_tree to detect the widened subtraction and use
3040 vect_convert_inputs to convert the inputs to the common type.
3041 (vect_handle_widen_op_by_const): Delete.
3042 (vect_recog_widen_op_pattern): New function.
3043 (vect_recog_widen_mult_pattern): Use it.
3044 (vect_recog_widen_shift_pattern): Likewise.
3045 (vect_recog_widen_sum_pattern): Use
3046 vect_look_through_possible_promotion to handle the promoted
3049 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
3051 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
3052 the containing gimple_seq *.
3053 * gimple-iterator.h (gsi_for_stmt): Declare it.
3054 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
3055 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
3056 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
3057 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
3058 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
3060 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
3062 (vect_mark_pattern_stmts): ...here. Handle cases in which the
3063 statement being replaced is part of an existing pattern
3064 definition sequence, inserting the new pattern statements before
3066 (vect_pattern_recog_1): Don't return a bool. If the statement
3067 is already part of a pattern, instead apply pattern matching
3068 to the pattern definition statements. Don't clear the
3069 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
3070 (vect_pattern_recog): Don't break after the first match;
3071 continue processing the pattern definition statements instead.
3072 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
3074 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
3076 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
3077 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
3078 (vect_recog_widen_sum_pattern): Use it.
3080 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
3082 * tree-vect-loop.c (vectorizable_reduction): Assert that the
3083 phi is not a pattern statement and has not been replaced by
3084 a pattern statement.
3085 * tree-vect-patterns.c (type_conversion_p): Don't check
3086 STMT_VINFO_IN_PATTERN_P.
3087 (vect_recog_vector_vector_shift_pattern): Likewise.
3088 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
3089 the pattern statement rather than the original statement; check
3090 directly for a WIDEN_MULT_EXPR here.
3091 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
3092 vect_is_simple_use to return the pattern statement rather
3093 than the original statement; use is_pattern_stmt_p to check
3094 for such a pattern statement.
3095 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
3096 to return the pattern statement rather than the original statement;
3097 don't do the same transformation here.
3098 (vect_is_simple_use): If the defining statement has been replaced
3099 by a pattern statement, return the pattern statement instead.
3100 Remove the corresponding (local) transformation from the vectype
3103 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
3105 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
3106 end and default to null.
3107 * tree-vect-loop.c (vect_create_epilog_for_reduction)
3108 (vectorizable_reduction): Update calls accordingly, dropping the
3109 gimple ** argument if the passed-back statement isn't needed.
3110 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
3111 (vect_recog_rotate_pattern): Likewise.
3112 (vect_recog_mask_conversion_pattern): Likewise.
3113 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
3114 (vect_mask_constant_operand_p): Likewise.
3115 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
3116 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
3117 (get_group_load_store_type, get_load_store_type): Likewise.
3118 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
3119 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
3120 (vectorizable_conversion, vectorizable_assignment): Likewise.
3121 (vectorizable_shift, vectorizable_operation): Likewise.
3122 (vectorizable_store, vect_is_simple_cond): Likewise.
3123 (vectorizable_condition, vectorizable_comparison): Likewise.
3124 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
3125 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
3126 and move it to the end. Cope with null def_stmt_outs.
3128 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
3130 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
3132 2018-06-29 Jeff Law <law@redhat.com>
3134 * config/v850/v850.c (v850_legitimate_address_p): Handle large
3135 displacements for TARGET_V850E2V3 and newer.
3136 (TARGET_LRA_P): Remove. Defaults to LRA now.
3137 * config/v850/v850.md (sign23byte_load): Remove.
3138 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
3139 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
3141 2018-06-29 Martin Liska <mliska@suse.cz>
3144 * coverage.c (coverage_init): Mangle full path name.
3145 * doc/invoke.texi: Document the change.
3146 * gcov-io.c (mangle_path): New.
3147 * gcov-io.h (mangle_path): Likewise.
3148 * gcov.c (mangle_name): Use mangle_path for path mangling.
3150 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3152 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
3153 if starting source register is not even.
3155 2018-06-29 Martin Liska <mliska@suse.cz>
3157 PR tree-optimization/86263
3158 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
3159 Make edge redirection.
3161 2018-06-29 David Malcolm <dmalcolm@redhat.com>
3163 * dumpfile.c (dump_loc): Add indentation based on scope depth.
3164 (dump_scope_depth): New variable.
3165 (get_dump_scope_depth): New function.
3166 (dump_begin_scope): New function.
3167 (dump_end_scope): New function.
3168 * dumpfile.h (get_dump_scope_depth): New declaration.
3169 (dump_begin_scope): New declaration.
3170 (dump_end_scope): New declaration.
3171 (class auto_dump_scope): New class.
3172 (AUTO_DUMP_SCOPE): New macro.
3173 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
3176 2018-06-29 Richard Biener <rguenther@suse.de>
3178 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
3179 compute_all_dependences succeeds.
3180 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
3181 exceed --param loop-max-datarefs-for-datadeps.
3183 2018-06-29 Jakub Jelinek <jakub@redhat.com>
3185 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
3187 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
3190 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
3191 alternative 0 in preferred_for_speed attribute.
3193 2018-06-28 Paul Koning <ni1d@arrl.net>
3195 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
3196 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
3197 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
3198 * config/pdp11/pdp11.md: Correct "length" attribute calculation
3199 for shift insn patterns.
3201 2018-06-28 David Malcolm <dmalcolm@redhat.com>
3203 * cgraph.c (cgraph_node::get_body): Replace assignments to
3204 "dump_file" with calls to set_dump_file.
3205 * dumpfile.c (alt_dump_file): Make static, and group with...
3206 (alt_flags): ...this definition.
3207 (dumps_are_enabled): New variable.
3208 (refresh_dumps_are_enabled): New function.
3209 (set_dump_file): New function.
3210 (set_alt_dump_file): New function.
3211 (gcc::dump_manager::dump_start): Replace assignments to
3212 "dump_file" and "alt_dump_file" with calls to set_dump_file and
3214 (gcc::dump_manager::dump_finish): Likewise.
3215 * dumpfile.h (alt_dump_file): Delete decl.
3216 (dumps_are_enabled): New variable decl.
3217 (set_dump_file): New function decl.
3218 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
3220 * tree-nested.c (lower_nested_functions): Replace assignments to
3221 "dump_file" with calls to set_dump_file.
3223 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
3225 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
3226 goto_locus of each outgoing edge of each basic block.
3228 2018-06-28 Richard Biener <rguenther@suse.de>
3230 * dwarf2out.c (decl_scope_table): Remove.
3231 (push_decl_scope): Likewise.
3232 (pop_decl_scope): Likewise.
3233 (gen_type_die_for_member): Do not call push/pop_decl_scope.
3234 (gen_struct_or_union_type_die): Likewise.
3235 (gen_tagged_type_die): Likewise.
3236 (dwarf2out_init): Do not initialize decl_scope_table.
3237 (dwarf2out_c_finalize): Do not free it.
3239 2018-06-28 Richard Biener <rguenther@suse.de>
3241 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
3242 deciding whether to not re-use a DIE.
3244 2018-06-28 Richard Biener <rguenther@suse.de>
3246 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
3247 DW_AT_abstract_origin attribute.
3249 2018-06-28 Martin Liska <mliska@suse.cz>
3251 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
3252 Use newly introduced constants.
3253 * tree-switch-conversion.h (struct jump_table_cluster):
3254 Define max_ratio_for_size and max_ratio_for_speed.
3256 2018-06-28 Martin Liska <mliska@suse.cz>
3258 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
3259 Add new checking assert to catch invalid state.
3260 (jump_table_cluster::can_be_handled): Handle single case
3262 (jump_table_cluster::is_beneficial): Bail out for such case.
3263 (bit_test_cluster::find_bit_tests):
3264 Add new checking assert to catch invalid state.
3265 (bit_test_cluster::can_be_handled): Handle single case
3267 (bit_test_cluster::is_beneficial): Bail out for such case.
3268 (switch_decision_tree::analyze_switch_statement):
3271 2018-06-28 Martin Liska <mliska@suse.cz>
3273 * common.opt: Introduce -completion option.
3274 * gcc.c (driver_handle_option): Handle it.
3275 (driver::main): Print completions if completion
3277 * opt-suggestions.c (option_proposer::get_completions):
3279 (option_proposer::suggest_completion): Likewise.
3280 (option_proposer::find_param_completions): Likewise.
3281 (verify_autocompletions): Likewise.
3282 (test_completion_valid_options): Likewise.
3283 (test_completion_valid_params): Likewise.
3284 (in_completion_p): Likewise.
3285 (empty_completion_p): Likewise.
3286 (test_completion_partial_match): Likewise.
3287 (test_completion_garbage): Likewise.
3288 (opt_proposer_c_tests): Likewise.
3289 * opt-suggestions.h: Declare new functions.
3290 * opts.c (common_handle_option): Handle OPT__completion_.
3291 * selftest-run-tests.c (selftest::run_tests): Add
3292 opt_proposer_c_tests.
3293 * selftest.c (assert_str_startswith): New.
3294 * selftest.h (assert_str_startswith): Likewise.
3295 (opt_proposer_c_tests): New.
3296 (ASSERT_STR_STARTSWITH): Likewise.
3298 2018-06-28 Martin Liska <mliska@suse.cz>
3300 * Makefile.in: Add opt-suggestions.o.
3301 * gcc-main.c: Include opt-suggestions.h.
3302 * gcc.c (driver::driver): Likewise.
3303 (driver::~driver): Remove m_option_suggestions.
3304 (driver::build_option_suggestions): Moved to option_proposer.
3305 (driver::suggest_option): Likewise.
3306 (driver::handle_unrecognized_options): Use option_proposer.
3307 * gcc.h (class driver): Add new memver m_option_proposer.
3308 * opt-suggestions.c: New file.
3309 * opt-suggestions.h: New file.
3311 2018-06-28 Martin Liska <mliska@suse.cz>
3313 * vec.h (class auto_string_vec): New (moved from auto_argvec).
3314 (auto_string_vec::~auto_string_vec): Likewise.
3316 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
3318 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
3319 prevent_decl_creation_for_types fields up and add reset_location field.
3320 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
3321 statement if id->reset_location is true.
3322 (copy_edges_for_bb): Do not set goto_locus on the new edges if
3323 id->reset_location is true.
3324 (copy_phis_for_bb): Force input_location on the arguments if
3325 id->reset_location is true.
3326 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
3327 is set on the function to be inlined.
3329 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
3331 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
3333 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
3335 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
3336 registers for Pmode.
3337 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
3338 hard registers for the clobbered pseudo.
3340 2018-06-27 Paul Koning <ni1d@arrl.net>
3342 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
3343 mutually exclusive options.
3344 * config/pdp11/constraints.md (h): New constraint.
3345 (O): Update definition to match shift code generation.
3346 (D): New constraint.
3347 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
3349 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
3351 (output_jump): Change arguments.
3352 (pdp11_fixed_cc_regs): New function.
3353 (pdp11_cc_mode): Ditto.
3354 (pdp11_expand_shift): Ditto.
3355 (pdp11_assemble_shift): Ditto.
3356 (pdp11_small_shift): Ditto.
3357 (pdp11_branch_cost): Remove.
3358 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
3360 (pdp11_register_move_cost): Update for CC registers.
3361 (pdp11_rtx_costs): Add case for LSHIFTRT.
3362 (pdp11_output_jump): Add CCNZ mode conditional branches.
3363 (notice_update_cc_on_set): Remove.
3364 (pdp11_cc_mode): New function.
3365 (simple_memory_operand): Correct pre/post decrement case.
3366 (no_side_effect_operand): New function.
3367 (pdp11_regno_reg_class): Add CC_REGS class.
3368 (pdp11_fixed_cc_regs): New function.
3369 (pdp11_small_shift): New function.
3370 (pdp11_expand_shift): New function to expand shift insns.
3371 (pdp11_assemble_shift): New function to output shifts.
3372 (pdp11_branch_cost): Remove.
3373 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
3374 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
3375 (WCHAR_TYPE): Ditto.
3376 (PTRDIFF_TYPE): Ditto.
3377 (ADJUST_INSN_LENGTH): New macro.
3378 (FIXED_REGISTERS): Add CC registers.
3379 (CALL_USED_REGISTERS): Ditto.
3381 (REG_CLASS_NAMES): Ditto.
3382 (REG_CLASS_CONTENTS): Ditto.
3383 (SELECT_CC_MODE): Use new function.
3384 (TARGET_FLAGS_REGNUM): New macro.
3385 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
3386 (cc0_reg_rtx): Remove.
3387 (CC_STATUS_MDEP): Remove.
3388 (CC_STATUS_MDEFP_INIT): Remove.
3389 (CC_IN_FPU): Remove.
3390 (NOTICE_UPDATE_CC): Remove.
3391 (REGISTER_NAMES): Add CC registers.
3392 (BRANCH_COST): Change to constant 1.
3393 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
3395 * config/pdp11/pdp11.opt (mbcopy): Remove.
3396 (mbcopy-builtin): Remove.
3397 (mbranch-cheap): Remove.
3398 (mbranch-expensive): Remove.
3399 * config/pdp11/predicates.md (expand_shift_operand): Update to
3400 match shift code generation.
3401 (ccnz_operator): New predicate.
3402 * doc/invoke.texi (PDP-11 Options): Remove deleted options
3403 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
3404 Remove non-existent option -mabshi, -mno-abshi. Document mutually
3406 * doc/md.texi (PDP-11): Document new D and h constraints. Update
3407 description of O constraint.
3409 2018-06-27 Jeff Law <law@redhat.com>
3410 Austin Law <austinklaw@gmail.com>
3412 * config/v850/v850.md (addsi3_set_flags): New pattern.
3413 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
3414 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
3415 (zero_extendhisi2_v850_set_flags): Likewise.
3416 (zero_extendqisi2_v850_set_flags): Likewise.
3417 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
3418 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
3419 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
3421 * config/v850/v850-protos.h (notice_update_cc): Remove.
3422 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
3423 (v850_print_operand): Handle 'D' and "d".
3424 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
3425 Add handling of arithmetic/logical operations compared against zero.
3426 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
3427 Do not look at v850_compare_op, instead get mode from last argument.
3428 (v850_gen_compare): Remove
3429 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
3430 after reload for prologue insns.
3431 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
3433 (construct_save_jarl): Likewise.
3434 (TARGET_FLAGS_REGNUM): Define.
3435 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
3436 (NOTICE_UPDATE_CC): Remove.
3437 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
3438 than cc0. Conditionalize on reload_completed.
3439 (cmpsi_insn, setfcc_insn): Likewise.
3440 (tst1 splitter): Turn into define_and_split which sets the flags
3442 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
3443 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
3444 (cstoresf4, cstoredf4): Clobber the flags.
3445 (cmpsi, cmpsf, cmpdf): Remove expanders.
3446 (setf_insn): Remove pattern.
3447 (addsi3): Turn into define_and_split which clobbers the flags after
3448 reload and a suitable pattern (addsi3_clobber_flags) for use after
3450 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
3451 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
3452 (ashrsi3, ashrsi3_v850e2): Likewise.
3453 (bins): Clobber the flags.
3454 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
3455 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
3456 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
3457 (call_value_internal_short, call_value_internal_long): Likewise.
3458 (callt_save_interrupt, callt_return_interrupt): Likewise.
3459 (save_interrupt, return_interrupt): Likewise.
3460 (callt_save_all_interrupt, save_all_interrupt): Likewise.
3461 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
3462 (restore_all_interrupt, _restore_all_interrupt): Likewise.
3463 (All FP comparisons): Only allow after reload has completed.
3465 (divh, divhu): Tweak output template.
3466 (branch_z_normal, branch_z_invert): Remove
3467 (branch_nz_normal, branch_nz_invert): Likewise.
3468 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
3470 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
3471 * config/v850/v850.c (notice_update_cc): Remove.
3472 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
3473 (CC_NO_CARRY): Likewise.
3474 (NOTICE_UPDATE_CC): Define to nothing.
3475 * config/v850/v850.md: Remove block comment on cc0 handling
3476 Remove "cc" attribute from all patterns. Remove cc_status handling
3477 from all patterns. Minor formatting fixes.
3479 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3481 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
3482 (cortex-a76.cortex-a55): Likewise.
3483 * config/aarch64/aarch64-tune.md: Regenerate.
3484 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
3485 cortex-a76.cortex-a55.
3487 2018-06-27 Jeff Law <law@redhat.com>
3489 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
3490 (MULTILIB_DIRNAMES): Similarly.
3492 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
3494 * gimple.h (gimple_return_retbnd): Delete.
3495 (gimple_return_set_retbnd): Likewise.
3496 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
3497 gimple_return_set_retbnd.
3498 * gimple-pretty-print.c (dump_gimple_return): Remove call to
3499 gimple_return_retbnd and adjust.
3500 * tree-inline.h (struct copy_body_data): Remove retbnd field.
3501 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
3502 Explicitly return NULL in a couple more cases. Move assertion
3503 on debug statements and remove unreachable code.
3504 (reset_debug_binding): Do not test id->retbnd.
3505 (expand_call_inline): Do not set it.
3507 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
3509 * configure.ac: Add --disable-gcov option.
3510 * configure: Regenerate.
3511 * Makefile.in: Honour @enable_gcov@.
3512 * doc/install.texi: Document --disable-gcov.
3514 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3516 * config/arm/arm-cpus.in (cortex-a76): New entry.
3517 (cortex-a76.cortex-a55): Likewise.
3518 * config/arm/arm-tables.opt: Regenerate.
3519 * config/arm/arm-tune.md: Likewise.
3520 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
3521 * doc/invoke.texi (ARM Options): Document cortex-a76 and
3522 cortex-a76.cortex-a55.
3524 2018-06-27 Tamar Christina <tamar.christina@arm.com>
3527 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
3529 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
3531 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
3533 (EPILOGUE_USES): Likewise.
3535 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
3537 * tree-inline.c (remap_location): New function extracted from...
3538 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
3539 (copy_phis_for_bb): ...here. Call remap_location.
3540 (copy_cfg_body): Adjust call to copy_edges_for_bb.
3542 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
3544 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
3545 unaligned vsx for 16B memset.
3547 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
3550 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
3551 ieee128_float_type_node to long_double_type_node unless
3552 TARGET_LONG_DOUBLE_128 is set.
3554 2018-06-26 David Malcolm <dmalcolm@redhat.com>
3556 * cfgloop.c (get_loop_location): Convert return type from
3557 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
3558 by implicit construction from rtx_insn *, and using
3559 dump_user_location_t::from_function_decl for the fallback case.
3560 * cfgloop.h (get_loop_location): Convert return type from
3561 location_t to dump_user_location_t.
3562 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
3563 dump_printf_loc to pass in a dump_location_t rather than a
3564 location_t, via the gimple stmt.
3565 * coverage.c (get_coverage_counts): Update calls to
3566 dump_printf_loc to pass in dump_location_t rather than a
3568 * doc/optinfo.texi (Dump types): Convert example of
3569 dump_printf_loc from taking "locus" to taking "insn". Update
3570 description of the "_loc" calls to cover dump_location_t.
3571 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
3573 (dump_user_location_t::dump_user_location_t): New constructors,
3574 from gimple *stmt and rtx_insn *.
3575 (dump_user_location_t::from_function_decl): New function.
3576 (dump_loc): Make static.
3577 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
3578 const dump_location_t &.
3579 (dump_generic_expr_loc): Delete.
3580 (dump_printf_loc): Convert param "loc" from location_t to
3581 const dump_location_t &.
3582 (selftest::test_impl_location): New function.
3583 (selftest::dumpfile_c_tests): New function.
3584 * dumpfile.h: Include "profile-count.h".
3585 (class dump_user_location_t): New class.
3586 (struct dump_impl_location_t): New struct.
3587 (class dump_location_t): New class.
3588 (dump_printf_loc): Convert 2nd param from source_location to
3589 const dump_location_t &.
3590 (dump_generic_expr_loc): Delete.
3591 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
3592 const dump_location_t &.
3593 * gimple-fold.c (fold_gimple_assign): Update call to
3594 dump_printf_loc to pass in a dump_location_t rather than a
3595 location_t, via the gimple stmt.
3596 (gimple_fold_call): Likewise.
3597 * gimple-loop-interchange.cc
3598 (loop_cand::analyze_iloop_reduction_var): Update for change to
3599 check_reduction_path.
3600 (tree_loop_interchange::interchange): Update for change to
3602 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
3603 change in return-type of find_loop_location.
3604 (graphite_regenerate_ast_isl): Likewise.
3605 * graphite-optimize-isl.c (optimize_isl): Likewise.
3606 * graphite.c (graphite_transform_loops): Likewise.
3607 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
3608 pass in a dump_location_t rather than a location_t, via the
3610 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3611 * ipa.c (walk_polymorphic_call_targets): Likewise.
3612 * loop-unroll.c (report_unroll): Convert "locus" param from
3613 location_t to dump_location_t.
3614 (decide_unrolling): Update for change to get_loop_location's
3616 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
3617 location_t to dump_user_location_t.
3618 (grid_find_single_omp_among_assignments_1): Updates calls to
3619 dump_printf_loc to pass in a dump_location_t rather than a
3620 location_t, via the gimple stmt.
3621 (grid_parallel_clauses_gridifiable): Convert "tloc" from
3622 location_t to dump_location_t. Updates calls to dump_printf_loc
3623 to pass in a dump_location_t rather than a location_t, via the
3625 (grid_inner_loop_gridifiable_p): Likewise.
3626 (grid_dist_follows_simple_pattern): Likewise.
3627 (grid_gfor_follows_tiling_pattern): Likewise.
3628 (grid_target_follows_gridifiable_pattern): Likewise.
3629 (grid_attempt_target_gridification): Convert initialization
3630 of local "grid" from memset to zero-initialization; FIXME: does
3631 this require C++11? Update call to dump_printf_loc to pass in a
3632 optinfo_location rather than a location_t, via the gimple stmt.
3633 * profile.c (read_profile_edge_counts): Updates call to
3634 dump_printf_loc to pass in a dump_location_t rather than a
3636 (compute_branch_probabilities): Likewise.
3637 * selftest-run-tests.c (selftest::run_tests): Call
3639 * selftest.h (dumpfile_c_tests): New decl.
3640 * tree-loop-distribution.c (pass_loop_distribution::execute):
3641 Update for change in return type of find_loop_location.
3642 * tree-parloops.c (parallelize_loops): Likewise.
3643 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
3644 "locus" from location_t to dump_user_location_t.
3645 (canonicalize_loop_induction_variables): Likewise.
3646 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
3647 for change in return type of find_loop_location.
3648 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
3649 to dump_printf_loc to pass in a dump_location_t rather than a
3650 location_t, via the stmt.
3651 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
3653 * tree-vect-loop-manip.c (find_loop_location): Convert return
3654 type from source_location to dump_user_location_t.
3655 (vect_do_peeling): Update for above change.
3656 (vect_loop_versioning): Update for change in type of
3658 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
3659 from location_t to dump_user_location_t.
3660 (vect_estimate_min_profitable_iters): Update for change in type
3662 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
3663 location_t to dump_location_t.
3664 (vect_slp_bb): Update for change in type of vect_location.
3665 * tree-vectorizer.c (vect_location): Convert from source_location
3666 to dump_user_location_t.
3667 (try_vectorize_loop_1): Update for change in vect_location's type.
3668 (vectorize_loops): Likewise.
3669 (increase_alignment): Likewise.
3670 * tree-vectorizer.h (vect_location): Convert from source_location
3671 to dump_user_location_t.
3672 (find_loop_location): Convert return type from source_location to
3673 dump_user_location_t.
3674 (check_reduction_path): Convert 1st param from location_t to
3675 dump_user_location_t.
3676 * value-prof.c (check_counter): Update call to dump_printf_loc to
3677 pass in a dump_user_location_t rather than a location_t; update
3678 call to error_at for change in type of "locus".
3679 (check_ic_target): Update call to dump_printf_loc to
3680 pass in a dump_user_location_t rather than a location_t, via the
3683 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
3685 * config/s390/s390.h (enum processor_flags): Do not use
3686 default tune parameter when -march was specified.
3688 2018-06-26 Jakub Jelinek <jakub@redhat.com>
3691 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
3692 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
3695 2018-06-26 Richard Biener <rguenther@suse.de>
3697 PR tree-optimization/86287
3699 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
3700 (vect_analyze_loop): Initialize n_stmts.
3702 2018-06-26 Richard Biener <rguenther@suse.de>
3705 * fold-const.c (fold_convertible_p): Pointer extension
3708 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
3711 * dwarf2out.c (loc_list_has_views): Adjust comments.
3712 (dw_loc_list): Split single cross-partition range with
3715 2018-06-25 Jeff Law <law@redhat.com>
3717 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
3718 on -mbig-switch by default.
3720 * config/v850/predicates.md (const_float_1_operand): Fix match_code
3722 (const_float_0_operand): Remove unused predicate.
3723 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
3724 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
3725 (recipsf2): New expander. Original pattern now called
3727 (recipdf2, recipdf2_insn): Similarly.
3728 (rsqrtsf2, rsqrtsf2_insn): Similarly
3729 (rsqrtdf2, rsqrtdf2_insn): Similarly
3731 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
3733 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
3734 Simplify logic for FreeBSD (twice).
3736 2018-06-25 Martin Sebor <msebor@redhat.com>
3738 PR tree-optimization/86204
3739 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
3740 a strnlen result if it's less than the length of the string.
3742 2018-06-25 Martin Sebor <msebor@redhat.com>
3744 PR tree-optimization/85700
3745 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
3746 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
3747 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
3749 2018-06-25 Martin Sebor <msebor@redhat.com>
3751 * doc/extend.texi (Zero-length arrays): Update and clarify.
3753 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
3755 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
3756 added IEEE/IBM long double multilib support on PowerPC little
3757 endian Linux systems.
3758 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
3759 (MULTILIB_DEFAULTS): Likewise.
3760 * config/rs6000/rs6000.c (rs6000_option_override_internal):
3762 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
3763 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
3764 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
3766 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
3769 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
3770 (REORDER_45): Likewise.
3772 2018-06-25 Jeff Law <law@redhat.com>
3774 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
3775 dividend to 32 bits. Adjust length.
3776 (udivmodhi4): Cleanup output template. Fix length.
3778 2018-06-25 Carl Love <cel@us.ibm.com>
3780 * config/rs6000/vsx.md: Change word selector to prefered location.
3782 2018-06-25 Richard Biener <rguenther@suse.de>
3784 PR tree-optimization/86304
3785 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
3786 epilogue-if-converted loops as well.
3788 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
3790 * lto-section-out.c (lto_begin_section): Do not print section
3791 name for noaddr and unnumbered dumps.
3793 2018-06-25 Richard Biener <rguenther@suse.de>
3795 * tree-vectorizer.h (struct vec_info_shared): New structure
3796 with parts split out from struct vec_info and loop_nest from
3797 struct _loop_vec_info.
3798 (struct vec_info): Adjust accordingly.
3799 (struct _loop_vec_info): Likewise.
3800 (LOOP_VINFO_LOOP_NEST): Adjust.
3801 (LOOP_VINFO_DATAREFS): Likewise.
3802 (LOOP_VINFO_DDRS): Likewise.
3803 (struct _bb_vec_info): Likewise.
3804 (BB_VINFO_DATAREFS): Likewise.
3805 (BB_VINFO_DDRS): Likewise.
3806 (struct _stmt_vec_info): Add dr_aux member.
3807 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
3808 (DR_MISALIGNMENT_UNINITIALIZED): New.
3809 (set_dr_misalignment): Adjust.
3810 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
3811 (vect_analyze_loop): Adjust prototype.
3812 (vect_analyze_loop_form): Likewise.
3813 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
3814 Compute dependences lazily.
3815 (vect_record_base_alignments): Use shared datarefs/ddrs.
3816 (vect_verify_datarefs_alignment): Likewise.
3817 (vect_analyze_data_refs_alignment): Likewise.
3818 (vect_analyze_data_ref_accesses): Likewise.
3819 (vect_analyze_data_refs): Likewise.
3820 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
3821 constructor parameter for shared part.
3822 (vect_analyze_loop_form): Pass in shared part and adjust.
3823 (vect_analyze_loop_2): Pass in storage for the number of
3824 stmts. Move loop nest finding to the caller. Compute
3826 (vect_analyze_loop): Pass in shared part.
3827 (vect_transform_loop): Verify shared datarefs are unchanged.
3828 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
3829 constructor parameter for shared part.
3830 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
3831 (vect_slp_bb): Verify shared datarefs are unchanged before
3833 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
3835 (new_stmt_vec_info): Initialize DR_AUX misalignment to
3836 DR_MISALIGNMENT_UNINITIALIZED.
3837 * tree-vectorizer.c (vec_info::vec_info): Add constructor
3838 parameter for shared part.
3839 (vec_info::~vec_info): Adjust.
3840 (vec_info_shared::vec_info_shared): New.
3841 (vec_info_shared::~vec_info_shared): Likewise.
3842 (vec_info_shared::save_datarefs): Likewise.
3843 (vec_info_shared::check_datarefs): Likewise.
3844 (try_vectorize_loop_1): Construct shared part live for analyses
3845 of a single loop for multiple vector sizes.
3846 * tree-parloops.c (gather_scalar_reductions): Adjust.
3848 2018-06-25 Richard Biener <rguenther@suse.de>
3850 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
3851 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
3852 (vect_analyze_data_refs): Remove similar code from here and
3853 simplify accordingly.
3855 2018-06-25 Richard Biener <rguenther@suse.de>
3857 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
3858 for reverse storage order accesses rather than asserting
3859 they cannot happen here.
3861 2018-06-25 Tom de Vries <tdevries@suse.de>
3864 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
3865 Use data16 instead of .byte for insn prefix.
3867 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
3870 * parser.c (make_char_string_pack): Pass this literal chars
3871 through cpp_interpret_string.
3872 (cp_parser_userdef_numeric_literal): Check the result of
3873 make_char_string_pack.
3875 2018-06-24 Maya Rashish <coypu@sdf.org>
3877 * ginclude/stddef.h: Simplify conditions around avoiding
3878 re-definition of __size_t.
3880 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
3882 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
3883 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
3885 2018-06-22 Maya Rashish <coypu@sdf.org>
3887 * doc/invoke.texi (mno-fancy-math-387): Update for changes
3888 made to OpenBSD and NetBSD through the years.
3890 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3892 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
3893 behavior of vec_pack (vector double, vector double) to match
3894 behavior of vec_float2 (vector double, vector double).
3896 2018-06-22 Olivier Hainque <hainque@adacore.com>
3898 * gimplify.c (gimplify_function_tree): Prevent creation
3899 of a trampoline for the address of the current function
3900 passed to entry/exit instrumentation hooks.
3902 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
3905 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
3908 2018-06-22 Martin Liska <mliska@suse.cz>
3910 PR tree-optimization/86263
3911 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
3912 Bail out if is_enabled is false.
3913 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
3915 (jump_table_cluster::is_enabled): New function.
3917 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
3919 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
3920 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
3921 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
3922 (lto_input_ts_binfo_tree_pointers): Likewise.
3923 * tree-streamer-out.c (streamer_write_tree_bitfields,
3924 write_ts_binfo_tree_pointers): Likewise.
3925 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
3927 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
3929 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
3931 2018-06-22 Martin Liska <mliska@suse.cz>
3933 * symbol-summary.h (get): Make it pure and inline move
3934 functionality from ::get function.
3935 (get): Remove and inline into ::get and ::get_create.
3936 (get_create): Move code from ::get function.
3938 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3941 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
3942 -x assembler-with-cpp.
3944 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3946 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
3947 _FILE_OFFSET_BITS=64 for C++.
3949 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
3951 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
3952 conversion insn that shows up when pr85657-3.c is compiled using
3953 IEEE 128-bit long double.
3954 (neg<mode>2_internal): Use the correct mode to check whether the
3955 mode is IBM extended.
3956 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
3957 multiply and divide external functions from being created more
3960 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
3962 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
3964 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
3965 the edge can be forwarded.
3966 (cfg_layout_merge_blocks): Likewise.
3968 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
3970 * except.c (finish_eh_generation): Commit edge insertions only after
3971 the EH edges have been redirected from post-landing to landing pads.
3973 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
3975 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
3976 create_tmp_var_for to create the FRAME decl.
3977 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
3979 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
3981 * tree-inline.c (copy_edges_for_bb): Minor tweak.
3982 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
3983 debug statement when resetting its value.
3984 (expand_call_inline): Copy the locus of the call onto the assignment
3985 of the return value, if any. Use local variable in more cases.
3987 2018-06-21 Martin Liska <mliska@suse.cz>
3989 * ipa-pure-const.c (propagate_nothrow): Use
3990 funct_state_summaries->get.
3991 (dump_malloc_lattice): Likewise.
3992 (propagate_malloc): Likewise.
3994 2018-06-21 Richard Biener <rguenther@suse.de>
3996 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
3997 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
3998 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
4000 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
4001 BLOCK_ABSTRACT_ORIGIN unconditionally.
4003 2018-06-21 David Malcolm <dmalcolm@redhat.com>
4005 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
4007 * ipa-reference.c (ipa_reference_c_finalize): Delete
4008 ipa_ref_opt_sum_summaries and set it to NULL.
4010 2018-06-21 Tom de Vries <tdevries@suse.de>
4012 PR tree-optimization/85859
4013 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
4014 test with comment from bb_no_side_effects_p.
4016 2018-06-21 Richard Biener <rguenther@suse.de>
4018 PR tree-optimization/86232
4019 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
4020 max for constant niter.
4022 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
4024 * config/aarch64/aarch64-simd.md
4025 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
4027 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
4029 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
4030 Make opernads of the unspec commutative.
4032 2018-06-21 Richard Biener <rguenther@suse.de>
4034 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
4035 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
4036 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
4037 (vect_analyze_data_ref_dependence): Re-order checks to deal with
4039 (vect_record_base_alignments): Do not record base alignment
4040 for gathers or scatters.
4041 (vect_compute_data_ref_alignment): Drop return value that is always
4042 true. Bail out early for gathers or scatters.
4043 (vect_enhance_data_refs_alignment): Bail out early for gathers
4045 (vect_find_same_alignment_drs): Likewise.
4046 (vect_analyze_data_refs_alignment): Remove dead code.
4047 (vect_slp_analyze_and_verify_node_alignment): Likewise.
4048 (vect_analyze_data_refs): For possible gathers or scatters do
4049 not create an alternate DR, just check their possible validity
4050 and mark them. Adjust DECL_NONALIASED handling to not rely
4052 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
4053 update inits of gathers or scatters.
4054 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
4055 Also copy gather/scatter flag to pattern vinfo.
4057 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
4059 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
4060 behavior of vec_packsu (vector unsigned long long, vector unsigned
4061 long long) to match behavior of vec_packs with same signature.
4063 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
4064 Thomas Schwinge <thomas@codesourcery.com>
4065 Cesar Philippidis <cesar@codesourcery.com>
4067 * gimplify.c (gimplify_scan_omp_clauses): Add support for
4068 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
4069 (gimplify_adjust_omp_clauses): Likewise.
4070 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
4071 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
4072 (gimplify_omp_target_update): Update handling of acc update and
4074 * omp-low.c (install_var_field): Remove unused parameter
4075 base_pointers_restrict.
4076 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
4077 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
4079 (omp_target_base_pointers_restrict_p): Delete.
4080 (scan_omp_target): Update call to scan_sharing_clauses.
4081 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
4083 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
4084 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
4085 (convert_local_omp_clauses): Likewise.
4086 * tree-pretty-print.c (dump_omp_clause): Likewise.
4087 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
4089 (omp_clause_code_name): Likewise.
4091 2018-06-20 Jakub Jelinek <jakub@redhat.com>
4094 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
4097 PR tree-optimization/86231
4098 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
4099 anti-range don't overwrite *vr0min before using it to compute *vr0max.
4101 2018-06-20 Tom de Vries <tdevries@suse.de>
4103 PR tree-optimization/86097
4104 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
4105 iv type if signedness of iv type is not the same as that of *nit.
4107 2018-06-20 Jakub Jelinek <jakub@redhat.com>
4109 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
4110 EDGE_EH edges, verify they are all EDGE_EH.
4112 2018-06-20 Maya Rashish <coypu@sdf.org>
4114 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
4116 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4118 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
4119 * config/aarch64/aarch64.c (xgene1_tunings): Add
4120 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
4121 (aarch64_mode_valid_for_sched_fusion_p):
4122 Allow 16-byte modes.
4123 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
4124 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
4126 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
4128 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
4129 * config/aarch64/iterators.md (VQ2): New mode iterator.
4131 2018-06-20 Martin Liska <mliska@suse.cz>
4133 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
4134 Change default ratio from 10 to 8.
4136 2018-06-20 Martin Liska <mliska@suse.cz>
4138 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
4140 (bit_test_cluster::find_bit_tests): Likewise.
4141 (switch_decision_tree::analyze_switch_statement): Find clusters.
4142 * tree-switch-conversion.h (struct jump_table_cluster): Document
4145 2018-06-20 Martin Liska <mliska@suse.cz>
4147 * tree-switch-conversion.c (switch_conversion::collect):
4148 Record m_uniq property.
4149 (switch_conversion::expand): Bail out for special conditions.
4150 (group_cluster::~group_cluster): New.
4151 (group_cluster::group_cluster): Likewise.
4152 (group_cluster::dump): Likewise.
4153 (jump_table_cluster::emit): New.
4154 (switch_decision_tree::fix_phi_operands_for_edges): New.
4155 (struct case_node): Remove struct.
4156 (jump_table_cluster::can_be_handled): New.
4157 (case_values_threshold): Moved to header.
4158 (reset_out_edges_aux): Likewise.
4159 (jump_table_cluster::is_beneficial): New.
4160 (bit_test_cluster::can_be_handled): Likewise.
4161 (add_case_node): Remove.
4162 (bit_test_cluster::is_beneficial): New.
4163 (case_bit_test::cmp): New.
4164 (bit_test_cluster::emit): New.
4165 (expand_switch_as_decision_tree_p): Remove.
4166 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
4167 (fix_phi_operands_for_edge): Likewise.
4168 (switch_decision_tree::analyze_switch_statement): New.
4169 (compute_cases_per_edge): Move ...
4170 (switch_decision_tree::compute_cases_per_edge): ... here.
4171 (try_switch_expansion): Likewise.
4172 (switch_decision_tree::try_switch_expansion): Likewise.
4173 (record_phi_operand_mapping): Likewise.
4174 (switch_decision_tree::record_phi_operand_mapping): Likewise.
4175 (emit_case_decision_tree): Likewise.
4176 (switch_decision_tree::emit): Likewise.
4177 (balance_case_nodes): Likewise.
4178 (switch_decision_tree::balance_case_nodes): Likewise.
4179 (dump_case_nodes): Likewise.
4180 (switch_decision_tree::dump_case_nodes): Likewise.
4181 (emit_jump): Likewise.
4182 (switch_decision_tree::emit_jump): Likewise.
4183 (emit_cmp_and_jump_insns): Likewise.
4184 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
4185 (emit_case_nodes): Likewise.
4186 (switch_decision_tree::emit_case_nodes): Likewise.
4187 (conditional_probability): Remove.
4188 * tree-switch-conversion.h (enum cluster_type): New.
4190 (struct cluster): Likewise.
4191 (cluster::cluster): Likewise.
4192 (struct simple_cluster): Likewise.
4193 (simple_cluster::simple_cluster): Likewise.
4194 (struct group_cluster): Likewise.
4195 (struct jump_table_cluster): Likewise.
4196 (struct bit_test_cluster): Likewise.
4197 (struct min_cluster_item): Likewise.
4198 (struct case_tree_node): Likewise.
4199 (case_tree_node::case_tree_node): Likewise.
4200 (jump_table_cluster::case_values_threshold): Likewise.
4201 (struct case_bit_test): Likewise.
4202 (struct switch_decision_tree): Likewise.
4203 (struct switch_conversion): Likewise.
4204 (switch_decision_tree::reset_out_edges_aux): Likewise.
4206 2018-06-20 Martin Liska <mliska@suse.cz>
4208 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
4209 (hoist_edge_and_branch_if_true): Likewise.
4210 (expand_switch_using_bit_tests_p): Likewise.
4211 (struct case_bit_test): Likewise.
4212 (case_bit_test_cmp): Likewise.
4213 (emit_case_bit_tests): Likewise.
4214 (switch_conversion::switch_conversion): New class.
4215 (struct switch_conv_info): Remove old struct.
4216 (collect_switch_conv_info): More to ...
4217 (switch_conversion::collect): ... this.
4218 (check_range): Likewise.
4219 (switch_conversion::check_range): Likewise.
4220 (check_all_empty_except_final): Likewise.
4221 (switch_conversion::check_all_empty_except_final): Likewise.
4222 (check_final_bb): Likewise.
4223 (switch_conversion::check_final_bb): Likewise.
4224 (create_temp_arrays): Likewise.
4225 (switch_conversion::create_temp_arrays): Likewise.
4226 (free_temp_arrays): Likewise.
4227 (gather_default_values): Likewise.
4228 (switch_conversion::gather_default_values): Likewise.
4229 (build_constructors): Likewise.
4230 (switch_conversion::build_constructors): Likewise.
4231 (constructor_contains_same_values_p): Likewise.
4232 (switch_conversion::contains_same_values_p): Likewise.
4233 (array_value_type): Likewise.
4234 (switch_conversion::array_value_type): Likewise.
4235 (build_one_array): Likewise.
4236 (switch_conversion::build_one_array): Likewise.
4237 (build_arrays): Likewise.
4238 (switch_conversion::build_arrays): Likewise.
4239 (gen_def_assigns): Likewise.
4240 (switch_conversion::gen_def_assigns): Likewise.
4241 (prune_bbs): Likewise.
4242 (switch_conversion::prune_bbs): Likewise.
4243 (fix_phi_nodes): Likewise.
4244 (switch_conversion::fix_phi_nodes): Likewise.
4245 (gen_inbound_check): Likewise.
4246 (switch_conversion::gen_inbound_check): Likewise.
4247 (process_switch): Use the newly created class.
4248 (switch_conversion::expand): New.
4249 (switch_conversion::~switch_conversion): New.
4250 * tree-switch-conversion.h: New file.
4252 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4254 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
4255 tree-vect-patterns.c.
4256 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
4257 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
4258 (vect_recog_sad_pattern): Likewise.
4259 (vect_recog_widen_sum_pattern): Likewise.
4260 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
4261 (vect_recog_widen_shift_pattern): Remove the type_in argument.
4262 (vect_recog_rotate_pattern): Likewise.
4263 (vect_recog_mult_pattern): Likewise.
4264 (vect_recog_vector_vector_shift_pattern): Likewise.
4265 (vect_recog_divmod_pattern): Likewise.
4266 (vect_recog_mixed_size_cond_pattern): Likewise.
4267 (vect_recog_bool_pattern): Likewise.
4268 (vect_recog_mask_conversion_pattern): Likewise.
4269 (vect_try_gather_scatter_pattern): Likewise.
4270 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
4271 (vect_recog_over_widening_pattern): Likewise.
4272 (vect_recog_gather_scatter_pattern): Likewise.
4273 (vect_recog_func_ptr): Move from tree-vectorizer.h
4274 (vect_vect_recog_func_ptrs): Move further down the file.
4275 (vect_recog_func): Likewise. Remove the third argument.
4276 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
4277 (vect_pattern_recog_1): Expect the pattern function to do any
4278 necessary target tests. Also expect it to provide a vector type.
4279 Remove the type_in handling.
4281 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4283 * tree-vect-patterns.c (vect_pattern_detected): New function.
4284 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
4285 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
4286 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
4287 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
4288 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
4289 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
4290 (vect_recog_mask_conversion_pattern)
4291 (vect_try_gather_scatter_pattern): Likewise.
4293 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4295 * tree-vect-patterns.c (vect_get_internal_def): New function.
4296 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
4297 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
4298 (search_type_for_mask_1): Use it.
4300 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4302 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
4303 redundant WIDEN_SUM_EXPR handling.
4304 (vect_recog_sad_pattern): Likewise.
4306 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4308 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
4309 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
4310 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
4311 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
4312 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
4314 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4316 * tree-vect-stmts.c (vectorizable_call): Make sure that we
4317 use the stmt_vec_info of the original bb statement for the
4318 new zero assignment, even if the call is part of a pattern.
4320 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4322 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
4323 that the sequence is attached to the original statement rather
4324 than the pattern statement.
4325 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
4326 PATTERN_DEF_SEQ from the original statement rather than
4327 the main pattern statement.
4328 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
4329 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
4330 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
4332 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
4334 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
4335 definition statements before the early exit for statements that aren't
4337 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
4339 (vect_transform_loop): ...here. Process pattern definition
4340 statements without first checking whether the main pattern
4341 statement is live or relevant.
4343 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
4345 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
4346 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
4348 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
4350 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
4351 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
4352 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
4353 (expand_block_compare): Change select_block_compare_mode call.
4354 (expand_strncmp_align_check): Use new functions, fix comment.
4355 (emit_final_str_compare_gpr): New function.
4356 (expand_strn_compare): Refactor and clean up code.
4357 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
4359 2018-06-19 Tony Reix <tony.reix@atos.com>
4360 Damien Bergamini <damien.bergamini@atos.com>
4361 David Edelsohn <dje.gcc@gmail.com>
4363 * collect2.c (static_obj): New variable.
4364 (static_libs): New variable.
4365 (is_in_list): Uncomment declaration.
4366 (main): Track AIX libraries linked statically.
4367 (is_in_list): Uncomment definition.
4368 (scan_prog_file): Don't add AIX shared libraries initializer
4369 to constructor list if linking statically.
4371 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
4373 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
4375 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
4377 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
4379 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
4382 2018-06-19 Martin Liska <mliska@suse.cz>
4384 * config/i386/i386.c (ix86_can_inline_p): Do not use
4385 ipa_fn_summaries::get_create.
4386 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
4388 (devirtualization_time_bonus): Likewise.
4389 (ipcp_propagate_stage): Likewise.
4390 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
4391 (edge_set_predicate): Likewise.
4392 (evaluate_conditions_for_known_args): Likewise.
4393 (evaluate_properties_for_edge): Likewise.
4394 (ipa_call_summary::reset): Tranform to ...
4395 (ipa_call_summary::~ipa_call_summary): ... this.
4396 (ipa_fn_summary::reset): Transform to ...
4397 (ipa_fn_summary::~ipa_fn_summary): ... this.
4398 (ipa_fn_summary_t::remove): Rename to ...
4399 (ipa_fn_summary_t::remove_callees): ... this.
4400 (ipa_fn_summary_t::duplicate): Use placement new
4401 instead of memory copy.
4402 (ipa_call_summary_t::duplicate): Likewise.
4403 (ipa_call_summary_t::remove): Remove.
4404 (dump_ipa_call_summary): Change get_create to get.
4405 (ipa_dump_fn_summary): Dump only when summary exists.
4406 (analyze_function_body): Use symbol_summary::get instead
4408 (compute_fn_summary): Likewise.
4409 (estimate_edge_devirt_benefit): Likewise.
4410 (estimate_edge_size_and_time): Likewise.
4411 (inline_update_callee_summaries): Likewise.
4412 (remap_edge_change_prob): Likewise.
4413 (remap_edge_summaries): Likewise.
4414 (ipa_merge_fn_summary_after_inlining): Likewise.
4415 (write_ipa_call_summary): Likewise.
4416 (ipa_fn_summary_write): Likewise.
4417 (ipa_free_fn_summary): Likewise.
4418 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
4419 (struct ipa_call_summary): Likewise.
4420 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
4422 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
4423 (estimate_size_after_inlining): Likewise.
4424 (estimate_growth): Likewise.
4425 (growth_likely_positive): Likewise.
4426 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
4427 (inline_call): Likewise.
4428 * ipa-inline.c (caller_growth_limits): Likewise.
4429 (can_inline_edge_p): Likewise.
4430 (can_inline_edge_by_limits_p): Likewise.
4431 (compute_uninlined_call_time): Likewise.
4432 (compute_inlined_call_time): Likewise.
4433 (want_inline_small_function_p): Likewise.
4434 (edge_badness): Likewise.
4435 (update_caller_keys): Likewise.
4436 (update_callee_keys): Likewise.
4437 (inline_small_functions): Likewise.
4438 (inline_to_all_callers_1): Likewise.
4439 (dump_overall_stats): Likewise.
4440 (early_inline_small_functions): Likewise.
4441 (early_inliner): Likewise.
4442 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
4443 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
4444 * ipa-pure-const.c (malloc_candidate_p): Likewise.
4445 * ipa-split.c (execute_split_functions): Likewise.
4446 * symbol-summary.h: Likewise.
4447 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
4449 2018-06-19 Richard Biener <rguenther@suse.de>
4451 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
4452 (vectorize_loops): ... here. Fix dbgcnt handling.
4453 (try_vectorize_loop): Wrap try_vectorize_loop_1.
4455 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
4458 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
4459 ieee128 argument takes up only one (vector) register, not two (floating
4462 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
4464 * gimplify.c (gimplify_init_constructor): Really never clear for an
4465 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
4467 2018-06-19 Richard Biener <rguenther@suse.de>
4469 PR tree-optimization/86179
4470 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
4471 after failed recognition.
4473 2018-06-18 Martin Sebor <msebor@redhat.com>
4476 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
4477 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
4478 Handle integer subtraction.
4479 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
4480 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
4482 2018-06-18 David Malcolm <dmalcolm@redhat.com>
4484 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
4485 param from rtx to rtx_insn *.
4486 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
4488 (frv_ifcvt_modify_insn): Likwise.
4489 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
4490 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
4491 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
4492 as_a <rtx_insn *> cast to local "unprotected_region" once
4493 it's been established that it's not NULL or pc_rtx.
4494 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
4495 param "sethi" from rtx to rtx_insn *.
4496 (nds32_group_float_insns): Likewise for param "insn".
4497 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
4499 (vax_output_int_subtract): Likewise.
4500 * config/vax/vax.c (vax_output_int_add): Likewise for param
4502 (vax_output_int_subtract): Likewise.
4503 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
4504 (emit_pattern_after): Likewise for param "after".
4505 (emit_insn_after): Likewise.
4506 (emit_jump_insn_after): Likewise.
4507 (emit_call_insn_after): Likewise.
4508 (emit_debug_insn_after): Likewise.
4509 (emit_pattern_before): Likewise for param "before".
4510 (emit_insn_before): Likewise.
4511 (emit_jump_insn_before): Likewise.
4512 * final.c (get_insn_template): Likewise for param "insn", removing
4514 * output.h (get_insn_template): Likewise for 2nd param.
4515 * rtl.h (emit_insn_before): Likewise.
4516 (emit_jump_insn_before): Likewise.
4517 (emit_debug_insn_before_noloc): Likewise.
4518 (emit_insn_after): Likewise.
4519 (emit_jump_insn_after): Likewise.
4520 (emit_call_insn_after): Likewise.
4521 (emit_debug_insn_after): Likewise.
4522 (set_insn_deleted): Likewise for param.
4524 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
4527 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
4528 floating point modes, so that IFmode is numerically greater than
4529 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
4530 to declare the ordering. This prevents IFmode from being
4531 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
4532 machine. Include rs6000-modes.h to share the fractional values
4533 between genmodes* and the rest of the compiler.
4537 * config/rs6000/rs6000-modes.h: New file.
4538 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
4539 meaning of rs6000_long_double_size so that 126..128 selects an
4540 appropriate 128-bit floating point type.
4541 (rs6000_option_override_internal): Likewise.
4542 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
4543 (TARGET_LONG_DOUBLE_128): Change the meaning of
4544 rs6000_long_double_size so that 126..128 selects an appropriate
4545 128-bit floating point type.
4546 (LONG_DOUBLE_TYPE_SIZE): Update comment.
4547 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
4548 source and destination to match the standard usage.
4549 (truncifkf2): Likewise.
4550 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
4551 ISA 2.07 to use an explicit clobber, instead of passing in a
4553 (copysign<mode>3_soft): Likewise.
4555 2018-06-18 David Malcolm <dmalcolm@redhat.com>
4557 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
4558 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
4559 (vect_slp_analyze_instance_dependence): Likewise.
4560 (vect_enhance_data_refs_alignment): Likewise.
4561 (vect_analyze_data_refs_alignment): Likewise.
4562 (vect_slp_analyze_and_verify_instance_alignment
4563 (vect_analyze_data_ref_accesses): Likewise.
4564 (vect_prune_runtime_alias_test_list): Likewise.
4565 (vect_analyze_data_refs): Likewise.
4566 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
4567 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
4568 (vect_analyze_scalar_cycles_1): Likewise.
4569 (vect_get_loop_niters): Likewise.
4570 (vect_analyze_loop_form_1): Likewise.
4571 (vect_update_vf_for_slp): Likewise.
4572 (vect_analyze_loop_operations): Likewise.
4573 (vect_analyze_loop): Likewise.
4574 (vectorizable_induction): Likewise.
4575 (vect_transform_loop): Likewise.
4576 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
4577 * tree-vect-slp.c (vect_analyze_slp): Likewise.
4578 (vect_make_slp_decision): Likewise.
4579 (vect_detect_hybrid_slp): Likewise.
4580 (vect_slp_analyze_operations): Likewise.
4581 (vect_slp_bb): Likewise.
4582 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
4583 (vectorizable_bswap): Likewise.
4584 (vectorizable_call): Likewise.
4585 (vectorizable_simd_clone_call): Likewise.
4586 (vectorizable_conversion): Likewise.
4587 (vectorizable_assignment): Likewise.
4588 (vectorizable_shift): Likewise.
4589 (vectorizable_operation): Likewise.
4590 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
4592 2018-06-18 Martin Sebor <msebor@redhat.com>
4594 PR tree-optimization/81384
4595 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
4596 * builtins.c (expand_builtin_strnlen): New function.
4597 (expand_builtin): Call it.
4598 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
4599 * builtins.def (BUILT_IN_STRNLEN): New.
4600 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
4601 Warn for bounds in excess of maximum object size.
4602 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
4603 single-value ranges. Handle strnlen.
4604 (handle_builtin_strlen): Handle strnlen.
4605 (strlen_check_and_optimize_stmt): Same.
4606 * doc/extend.texi (Other Builtins): Document strnlen.
4608 2018-06-18 Maya Rashish <coypu@sdf.org>
4610 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
4611 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
4612 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
4614 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
4616 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
4618 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4620 * tree.c (escaped_string::escape): Replace cast to char * by
4621 const_cast<char *> (unescaped).
4623 2018-06-18 Nick Clifton <nickc@redhat.com>
4626 * tree.c (escaped_string): New class. Converts an unescaped
4627 string into its escaped equivalent.
4628 (warn_deprecated_use): Use the new class to convert the
4629 deprecation message, if present.
4630 (test_escaped_strings): New self test.
4631 (test_c_tests): Add test_escaped_strings.
4632 * doc/extend.texi (deprecated): Add a note that the
4633 deprecation message is affected by the -fmessage-length
4634 option, and that control characters will be escaped.
4635 (#pragma GCC error): Document this pragma.
4636 (#pragma GCC warning): Likewise.
4637 * doc/invoke.texi (-fmessage-length): Document this option's
4638 effect on the #warning and #error preprocessor directives and
4639 the deprecated attribute.
4641 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
4643 * tree.c (decl_value_expr_lookup): Revert latest change.
4644 (decl_value_expr_insert): Likewise.
4646 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
4648 * gimplify.c (nonlocal_vlas): Delete.
4649 (nonlocal_vla_vars): Likewise.
4650 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
4652 (gimplify_body): Do not create and destroy nonlocal_vlas.
4653 * tree-nested.c: Include diagnostic.h.
4654 (use_pointer_in_frame): Tweak.
4655 (lookup_field_for_decl): Add assertion and declare the transformation.
4656 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
4657 internal error when the reference is in a wrong context. Do not
4658 create a debug decl by default.
4659 (note_nonlocal_block_vlas): Delete.
4660 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
4661 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
4662 create a debug decl by default.
4663 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
4664 call is in a wrong context.
4665 (fixup_vla_decls): New function.
4666 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
4667 debug variables were created.
4668 * tree.c (decl_value_expr_lookup): Add checking assertion.
4669 (decl_value_expr_insert): Likewise.
4671 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
4674 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
4675 * tree-scalar-evolution.c (interpret_expr): Likewise.
4676 (expression_expensive_p): Likewise.
4677 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
4678 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
4679 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
4680 (ssa_defined_by_minus_one_stmt_p): New.
4682 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
4685 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
4686 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
4687 * dojump.c (do_jump): Likewise.
4688 * expr.c (expand_expr_real_2): Check operand type's sign.
4689 * fold-const.c (const_unop): Handle ABSU_EXPR.
4690 (fold_abs_const): Likewise.
4691 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
4692 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
4693 (strip_sign_op_1): Likesise.
4694 * match.pd: Add new pattern to generate ABSU_EXPR.
4695 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
4696 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
4697 * tree-eh.c (operation_could_trap_helper_p): Likewise.
4698 * tree-inline.c (estimate_operator_cost): Likewise.
4699 * tree-pretty-print.c (dump_generic_node): Likewise.
4700 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
4701 * tree.def (ABSU_EXPR): New.
4703 2018-06-16 Jakub Jelinek <jakub@redhat.com>
4706 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
4707 documented as preserved for backward compatibility only.
4708 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
4710 PR rtl-optimization/86108
4711 * bb-reorder.c (create_forwarder_block): Renamed to ...
4712 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
4713 jump from new landing pad to the second part.
4714 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
4717 2018-06-15 Jakub Jelinek <jakub@redhat.com>
4720 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
4721 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
4722 Only call store_expr for halves if the mode is the same.
4725 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
4726 Fix up comment formatting.
4728 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
4730 * typed-splay-tree.h (typed_splay_tree::remove): New function.
4731 (typed_splay_tree::closure,
4732 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
4733 (typed_splay_tree::typed_splay_tree,
4734 typed_splay_tree::operator =): Declared private.
4735 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
4736 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
4737 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
4738 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
4739 typed_splay_tree::splay_tree_splay,
4740 typed_splay_tree::splay_tree_foreach_helper,
4741 typed_splay_tree::splay_tree_insert,
4742 typed_splay_tree::splay_tree_remove,
4743 typed_splay_tree::splay_tree_lookup,
4744 typed_splay_tree::splay_tree_predecessor,
4745 typed_splay_tree::splay_tree_successor,
4746 typed_splay_tree::splay_tree_min,
4747 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
4748 (typed_splay_tree::root, typed_splay_tree::comp,
4749 typed_splay_tree::delete_key,
4750 typed_splay_tree::delete_value): New data members.
4751 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
4752 typed_splay_tree::remove.
4754 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
4756 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
4757 -mginv and -mno-ginv to the assembler.
4758 * config/mips/mips.opt (-mcrc): New option.
4760 * doc/invoke.text (-mcrc): Document.
4763 2018-06-15 Nick Clifton <nickc@redhat.com>
4766 * tree.c (escaped_string): New class. Converts an unescaped
4767 string into its escaped equivalent.
4768 (warn_deprecated_use): Use the new class to convert the
4769 deprecation message, if present.
4770 (test_escaped_strings): New self test.
4771 (test_c_tests): Add test_escaped_strings.
4772 * doc/extend.texi (deprecated): Add a note that the
4773 deprecation message is affected by the -fmessage-length
4774 option, and that control characters will be escaped.
4775 (#pragma GCC error): Document this pragma.
4776 (#pragma GCC warning): Likewise.
4777 * doc/invoke.texi (-fmessage-length): Document this option's
4778 effect on the #warning and #error preprocessor directives and
4779 the deprecated attribute.
4781 2018-06-15 Richard Biener <rguenther@suse.de>
4783 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
4784 here, also noting vector size used.
4785 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
4786 size used in MSG_OPTIMIZED_LOCATIONS dump.
4787 (pass_slp_vectorize::execute): Adjust.
4789 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
4792 * config/arc/arc.c (arc_return_address_register): Fix
4795 2018-06-15 Richard Biener <rguenther@suse.de>
4798 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
4799 leave useless conversion stripping to force_gimple_operand_gsi.
4800 (gimplify_build2): Likewise.
4801 (gimplify_build1): Likewise.
4803 2018-06-15 Richard Biener <rguenther@suse.de>
4806 * tree-cfg.c (move_stmt_op): unshare invariant addresses
4807 before adjusting their block.
4809 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
4811 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
4812 multilibs for *-*-rtems*.
4813 * config/riscv/t-rtems: New file.
4815 2018-06-14 Jakub Jelinek <jakub@redhat.com>
4818 * match.pd ((A +- CST1) +- CST2): Punt if last resort
4819 unsigned_type_for returns NULL.
4822 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
4823 subregs of multi-word pseudos unless the float mode has word size.
4825 2018-06-14 Richard Biener <rguenther@suse.de>
4828 * tree-vect-generic.c (build_word_mode_vector_type): Remove
4829 duplicate and harmful type_hash_canon.
4830 * tree.c (type_hash_canon): Assert we didn't find ourselves.
4832 2018-06-14 Richard Biener <rguenther@suse.de>
4835 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
4838 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
4840 * config/rtems.h (STDINT_LONG32): Define.
4842 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
4843 Prachi Godbole <prachi.godbole@imgtec.com>
4845 * config/mips/mips-cpus.def: Define P6600.
4846 * config/mips/mips-tables.opt: Regenerate.
4847 * config/mips/mips.c (mips_ucbranch_type): New enum.
4848 (mips_rtx_cost_data): Add support for P6600.
4849 (mips_issue_rate): Likewise.
4850 (mips_multipass_dfa_lookahead): Likewise.
4851 (mips_avoid_hazard): Likewise.
4852 (mips_reorg_process_insns): Likewise.
4853 (mips_classify_branch_p6600): New function.
4854 * config/mips/mips.h (TUNE_P6600): New define.
4855 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
4856 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
4857 * config/mips/mips.md: Include p6600.md.
4858 (processor): Add p6600.
4859 * config/mips/p6600.md: New file.
4860 * doc/invoke.texi: Add p6600 to supported architectures.
4862 2018-06-13 Martin Sebor <msebor@redhat.com>
4864 PR tree-optimization/86114
4865 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
4867 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
4869 2018-06-13 Richard Biener <rguenther@suse.de>
4871 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
4872 Properly set vector type of the intermediate stmt.
4873 * tree-vect-stmts.c (vectorizable_operation): The destination
4874 var always has vectype_out type.
4876 2018-06-13 Jeff Law <law@redhat.com>
4878 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
4879 integer 0 for argument to print_rtl_with_bb.
4880 (rl78_reorg): Likewise.
4882 2018-06-13 David Malcolm <dmalcolm@redhat.com>
4884 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
4885 from rtx to rtx_insn *.
4886 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
4888 (add_sched_insns_for_speculation): Likewise for local "target",
4889 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
4890 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
4891 from rtx_insn ** to rtx_code_label **.
4892 (reorg_emit_nops): Likewise.
4893 (c6x_reorg): Likewise for local "call_labels".
4894 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
4896 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
4897 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
4898 the loops over LABEL_REFS.
4899 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
4901 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
4902 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
4903 (split_branches): Strengthen local "olabel" from rtx to
4904 rtx_insn *, adding a safe_as_a cast.
4905 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
4907 (add_insn_after): Likewise for first two params.
4908 (add_insn_before): Likewise.
4909 (remove_insn): Likewise for param.
4910 (emit_pattern_before_noloc): Likewise for second and third params.
4911 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
4912 (emit_call_insn_before_noloc): Likewise.
4913 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
4915 (emit_barrier_before): Likewise.
4916 (emit_label_before): Strengthen "label" param from "rtx" to
4917 "rtx_code_label *". Strengthen "before" param from "rtx" to
4919 (emit_insn_after_1): Strengthen "after" param from "rtx" to
4921 (emit_pattern_after_noloc): Likewise.
4922 (emit_insn_after_noloc): Likewise.
4923 (emit_jump_insn_after_noloc): Likewise.
4924 (emit_call_insn_after_noloc): Likewise.
4925 (emit_debug_insn_after_noloc): Likewise.
4926 (emit_barrier_after): Likewise.
4927 (emit_label_after): Likewise for both params.
4928 (emit_pattern_after_setloc): Likewise for "after" param. Convert
4929 "loc" param from "int" to "location_t".
4930 (emit_insn_after_setloc): Likewise.
4931 (emit_jump_insn_after_setloc): Likewise.
4932 (emit_call_insn_after_setloc): Likewise.
4933 (emit_debug_insn_after_setloc): Likewise.
4934 (emit_pattern_before_setloc): Likewise for "before" param. Convert
4935 "loc" param from "int" to "location_t".
4936 (emit_pattern_before): Convert NULL_RTX to NULL.
4937 (emit_insn_before_setloc): Convert "loc" param from "int" to
4939 (emit_jump_insn_before_setloc): Likewise.
4940 (emit_call_insn_before_setloc): Likewise.
4941 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
4942 rtx_insn *. Convert "loc" param from "int" to "location_t".
4943 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
4944 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
4945 Convert 3rd param from "int" to "location_t".
4946 (emit_barrier_before, emit_barrier_after, next_real_insn):
4947 Strengthen param from rtx to rtx_insn *.
4948 (emit_label_before): Strengthen 1st param from "rtx" to
4949 "rtx_code_label *". Strengthen 2nd param from "rtx" to
4951 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
4952 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
4953 Strengthen 2nd param from "rtx" to "rtx_insn *".
4954 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
4955 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
4956 Likewise. Convert 3rd param from "int" to "location_t".
4957 (emit_label_after): Strengthen 1st param from "rtx" to
4959 (next_real_insn, remove_insn): Strengthen param from "rtx" to
4961 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
4962 from "rtx" to "rtx_insn *".
4964 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
4966 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
4967 bodies streamed in with -Q.
4968 * dumpfile.c (dump_files): Add lto-stream-out dump file.
4969 * dumpfile.h (tree_dump_index): Add lto_stream_out.
4970 * gimple-streamer-out.c: Include gimple-pretty-print.h
4971 (output_bb): Dump stmts streamed.
4972 * lto-section-out.c: Include print-tree.h
4973 (lto_begin_section): Dump sections created.
4974 (lto_output_decl_index): Dump decl encoded.
4975 * lto-streamer-out.c: Include print-tree.h
4976 (create_output_block): Dump output block created.
4977 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
4978 (output_function): Dump function output.
4979 (output_constructor): Dump constructor streamed.
4980 (write_global_stream): Output indexes encoded.
4981 (produce_asm_for_decls): Dump streams encoded.
4982 * lto-streamer.c (streamer_dump_file): New global var.
4983 * lto-streamer.h (streamer_dump_file): Declare.
4984 * passes.c (ipa_write_summaries): Initialize streamer dump.
4985 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
4988 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
4991 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
4992 offsets for register save directives. Emit a second batch of save
4993 directives, if need be, when the function accesses prior frames.
4995 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
4997 * config/arc/fpu.md (fmasf4): Force operand to register.
4998 (fnmasf4): Likewise.
5000 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
5002 * config/arc/arc-protos.h (arc_pad_return): Remove.
5003 * config/arc/arc.c (machine_function): Remove force_short_suffix
5005 (arc_print_operand): Adjust printing of '&'.
5006 (arc_verify_short): Remove conditional printing of short suffix.
5007 (arc_final_prescan_insn): Remove reference to size_reason.
5008 (pad_return): New function.
5009 (arc_reorg): Call pad_return.
5010 (arc_pad_return): Remove.
5011 (arc_init_machine_status): Remove reference to force_short_suffix.
5012 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
5013 (attr length): When attribute iscompact is true force to 2
5014 regardless; in the case of maybe check if we want to force the
5015 instruction to have 4 bytes length.
5016 (nopv): Change it to generate 4 byte long nop as well.
5017 (blockage): New pattern.
5018 (simple_return): Remove call to arc_pad_return.
5019 (p_return_i): Likewise.
5021 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
5023 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
5025 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
5027 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
5030 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
5032 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
5033 for ARC700 and ARCv2.
5035 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
5038 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
5039 operands[2] instead of operands[1].
5042 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
5044 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
5045 case, check whether the outer register overlaps an unallocatable
5046 register, not just whether it fits the required class.
5048 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
5050 * poly-int.h (can_div_trunc_p): Add new overload in which all values
5052 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
5053 (memrefs_conflict_p): Likewise.
5054 (init_alias_analysis): Likewise.
5055 * cfgexpand.c (expand_debug_expr): Likewise.
5056 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
5057 * cse.c (fold_rtx): Likewise.
5058 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
5059 * expr.c (emit_block_move_hints): Likewise.
5060 (clear_storage_hints, push_block, emit_push_insn): Likewise.
5061 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
5062 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
5063 (emit_group_store): Likewise.
5064 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
5065 to read the PRE/POST_MODIFY increment.
5066 * calls.c (store_one_arg): Use strip_offset.
5067 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
5069 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
5071 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
5072 (simplify_binary_operation_1): Extend CONST_INT handling to
5074 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
5075 than a HOST_WIDE_INT.
5076 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
5078 (adjust_mems, add_stores): Update accodingly.
5079 (vt_canonicalize_addr): Track polynomial offsets.
5080 (emit_note_insn_var_location): Likewise.
5081 (vt_add_function_parameter): Likewise.
5082 (vt_initialize): Likewise.
5084 2018-06-12 Jeff Law <law@redhat.com>
5086 * config.gcc (alpha*-*-freebsd*): Remove.
5087 * config/alpha/freebsd.h: Remove.
5089 2018-06-12 David Malcolm <dmalcolm@redhat.com>
5092 * spellcheck-tree.c (levenshtein_distance): Rename to...
5093 (get_edit_distance): ...this, and update for underlying renaming.
5094 * spellcheck-tree.h (levenshtein_distance): Rename to...
5095 (get_edit_distance): ...this.
5096 * spellcheck.c (levenshtein_distance): Rename to...
5097 (get_edit_distance): ...this. Convert from Levenshtein distance
5098 to Damerau-Levenshtein distance by supporting transpositions of
5099 adjacent characters. Rename "v1" to "v_next" and "v0" to
5101 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
5102 (selftest::test_edit_distance_unit_test_oneway): ...this, and
5103 update for underlying renaming.
5104 (selftest::levenshtein_distance_unit_test): Rename to...
5105 (selftest::test_get_edit_distance_unit): ...this, and update for
5106 underlying renaming.
5107 (selftest::test_find_closest_string): Add example from PR 69968
5108 where transposition helps
5109 (selftest::test_metric_conditions): Update for renaming.
5110 (selftest::test_metric_conditions): Likewise.
5111 (selftest::spellcheck_c_tests): Likewise.
5112 * spellcheck.h (levenshtein_distance): Rename both overloads to...
5113 (get_edit_distance): ...this.
5114 (best_match::consider): Update for renaming.
5116 2018-06-12 Martin Sebor <msebor@redhat.com>
5118 PR tree-optimization/85259
5119 * builtins.c (compute_objsize): Handle constant offsets.
5120 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
5121 true iff a warning has been issued.
5122 * gimple.h (gimple_nonartificial_location): New function.
5123 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
5124 gimple_nonartificial_location and handle -Wno-system-headers.
5125 (handle_builtin_stxncpy): Same.
5127 2018-06-12 Martin Sebor <msebor@redhat.com>
5130 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
5132 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
5134 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
5135 BUILTIN_VEC_XST entries for pointer to double and long long.
5137 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
5140 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
5142 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
5145 2018-06-12 Martin Liska <mliska@suse.cz>
5147 * doc/options.texi: Document IntegerRange.
5149 2018-06-12 Martin Liska <mliska@suse.cz>
5151 * config/i386/i386.opt: Make MPX-related options as Deprecated.
5152 * opt-functions.awk: Handle Deprecated flag.
5153 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
5155 (read_cmdline_option): Report warning for a deprecated option.
5156 * opts.h (struct cl_option): Add new field cl_deprecated.
5157 (CL_ERR_DEPRECATED): New.
5159 2018-06-12 Martin Liska <mliska@suse.cz>
5161 * doc/options.texi: Document Deprecated option flag.
5163 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
5165 * config/arc/arc-arch.h (arc_extras): New enum.
5166 (arc_cpu_t):Add field extra.
5167 (arc_cpu_types): Consider the extras.
5168 * config/arc/arc-cpus.def: Add extras info.
5169 * config/arc/arc-opts.h (processor_type): Consider extra field.
5170 * config/arc/arc.c (arc_override_options): Handle extra field.
5172 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
5174 * config/arc/arc-arch.h: Update ARC_OPTX macro.
5175 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
5177 * config/arc/arc.c (arc_init): Update pic warning.
5178 (irq_range): Update irq range parsing warnings.
5179 (arc_override_options): Update various warning messages.
5180 (arc_handle_aux_attribute): Likewise.
5182 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
5184 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
5186 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5188 * doc/sourcebuild.texi: Document usage of line number 0 in verify
5189 compiler messages directives.
5191 2018-06-12 Matthew Fortune <mfortune@gmail.com>
5193 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
5194 * config/mips/mips-tables.opt: Regenerate.
5195 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
5197 * doc/invoke.texi: Document -march=i6500.
5199 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
5201 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
5202 (i6400_gpmul): Add cpu_unit.
5203 (i6400_gpdiv): Likewise.
5204 (i6400_msa_add_d): Update reservations.
5205 (i6400_msa_int_add) Likewise.
5206 (i6400_msa_short_logic3) Likewise.
5207 (i6400_msa_short_logic2) Likewise.
5208 (i6400_msa_short_logic) Likewise.
5209 (i6400_msa_move) Likewise.
5210 (i6400_msa_cmp) Likewise.
5211 (i6400_msa_short_float2) Likewise.
5212 (i6400_msa_div_d) Likewise.
5213 (i6400_msa_long_logic1) Likewise.
5214 (i6400_msa_long_logic2) Likewise.
5215 (i6400_msa_mult) Likewise.
5216 (i6400_msa_long_float2) Likewise.
5217 (i6400_msa_long_float4) Likewise.
5218 (i6400_msa_long_float5) Likewise.
5219 (i6400_msa_long_float8) Likewise.
5220 (i6400_fpu_fadd): Include frint type.
5221 (i6400_fpu_store): New define_insn_reservation.
5222 (i6400_fpu_load): Likewise.
5223 (i6400_fpu_move): Likewise.
5224 (i6400_fpu_fcmp): Likewise.
5225 (i6400_fpu_fmadd): Likewise.
5226 (i6400_int_mult): Include imul3nc type and update reservation.
5227 (i6400_int_div): Include idiv3 type and update reservation.
5228 (i6400_int_load): Update to check type not move_type.
5229 (i6400_int_store): Likewise.
5230 (i6400_int_prefetch): Set zero latency.
5232 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
5234 * gcc.c: Document new %@{...} sequence.
5235 (LINK_COMMAND_SPEC): Use it for the -L switches.
5236 (cpp_unique_options): Use it for the -I switches.
5237 (at_file_argbuf): New global variable.
5238 (in_at_file): Likewise.
5239 (alloc_args): Create at_file_argbuf.
5240 (clear_args): Truncate at_file_argbuf.
5241 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
5242 (open_at_file): New function.
5243 (close_at_file): Likewise.
5244 (create_at_file): Delete.
5245 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
5248 (validate_switches_from_spec): Deal with %@{...} sequence.
5249 (validate_switches): Likewise.
5250 (driver::finalize): Call clear_args.
5252 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
5254 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
5256 2018-06-11 Martin Sebor <msebor@redhat.com>
5258 * doc/invoke.texi (-Wall): List -Wc++17-compat.
5259 (Wno-class-memaccess): Add @opindex.
5260 (Wno-templates, Wno-multiple-inheritance): Same.
5261 (Wno-virtual-inheritance, Wno-namespaces): Same.
5262 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
5263 (Wno-format-overflow, Wno-format-truncation): Same.
5264 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
5265 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
5266 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
5267 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
5268 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
5269 (Wno-misspelled-isr): Same.
5271 2018-06-11 Martin Sebor <msebor@redhat.com>
5273 * PR tree-optimization/86083
5274 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
5276 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
5278 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
5280 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
5283 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
5284 on the correct operand.
5285 (*movdi_internal64): Ditto.
5287 2018-06-11 Martin Liska <mliska@suse.cz>
5289 PR tree-optimization/86089
5290 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
5292 2018-06-11 Julia Koval <julia.koval@intel.com>
5294 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
5295 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
5296 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
5298 2018-06-11 Olivier Hainque <hainque@adacore.com>
5300 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
5301 for Ada with strict dwarf2.
5303 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
5306 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
5309 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
5311 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
5313 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
5315 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
5318 2018-06-08 Martin Liska <mliska@suse.cz>
5320 * tree-cfg.h (debug_function): Fix argument type to match
5323 2018-06-08 Martin Liska <mliska@suse.cz>
5325 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
5326 Remove usage of MPX-related (and removed) fields.
5327 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
5329 2018-06-08 David Malcolm <dmalcolm@redhat.com>
5331 * cfg.c (debug): Use TDF_NONE rather than 0.
5332 * cfghooks.c (debug): Likewise.
5333 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
5334 (struct dump_option_value_info): Convert to...
5335 (struct kv_pair): ...this template type.
5336 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
5338 (optinfo_verbosity_options): Likewise.
5339 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
5341 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
5342 than int for "optgroup_flags" param.
5343 (dump_generic_expr_loc): Use dump_flags_t rather than int for
5345 (dump_dec): Likewise.
5346 (dump_finish): Use TDF_NONE rather than 0.
5347 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
5348 rather than int for "optgroup_flags" param. Use TDF_NONE rather
5349 than 0. Update for change to option_ptr.
5350 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
5351 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
5352 0. Update for changes to optinfo_verbosity_options and
5354 (opt_info_switch_p): Convert optgroup_flags from int to
5356 (dump_basic_block): Use dump_flags_t rather than int
5357 for "dump_kind" param.
5358 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
5359 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
5360 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
5361 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
5362 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
5363 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
5364 TDF_NONE): Convert from macros to...
5365 (enum dump_flag): ...this new enum.
5366 (dump_flags_t): Update to use enum.
5367 (operator|, operator&, operator~, operator|=, operator&=):
5368 Implement for dump_flags_t.
5369 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
5370 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
5371 Convert from macros to...
5372 (enum optgroup_flag): ...this new enum.
5373 (optgroup_flags_t): New typedef.
5374 (operator|, operator|=): Implement for optgroup_flags_t.
5375 (struct dump_file_info): Convert field "alt_flags" to
5376 dump_flags_t. Convert field "optgroup_flags" to
5378 (dump_basic_block): Use dump_flags_t rather than int for param.
5379 (dump_generic_expr_loc): Likewise.
5380 (dump_dec): Likewise.
5381 (dump_register): Convert param "optgroup_flags" to
5383 (opt_info_enable_passes): Likewise.
5384 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
5386 * gimple-pretty-print.c (debug): Likewise.
5387 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
5388 (merged_store_group::apply_stores): Likewise.
5389 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
5390 * gimple.c (verify_gimple_pp): Likewise.
5391 * graphite-poly.c (print_pbb_body): Likewise.
5392 * passes.c (pass_manager::register_one_dump_file): Convert
5393 local "optgroup_flags" to optgroup_flags_t.
5394 * print-tree.c (print_node): Use TDF_NONE rather than 0.
5396 (debug_body): Likewise.
5397 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
5398 to optgroup_flags_t.
5399 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
5401 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
5402 (convert_mult_to_fma): Likewise.
5403 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
5404 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
5405 * tree-vect-data-refs.c (dump_lower_bound): Convert param
5406 "dump_kind" to dump_flags_t.
5408 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
5410 * config/rs6000/rs6000.c (min, max): Delete.
5412 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
5414 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
5415 -mabi=spe and -mabi=no-spe.
5417 2018-06-08 Martin Liska <mliska@suse.cz>
5419 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
5420 where we expect an existing summary.
5422 2018-06-08 Martin Liska <mliska@suse.cz>
5424 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
5425 * ipa-inline.h (estimate_edge_growth): Likewise.
5427 2018-06-08 Martin Liska <mliska@suse.cz>
5429 * cgraph.c (function_version_hasher::hash): Use
5430 cgraph_node::get_uid ().
5431 (function_version_hasher::equal):
5432 * cgraph.h (cgraph_node::get_uid): New method.
5433 * ipa-inline.c (update_caller_keys): Use
5434 cgraph_node::get_uid ().
5435 (update_callee_keys): Likewise.
5436 * ipa-utils.c (searchc): Likewise.
5437 (ipa_reduced_postorder): Likewise.
5438 * lto-cgraph.c (input_node): Likewise.
5439 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
5440 * symbol-summary.h (symtab_insertion): Likewise.
5441 (symtab_removal): Likewise.
5442 (symtab_duplication): Likewise.
5443 * tree-pretty-print.c (dump_function_header): Likewise.
5444 * tree-sra.c (convert_callers_for_node): Likewise.
5446 2018-06-08 Martin Liska <mliska@suse.cz>
5448 * cgraph.c (symbol_table::create_edge): Always assign a new
5450 (symbol_table::free_edge): Do not recycle numbers.
5451 * cgraph.h (cgraph_edge::get): New method.
5452 * symbol-summary.h (symtab_removal): Use it.
5453 (symtab_duplication): Likewise.
5454 (call_summary::hashable_uid): Remove.
5456 2018-06-08 Martin Liska <mliska@suse.cz>
5458 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
5459 (initialize_growth_caches): Remove.
5460 (free_growth_caches): Likewise.
5461 (do_estimate_edge_time): Use edge_growth_cache.
5462 (do_estimate_edge_size): Likewise.
5463 (do_estimate_edge_hints): Likewise.
5464 * ipa-inline.c (reset_edge_caches): Likewise.
5465 (recursive_inlining): Likewise.
5466 (inline_small_functions): Likewise.
5467 * ipa-inline.h (initialize_growth_caches): Remove.
5468 (estimate_edge_size): Likewise.
5469 (estimate_edge_time): Likewise.
5470 (estimate_edge_hints): Likewise.
5471 (reset_edge_growth_cache): Likewise.
5472 * symbol-summary.h (call_summary::remove): New method.
5474 2018-06-08 Martin Liska <mliska@suse.cz>
5476 * ipa-cp.c (class edge_clone_summary): New summary.
5477 (grow_edge_clone_vectors): Remove.
5478 (ipcp_edge_duplication_hook): Remove.
5479 (class edge_clone_summary_t): New call_summary class.
5480 (ipcp_edge_removal_hook): Remove.
5481 (edge_clone_summary_t::duplicate): New function.
5482 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
5483 (create_specialized_node): Likewise.
5484 (ipcp_driver): Initialize edge_clone_summaries and do not
5487 2018-06-08 Martin Liska <mliska@suse.cz>
5489 * symbol-summary.h (get): New function.
5490 (call_summary::m_initialize_when_cloning): New class member.
5492 2018-06-08 Martin Liska <mliska@suse.cz>
5494 * cgraph.c (cgraph_node::remove): Do not recycle uid.
5495 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
5496 (symbol_table::allocate_cgraph_symbol): Do not set uid.
5497 * passes.c (uid_hash_t): Record removed_nodes by their uids.
5498 (remove_cgraph_node_from_order): Use the removed_nodes set.
5499 (do_per_function_toporder): Likwise.
5500 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
5501 instead of summary_uid.
5502 (symtab_removal): Likewise.
5503 (symtab_duplication): Likewise.
5505 2018-06-08 Martin Liska <mliska@suse.cz>
5507 * ipa-cp.c (ipcp_store_bits_results): Use
5508 ipcp_transformation_sum.
5509 (ipcp_store_vr_results): Likewise.
5510 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
5512 (ipcp_transformation_initialize): ... this.
5513 (ipa_set_node_agg_value_chain):
5514 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
5515 (write_ipcp_transformation_info): Likewise.
5516 (read_ipcp_transformation_info): Likewise.
5517 (ipcp_update_bits): Likewise.
5518 (ipcp_update_vr): Likewise.
5519 (ipcp_transform_function): Likewise.
5520 * ipa-prop.h: Rename ipcp_transformation_summary to
5521 ipcp_transformation.
5522 (class ipcp_transformation_t): New function summary.
5523 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
5524 (ipa_get_agg_replacements_for_node): Likewise.
5526 2018-06-08 Martin Liska <mliska@suse.cz>
5528 * ipa-pure-const.c (struct funct_state_d): Do it class instead
5530 (class funct_state_summary_t): New function_summary class.
5531 (has_function_state): Remove.
5532 (get_function_state): Likewise.
5533 (set_function_state): Likewise.
5534 (add_new_function): Likewise.
5535 (funct_state_summary_t::insert): New function.
5536 (duplicate_node_data): Remove.
5537 (remove_node_data): Remove.
5538 (funct_state_summary_t::duplicate): New function.
5539 (register_hooks): Create new funct_state_summaries.
5540 (pure_const_generate_summary): Use it.
5541 (pure_const_write_summary): Likewise.
5542 (pure_const_read_summary): Likewise.
5543 (propagate_pure_const): Likewise.
5544 (propagate_nothrow): Likewise.
5545 (dump_malloc_lattice): Likewise.
5546 (propagate_malloc): Likewise.
5547 (execute): Do not register hooks, just remove summary
5549 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
5552 2018-06-08 Martin Liska <mliska@suse.cz>
5554 * ipa-reference.c (remove_node_data): Remove.
5555 (duplicate_node_data): Likewise.
5556 (class ipa_ref_var_info_summary_t): New class.
5557 (class ipa_ref_opt_summary_t): Likewise.
5558 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
5559 (get_reference_optimization_summary): Use
5560 ipa_ref_opt_sum_summaries.
5561 (set_reference_vars_info): Remove.
5562 (set_reference_optimization_summary): Likewise.
5563 (ipa_init): Create summaries.
5564 (init_function_info): Use function summary.
5565 (ipa_ref_opt_summary_t::duplicate): New function.
5566 (ipa_ref_opt_summary_t::remove): New function.
5567 (get_read_write_all_from_node): Fix GNU coding style.
5568 (propagate): Use function summary.
5569 (write_node_summary_p): Fix GNU coding style.
5570 (stream_out_bitmap): Likewise.
5571 (ipa_reference_read_optimization_summary): Use function summary.
5572 (ipa_reference_c_finalize): Do not release hooks.
5574 2018-06-08 Martin Liska <mliska@suse.cz>
5576 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
5577 (analyze_function_body): Extract multiple calls of get_create.
5578 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
5579 * ipa-inline.c (recursive_inlining): Use ::get method.
5580 * ipa-inline.h (estimate_edge_growth): Likewise.
5582 2018-06-08 Martin Liska <mliska@suse.cz>
5584 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
5586 (hsa_function_summary::hsa_function_summary): Use the new enum
5588 (hsa_gpu_implementation_p): Use hsa_summaries::get.
5589 * hsa-gen.c (hsa_get_host_function): Likewise.
5590 (get_brig_function_name): Likewise.
5591 * ipa-hsa.c (process_hsa_functions): Likewise.
5592 (ipa_hsa_write_summary): Likewise.
5593 * symbol-summary.h (symtab_duplication): Use ::get function/
5594 (get): New function.
5596 2018-06-08 Martin Liska <mliska@suse.cz>
5598 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
5600 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
5601 (hsa_register_kernel): Likewise.
5602 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
5603 * hsa-gen.c (hsa_get_host_function): Likewise.
5604 (get_brig_function_name): Likewise.
5605 (generate_hsa): Likewise.
5606 (pass_gen_hsail::execute): Likewise.
5607 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
5608 (devirtualization_time_bonus): Likewise.
5609 (ipcp_propagate_stage): Likewise.
5610 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
5611 (edge_set_predicate): Likewise.
5612 (evaluate_conditions_for_known_args): Likewise.
5613 (evaluate_properties_for_edge): Likewise.
5614 (ipa_fn_summary::reset): Likewise.
5615 (ipa_fn_summary_t::duplicate): Likewise.
5616 (dump_ipa_call_summary): Likewise.
5617 (ipa_dump_fn_summary): Likewise.
5618 (analyze_function_body): Likewise.
5619 (compute_fn_summary): Likewise.
5620 (estimate_edge_devirt_benefit): Likewise.
5621 (estimate_edge_size_and_time): Likewise.
5622 (estimate_calls_size_and_time): Likewise.
5623 (estimate_node_size_and_time): Likewise.
5624 (inline_update_callee_summaries): Likewise.
5625 (remap_edge_change_prob): Likewise.
5626 (remap_edge_summaries): Likewise.
5627 (ipa_merge_fn_summary_after_inlining): Likewise.
5628 (ipa_update_overall_fn_summary): Likewise.
5629 (read_ipa_call_summary): Likewise.
5630 (inline_read_section): Likewise.
5631 (write_ipa_call_summary): Likewise.
5632 (ipa_fn_summary_write): Likewise.
5633 (ipa_free_fn_summary): Likewise.
5634 * ipa-hsa.c (process_hsa_functions): Likewise.
5635 (ipa_hsa_write_summary): Likewise.
5636 (ipa_hsa_read_section): Likewise.
5637 * ipa-icf.c (sem_function::merge): Likewise.
5638 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
5639 (do_estimate_edge_time): Likewise.
5640 (estimate_size_after_inlining): Likewise.
5641 (estimate_growth): Likewise.
5642 (growth_likely_positive): Likewise.
5643 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
5644 (inline_call): Likewise.
5645 * ipa-inline.c (caller_growth_limits): Likewise.
5646 (can_inline_edge_p): Likewise.
5647 (can_inline_edge_by_limits_p): Likewise.
5648 (compute_uninlined_call_time): Likewise.
5649 (compute_inlined_call_time): Likewise.
5650 (want_inline_small_function_p): Likewise.
5651 (edge_badness): Likewise.
5652 (update_caller_keys): Likewise.
5653 (update_callee_keys): Likewise.
5654 (recursive_inlining): Likewise.
5655 (inline_small_functions): Likewise.
5656 (inline_to_all_callers_1): Likewise.
5657 (dump_overall_stats): Likewise.
5658 (early_inline_small_functions): Likewise.
5659 (early_inliner): Likewise.
5660 * ipa-inline.h (estimate_edge_growth): Likewise.
5661 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
5662 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
5663 * ipa-prop.h (IPA_NODE_REF): Likewise.
5664 (IPA_EDGE_REF): Likewise.
5665 * ipa-pure-const.c (malloc_candidate_p): Likewise.
5666 (propagate_malloc): Likewise.
5667 * ipa-split.c (execute_split_functions): Likewise.
5668 * symbol-summary.h: Rename get to get_create.
5670 (get_create): Likewise.
5671 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
5673 2018-06-08 Martin Liska <mliska@suse.cz>
5675 * symbol-summary.h (release): Move definition out of class
5677 (symtab_removal): Likewise.
5678 (symtab_duplication): Likewise.
5680 2018-06-08 Martin Liska <mliska@suse.cz>
5682 * symbol-summary.h (function_summary): Move constructor
5683 implementation out of class declaration.
5684 (release): Likewise.
5685 (symtab_insertion): Likewise.
5686 (symtab_removal): Likewise.
5687 (symtab_duplication): Likewise.
5690 2018-06-08 Martin Liska <mliska@suse.cz>
5692 * Makefile.in: Remove support for MPX (macros, related functions,
5693 fields in cgraph_node, ...).
5694 * builtin-types.def (BT_BND): Likewise.
5695 (BT_FN_BND_CONST_PTR): Likewise.
5696 (BT_FN_CONST_PTR_BND): Likewise.
5697 (BT_FN_VOID_PTR_BND): Likewise.
5698 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
5699 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
5700 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
5701 (expand_builtin_mempcpy_with_bounds): Likewise.
5702 (expand_builtin_memset_with_bounds): Likewise.
5703 (expand_builtin_memset_args): Likewise.
5704 (std_expand_builtin_va_start): Likewise.
5705 (expand_builtin): Likewise.
5706 (expand_builtin_with_bounds): Likewise.
5707 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
5708 (DEF_LIB_BUILTIN_CHKP): Likewise.
5709 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
5710 (DEF_CHKP_BUILTIN): Likewise.
5711 (BUILT_IN_MEMCPY): Likewise.
5712 (BUILT_IN_MEMMOVE): Likewise.
5713 (BUILT_IN_MEMPCPY): Likewise.
5714 (BUILT_IN_MEMSET): Likewise.
5715 (BUILT_IN_STPCPY): Likewise.
5716 (BUILT_IN_STRCAT): Likewise.
5717 (BUILT_IN_STRCHR): Likewise.
5718 (BUILT_IN_STRCPY): Likewise.
5719 (BUILT_IN_STRLEN): Likewise.
5720 (BUILT_IN_MEMCPY_CHK): Likewise.
5721 (BUILT_IN_MEMMOVE_CHK): Likewise.
5722 (BUILT_IN_MEMPCPY_CHK): Likewise.
5723 (BUILT_IN_MEMSET_CHK): Likewise.
5724 (BUILT_IN_STPCPY_CHK): Likewise.
5725 (BUILT_IN_STRCAT_CHK): Likewise.
5726 (BUILT_IN_STRCPY_CHK): Likewise.
5727 * calls.c (store_bounds): Likewise.
5728 (emit_call_1): Likewise.
5729 (special_function_p): Likewise.
5730 (maybe_warn_nonstring_arg): Likewise.
5731 (initialize_argument_information): Likewise.
5732 (finalize_must_preallocate): Likewise.
5733 (compute_argument_addresses): Likewise.
5734 (expand_call): Likewise.
5735 * cfgexpand.c (expand_call_stmt): Likewise.
5736 (expand_return): Likewise.
5737 (expand_gimple_stmt_1): Likewise.
5738 (pass_expand::execute): Likewise.
5739 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
5740 (cgraph_node::remove): Likewise.
5741 (cgraph_node::dump): Likewise.
5742 (cgraph_node::verify_node): Likewise.
5743 * cgraph.h (chkp_function_instrumented_p): Likewise.
5744 (symtab_node::get_alias_target): Likewise.
5745 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
5746 (cgraph_local_p): Likewise.
5747 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
5748 (cgraph_edge::rebuild_references): Likewise.
5749 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
5750 (walk_polymorphic_call_targets): Likewise.
5751 (cgraph_node::expand_thunk): Likewise.
5752 (symbol_table::output_weakrefs): Likewise.
5753 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
5754 (ix86_handle_option): Likewise.
5755 * config/i386/constraints.md: Likewise.
5756 * config/i386/i386-builtin-types.def (BND): Likewise.
5760 * config/i386/i386-builtin.def (BDESC_END): Likewise.
5761 (BDESC_FIRST): Likewise.
5763 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
5764 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
5765 * config/i386/i386.c (enum reg_class): Likewise.
5766 (ix86_target_string): Likewise.
5767 (ix86_option_override_internal): Likewise.
5768 (ix86_conditional_register_usage): Likewise.
5769 (ix86_valid_target_attribute_inner_p): Likewise.
5770 (ix86_set_indirect_branch_type): Likewise.
5771 (ix86_set_current_function): Likewise.
5772 (ix86_function_arg_regno_p): Likewise.
5773 (init_cumulative_args): Likewise.
5774 (ix86_function_arg_advance): Likewise.
5775 (ix86_function_arg): Likewise.
5776 (ix86_pass_by_reference): Likewise.
5777 (ix86_function_value_regno_p): Likewise.
5778 (ix86_function_value_1): Likewise.
5779 (ix86_function_value_bounds): Likewise.
5780 (ix86_return_in_memory): Likewise.
5781 (ix86_setup_incoming_vararg_bounds): Likewise.
5782 (ix86_va_start): Likewise.
5783 (indirect_thunk_need_prefix): Likewise.
5784 (print_reg): Likewise.
5785 (ix86_print_operand): Likewise.
5786 (ix86_expand_call): Likewise.
5787 (ix86_output_function_return): Likewise.
5788 (reg_encoded_number): Likewise.
5789 (BDESC_VERIFYS): Likewise.
5790 (ix86_init_mpx_builtins): Likewise.
5791 (ix86_init_builtins): Likewise.
5792 (ix86_emit_cmove): Likewise.
5793 (ix86_emit_move_max): Likewise.
5794 (ix86_expand_builtin): Likewise.
5795 (ix86_builtin_mpx_function): Likewise.
5796 (ix86_get_arg_address_for_bt): Likewise.
5797 (ix86_load_bounds): Likewise.
5798 (ix86_store_bounds): Likewise.
5799 (ix86_load_returned_bounds): Likewise.
5800 (ix86_store_returned_bounds): Likewise.
5801 (ix86_class_likely_spilled_p): Likewise.
5802 (ix86_hard_regno_mode_ok): Likewise.
5803 (x86_order_regs_for_local_alloc): Likewise.
5804 (ix86_mitigate_rop): Likewise.
5805 (ix86_bnd_prefixed_insn_p): Likewise.
5806 (ix86_mpx_bound_mode): Likewise.
5807 (ix86_make_bounds_constant): Likewise.
5808 (ix86_initialize_bounds): Likewise.
5809 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
5810 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
5811 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
5812 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
5813 (TARGET_CHKP_BOUND_MODE): Likewise.
5814 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
5815 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
5816 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
5817 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
5818 * config/i386/i386.h (TARGET_MPX): Likewise.
5819 (TARGET_MPX_P): Likewise.
5820 (VALID_BND_REG_MODE): Likewise.
5821 (FIRST_BND_REG): Likewise.
5822 (LAST_BND_REG): Likewise.
5823 (enum reg_class): Likewise.
5824 (BND_REG_P): Likewise.
5825 (BND_REGNO_P): Likewise.
5826 (BNDmode): Likewise.
5827 (ADJUST_INSN_LENGTH): Likewise.
5828 * config/i386/i386.md: Likewise.
5829 * config/i386/i386.opt: Likewise.
5830 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
5831 (defined): Likewise.
5832 (LINK_MPX): Likewise.
5833 (MPX_SPEC): Likewise.
5834 (LIBMPX_SPEC): Likewise.
5835 (LIBMPXWRAPPERS_SPEC): Likewise.
5836 (CHKP_SPEC): Likewise.
5837 * config/i386/predicates.md: Likewise.
5838 * dbxout.c (dbxout_type): Likewise.
5839 * doc/extend.texi: Likewise.
5840 * doc/invoke.texi: Likewise.
5841 * doc/md.texi: Likewise.
5842 * doc/tm.texi: Likewise.
5843 * doc/tm.texi.in: Likewise.
5844 * dwarf2out.c (is_base_type): Likewise.
5845 (gen_formal_types_die): Likewise.
5846 (gen_subprogram_die): Likewise.
5847 (gen_type_die_with_usage): Likewise.
5848 (gen_decl_die): Likewise.
5849 (dwarf2out_late_global_decl): Likewise.
5850 * expr.c (expand_assignment): Likewise.
5851 (emit_storent_insn): Likewise.
5852 (store_expr_with_bounds): Likewise.
5853 (store_expr): Likewise.
5854 (expand_expr_real_1): Likewise.
5855 * expr.h (store_expr_with_bounds): Likewise.
5856 * function.c (use_register_for_decl): Likewise.
5857 (struct bounds_parm_data): Likewise.
5858 (assign_parms_augmented_arg_list): Likewise.
5859 (assign_parm_find_entry_rtl): Likewise.
5860 (assign_parm_is_stack_parm): Likewise.
5861 (assign_parm_load_bounds): Likewise.
5862 (assign_bounds): Likewise.
5863 (assign_parms): Likewise.
5864 (expand_function_start): Likewise.
5865 * gcc.c (CHKP_SPEC): Likewise.
5866 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
5867 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
5868 (wrestrict_dom_walker::check_call): Likewise.
5869 * gimple.c (gimple_build_call_from_tree): Likewise.
5870 * gimple.h (enum gf_mask): Likewise.
5871 (gimple_call_with_bounds_p): Likewise.
5872 (gimple_call_set_with_bounds): Likewise.
5873 * gimplify.c (gimplify_init_constructor): Likewise.
5874 * ipa-cp.c (initialize_node_lattices): Likewise.
5875 (propagate_constants_across_call): Likewise.
5876 (find_more_scalar_values_for_callers_subset): Likewise.
5877 * ipa-hsa.c (process_hsa_functions): Likewise.
5878 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
5879 * ipa-icf.c (sem_function::merge): Likewise.
5880 * ipa-inline.c (early_inliner): Likewise.
5881 * ipa-pure-const.c (warn_function_noreturn): Likewise.
5882 (warn_function_cold): Likewise.
5883 (propagate_pure_const): Likewise.
5884 * ipa-ref.h (enum GTY): Likewise.
5885 * ipa-split.c (find_retbnd): Likewise.
5886 (consider_split): Likewise.
5887 (split_function): Likewise.
5888 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
5889 * ipa.c (walk_polymorphic_call_targets): Likewise.
5890 (symbol_table::remove_unreachable_nodes): Likewise.
5891 (process_references): Likewise.
5892 (cgraph_build_static_cdtor_1): Likewise.
5893 * lto-cgraph.c (lto_output_node): Likewise.
5894 (output_refs): Likewise.
5895 (compute_ltrans_boundary): Likewise.
5896 (input_overwrite_node): Likewise.
5897 (input_node): Likewise.
5898 (input_cgraph_1): Likewise.
5899 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
5900 * passes.c (pass_manager::execute_early_local_passes): Likewise.
5901 (class pass_chkp_instrumentation_passes): Likewise.
5902 (make_pass_chkp_instrumentation_passes): Likewise.
5903 * passes.def: Likewise.
5904 * rtl.h (struct GTY): Likewise.
5905 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
5906 * stor-layout.c (layout_type): Likewise.
5907 * symtab.c: Likewise.
5908 * target.def: Likewise.
5909 * targhooks.c (default_chkp_bound_type): Likewise.
5910 (default_chkp_bound_mode): Likewise.
5911 (default_builtin_chkp_function): Likewise.
5912 (default_chkp_function_value_bounds): Likewise.
5913 (default_chkp_make_bounds_constant): Likewise.
5914 (default_chkp_initialize_bounds): Likewise.
5915 * targhooks.h (default_chkp_bound_type): Likewise.
5916 (default_chkp_bound_mode): Likewise.
5917 (default_builtin_chkp_function): Likewise.
5918 (default_chkp_function_value_bounds): Likewise.
5919 (default_chkp_make_bounds_constant): Likewise.
5920 (default_chkp_initialize_bounds): Likewise.
5921 * toplev.c (compile_file): Likewise.
5922 (process_options): Likewise.
5923 * tree-core.h (DEF_BUILTIN): Likewise.
5924 (DEF_BUILTIN_CHKP): Likewise.
5925 * tree-inline.c (declare_return_variable): Likewise.
5926 (remap_gimple_stmt): Likewise.
5927 (copy_bb): Likewise.
5928 (initialize_inlined_parameters): Likewise.
5929 (expand_call_inline): Likewise.
5930 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
5931 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
5932 (make_pass_ipa_chkp_produce_thunks): Likewise.
5933 (make_pass_chkp): Likewise.
5934 (make_pass_chkp_opt): Likewise.
5935 (make_pass_chkp_instrumentation_passes): Likewise.
5936 * tree-pretty-print.c (dump_generic_node): Likewise.
5937 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
5938 * tree-ssa-dce.c (propagate_necessity): Likewise.
5939 (eliminate_unnecessary_stmts): Likewise.
5940 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
5941 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
5942 * tree-ssa-sccvn.h: Likewise.
5943 * tree-ssa-strlen.c (get_string_length): Likewise.
5944 (valid_builtin_call): Likewise.
5945 (adjust_last_stmt): Likewise.
5946 (handle_builtin_strchr): Likewise.
5947 (handle_builtin_strcpy): Likewise.
5948 (handle_builtin_stxncpy): Likewise.
5949 (handle_builtin_memcpy): Likewise.
5950 (handle_builtin_strcat): Likewise.
5951 (strlen_check_and_optimize_stmt): Likewise.
5952 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
5953 * tree-streamer-in.c: Likewise.
5954 * tree-streamer.c (record_common_node): Likewise.
5955 * tree.c (tree_code_size): Likewise.
5956 (wide_int_to_tree_1): Likewise.
5957 (type_contains_placeholder_1): Likewise.
5958 (build_common_tree_nodes): Likewise.
5959 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
5960 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
5961 (POINTER_BOUNDS_P): Likewise.
5962 (BOUNDED_TYPE_P): Likewise.
5963 (BOUNDED_P): Likewise.
5964 (CALL_WITH_BOUNDS_P): Likewise.
5965 (pointer_bounds_type_node): Likewise.
5966 * value-prof.c (gimple_ic): Likewise.
5967 * var-tracking.c (vt_add_function_parameters): Likewise.
5968 * varasm.c (make_decl_rtl): Likewise.
5969 (assemble_start_function): Likewise.
5970 (output_constant): Likewise.
5971 (maybe_assemble_visibility): Likewise.
5972 * varpool.c (ctor_for_folding): Likewise.
5973 * chkp-builtins.def: Remove.
5974 * ipa-chkp.c: Remove.
5975 * ipa-chkp.h: Remove.
5976 * rtl-chkp.c: Remove.
5977 * rtl-chkp.h: Remove.
5978 * tree-chkp-opt.c: Remove.
5979 * tree-chkp.c: Remove.
5980 * tree-chkp.h: Remove.
5982 2018-06-07 Carl Love <cel@us.ibm.com>
5984 * config/rs6000/vsx.md (vextract_fp_from_shorth,
5985 vextract_fp_from_shortl): Add BE support.
5987 2018-06-07 Paul Koning <ni1d@arrl.net>
5989 * compare-elim.c (try_merge_compare): Don't merge compare if
5990 address contains a side effect.
5991 (try_eliminate_compare): Likewise.
5993 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
5995 * config.gcc: Support "tremont".
5996 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
5997 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
5999 * config/i386/i386.c (m_TREMONT): Define.
6000 (processor_target_table): Add "tremont".
6001 (PTA_TREMONT): Define.
6002 (ix86_lea_outperforms): Add TARGET_TREMONT.
6003 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
6004 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
6005 and M_INTEL_GOLDMONT_PLUS.
6006 (fold_builtin_cpu): Add "tremont".
6007 (ix86_add_stmt_cost): Add TARGET_TREMONT.
6008 (ix86_option_override_internal): Add "tremont".
6009 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
6010 (processor_type): Add PROCESSOR_TREMONT.
6011 * config/i386/x86-tune.def: Add m_TREMONT.
6012 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
6014 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
6016 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
6017 symbol defined for msp430i* devices to be lower case.
6019 2018-06-07 Richard Biener <rguenther@suse.de>
6021 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
6022 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
6023 Properly wrap signed arithmetic if overflow wraps.
6025 2018-06-07 Jakub Jelinek <jakub@redhat.com>
6027 PR tree-optimization/69615
6028 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
6029 of a cast from a same precision integral SSA_NAME in a bb dominated
6030 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
6031 cast to utype if rhs2 has already a compatible type.
6033 2018-06-07 Richard Biener <rguenther@suse.de>
6035 PR tree-optimization/85935
6036 * graphite-scop-detection.c (find_params_in_bb): Analyze
6037 condition operands with respect to the correct loop. Assert
6038 the analysis doesn't fail.
6040 2018-06-04 Carl Love <cel@us.ibm.com>
6042 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
6043 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
6044 as it is slightly cheaper.
6045 (first_match_or_eos_index_<mode>):
6046 Calculate index using natural element order.
6047 (first_match_index_<mode>):
6048 Calculate index using natural element order.
6049 (first_match_or_eos_index_<mode>):
6050 Calculate index using natural order.
6051 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
6052 for BE and LE modes.
6053 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
6054 P9V_BUILTIN_VCLZLSBB_V16QI.
6055 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
6058 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
6060 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
6061 indentation and line wrap for many prototypes. Add missing
6062 @smallexample directives around block of prototypes for vec_xl and
6065 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
6067 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
6068 track if we pass or return IEEE 128-bit floating point.
6069 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
6070 C++ mangling that is compatible with GCC 8.1.
6071 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
6072 (init_cumulative_args): Note if we pass or return IEEE 128-bit
6073 floating point types.
6074 (rs6000_function_arg_advance_1): Likewise.
6075 (rs6000_mangle_type): Optionally generate mangled names that match
6076 what GCC 8.1 generated for IEEE 128-bit floating point types.
6077 (rs6000_globalize_decl_name): If we have an external function that
6078 passes or returns IEEE 128-bit types, generate a weak reference
6079 from the mangled name used in GCC 8.1 to the current mangled
6081 (rs6000_init_builtins): Make __ibm128 use the long double type if
6082 long double is IBM extended double. Make __float128 use the long
6083 double type if long double is IEEE 128-bit.
6086 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
6087 macro for __ibm128 built-in functions.
6088 (PACK_IF): Add __ibm128 pack/unpack functions.
6089 (UNPACK_IF): Likewise.
6090 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
6091 enable long double built-in functions if long double is IEEE
6092 128-bit floating point.
6093 (rs6000_invalid_builtin): Update long double built-in function
6095 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
6096 functions, adjust the built-in function to use the long double
6097 built-in function if __ibm128 and long double are the same type.
6098 * doc/extend.texi (PowerPC builtins): Update documention for
6099 __builtin_{,un}pack_longdouble. Add documentation for
6100 __builtin_{,un}pack_ibm128.
6102 2018-06-06 Jim Wilson <jimw@sifive.com>
6104 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
6105 (struct machine_function): New field interrupt_mode.
6106 (riscv_handle_type_attribute): New function. Add forward declaration.
6107 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
6108 (riscv_expand_epilogue): Check interrupt_mode field.
6109 (riscv_set_current_function): Check interrupt attribute args and
6110 set interrupt_mode field.
6111 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
6112 (riscv_sret, riscv_uret): New.
6113 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
6114 new arguments to interrupt attribute.
6116 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
6119 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
6120 Don't handle -mcpu=power8 if -mpower9-vector is also used.
6122 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
6124 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
6125 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
6126 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
6127 several redundant entries.
6129 2018-06-06 David Malcolm <dmalcolm@redhat.com>
6131 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
6132 type from "rtx" to "rtx_insn *".
6133 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
6134 for local "call_insn", removing cast.
6135 (ix86_expand_call): Likewise, introducing a "call_insn" local.
6137 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
6139 PR tree-optimization/86066
6140 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
6141 for BIT_INSERT_EXPR stores.
6143 2018-06-06 Richard Biener <rguenther@suse.de>
6145 PR tree-optimization/86062
6146 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
6147 component refs ontop
6148 of to be offsetted base.
6150 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
6152 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
6153 to be static and remove check on interrupt attribute name.
6155 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
6157 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
6158 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
6160 2018-06-05 Steve Ellcey <sellcey@cavium.com>
6163 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
6165 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
6166 Remove second argument, change how error is called.
6167 (aarch64_layout_arg): Remove second argument from
6168 aarch64_err_no_fpadvsimd call.
6169 (aarch64_init_cumulative_args): Ditto.
6170 (aarch64_gimplify_va_arg_expr): Ditto.
6171 * config/aarch64/aarch64.md (mov<mode>): Ditto.
6173 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
6175 * config/i386/i386.md (simple_return_indirect_internal): New expander.
6176 (*simple_return_indirect_internal<mode>): Rename from
6177 simple_return_indirect_internal. Use W mode iterator.
6178 (rstorssp): New expander.
6179 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
6180 (clrssbsy): New expander.
6181 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
6183 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
6185 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
6187 (cmse_check_pointed_object): Likewise.
6189 2018-06-05 Martin Liska <mliska@suse.cz>
6191 PR gcov-profile/47618
6192 * doc/invoke.texi: Document how -fprofile-dir format
6195 2018-06-05 Richard Biener <rguenther@suse.de>
6197 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
6198 removal pretend DOM info isn't available so we do not update
6199 it and only remove edges, not dominated blocks. Actually free
6200 DOM info in case we removed something. Remove unreachable blocks.
6201 (mfb_keep_latches): Work with either DOM info or marked backedges.
6202 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
6203 first. Mark backedges if DOM info isn't available.
6204 (Re-)compute DOM info after cleanup_control_flow_pre.
6206 2018-06-05 Richard Biener <rguenther@suse.de>
6208 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
6209 (locus_discrim_hasher::hash): Adjust.
6210 (locus_discrim_hasher::equal): Likewise.
6211 (next_discriminator_for_locus): Work on line directly.
6212 (same_line_p): Pass in expanded locus1 as well.
6213 (assign_discriminators): Avoid redundant location expansions.
6215 2018-06-05 Richard Biener <rguenther@suse.de>
6217 PR tree-optimization/86046
6218 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
6219 if required after clearing TREE_ADDRESSABLE.
6221 2018-06-05 Richard Biener <rguenther@suse.de>
6223 PR tree-optimization/86047
6224 * tree-ssa-loop.c (for_each_index): Glob handling of all
6225 decls and constants and really handle all of them.
6227 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6230 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
6231 qualifier_void_pointer and qualifier_const_void_pointer.
6232 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
6233 (arm_init_builtins): Handle the above.
6234 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
6235 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
6236 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
6239 2018-06-05 Martin Liska <mliska@suse.cz>
6241 * auto-profile.c (read_autofdo_file): Do not use
6242 gcov_ctr_summary struct.
6243 (afdo_callsite_hot_enough_for_early_inline): Likewise.
6244 * coverage.c (struct counts_entry): Likewise.
6245 (read_counts_file): Read just single summary entry.
6246 (get_coverage_counts): Use gcov_summary struct.
6247 * coverage.h (get_coverage_counts): Likewise.
6248 * gcov-dump.c (dump_working_sets): Likewise.
6249 (tag_summary): Dump just single summary.
6250 * gcov-io.c (gcov_write_summary): Write just histogram
6252 (gcov_read_summary): Read just single summary.
6253 (compute_working_sets): Use gcov_summary struct.
6254 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
6255 of GCOV_COUNTERS_SUMMABLE.
6256 (GCOV_COUNTERS_SUMMABLE): Remove.
6257 (GCOV_FIRST_VALUE_COUNTER): Replace with
6258 GCOV_COUNTER_V_INTERVAL.
6259 (struct gcov_ctr_summary): Remove.
6260 (struct gcov_summary): Directly use fields of former
6262 (compute_working_sets): Use gcov_summary struct.
6263 * gcov.c (read_count_file): Do not use ctrs fields.
6264 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
6266 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
6268 * profile.c: Likewise.
6269 * profile.h: Likewise.
6271 2018-06-05 Martin Liska <mliska@suse.cz>
6273 PR gcov-profile/84846
6274 * gcov.c (output_lines): Print working directory only
6275 in intermediate format.
6277 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
6279 * config/s390/s390-builtin-types.def: Add void function type.
6280 * config/s390/s390-builtins.def: Use the function type for the
6283 2018-06-04 Jim Wilson <jimw@sifive.com>
6285 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
6287 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
6288 and maybe_eh_return. Change regno to unsigned int. Use new args to
6289 handle EH_RETURN_DATA_REGNO registers properly.
6290 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
6291 (riscv_expand_epilogue): Update comment. Change argument name and
6292 type. Update code to use new name and type. Pass new args to
6293 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
6295 * config/riscv/riscv.md (NORMAL_RETURN): New.
6296 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
6297 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
6298 (eh_return): Call gen_eh_return_internal and emit barrier.
6299 (eh_return_internal): Call riscv_expand_epilogue.
6301 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
6303 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
6304 bit_insertion field and declare can_be_merged_into method.
6305 (merged_store_group::can_be_merged_into): New method.
6306 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
6307 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
6308 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
6310 2018-06-04 Richard Biener <rguenther@suse.de>
6312 PR tree-optimization/85955
6313 * builtins.c (fold_builtin_sincos): Convert pointers to
6314 destination to appropriate type before dereferencing.
6316 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
6318 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
6320 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
6322 * expr.c (expand_expr_real_1): Force the operand into memory if
6323 its TYPE_MODE is BLKmode and if there is no integer mode for
6324 the number of bits being extracted.
6326 2018-06-04 Jakub Jelinek <jakub@redhat.com>
6330 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
6331 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
6333 2018-06-04 Richard Biener <rguenther@suse.de>
6335 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
6336 (cleanup_tree_cfg_noloop): ... single caller. Do
6337 start_recording_case_labels later.
6339 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
6341 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
6342 to _IMMINTRIN_H_INCLUDED.
6343 * config/i386/pconfigintrin.h: Ditto.
6344 * config/i386/waitpkgintrin.h: Ditto.
6345 * config/i386/immintrin.h: Add includes for sgxintrin.h,
6346 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
6347 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
6348 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
6349 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
6350 waitpkgintrin.h and cldemoteintrin.h.
6352 2018-06-04 Richard Biener <rguenther@suse.de>
6354 PR tree-optimization/86038
6355 * tracer.c (find_best_successor): Check probability for
6356 being initialized, bail out if not.
6358 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
6361 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
6362 of bits to ignore when comparing architectures.
6364 2018-06-04 Jakub Jelinek <jakub@redhat.com>
6366 PR tree-optimization/69615
6367 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
6368 maximum or minimum of the type, try to merge it also as if
6369 range1 is + [-, x - 1] or + [x + 1, -].
6372 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
6374 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
6376 PR tree-optimization/86034
6377 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
6378 the unsigned bitfield type in a bit insertion sequence if it does not
6379 have a larger precision than the bitfield size.
6380 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
6382 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
6384 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
6386 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
6388 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
6389 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
6390 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
6391 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
6393 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
6395 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
6396 Disable -fdelete-null-pointer-checks for ELF toolchain.
6398 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
6399 Kito Cheng <kito.cheng@gmail.com>
6401 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
6402 (nds32le-*-*, nds32be-*-*): Integrate checking process.
6403 (nds32*-*-*): Add glibc and uclibc conditions.
6404 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
6405 (TARGET_EXCEPT_UNWIND_INFO): Define.
6406 * config/nds32/elf.h: New file.
6407 * config/nds32/linux.h: New file.
6408 * config/nds32/nds32-elf.opt: New file.
6409 * config/nds32/nds32-linux.opt: New file.
6410 * config/nds32/nds32-fp-as-gp.c
6411 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
6412 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
6414 (nds32_asm_file_end): Ditto.
6415 (nds32_print_operand): Ditto.
6416 (nds32_insert_attributes): Ditto.
6417 (nds32_init_libfuncs): New function.
6418 (TARGET_HAVE_TLS): Define.
6419 (TARGET_INIT_LIBFUNCS): Define.
6420 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
6422 (TARGET_ELF): Apply different mcmodel setting.
6423 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
6424 been migrated into elf.h and linux.h files.
6425 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
6426 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
6427 (mcmodel): The content has been migrated into nds32-elf.opt and
6428 nds32-linux.opt files.
6429 * config/nds32/t-elf: New file.
6430 * config/nds32/t-linux: New file.
6432 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
6433 Shiva Chen <shiva0217@gmail.com>
6435 * config/nds32/constants.md (unspec_volatile_element): Add
6436 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
6437 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
6439 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
6440 (make_pass_nds32_fp_as_gp): Declare.
6441 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
6443 (nds32_asm_function_end_prologue): Remove unused asm output.
6444 (nds32_asm_function_begin_epilogue): Remove unused asm output.
6445 (nds32_asm_file_start): Output necessary fp_as_gp information.
6446 (nds32_option_override): Adjust register usage.
6447 (nds32_expand_prologue): Consider fp_as_gp situation.
6448 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
6449 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
6452 (simple_return): Ditto.
6453 (omit_fp_begin): Output special directive for fp_as_gp.
6454 (omit_fp_end): Output special directive for fp_as_gp.
6455 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
6456 mforbid-fp-as-gp): New options.
6458 2018-06-01 Mark Wielaard <mark@klomp.org>
6460 * dwarf2out.c (dwarf2out_finish): Remove generation of
6461 DW_AT_loclists_base.
6463 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
6465 * gimple-ssa-store-merging.c: Include gimple-fold.h.
6466 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
6467 (struct merged_store_group): Add bit_insertion field.
6468 (dump_char_array): Use standard hexadecimal format.
6469 (merged_store_group::merged_store_group): Set bit_insertion to false.
6470 (merged_store_group::apply_stores): Use optimal buffer size. Deal
6471 with BIT_INSERT_EXPR stores. Move up code updating the mask and
6472 also print the mask in the dump file.
6473 (pass_store_merging::gate): Minor tweak.
6474 (imm_store_chain_info::coalesce_immediate): Fix wrong association
6475 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
6476 stores with INTEGER_CST stores.
6477 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
6478 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
6479 and use it throughout. Generate bit insertion sequences if need be.
6480 (pass_store_merging::process_store): Remove redundant condition.
6481 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
6483 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
6485 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
6486 the 128-bit floating point types. Fix function comment.
6488 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6490 * config/aarch64/aarch64-simd.md
6491 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
6493 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
6496 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
6498 PR tree-optimization/85989
6499 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
6501 (backprop::intersect_uses): Check it when deciding whether this
6502 is a backedge reference.
6503 (backprop::process_block): Add each phi to m_visited_phis
6504 after visiting it, then clear it at the end.
6506 2018-06-01 Richard Biener <rguenther@suse.de>
6508 * tree-vectorizer.h (vect_dr_stmt): New function.
6509 (vect_get_load_cost): Adjust.
6510 (vect_get_store_cost): Likewise.
6511 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
6512 Use vect_dr_stmt instead of DR_SMTT.
6513 (vect_record_base_alignments): Likewise.
6514 (vect_calculate_target_alignment): Likewise.
6515 (vect_compute_data_ref_alignment): Likewise and make static.
6516 (vect_update_misalignment_for_peel): Likewise.
6517 (vect_verify_datarefs_alignment): Likewise.
6518 (vector_alignment_reachable_p): Likewise.
6519 (vect_get_data_access_cost): Likewise. Pass down
6520 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
6521 (vect_get_peeling_costs_all_drs): Likewise.
6522 (vect_peeling_hash_get_lowest_cost): Likewise.
6523 (vect_enhance_data_refs_alignment): Likewise.
6524 (vect_find_same_alignment_drs): Likewise.
6525 (vect_analyze_data_refs_alignment): Likewise.
6526 (vect_analyze_group_access_1): Likewise.
6527 (vect_analyze_group_access): Likewise.
6528 (vect_analyze_data_ref_access): Likewise.
6529 (vect_analyze_data_ref_accesses): Likewise.
6530 (vect_vfa_segment_size): Likewise.
6531 (vect_small_gap_p): Likewise.
6532 (vectorizable_with_step_bound_p): Likewise.
6533 (vect_prune_runtime_alias_test_list): Likewise.
6534 (vect_analyze_data_refs): Likewise.
6535 (vect_supportable_dr_alignment): Likewise.
6536 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
6537 (vect_gen_prolog_loop_niters): Likewise.
6538 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
6539 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
6541 (vect_recog_mask_conversion_pattern): Likewise.
6542 (vect_try_gather_scatter_pattern): Likewise.
6543 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
6544 to vect_get_store_cost.
6545 (vect_get_store_cost): Get stmt_info instead of DR.
6546 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
6547 (vect_get_load_cost): Get stmt_info instead of DR.
6549 2018-06-01 Richard Biener <rguenther@suse.de>
6552 * gimple-fold.c (var_decl_component_p): Also allow offsetted
6553 vars wrapped in MEM_REFs.
6555 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
6557 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
6558 Fix subreg tests so that we only return a choice between
6559 GENERAL_REGS and FP_REGS if the original classes included both.
6561 2018-06-01 Richard Biener <rguenther@suse.de>
6564 * tree-ssa-structalias.c (get_function_part_constraint):
6565 Handle NULL fi->decl.
6566 (find_func_aliases_for_call): Properly handle indirect
6567 fi from direct call.
6568 (find_func_clobbers): Likewise.
6569 (ipa_pta_execute): Likewise.
6570 (create_variable_info_for): For functions that are ifunc_resolver
6571 resolve to a varinfo that contains the result of the resolver call.
6572 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
6575 2018-05-31 Michael Collison <michael.collison@arm.com>
6577 * config/aarch64/aarch64.md:
6578 (*fix_to_zero_extenddfdi2): New pattern.
6579 * gcc.target/aarch64/fix_extend1.c: New testcase.
6581 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
6585 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
6586 and BUILT_IN_STRNCMP_EQ.
6587 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
6588 BUILT_IN_STRNCMP_EQ.
6589 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
6590 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
6591 (gimple_fold_builtin): Likewise.
6592 * tree-ssa-strlen.c (compute_string_length): New function.
6593 (determine_min_obsize): New function.
6594 (handle_builtin_string_cmp): New function to handle calls to
6595 string compare functions.
6596 (strlen_optimize_stmt): Add handling to builtin string compare
6598 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
6599 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
6600 * tree.c (build_common_builtin_nodes): Add new defines of
6601 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
6603 2018-05-31 Jakub Jelinek <jakub@redhat.com>
6606 * bb-reorder.c (pass_partition_blocks::gate): Return false for
6607 functions with naked attribute.
6609 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
6611 * config/i386/sse.md (avx_vec_concat<mode>):
6612 Substitute concat_tg_mode mode attribute with xtg_mode.
6613 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
6614 (concat_tg_mode): Remove mode attribute.
6616 2018-05-31 Martin Sebor <msebor@redhat.com>
6619 * calls.c (alloc_max_size): Correct a logic error/typo.
6620 Treat excessive arguments as infinite. Warn for invalid arguments.
6621 * doc/invoke.texi (-Walloc-size-larger-than): Update.
6623 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
6626 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
6627 and movx for Haswell.
6629 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
6630 Cesar Philippidis <cesar@codesourcery.com>
6633 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
6634 when emitting error on private/firstprivate reductions.
6635 * omp-low.c (lower_omp_target): Avoid reference-type processing
6636 on pointers for firstprivate clause.
6638 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
6640 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
6643 * config/aarch64/aarch64-simd.md
6644 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
6645 (aarch64_ld1_x3_<mode>): Likewise
6646 (aarch64_st1x2<VALLDIF:mode>): Likewise
6647 (aarch64_st1_x2_<mode>): Likewise
6648 (aarch64_st1x3<VALLDIF:mode>): Likewise
6649 (aarch64_st1_x3_<mode>): Likewise
6650 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
6651 (vld1_s8_x3): Likewise.
6652 (vld1_u16_x3): Likewise.
6653 (vld1_s16_x3): Likewise.
6654 (vld1_u32_x3): Likewise.
6655 (vld1_s32_x3): Likewise.
6656 (vld1_u64_x3): Likewise.
6657 (vld1_s64_x3): Likewise.
6658 (vld1_f16_x3): Likewise.
6659 (vld1_f32_x3): Likewise.
6660 (vld1_f64_x3): Likewise.
6661 (vld1_p8_x3): Likewise.
6662 (vld1_p16_x3): Likewise.
6663 (vld1_p64_x3): Likewise.
6664 (vld1q_u8_x3): Likewise.
6665 (vld1q_s8_x3): Likewise.
6666 (vld1q_u16_x3): Likewise.
6667 (vld1q_s16_x3): Likewise.
6668 (vld1q_u32_x3): Likewise.
6669 (vld1q_s32_x3): Likewise.
6670 (vld1q_u64_x3): Likewise.
6671 (vld1q_s64_x3): Likewise.
6672 (vld1q_f16_x3): Likewise.
6673 (vld1q_f32_x3): Likewise.
6674 (vld1q_f64_x3): Likewise.
6675 (vld1q_p8_x3): Likewise.
6676 (vld1q_p16_x3): Likewise.
6677 (vld1q_p64_x3): Likewise.
6678 (vst1_s64_x2): Likewise.
6679 (vst1_u64_x2): Likewise.
6680 (vst1_f64_x2): Likewise.
6681 (vst1_s8_x2): Likewise.
6682 (vst1_p8_x2): Likewise.
6683 (vst1_s16_x2): Likewise.
6684 (vst1_p16_x2): Likewise.
6685 (vst1_s32_x2): Likewise.
6686 (vst1_u8_x2): Likewise.
6687 (vst1_u16_x2): Likewise.
6688 (vst1_u32_x2): Likewise.
6689 (vst1_f16_x2): Likewise.
6690 (vst1_f32_x2): Likewise.
6691 (vst1_p64_x2): Likewise.
6692 (vst1q_s8_x2): Likewise.
6693 (vst1q_p8_x2): Likewise.
6694 (vst1q_s16_x2): Likewise.
6695 (vst1q_p16_x2): Likewise.
6696 (vst1q_s32_x2): Likewise.
6697 (vst1q_s64_x2): Likewise.
6698 (vst1q_u8_x2): Likewise.
6699 (vst1q_u16_x2): Likewise.
6700 (vst1q_u32_x2): Likewise.
6701 (vst1q_u64_x2): Likewise.
6702 (vst1q_f16_x2): Likewise.
6703 (vst1q_f32_x2): Likewise.
6704 (vst1q_f64_x2): Likewise.
6705 (vst1q_p64_x2): Likewise.
6706 (vst1_s64_x3): Likewise.
6707 (vst1_u64_x3): Likewise.
6708 (vst1_f64_x3): Likewise.
6709 (vst1_s8_x3): Likewise.
6710 (vst1_p8_x3): Likewise.
6711 (vst1_s16_x3): Likewise.
6712 (vst1_p16_x3): Likewise.
6713 (vst1_s32_x3): Likewise.
6714 (vst1_u8_x3): Likewise.
6715 (vst1_u16_x3): Likewise.
6716 (vst1_u32_x3): Likewise.
6717 (vst1_f16_x3): Likewise.
6718 (vst1_f32_x3): Likewise.
6719 (vst1_p64_x3): Likewise.
6720 (vst1q_s8_x3): Likewise.
6721 (vst1q_p8_x3): Likewise.
6722 (vst1q_s16_x3): Likewise.
6723 (vst1q_p16_x3): Likewise.
6724 (vst1q_s32_x3): Likewise.
6725 (vst1q_s64_x3): Likewise.
6726 (vst1q_u8_x3): Likewise.
6727 (vst1q_u16_x3): Likewise.
6728 (vst1q_u32_x3): Likewise.
6729 (vst1q_u64_x3): Likewise.
6730 (vst1q_f16_x3): Likewise.
6731 (vst1q_f32_x3): Likewise.
6732 (vst1q_f64_x3): Likewise.
6733 (vst1q_p64_x3): Likewise.
6735 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
6737 * config/msp430/msp430.c (msp430_output_labelref): Prepend
6738 user_label_prefix to name.
6740 * tree-core.h: Update comment about the format of NAME string
6741 passed to handler in attribute_spec.
6743 * config/msp430/msp430.md: Remove erroneous subreg expression from
6744 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
6745 zero_extend{q,h}isi2.
6747 2018-05-30 Borislav Petkov <bp@suse.de>
6749 * doc/extend.texi: Document some architecture specific
6750 constraints and sort entries.
6752 2018-05-30 Martin Sebor <msebor@redhat.com>
6755 * builtins.c (expand_builtin_stpcpy_1): New function.
6756 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
6757 only if the former succeeds.
6759 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
6761 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
6764 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
6766 * doc/invoke.texi (-flinker-output): Document
6768 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
6770 * passes.c (ipa_write_summaries): Only modify statements if body
6772 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
6773 incrementally linking.
6774 (ipa_passes): Likewise.
6775 * lto-cgraph.c (lto_output_node): When incrementally linking do not
6776 pass down resolution info.
6777 * common.opt (flag_incremental_link): Update info.
6778 * gcc.c (plugin specs): Turn flinker-output=* to
6779 -plugin-opt=-linker-output-known
6780 * toplev.c (compile_file): Also cut compilation when doing incremental
6782 * flag-types. (enum lto_partition_model): Add
6783 LTO_LINKER_OUTPUT_NOLTOREL.
6784 (invoke.texi): Add -flinker-output docs.
6785 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
6786 link same way as WPA; do not stream in dead initializers.
6788 * dwarf2out.c (dwarf2out_die_ref_for_decl,
6789 darf2out_register_external_decl): Support incremental link.
6791 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
6793 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
6795 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
6797 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
6798 it down to simple_object_copy_lto_debug_sections.
6799 (run_gcc): Determine incremental LTO link time and configure
6800 lto1 into non-wpa mode, disable renaming of debug sections.
6802 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
6804 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
6805 descriptions of various incorrectly documented functions.
6807 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
6810 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
6811 address check not strict.
6813 2018-05-30 Richard Biener <rguenther@suse.de>
6815 PR tree-optimization/85964
6816 * tracer.c (better_p): Drop initialized count check, we only
6817 call the function with initialized counts now.
6818 (find_best_successor): Do find a best edge if one
6819 has uninitialized count.
6820 (find_best_predecessor): Likewise. Do BB frequency check only
6821 if count is initialized.
6823 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
6825 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
6826 (aarch64_ldrstr_offset_compare): New.
6827 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
6828 load/store orderings.
6829 (aarch64_gen_adjusted_ldpstp): Likewise.
6831 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
6833 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
6834 Check for subset of GENERAL_REGS and FP_REGS.
6835 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
6838 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
6840 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
6841 and wi::to_poly_offset. Add the current offset and then check
6842 whether the sum fits, rather than using an unchecked addition of
6843 a checked term. Check for a shwi rather than a uhwi.
6844 * expr.c (get_bit_range): Use tree_to_poly_uint64.
6845 (store_constructor): Use poly_int_tree_p.
6846 (expand_expr_real_1): Likewise.
6847 * function.c (assign_temp): Likewise.
6848 * fold-const.c (const_binop): Use poly_int_tree_p and
6850 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
6852 * ipa-icf-gimple.c (func_checker::compare_operand): Use
6853 to_poly_offset for MEM offsets.
6854 * ipa-icf.c (sem_variable::equals): Likewise.
6855 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
6856 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
6857 wi::to_poly_offset for BIT_FIELD_REF offsets.
6858 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
6860 * var-tracking.c (emit_note_insn_var_location): Use
6861 tree_to_poly_uint64.
6863 2018-05-29 Jim Wilson <jimw@sifive.com>
6865 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
6867 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
6870 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
6871 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
6873 (sse4_1_round<mode>2): Use nonimmediate_operand
6874 for operand 1 predicate.
6876 2018-05-29 Martin Sebor <msebor@redhat.com>
6877 Richard Biener <rguenther@suse.de>
6880 * calls.c (get_size_range): Call determine_value_range instead
6881 of get_value_range..
6882 * tree-vrp.h (determine_value_range): Declared new function.
6883 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
6885 2018-05-29 Richard Biener <rguenther@suse.de>
6887 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
6888 sure to use non-pattern stmts for get_earlier_stmt arguments.
6889 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
6890 called on pattern stmts.
6891 (get_later_stmt): Likewise.
6893 2018-05-29 Martin Liska <mliska@suse.cz>
6895 PR gcov-profile/85759
6896 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
6899 2018-05-29 Jakub Jelinek <jakub@redhat.com>
6901 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
6903 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
6907 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
6908 VEC_PACK_FLOAT_EXPR): New tree codes.
6909 * tree-pretty-print.c (op_code_prio): Handle
6910 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
6911 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
6912 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
6913 * tree-inline.c (estimate_operator_cost): Likewise.
6914 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
6915 * fold-const.c (const_binop): Likewise.
6916 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
6917 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
6918 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
6919 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
6920 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
6921 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
6922 * expr.c (expand_expr_real_2): Likewise.
6923 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
6924 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
6925 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
6927 * optabs.c (expand_widen_pattern_expr): For
6928 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
6929 sign from result type rather than operand's type.
6930 (expand_binop_directly): For vec_packu_float_optab and
6931 vec_packs_float_optab allow result type to be different from operand's
6933 * optabs-tree.c (optab_for_tree_code): Handle
6934 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
6935 VEC_PACK_FLOAT_EXPR. Formatting fixes.
6936 * tree-vect-generic.c (expand_vector_operations_1): Handle
6937 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
6938 VEC_PACK_FLOAT_EXPR.
6939 * tree-vect-stmts.c (supportable_widening_operation): Handle
6941 (supportable_narrowing_operation): Handle FLOAT_EXPR.
6942 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
6943 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
6944 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
6945 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
6947 (vec_pack<floatprefix>_float_<mode>): New expander.
6948 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
6950 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
6951 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
6952 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
6953 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
6954 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
6956 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
6957 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
6958 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
6959 VEC_PACK_FLOAT_EXPR): Document.
6961 2018-05-29 Richard Biener <rguenther@suse.de>
6963 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
6965 (stmt_vec_info_vec): Make pointer.
6966 (init_stmt_vec_info_vec): Remove.
6967 (free_stmt_vec_info_vec): Likewise.
6968 (set_stmt_vec_info_vec): New function.
6969 (free_stmt_vec_infos): Likewise.
6970 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
6971 (set_vinfo_for_stmt): Likewise.
6972 (get_earlier_stmt): Likewise.
6973 (get_later_stmt): Likewise.
6974 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
6975 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
6976 (vec_info::~vec_info): Free stmt_vec_infos.
6977 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
6978 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
6979 (pass_slp_vectorize::execute): Likewise.
6980 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
6981 (free_stmt_vec_info_vec): Likewise.
6982 (set_stmt_vec_info_vec): New function.
6983 (free_stmt_vec_infos): Likewise.
6984 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
6985 the global stmt_vec_info_vec.
6986 * tree-parloops.c (gather_scalar_reductions): Use
6987 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
6990 2018-05-29 Richard Biener <rguenther@suse.de>
6992 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
6994 2018-05-29 Martin Liska <mliska@suse.cz>
6995 David Malcolm <dmalcolm@redhat.com>
6997 * vec.c (test_reverse): New.
6998 (vec_c_tests): Add new test.
6999 * vec.h (vl_ptr>::reverse): New function.
7001 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
7003 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
7005 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
7008 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
7010 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
7012 2018-05-28 Richard Biener <rguenther@suse.de>
7014 PR tree-optimization/85933
7015 * tree-vect-data-refs.c (vect_record_base_alignments): Only
7016 look at stmts marked as vectorizable.
7018 2018-05-28 Richard Biener <rguenther@suse.de>
7020 PR tree-optimization/85934
7021 * tree-vect-generic.c (expand_vector_operations_1): Hoist
7022 vector boolean check before scalar optimization.
7024 2018-05-28 Jakub Jelinek <jakub@redhat.com>
7026 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
7029 2018-05-28 Mark Wielaard <mark@klomp.org>
7031 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
7032 if it is an expression containing a minus sign.
7034 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
7036 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
7038 2018-05-27 Paul Koning <ni1d@arrl.net>
7040 * config/pdp11/pdp11.md (truncsihi2): Remove.
7042 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
7043 Chung-Ju Wu <jasonwucj@gmail.com>
7045 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
7047 (unaligned_store_dw): Ditto.
7048 * config/nds32/nds32-memory-manipulation.c
7049 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
7050 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
7051 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
7052 (emit_setmem_word_loop): Rename to ...
7053 (emit_setmem_doubleword_loop): ... this.
7054 (nds32_gen_dup_4_byte_to_word_value): New function.
7055 (nds32_gen_dup_8_byte_to_double_word_value): New function.
7056 (nds32_expand_setmem_loop): Refine implementation.
7057 (nds32_expand_setmem_loop_v3m): Ditto.
7058 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
7061 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
7063 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
7065 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
7067 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
7068 (nds32_init_machine_status): Initialize machine->attr_naked_p and
7069 machine->attr_no_prologue_p.
7070 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
7071 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
7072 (nds32_expand_epilogue): Consider attr_naked_p.
7073 (nds32_expand_epilogue_v3pop): Likewise.
7074 (nds32_can_use_return_insn): Likewise.
7075 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
7076 attr_no_prologue_p fields.
7077 * config/nds32/nds32.opt (mret-in-naked-func): New option.
7079 2018-05-27 Jakub Jelinek <jakub@redhat.com>
7082 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
7084 * config/i386/sse.md
7085 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
7087 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
7089 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
7091 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
7093 (*<floatsuffix>floatv2div2sf2): Rename to ...
7094 (*float<floatunssuffix>v2div2sf2): ... this.
7095 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
7096 (float<floatunssuffix>v2div2sf2_mask): ... this.
7097 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
7098 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
7099 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
7101 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
7103 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
7105 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
7107 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
7109 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
7111 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
7112 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
7113 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
7114 gen_ufix_truncv8dfv8si2.
7115 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
7116 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
7117 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
7118 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
7119 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
7120 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
7121 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
7122 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
7124 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
7128 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
7130 2018-05-25 Jim Wilson <jimw@sifive.com>
7132 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
7133 * config/riscv/riscv.c (struct machine_function): Add
7134 interrupt_handler_p and attribute_checked_p fields.
7135 (riscv_attribute_table): Add interrupt.
7136 (riscv_interrupt_type_p): New.
7137 (riscv_save_reg_p): Save extra regs for interrupt handler.
7138 (riscv_use_save_libcall): Return false for interrupt handler.
7139 (riscv_first_stack_step): Add forward declaration.
7140 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
7141 for interrupt handler with large frame. Use it for saved reg list.
7142 (riscv_expand_prologue): Move flag_stack_usage_info support to
7143 eliminate duplication.
7144 (riscv_expand_epilogue): Generate mret for interrupt handler.
7145 (riscv_epilogue_uses): New.
7146 (riscv_can_use_return_insn): Return false for interrupt handler.
7147 (riscv_function_ok_for_sibcall): Likewise.
7148 (riscv_set_current_function): Add interrupt handler support.
7149 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
7150 * config/riscv/riscv.md (UNSPECV_MRET): New.
7152 (riscv_frflags, riscv_fsflags): Use tab after opcode.
7154 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
7156 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
7158 PR tree-optimization/85712
7159 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
7160 this candidate has already been replaced in-situ by a copy.
7162 2018-05-25 Jason Merrill <jason@redhat.com>
7164 PR c++/80485 - inline function non-zero address.
7165 * symtab.c (nonzero_address): Check DECL_COMDAT.
7167 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
7170 * config/alpha/alpha.md (ashlsi3): New insn pattern.
7171 (*ashlsi_se): Rename from *ashldi_se. Define as sign
7172 extension of SImode operation. Use const123_operand predicate.
7173 (*saddsi_1): Remove.
7174 (*saddl_se_1): Ditto.
7176 (*ssubl_se_1): Ditto.
7177 * config/alpha/predicates.md (const123_operand): New predicate.
7178 * config/alpha/constraints.md (P): Use IN_RANGE.
7180 2018-05-25 Richard Biener <rguenther@suse.de>
7182 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
7184 (ref_maybe_used_by_stmt_p): Likewise.
7185 (stmt_may_clobber_ref_p): Likewise.
7186 (stmt_may_clobber_ref_p_1): Likewise.
7187 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
7189 (ref_maybe_used_by_stmt_p): Likewise.
7190 (stmt_may_clobber_ref_p): Likewise.
7191 (stmt_may_clobber_ref_p_1): Likewise.
7192 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
7193 the alias oracle to disambiguate DRs with stmts DR analysis
7195 (vect_analyze_data_refs): Do not give up on not analyzable
7196 DRs for BB vectorization. Remove code truncating the dataref
7199 2018-05-25 Jakub Jelinek <jakub@redhat.com>
7202 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
7203 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
7204 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
7206 2018-05-25 Richard Biener <rguenther@suse.de>
7208 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
7209 function, combining stmt data ref gathering and fatal analysis
7211 (vect_analyze_data_refs): Remove now redudnant code and simplify.
7212 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
7213 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
7214 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
7215 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
7217 2018-05-25 Bin Cheng <bin.cheng@arm.com>
7219 PR tree-optimization/85720
7220 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
7221 SCC if all partitions are builtins.
7222 (version_loop_by_alias_check): New parameter. Generate cancelable
7223 runtime alias check if all partitions are builtins.
7224 (distribute_loop): Update call to above function.
7226 2018-05-25 Bin Cheng <bin.cheng@arm.com>
7228 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
7229 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
7230 (parm_default_def_partition_arg): Ditto.
7231 (set_parm_default_def_partition): Ditto.
7232 (get_parm_default_def_partitions): Ditto and make it static.
7233 (get_undefined_value_partitions): Ditto and make it static.
7234 (remove_ssa_form): Refactor call to init_var_map here.
7235 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
7236 computation for loop region.
7237 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
7238 (register_default_def): Delete.
7239 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
7240 (parm_default_def_partition_arg): Ditto.
7241 (set_parm_default_def_partition): Ditto.
7242 (get_parm_default_def_partitions): Ditto and make it static.
7243 (get_undefined_value_partitions): Ditto and make it static.
7244 (coalesce_with_default, coalesce_with_default): Update comment.
7245 (create_coalesce_list_for_region): New func factored out from
7246 create_outofssa_var_map.
7247 (populate_coalesce_list_for_outofssa): New func factored out from
7248 create_outofssa_var_map and coalesce_ssa_name.
7249 (create_outofssa_var_map): Delete.
7250 (coalesce_ssa_name): Refactor to support live range computation.
7251 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
7252 (get_parm_default_def_partitions): Delete.
7253 (get_undefined_value_partitions): Ditto.
7254 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
7255 computation for loop region.
7256 (new_tree_live_info, loe_visit_block): Ditto.
7257 (live_worklist, set_var_live_on_entry): Ditto.
7258 (calculate_live_on_exit, verify_live_on_entry): Ditto.
7259 * tree-ssa-live.h (struct _var_map): New fields.
7260 (init_var_map): Change decl.
7261 (region_contains_p): New.
7263 2018-05-25 Bin Cheng <bin.cheng@arm.com>
7265 * tree-ssa-live.h (live_merge_and_clear): Delete.
7267 2018-05-25 Richard Biener <rguenther@suse.de>
7270 * tree-dump.c (dequeue_and_dump): Remove access to removed
7271 operand 2 of a SWITCH_EXPR.
7273 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
7275 * doc/sourcebuild.texi (vect_double_cond_arith): Include
7276 multiplication and division.
7277 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
7278 (cond_udiv@var{m}, cond_umod@var{m}): Document.
7279 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
7280 (cond_udiv_optab, cond_umod_optab): New optabs.
7281 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
7282 (IFN_COND_RDIV): New internal functions.
7283 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
7284 TRUNC_MOD_EXPR and RDIV_EXPR.
7285 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
7286 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
7288 (SVE_INT_BINARY): Include mult.
7289 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
7290 (optab, sve_int_op): Handle mult.
7291 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
7293 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
7294 for SVE_INT_BINARY_SD.
7296 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
7298 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
7299 (optab, sve_int_op): Handle div and udiv.
7300 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
7301 for SVE_INT_BINARY_SD.
7302 (*<optab><mode>3): New insn for the same.
7304 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
7306 * tree-vect-patterns.c: Include predict.h.
7307 (vect_recog_divmod_pattern): Restrict check for division support
7308 to when optimizing for size.
7310 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
7312 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
7313 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
7314 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
7315 (gimple_match_op::set_op): Likewise.
7316 (gimple_resimplify4): Declare.
7317 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
7318 (expr::gen_transform): Likewise.
7319 (decision_tree::gen): Generate a simplification routine for 4 operands.
7320 * gimple-match-head.c (gimple_simplify): Add an overload for
7321 4 operands. In the top-level function, handle up to 4 call
7322 arguments and call gimple_resimplify4.
7323 (gimple_resimplify4): New function.
7324 (build_call_internal): Pass a fourth operand.
7325 (maybe_push_to_seq): Likewise.
7326 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
7327 Fold VEC_COND_EXPRs of an operation and a default value into
7328 an IFN_COND_* function if possible.
7329 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
7331 (SVE_COND_FP_BINARY): Include them.
7332 (optab, sve_fp_op): Handle them.
7333 (SVE_INT_BINARY_REV): New code iterator.
7334 (SVE_COND_FP_BINARY_REV): New int iterator.
7335 (commutative): New int attribute.
7336 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
7338 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
7340 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
7341 (*cond_<optab><mode>): New patterns for reversed operands.
7343 2018-05-25 Richard Biener <rguenther@suse.de>
7345 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
7346 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
7347 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
7348 (STMT_VINFO_GROUPED_ACCESS): Adjust.
7349 * tree-vect-data-refs.c (everywhere): Adjust users.
7350 * tree-vect-loop.c (everywhere): Likewise.
7351 * tree-vect-slp.c (everywhere): Likewise.
7352 * tree-vect-stmts.c (everywhere): Likewise.
7353 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
7355 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7357 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
7359 (gcc_cv_as_section_exclude): ... this.
7360 Try Solaris as #exclude syntax.
7361 * configure: Regenerate.
7362 * config.in: Regenerate.
7363 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
7365 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
7366 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
7368 * varasm.c (default_elf_asm_named_section): Don't check if
7369 HAVE_GAS_SECTION_EXCLUDE is defined.
7371 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
7373 * doc/md.texi: Update the documentation of the cond_* optabs
7374 to mention the new final operand. Fix GET_MODE_NUNITS call.
7375 Describe the scalar case too.
7376 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
7377 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
7379 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
7380 (get_conditional_internal_fn): Update comment.
7381 * tree-vect-loop.c (vectorizable_reduction): Pass the original
7382 accumulator value as a final argument to conditional functions.
7383 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
7384 a define_expand and add an "else" operand. Assert for now that
7385 the else operand is equal to operand 2. Use SVE_INT_BINARY and
7386 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
7387 (*cond_<optab><mode>): New patterns.
7388 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
7389 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
7390 (UNSPEC_COND_EOR): Delete.
7391 (optab): Remove associated mappings.
7392 (SVE_INT_BINARY): New code iterator.
7393 (sve_int_op): Remove int attribute and add "minus" to the code
7395 (SVE_COND_INT_OP): Delete.
7396 (SVE_COND_FP_OP): Rename to...
7397 (SVE_COND_FP_BINARY): ...this.
7399 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
7401 * optabs.c (can_reuse_operands_p): New function.
7402 (maybe_legitimize_operands): Try to reuse the results for
7405 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
7407 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
7408 Add {q} suffix to insn mnemonic.
7410 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
7412 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
7413 (msp430_warn_func_return): New.
7415 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
7417 * fold-const.c (tree_nonzero_bits): New function.
7418 * fold-const.h (tree_nonzero_bits): Likewise.
7419 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
7420 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
7422 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
7426 * varasm.c (assemble_alias): Check ifunc_resolver only on
7429 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
7432 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
7433 when memory input operand is handled.
7435 2018-05-24 Luis Machado <luis.machado@linaro.org>
7437 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
7439 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
7441 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
7443 * match.pd: Delay FMA folds until after vectorization.
7445 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
7448 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
7449 address check not strict.
7451 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
7453 * gimple-match.h (gimple_match_op): New class.
7454 (mprts_hook): Replace parameters with a gimple_match_op *.
7455 (maybe_build_generic_op): Likewise.
7456 (gimple_simplified_result_is_gimple_val): Replace parameters with
7457 a const gimple_match_op *.
7458 (gimple_simplify): Replace code_helper * and tree * parameters with
7459 a gimple_match_op * parameter.
7460 (gimple_resimplify1): Replace code_helper *, tree and tree *
7461 parameters with a gimple_match_op * parameter.
7462 (gimple_resimplify2): Likewise.
7463 (gimple_resimplify3): Likewise.
7464 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
7465 parameters with a gimple_match_op * parameter.
7466 * gimple-match-head.c (gimple_simplify): Change prototypes of
7467 auto-generated functions to take a gimple_match_op * instead of
7468 separate code_helper * and tree * parameters. Make the same
7469 change in the top-level overload and update calls to the
7470 gimple_resimplify routines. Update calls to the auto-generated
7471 functions and to maybe_push_res_to_seq in the publicly-facing
7472 operation-specific gimple_simplify overloads.
7473 (gimple_match_op::MAX_NUM_OPS): Define.
7474 (gimple_resimplify1): Replace rcode and ops with a single res_op
7475 parameter. Update call to gimple_simplify.
7476 (gimple_resimplify2): Likewise.
7477 (gimple_resimplify3): Likewise.
7478 (mprts_hook): Replace parameters with a gimple_match_op *.
7479 (maybe_build_generic_op): Likewise.
7480 (build_call_internal): Replace type, nargs and ops with
7481 a gimple_match_op *.
7482 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
7483 with a single gimple_match_op *. Update calls to mprts_hook,
7484 build_call_internal and gimple_simplified_result_is_gimple_val.
7485 Factor out code that is common to the tree_code and combined_fn cases.
7486 * genmatch.c (expr::gen_transform): Replace tem_code and
7487 tem_ops with a gimple_match_op called tem_op. Update calls
7488 to the gimple_resimplify functions and maybe_push_res_to_seq.
7489 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
7490 res_ops. Update call to the gimple_resimplify functions.
7491 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
7492 (decision_tree::gen): Make the functions take a gimple_match_op *
7493 called res_op instead of separate res_code and res_ops parameters.
7494 Update call accordingly.
7495 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
7496 and ops with a single res_op parameter. Update calls to
7497 maybe_build_generic_op and maybe_push_res_to_seq.
7498 (fold_stmt_1): Update calls to gimple_simplify and
7499 replace_stmt_with_simplification.
7500 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
7501 and gimple_simplified_result_is_gimple_val.
7502 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
7504 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
7505 with a gimple_match_op *.
7506 (vn_nary_build_or_lookup): Likewise. Update call to
7507 vn_nary_build_or_lookup_1.
7508 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
7509 gimple_match_op *. Update calls to the gimple_resimplify routines
7510 and to gimple_simplified_result_is_gimple_val.
7511 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
7512 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
7513 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
7514 (visit_nary_op): Likewise.
7515 (visit_reference_op_load): Likewise.
7517 2018-05-23 Luis Machado <luis.machado@linaro.org>
7519 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
7520 modifier for printing the step amount.
7522 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
7525 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
7528 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
7530 * doc/sourcebuild.texi (Endianness): New subsubsection.
7532 2018-05-23 Luis Machado <luis.machado@linaro.org>
7534 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7535 <prefetch_dynamic_strides>: New const bool field.
7536 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7537 prefetch_dynamic_strides.
7538 (exynosm1_prefetch_tune): Likewise.
7539 (thunderxt88_prefetch_tune): Likewise.
7540 (thunderx_prefetch_tune): Likewise.
7541 (thunderx2t99_prefetch_tune): Likewise.
7542 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
7544 (aarch64_override_options_internal): Update to set
7545 PARAM_PREFETCH_DYNAMIC_STRIDES.
7546 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
7547 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
7548 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
7549 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
7550 prefetch-dynamic-strides setting.
7552 2018-05-23 Luis Machado <luis.machado@linaro.org>
7554 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7555 <minimum_stride>: New const int field.
7556 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7557 minimum_stride field defaulting to -1.
7558 (exynosm1_prefetch_tune): Likewise.
7559 (thunderxt88_prefetch_tune): Likewise.
7560 (thunderx_prefetch_tune): Likewise.
7561 (thunderx2t99_prefetch_tune): Likewise.
7562 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
7563 <default_opt_level>: Set to 3.
7564 (aarch64_override_options_internal): Update to set
7565 PARAM_PREFETCH_MINIMUM_STRIDE.
7566 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
7567 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
7568 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
7569 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
7570 stride is constant and is below the minimum stride threshold.
7572 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7574 * config/arm/arm-cpus.in (mode26): Delete.
7575 (armv4): Delete mode26 reference.
7576 * config/arm/arm.c (arm_configure_build_target): Delete use of
7579 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
7581 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
7583 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
7584 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
7586 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
7587 (floatunsdidf2): Ditto.
7589 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
7591 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
7592 (fixuns_trunc<mode>si2_avx512f): Ditto.
7593 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
7594 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
7595 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
7597 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
7599 PR rtl-optimization/79985
7600 * df-scan.c (df_insn_refs_collect): Remove special case for
7601 global registers and asm statements.
7603 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
7605 * extend.texi (Global Register Variables): Rewrite the bullet list.
7606 Note that the register is available for allocation. Note that access
7607 via inline asm must use constraints. Add note about async-signal
7608 handlers. Remove paragraph about automagic register selection.
7610 2018-05-23 Richard Biener <rguenther@suse.de>
7612 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
7613 of fixed offset from memset VN.
7615 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
7617 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
7619 (alloc_cand_and_find_basis): Initialize first_interp field.
7620 (slsr_process_mul): Modify first_interp field.
7621 (slsr_process_add): Likewise.
7622 (slsr_process_cast): Modify first_interp field for each new
7624 (slsr_process_copy): Likewise.
7625 (dump_candidate): Dump first_interp field.
7626 (replace_mult_candidate): Process all interpretations, not just
7628 (replace_rhs_if_not_dup): Likewise.
7629 (replace_one_candidate): Likewise.
7631 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
7633 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
7635 (aarch64_needs_frame_chain): New function.
7636 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
7638 2018-05-23 Sudakshina Das <sudi.das@arm.com>
7641 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
7642 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
7643 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
7644 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
7645 as true for strict-align.
7646 (aarch64_can_inline_p): Perform checks even when callee has no
7647 attributes to check for strict alignment.
7648 * doc/extend.texi (AArch64 Function Attributes): Document
7650 * doc/invoke.texi: (AArch64 Options): Likewise.
7652 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
7654 PR tree-optimization/85853
7655 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
7656 the handling of the root of the node to...
7657 (vect_slp_analyze_node_operations_1): ...this new function,
7658 and run the whole thing with the child nodes' def types
7659 set according to their SLP node's def type.
7661 2018-05-23 Richard Biener <rguenther@suse.de>
7664 * tree-data-ref.c (create_runtime_alias_checks): Defer
7665 and ignore overflow warnings.
7667 2018-05-23 Yury Gribov <tetra2005@gmail.com>
7669 PR tree-optimization/85822
7670 * tree-vrp.c (is_masked_range_test): Fix handling of negative
7673 2018-05-23 Richard Biener <rguenther@suse.de>
7675 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
7676 memset constants via native_interpret_expr.
7678 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
7681 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
7683 (cgraph_node::create_alias): Likewise.
7684 (cgraph_node::get_availability): Check ifunc_resolver instead
7685 of looking up ifunc attribute.
7686 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
7687 * varasm.c (do_assemble_alias): Likewise.
7688 (assemble_alias): Likewise.
7689 (default_binds_local_p_3): Likewise.
7690 * cgraph.h (cgraph_node): Add ifunc_resolver.
7691 (cgraph_node::only_called_directly_or_aliased_p): Return false
7693 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
7695 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
7696 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
7697 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
7698 instead of looking up ifunc attribute.
7700 2018-05-22 Luis Machado <luis.machado@linaro.org>
7702 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
7704 2018-05-22 Martin Sebor <msebor@redhat.com>
7707 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
7708 only when expasion succeeds.
7709 (expand_builtin_strcmp): Same.
7710 (expand_builtin_strncmp): Same.
7712 2018-05-22 Martin Sebor <msebor@redhat.com>
7714 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
7716 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
7717 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7719 * config/aarch64/aarch64-ldpstp.md: Replace uses of
7720 aarch64_mem_pair_operand with memory_operand and delete operand swapping
7722 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
7723 Add check for legitimate_address.
7724 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
7725 (aarch64_swap_ldrstr_operands): New.
7726 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
7729 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
7730 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7732 * config/aarch64/aarch64.md: New patterns to generate stp
7734 (store_pair_sw, store_pair_dw): New patterns to generate stp for
7735 single words and double words.
7736 (load_pair_sw, load_pair_dw): Likewise.
7737 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
7739 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
7741 * config/aarch64/aarch64-ldpstp.md: Modify peephole
7742 for different mode ldpstp and add peephole for merged zero stores.
7744 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
7746 (aarch64_gen_store_pair): Rename calls to match new patterns.
7747 (aarch64_gen_load_pair): Rename calls to match new patterns.
7748 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
7749 (load_pair<DREG:mode><DREG2:mode>): ... This.
7750 (store_pair<mode>): Rename to...
7751 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
7752 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
7754 (V_INT_EQUIV): Handle SImode.
7755 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
7758 2018-05-22 Martin Sebor <msebor@redhat.com>
7761 * calls.c (maybe_warn_nonstring_arg): Use string length to set
7762 or ajust the presumed bound on an operation to avoid unnecessary
7765 2018-05-22 Martin Sebor <msebor@redhat.com>
7767 PR tree-optimization/85826
7768 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
7769 assuming that a DECL necesarily has a constant size.
7771 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
7774 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
7776 2018-05-22 Richard Biener <rguenther@suse.de>
7778 PR tree-optimization/85834
7779 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
7780 non-constant and non-zero memset arguments.
7782 2018-05-22 Martin Liska <mliska@suse.cz>
7785 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
7787 2018-05-22 Richard Biener <rguenther@suse.de>
7789 PR tree-optimization/85863
7790 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
7791 comparisons when vectype is specified.
7792 (vectorizable_condition): Do not specify vectype for
7793 vect_is_simple_cond when SLP vectorizing.
7795 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
7798 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
7799 define __ibm128 as long double.
7800 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
7801 as a distinct type when IEEE 128-bit support is enabled.
7802 (init_float128_ieee): Fix up conversions between IFmode and IEEE
7803 128-bit types to use the correct functions.
7804 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
7805 convert between 128-bit floating point types that have different
7806 modes but the same representation, instead of using gen_lowpart to
7808 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
7810 (IFKF_reg): New attributes to give the register constraints for
7812 (extend<mode>tf2_internal): New insns to mark an explicit
7813 conversion between 128-bit floating point types that have a
7814 different mode but share the same representation.
7816 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
7818 PR tree-optimization/85814
7819 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
7820 a null return from get_strinfo when unsharing the next
7821 strinfo in the chain.
7823 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
7826 * varasm.c (weak_finish): Clean up weak_decls.
7828 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7830 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
7831 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
7832 UNSPEC_UADALP values.
7833 * config/aarch64/iterators.md (ABAL): New int iterator.
7836 (sur): Add mappings for the above.
7837 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
7839 (aarch64_<sur>abal<mode>_4): Likewise.
7840 (aarch64_<sur>adalp<mode>_3): Likewise.
7841 (<sur>sadv16qi): New define_expand.
7843 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
7845 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
7846 (*movdf_internal): Ditto.
7847 (*rcpsf2_sse): Ditto.
7848 (*rsqrtsf2_sse): Ditto.
7849 (*sqrt<mode>2_sse): Ditto.
7851 2018-05-21 Tamar Christina <tamar.christina@arm.com>
7853 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
7855 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
7856 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
7857 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
7858 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
7860 * config/aarch64/arm_neon.h: Likewise.
7861 * config/aarch64/iterators.md (VQ_I): New.
7863 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
7865 * config.gcc: Add arc/t-multilib-linux to tmake_file for
7867 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
7870 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
7872 * config/nds32/constraints.md (S): New constraint.
7873 * config/nds32/nds32.md (call_internal): Use constraint S.
7874 (call_value_internal): Likewise.
7875 (sibcall_internal): Likewise.
7876 (sibcall_value_internal): Likewise.
7878 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
7879 Chung-Ju Wu <jasonwucj@gmail.com>
7881 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
7884 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
7885 Chung-Ju Wu <jasonwucj@gmail.com>
7887 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
7888 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
7889 (nds32_rtx_costs_impl): Simplify.
7890 (nds32_address_cost_impl): Simplify.
7891 (nds32_init_rtx_costs): New function.
7892 (nds32_rtx_costs_speed_prefer): Likewise.
7893 (nds32_rtx_costs_size_prefer): Likewise.
7894 (nds32_address_cost_speed_prefer): Likewise.
7895 (nds32_address_cost_speed_fwprop): Likewise.
7896 (nds32_address_cost_size_prefer): Likewise.
7897 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
7898 * config/nds32/nds32.c (nds32_option_override): Use
7899 nds32_init_rtx_costs function.
7901 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
7903 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
7904 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
7905 (TARGET_PIPELINE_N8): Likewise.
7906 (TARGET_PIPELINE_N10): Likewise.
7907 (TARGET_PIPELINE_N13): Likewise.
7908 (TARGET_PIPELINE_GRAYWOLF): Likewise.
7910 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
7912 * config/nds32/nds32-fpu.md: Update copyright year.
7914 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
7916 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
7918 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
7920 * config/nds32/nds32.c
7921 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
7922 * config/nds32/nds32.opt (minline-asm-r15): New option.
7924 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
7926 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
7928 * config/nds32/nds32.md (abssi2): New pattern.
7930 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
7932 * config/i386/i386.md (rex64namesuffix): New mode attribute.
7933 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
7934 Merge insn pattern from sse_cvtsi2ss<round_name> and
7935 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
7936 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
7937 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
7938 using SWI48 mode iterator.
7939 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
7940 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
7941 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
7942 pattern from sse_cvttss2si<round_saeonly_name>
7943 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
7944 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
7945 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
7946 using SWI48 mode iterator.
7947 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
7948 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
7949 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
7950 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
7951 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
7952 using SWI48 mode iterator.
7953 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
7954 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
7955 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
7956 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
7957 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
7958 SWI48 mode iterator.
7959 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
7960 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
7961 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
7962 pattern from sse_cvttsd2si<round_saeonly_name>
7963 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
7965 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
7967 * config/nds32/nds32-md-auxiliary.c
7968 (nds32_valid_smw_lwm_base_p): Refine.
7969 (nds32_output_smw_single_word): Refine.
7970 (nds32_output_smw_double_word): New.
7971 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
7973 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
7975 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
7976 (nds32_output_stack_pop): Refine.
7977 (nds32_expand_unaligned_load): Refine.
7978 (nds32_expand_unaligned_store): Refine.
7980 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7981 Chung-Ju Wu <jasonwucj@gmail.com>
7983 * config/nds32/constants.md: Add TP_REGNUM constant.
7984 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
7985 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
7987 * config/nds32/nds32-doubleword.md: Consider flag_pic.
7988 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
7989 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
7990 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
7991 and PIC code generation.
7992 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
7994 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
7996 * config/nds32/nds32.md: Support TLS and PIC.
7997 * config/nds32/nds32.c: Support TLS and PIC.
7998 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
7999 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
8002 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
8004 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
8005 mode with E_ prefix.
8007 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
8008 Chung-Ju Wu <jasonwucj@gmail.com>
8010 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
8011 * config/nds32/nds32-md-auxiliary.c
8012 (symbolic_reference_mentioned_p): New.
8013 (nds32_legitimize_ict_address): New.
8014 (nds32_expand_ict_move): New.
8015 (nds32_indirect_call_referenced_p): New.
8016 (nds32_symbol_binds_local_p): Delete.
8017 (nds32_long_call_p): Modify.
8018 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
8019 * config/nds32/nds32-protos.h
8020 (symbolic_reference_mentioned_p): Declare.
8021 (nds32_legitimize_ict_address): Declare.
8022 (nds32_expand_ict_move): Declare.
8023 (nds32_indirect_call_referenced_p): Declare.
8024 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
8025 (nds32_relax_group): Use nds32_ict_const_p as condition.
8026 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
8027 (nds32_asm_file_start): Output ict_model directive in asm code.
8028 (nds32_legitimate_address_p): Consider indirect call.
8029 (nds32_print_operand): Consider indirect call.
8030 (nds32_print_operand_address): Consider indirect call.
8031 (nds32_insert_attributes): Handle "indirect_call" attribute.
8032 (TARGET_LEGITIMATE_ADDRESS_P): Define.
8033 (TARGET_LEGITIMATE_CONSTANT_P): Define.
8034 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
8035 (TARGET_DELEGITIMIZE_ADDRESS): Define.
8036 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
8037 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
8038 (TARGET_ICT_MODEL_SMALL): Define.
8039 (TARGET_ICT_MODEL_LARGE): Define.
8040 * config/nds32/nds32.md (movsi): Consider ict model.
8041 (call, call_value): Consider ict model.
8042 (sibcall, sibcall_value): Consider ict model.
8043 * config/nds32/nds32.opt (mict-model): New option.
8044 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
8047 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
8048 Monk Chiang <sh.chiang04@gmail.com>
8049 Jim Wilson <jimw@sifive.com>
8051 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
8052 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
8053 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
8054 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
8055 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
8056 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
8057 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
8058 compute save_libcall_adjustment properly.
8059 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
8060 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
8061 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
8062 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
8063 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
8064 (ABI_SPEC): Handle mabi=ilp32e.
8065 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
8066 (RVE): Add RVE mask.
8067 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
8068 <-march>: Add rv32e as an example.
8070 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
8073 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
8074 (intra_create_variable_infos): Handle C++ constructors.
8076 2018-05-18 Martin Liska <mliska@suse.cz>
8078 * passes.def: Remove a redundant pass.
8080 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
8083 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
8085 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8087 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
8089 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
8091 (armv2, armv2a, armv3, armv3m): Delete architectures.
8092 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
8093 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
8094 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
8096 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
8097 (*mulsidi3adddi): Likewise.
8098 (mulsidi3): Likewise.
8099 (*mulsidi3_nov6): Likewise.
8100 (umulsidi3): Likewise.
8101 (umulsidi3_nov6): Likewise.
8102 (umaddsidi4): Likewise.
8103 (*umulsidi3adddi): Likewise.
8104 (smulsi3_highpart): Likewise.
8105 (*smulsi3_highpart_nov6): Likewise.
8106 (umulsi3_highpart): Likewise.
8107 (*umulsi3_highpart_nov6): Likewise.
8108 * config/arm/arm.h (arm_arch3m): Delete.
8109 * config/arm/arm.c (arm_arch3m): Delete.
8110 (arm_option_override_internal): Update armv3-related comment.
8111 (arm_configure_build_target): Delete use of isa_bit_mode32.
8112 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
8113 (arm_rtx_costs_internal): Delete check of arm_arch3m.
8114 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
8116 (mulusa3): Likewise.
8117 * config/arm/arm-protos.h (arm_arch3m): Delete.
8118 * config/arm/arm-tables.opt: Regenerate.
8119 * config/arm/arm-tune.md: Likewise.
8120 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
8121 deleted architectures.
8123 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8125 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
8126 (armv5t, armv5te): New features.
8127 (ARMv5, ARMv5e): Delete fgroups.
8128 (ARMv5t, ARMv5te): Adjust for above changes.
8130 (armv5, armv5e): Delete arches.
8131 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
8133 (*call_reg_arm): Likewise.
8134 (*call_value_reg_armv5): Likewise.
8135 (*call_value_reg_arm): Likewise.
8136 (*call_symbol): Likewise.
8137 (*call_value_symbol): Likewise.
8138 (*sibcall_insn): Likewise.
8139 (*sibcall_value_insn): Likewise.
8141 (prefetch): Likewise.
8142 (define_split and define_peephole2 dependent on arm_arch5):
8144 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
8146 (TARGET_ARM_QBIT): Likewise.
8147 (TARGET_DSP_MULTIPLY): Likewise.
8148 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
8149 (arm_arch5, arm_arch5e): Delete.
8150 (arm_arch5t, arm_arch5te): Declare.
8151 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
8152 (arm_arch5t): Declare.
8153 (arm_option_reconfigure_globals): Update for the above.
8154 (arm_options_perform_arch_sanity_checks): Update comment, replace
8155 use of arm_arch5 with arm_arch5t.
8156 (use_return_insn): Likewise.
8157 (arm_emit_call_insn): Likewise.
8158 (output_return_instruction): Likewise.
8159 (arm_final_prescan_insn): Likewise.
8160 (arm_coproc_builtin_available): Likewise.
8161 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
8162 arm_arch5e with arm_arch5t and arm_arch5te.
8163 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
8164 (arm_arch5t, arm_arch5te): Declare.
8165 * config/arm/arm-tables.opt: Regenerate.
8166 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
8167 * config/arm/t-multilib: Likewise.
8168 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
8169 instead of arm_arch5.
8170 (*call_reg_thumb1): Likewise.
8171 (*call_value_reg_thumb1_v5): Likewise.
8172 (*call_value_reg_thumb1): Likewise.
8173 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
8175 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
8177 2018-05-18 Martin Liska <mliska@suse.cz>
8179 PR gcov-profile/84846
8180 * doc/gcov.texi: Document -t option of gcov tool.
8182 2018-05-18 Martin Liska <mliska@suse.cz>
8184 PR gcov-profile/84846
8185 * gcov.c (print_usage): Add new -t option.
8186 (process_args): Handle the option.
8187 (generate_results): Use stdout as output when requested by
8190 2018-05-18 Martin Liska <mliska@suse.cz>
8192 PR gcov-profile/84846
8193 * coverage.c (coverage_init): Write PWD to .gcno file.
8194 * doc/gcov.texi: Document how working directory is printed.
8195 * gcov-dump.c (dump_gcov_file): Print PWD.
8196 * gcov.c (output_intermediate_file): Likewise.
8197 (read_graph_file): Read PWD string.
8198 (output_lines): Print PWD.
8200 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8203 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
8204 for retval and return false if all args to phi are zero.
8206 2018-05-18 Richard Biener <rguenther@suse.de>
8208 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
8210 (evrp_dom_walker::before_dom_children): Call it.
8212 2018-05-18 Richard Biener <rguenther@suse.de>
8214 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
8215 results when processing array refs with variable index.
8217 2018-05-18 Toon Moene <toon@moene.org>
8219 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
8220 directly after that of -floop-interchange. Indicate that both
8221 options are enabled by default when specifying -O3.
8223 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8225 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
8226 iterator. Delete separate integer-mode vec_set<mode> expander.
8227 (aarch64_simd_vec_setv2di): Delete.
8228 (vec_setv2di): Delete.
8229 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
8230 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
8231 the "w, r" alternative.
8233 2018-05-18 Martin Liska <mliska@suse.cz>
8235 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
8236 * tree-pass.h (make_pass_lower_switch_O0): New function.
8237 * tree-switch-conversion.c (node_has_low_bound): Remove.
8238 (node_has_high_bound): Likewise.
8239 (node_is_bounded): Likewise.
8240 (class pass_lower_switch): Make it a template type and create
8242 (pass_lower_switch::execute): Add template argument.
8243 (make_pass_lower_switch): New function.
8244 (make_pass_lower_switch_O0): New function.
8245 (do_jump_if_equal): Remove.
8246 (emit_case_nodes): Simplify to just handle all 3 cases and leave
8247 all the hard work to tree optimization passes.
8249 2018-05-18 Martin Liska <mliska@suse.cz>
8251 * dbgcnt.c (limit_low): Renamed from limit.
8252 (limit_high): New variable.
8253 (dbg_cnt_is_enabled): Check for upper limit.
8254 (dbg_cnt): Adjust dumping.
8255 (dbg_cnt_set_limit_by_index): Add new argument for high
8257 (dbg_cnt_set_limit_by_name): Likewise.
8258 (dbg_cnt_process_single_pair): Parse new format.
8259 (dbg_cnt_process_opt): Use strtok.
8260 (dbg_cnt_list_all_counters): Remove 'value' and add
8262 * doc/invoke.texi: Document changes.
8264 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
8266 * doc/sourcebuild.texi (scalar_all_fma): Document.
8267 * tree.def (FMA_EXPR): Delete.
8268 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
8269 * internal-fn.c (ternary_direct): New macro.
8270 (expand_ternary_optab_fn): Likewise.
8271 (direct_ternary_optab_supported_p): Likewise.
8272 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
8273 * builtins.c (fold_builtin_fma): Delete.
8274 (fold_builtin_3): Don't call it.
8275 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
8276 * expr.c (expand_expr_real_2): Likewise.
8277 * fold-const.c (operand_equal_p): Likewise.
8278 (fold_ternary_loc): Likewise.
8279 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
8280 * gimple.c (DEFTREECODE): Likewise.
8281 * gimplify.c (gimplify_expr): Likewise.
8282 * optabs-tree.c (optab_for_tree_code): Likewise.
8283 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
8284 * tree-eh.c (operation_could_trap_p): Likewise.
8285 (stmt_could_throw_1_p): Likewise.
8286 * tree-inline.c (estimate_operator_cost): Likewise.
8287 * tree-pretty-print.c (dump_generic_node): Likewise.
8288 (op_code_prio): Likewise.
8289 * tree-ssa-loop-im.c (stmt_cost): Likewise.
8290 * tree-ssa-operands.c (get_expr_operands): Likewise.
8291 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
8292 * fold-const-call.h (fold_fma): Delete.
8293 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
8294 CFN_FNMA and CFN_FNMS.
8296 * genmatch.c (combined_fn): New enum.
8297 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
8298 (commutative_op): New function.
8299 (commutate): Use it. Handle more than 2 operands.
8300 (dt_operand::gen_gimple_expr): Use commutative_op.
8301 (parser::parse_expr): Allow :c to be used with non-binary
8302 operators if the commutative operand is known.
8303 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
8304 CFN_FMS, CFN_FNMA and CFN_FNMS.
8305 (backprop::process_assign_use): Remove FMA_EXPR handling.
8306 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
8307 (gen_hsa_fma): New function.
8308 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
8309 IFN_FNMA and IFN_FNMS.
8310 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
8311 * gimple-fold.h (follow_all_ssa_edges): Declare.
8312 * gimple-fold.c (follow_all_ssa_edges): New function.
8313 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
8314 gimple_build interface and use follow_all_ssa_edges to fold the result.
8315 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
8316 instead of checking for optabs directly.
8317 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
8318 rather than FMA_EXPRs.
8319 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
8320 call to IFN_FMA instead of an FMA_EXPR.
8322 2018-05-17 Jim Wilson <jimw@sifive.com>
8324 * expr.c (do_tablejump): When converting index to Pmode, if we have a
8325 sign extended promoted subreg, and the range does not have the sign bit
8326 set, then do a sign extend.
8328 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
8329 test, check for sign extended subreg and/or constant operands, and
8330 do a sign extend in that case.
8332 2018-05-17 Steve Ellcey <sellcey@cavium.com>
8334 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
8335 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
8337 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
8338 Change logics_shift_reg to logics_shift_imm.
8339 (thunderx2t99_fp_loadpair_basic): Delete.
8340 (thunderx2t99_fp_storepair_basic): Delete.
8341 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
8342 (thunderx2t99_asimd_polynomial): Delete.
8343 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
8344 and neon_fp_mul_d_scalar_q.
8345 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
8346 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
8347 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
8348 (thunderx2t99_asimd_lut): Add missing tbl types.
8349 (thunderx2t99_asimd_ext): Delete.
8350 (thunderx2t99_asimd_load1_1_mult): Delete.
8351 (thunderx2t99_asimd_load1_2_mult): Delete.
8352 (thunderx2t99_asimd_load1_ldp): New.
8353 (thunderx2t99_asimd_load1): New.
8354 (thunderx2t99_asimd_load2): Add missing *load2* types.
8355 (thunderx2t99_asimd_load3): New.
8356 (thunderx2t99_asimd_load4): New.
8357 (thunderx2t99_asimd_store1_1_mult): Delete.
8358 (thunderx2t99_asimd_store1_2_mult): Delete.
8359 (thunderx2t99_asimd_store2_mult): Delete.
8360 (thunderx2t99_asimd_store2_onelane): Delete.
8361 (thunderx2t99_asimd_store_stp): New.
8362 (thunderx2t99_asimd_store1): New.
8363 (thunderx2t99_asimd_store2): New.
8364 (thunderx2t99_asimd_store3): New.
8365 (thunderx2t99_asimd_store4): New.
8367 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
8369 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
8370 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
8372 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
8373 Segher Boessenkool <segher@kernel.crashing.org>
8376 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
8379 2018-05-17 Richard Biener <rguenther@suse.de>
8381 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
8382 for pruning loop and prune defs feeding only already visited PHIs.
8384 2018-05-17 Richard Biener <rguenther@suse.de>
8386 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
8388 2018-05-17 Bin Cheng <bin.cheng@arm.com>
8389 Richard Biener <rguenther@suse.de>
8391 PR tree-optimization/85793
8392 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
8393 for VMAT_ELEMENTWISE.
8395 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
8397 * internal-fn.h (lookup_internal_fn): Declare
8398 * internal-fn.c (lookup_internal_fn): New function.
8399 * gimple.c (gimple_build_call_from_tree): Handle calls to
8401 * gimple-pretty-print.c (dump_gimple_call): Print "." before
8402 internal function names.
8403 * tree-pretty-print.c (dump_generic_node): Likewise.
8404 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
8406 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
8408 * gimple-fold.h (gimple_build): Make the function forms take
8409 combined_fn rather than built_in_function.
8410 (gimple_simplify): Likewise.
8411 * gimple-match-head.c (gimple_simplify): Likewise.
8412 * gimple-fold.c (gimple_build): Likewise.
8413 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
8414 rather than gimple_build_call_internal.
8415 (get_initial_defs_for_reduction): Likewise.
8416 (vect_create_epilog_for_reduction): Likewise.
8417 (vectorizable_live_operation): Likewise.
8419 2018-05-17 Martin Liska <mliska@suse.cz>
8421 * gimple-ssa-sprintf.c (format_directive): Do not use
8422 space in between 'G_' and '('.
8424 2018-05-17 Jakub Jelinek <jakub@redhat.com>
8427 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
8428 even if the mask is not all ones.
8431 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
8433 (ix86_gimple_fold_builtin): Likewise.
8436 * config/i386/i386.c: Include tree-vector-builder.h.
8437 (ix86_vector_shift_count): New function.
8438 (ix86_fold_builtin): Fold shift builtins by scalar count.
8439 (ix86_gimple_fold_builtin): Likewise.
8441 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
8442 _mm512_setzero): New intrinsics.
8444 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
8445 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8447 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
8448 code generation for cases where splatting a value is not useful.
8449 * simplify-rtx.c (simplify_ternary_operation): Simplify
8450 vec_merge across a vec_duplicate and a paradoxical subreg forming
8451 a vector mode to a vec_concat.
8453 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
8455 * config.gcc: Support "goldmont-plus".
8456 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
8458 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
8459 PROCESSOR_GOLDMONT_PLUS.
8460 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
8461 (processor_target_table): Add "goldmont-plus".
8462 (PTA_GOLDMONT_PLUS): Define.
8463 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
8464 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
8465 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
8466 (fold_builtin_cpu): Add "goldmont-plus".
8467 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
8468 (ix86_option_override_internal): Add "goldmont-plus".
8469 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
8470 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
8471 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
8472 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
8474 2018-05-17 Richard Biener <rguenther@suse.de>
8476 PR tree-optimization/85757
8477 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
8478 remove defs that only feed that PHI from further processing.
8480 2018-05-16 Jim Wilson <jimw@sifive.com>
8482 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
8484 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
8486 2018-05-16 Mark Wielaard <mark@klomp.org>
8488 * dwarf2out.c (count_index_strings): New function.
8489 (output_indirect_strings): Call count_index_strings and generate
8490 header for dwarf_version >= 5.
8492 2018-05-16 Mark Wielaard <mark@klomp.org>
8494 * dwarf2out.c (dwarf_FORM): New function.
8495 (set_indirect_string): Use dwarf_FORM.
8496 (reset_indirect_string): Likewise.
8497 (size_of_die): Likewise.
8498 (value_format): Likewise.
8499 (output_die): Likewise.
8500 (add_skeleton_AT_string): Likewise.
8501 (output_macinfo_op): Likewise.
8502 (index_string): Likewise.
8503 (output_index_string_offset): Likewise.
8504 (output_index_string): Likewise.
8505 (count_index_strings): Likewise.
8507 2018-05-16 Carl Love <cel@us.ibm.com>
8509 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
8510 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
8512 2018-05-16 Martin Jambor <mjambor@suse.cz>
8514 * ipa-prop.c (ipa_free_all_edge_args): Remove.
8515 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
8517 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
8519 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
8520 (fnma<mode>4): Likewise.
8521 (fms<mode>4): Likewise.
8522 (fnms<mode>4): Likewise.
8523 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
8524 (aarch64_fnma<mode>4): Likewise.
8525 (aarch64_fms<mode>4): Likewise.
8526 (aarch64_fnms<mode>4): Likewise.
8527 (aarch64_fnmadd<mode>4): Likewise.
8529 2018-05-16 Jason Merrill <jason@redhat.com>
8531 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
8533 2018-05-16 Richard Biener <rguenther@suse.de>
8535 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
8536 (dump_stmt_cost): Declare.
8537 (add_stmt_cost): Dump cost we add.
8538 (add_stmt_costs): New function.
8539 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
8541 (vect_analyze_stmt): Adjust prototype.
8542 (vectorizable_condition): Likewise.
8543 (vectorizable_live_operation): Likewise.
8544 (vectorizable_reduction): Likewise.
8545 (vectorizable_induction): Likewise.
8546 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
8547 cost vector to pass to vectorizable_ and record afterwards.
8548 (vect_model_reduction_cost): Take cost vector argument and adjust.
8549 (vect_model_induction_cost): Likewise.
8550 (vectorizable_reduction): Likewise.
8551 (vectorizable_induction): Likewise.
8552 (vectorizable_live_operation): Likewise.
8553 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
8554 SLP_TREE_NUMBER_OF_VEC_STMTS.
8555 (vect_analyze_slp_cost_1): Remove.
8556 (vect_analyze_slp_cost): Likewise.
8557 (vect_slp_analyze_node_operations): Take visited args and
8558 a target cost vector. Avoid processing already visited stmt sets.
8559 (vect_slp_analyze_operations): Use a local cost vector to gather
8560 costs and register those of non-discarded instances.
8561 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
8562 (vect_schedule_slp_instance): Remove copying of
8563 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
8565 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
8566 adding cost. Record cost entry location.
8567 (vect_prologue_cost_for_slp_op): Function to compute cost of
8568 a constant or invariant generated for SLP vect in the prologue,
8569 split out from vect_analyze_slp_cost_1.
8570 (vect_model_simple_cost): Make static. Adjust for SLP costing.
8571 (vect_model_promotion_demotion_cost): Likewise.
8572 (vect_model_store_cost): Likewise, make static.
8573 (vect_model_load_cost): Likewise.
8574 (vectorizable_bswap): Add cost vector arg and adjust.
8575 (vectorizable_call): Likewise.
8576 (vectorizable_simd_clone_call): Likewise.
8577 (vectorizable_conversion): Likewise.
8578 (vectorizable_assignment): Likewise.
8579 (vectorizable_shift): Likewise.
8580 (vectorizable_operation): Likewise.
8581 (vectorizable_store): Likewise.
8582 (vectorizable_load): Likewise.
8583 (vectorizable_condition): Likewise.
8584 (vectorizable_comparison): Likewise.
8585 (can_vectorize_live_stmts): Likewise.
8586 (vect_analyze_stmt): Likewise.
8587 (vect_transform_stmt): Adjust calls to vectorizable_*.
8588 * tree-vectorizer.c: Include gimple-pretty-print.h.
8589 (dump_stmt_cost): New function.
8591 2018-05-16 Richard Biener <rguenther@suse.de>
8593 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
8594 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
8595 * tree-ssa-dse.c: Include tree-ssa-loop.h.
8596 (check_name): New callback.
8597 (dse_classify_store): Track cycles via a visited bitmap of PHI
8598 defs and simplify handling of in-loop and across loop dead stores
8599 and properly fail for loop-variant refs. Handle byte-tracking with
8600 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
8603 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
8605 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
8606 (vect_get_mask_type_for_stmt): Likewise.
8607 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
8609 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
8610 to determine the statement's vector type and the vector type that
8611 should be used for calculating nunits. Deal with cases in which
8612 the type has to be deferred.
8613 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
8614 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
8615 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
8616 (vect_determine_vf_for_stmt): New functions, split out from...
8617 (vect_determine_vectorization_factor): ...here.
8618 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
8619 (vect_get_mask_type_for_stmt): New functions, split out from
8620 vect_determine_vectorization_factor.
8622 2018-05-16 Richard Biener <rguenther@suse.de>
8624 * tree-cfg.c (verify_gimple_assign_ternary): Properly
8625 verify the [VEC_]COND_EXPR embedded comparison.
8627 2018-05-15 Martin Sebor <msebor@redhat.com>
8629 PR tree-optimization/85753
8630 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
8631 RECORD_TYPE in addition to ARRAY_TYPE.
8633 2018-05-15 Martin Sebor <msebor@redhat.com>
8636 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
8638 2018-05-15 Richard Biener <rguenther@suse.de>
8640 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
8641 add by_clobber_p one. Change algorithm to collect all defs
8642 representing uses we need to walk and try reducing them to
8643 a single one before failing.
8644 (dse_dom_walker::dse_optimize_stmt): Adjust.
8646 2018-05-13 Mark Wielaard <mark@klomp.org>
8648 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
8649 (size_of_loc_descr): Likewise.
8650 (output_loc_operands): Likewise.
8651 (output_loc_operands_raw): Likewise.
8652 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
8653 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
8654 (hash_loc_operands): Likewise.
8655 (compare_loc_operands): Likewise.
8657 2018-05-14 Mark Wielaard <mark@klomp.org>
8659 * dwarf2out.c (count_index_addrs): New function.
8660 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
8662 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8664 PR tree-optimization/83648
8665 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
8666 return value as malloc candidate.
8668 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8671 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
8672 param as true in call to suggest_attribute.
8674 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
8676 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
8677 -mreadonly-in-sdata.
8679 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8681 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
8683 (aarch64_crypto_aesd_fused): Likewise.
8685 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
8687 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
8688 (movsi_aarch64): Likewise.
8689 (load_pairsi): Likewise.
8690 (load_pairdi): Likewise.
8691 (store_pairsi): Likewise.
8692 (store_pairdi): Likewise.
8693 (load_pairsf): Likewise.
8694 (load_pairdf): Likewise.
8695 (store_pairsf): Likewise.
8696 (store_pairdf): Likewise.
8697 (zero_extend): Likewise.
8698 (trunc): Swap alternatives.
8699 (fcvt_target): Add '?' to prefer w over r.
8701 2018-05-14 Jakub Jelinek <jakub@redhat.com>
8704 * config/i386/i386.md: Disallow non-commutative arithmetics in
8705 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
8706 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
8707 in the peephole2 before it.
8709 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
8711 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
8712 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
8713 (ix86_handle_option): Handle -mcldemote.
8714 * config.gcc: New header.
8715 * config/i386/cldemoteintrin.h: New file.
8716 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
8717 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
8719 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
8720 OPTION_MASK_ISA_CLDEMOTE.
8721 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
8722 (ix86_valid_target_attribute_inner_p): Ditto.
8723 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
8724 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
8725 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
8726 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
8727 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
8729 * config/i386/i386.opt: Add -mcldemote.
8730 * config/i386/x86intrin.h: New header.
8731 * doc/invoke.texi: Add -mcldemote.
8733 2018-05-14 Richard Biener <rguenther@suse.de>
8735 * doc/match-and-simplify.texi: Adjust :s documentation.
8737 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
8739 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
8740 intended memcpy size.
8741 (REORDER_45): Likewise.
8743 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
8745 * sort.cc: New file.
8746 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
8747 * vec.c (qsort_chk): Use gcc_qsort.
8748 * Makefile.in (OBJS-libcommon): Add sort.o.
8749 (build/sort.o): New target. Use it...
8750 (BUILD_RTL): ... here, and...
8751 (build/gencfn-macros): ... here, and...
8752 (build/genmatch): ... here.
8754 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
8755 Chung-Ju Wu <jasonwucj@gmail.com>
8757 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
8758 * config/nds32/nds32-graywolf.md: New file.
8759 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
8760 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
8762 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
8763 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
8764 * config/nds32/nds32.md (pipeline_model): Add graywolf.
8765 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
8766 * config/nds32/pipelines.md: Include n15 settings.
8768 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
8769 Chung-Ju Wu <jasonwucj@gmail.com>
8771 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
8772 * config/nds32/nds32-n13.md: New file.
8773 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
8774 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
8776 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
8777 * config/nds32/nds32.md (pipeline_model): Add n13.
8778 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
8779 * config/nds32/pipelines.md: Include n13 settings.
8781 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
8782 Chung-Ju Wu <jasonwucj@gmail.com>
8784 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
8785 * config/nds32/nds32-n10.md: New file.
8786 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
8787 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
8789 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
8790 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
8791 * config/nds32/nds32.md (pipeline_model): Add n10.
8792 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
8793 * config/nds32/pipelines.md: Include n10 settings.
8795 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
8796 Kito Cheng <kito.cheng@gmail.com>
8797 Chung-Ju Wu <jasonwucj@gmail.com>
8799 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
8800 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
8801 Add enum values for DSP extension instructions.
8802 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
8804 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
8805 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
8807 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
8808 * config/nds32/nds32-dspext.md: New file for DSP implementation.
8809 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
8810 * config/nds32/nds32-intrinsic.md: Likewise.
8811 * config/nds32/nds32_intrinsic.h: Likewise.
8812 * config/nds32/nds32-md-auxiliary.c: Likewise.
8813 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
8814 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
8815 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
8816 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
8817 * config/nds32/nds32-protos.h: New declarations for DSP extension.
8818 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
8819 TYPE_DMAC in switch statement.
8820 * config/nds32/nds32.c: New checking and implementation for DSP
8821 extension instructions.
8822 * config/nds32/nds32.h: Likewise.
8823 * config/nds32/nds32.md: Likewise.
8824 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
8825 * config/nds32/predicates.md: Implement new predicates for DSP
8828 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
8830 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
8831 Reformat alternatives and attributes so it is easier to identify
8832 which constraints/attributes go with which instruction.
8833 (mov<mode>_hardfloat32, FMOVE64): Likewise.
8834 (mov<mode>_softfloat32, FMOVE64): Likewise.
8835 (mov<mode>_hardfloat64, FMOVE64): Likewise.
8836 (mov<mode>_softfloat64, FMOVE64): Likewise.
8838 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
8840 * doc/extend.texi (PowerPC Built-in Functions): Rename this
8842 (Basic PowerPC Built-in Functions): The new name of the
8843 subsection previously known as "PowerPC Built-in Functions".
8844 (Basic PowerPC Built-in Functions Available on all Configurations):
8846 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
8847 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
8848 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
8849 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
8851 2018-05-11 Martin Jambor <mjambor@suse.cz>
8854 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
8857 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
8860 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
8862 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
8864 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
8865 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
8866 (ix86_handle_option): Handle -mwaitpkg.
8867 * config.gcc: New header.
8868 * config/i386/cpuid.h (bit_WAITPKG): New bit.
8869 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
8870 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
8872 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
8873 OPTION_MASK_ISA_WAITPKG.
8874 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
8875 (ix86_option_override_internal): Add PTA_WAITPKG.
8876 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
8877 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
8878 IX86_BUILTIN_TPAUSE.
8879 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
8880 __builtin_ia32_umwait and __builtin_ia32_tpause.
8881 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
8882 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
8883 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
8884 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
8885 UNSPECV_TPAUSE): New.
8886 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
8887 * config/i386/i386.opt: Add -mwaitpkg.
8888 * config/i386/waitpkgintrin.h: New file.
8889 * config/i386/x86intrin.h: New header.
8890 * doc/invoke.texi: Add -mwaitpkg.
8892 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
8895 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
8897 (cortex-m0): Use armv6s-m isa.
8898 (cortex-m0plus): Likewise.
8899 (cortex-m1): Likewise.
8900 (cortex-m0.small-multiply): Likewise.
8901 (cortex-m0plus.small-multiply): Likewise.
8902 (cortex-m1.small-multiply): Likewise.
8904 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
8905 Jakub Jelinek <jakub@redhat.com>
8907 PR tree-optimization/85692
8908 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
8909 source permute as well.
8911 2018-05-11 Martin Liska <mliska@suse.cz>
8914 * doc/extend.texi: Document LLVM style format for no_sanitize
8917 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
8919 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
8920 mode_supports_vsx_dform_quad to mode_supports_dq_form.
8921 (mode_supports_vsx_dform_quad): Likewise.
8922 (mode_supports_vmx_dform): Move these functions to be next to the
8923 other mode_supports functions.
8924 (mode_supports_dq_form): Likewise.
8925 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
8926 mode_supports_dq_form.
8927 (reg_offset_addressing_ok_p): Likewise.
8928 (offsettable_ok_by_alignment): Likewise.
8929 (rs6000_legitimate_offset_address_p): Likewise.
8930 (legitimate_lo_sum_address_p): Likewise.
8931 (rs6000_legitimize_address): Likewise.
8932 (rs6000_legitimize_reload_address): Likewise.
8933 (rs6000_secondary_reload_inner): Likewise.
8934 (rs6000_preferred_reload_class): Likewise.
8935 (rs6000_output_move_128bit): Likewise.
8937 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
8939 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
8940 Generate SImode target register for null target.
8941 <case IX86_BUILTIN_XGETBV>: Ditto.
8942 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
8943 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
8945 2018-05-10 Carl Love <cel@us.ibm.com>
8947 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
8948 dcbtt and dcbtstt if operands[2] is 0.
8950 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
8953 * config/i386/sse.md (usadv64qi): New expander.
8955 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
8957 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
8958 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
8959 -maltivec=be support.
8960 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
8961 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
8962 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
8963 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
8964 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
8965 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
8966 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
8967 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
8968 altivec_vsumsws): Adjust.
8969 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
8970 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
8971 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
8973 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
8974 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
8975 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
8976 (altivec_lve<VI_char>x): Delete expand.
8977 (*altivec_lve<VI_char>x_internal): Rename to...
8978 (altivec_lve<VI_char>x): ... this.
8979 (altivec_lvxl_<mode>): Delete expand.
8980 (*altivec_lvxl_<mode>_internal): Rename to ...
8981 (altivec_lvxl_<mode>): ... this.
8982 (altivec_stvxl_<mode>): Delete expand.
8983 (*altivec_stvxl_<mode>_internal): Rename to ...
8984 (altivec_stvxl_<mode>): ... this.
8985 (altivec_stve<VI_char>x): Delete expand.
8986 (*altivec_stve<VI_char>x_internal): Rename to ...
8987 (altivec_stve<VI_char>x): ... this.
8988 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
8989 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
8990 reduc_plus_scal_<mode>): Adjust.
8991 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
8993 (rs6000_cpu_cpp_builtins): Adjust.
8994 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
8995 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
8996 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
8997 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
8998 -maltivec=be support.
8999 (rs6000_split_vec_extract_var): Adjust.
9000 (rs6000_split_v4si_init): Adjust.
9001 (swap_selector_for_mode): Delete.
9002 (altivec_expand_lvx_be, altivec_expand_stvx_be,
9003 altivec_expand_stvex_be): Delete.
9004 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
9005 -maltivec=be support.
9006 (rs6000_gimple_fold_builtin): Ditto.
9007 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
9009 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
9010 (TARGET_DIRECT_MOVE_64BIT): Adjust.
9011 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
9012 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
9013 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
9014 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
9015 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
9016 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
9017 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
9018 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
9019 anonymous split): Adjust.
9020 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
9021 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
9023 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
9025 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
9026 when --with-gxx-include-dir is also specified.
9027 * configure: Regenerate.
9029 2018-05-09 Jim Wilson <jimw@sifive.com>
9032 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
9033 * config/riscv/t-withmultilib: New.
9034 * config/riscv/withmultilib.h: New.
9035 * doc/install.texi: Document RISC-V --with-multilib-list support.
9037 2018-05-09 Richard Biener <rguenther@suse.de>
9039 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
9041 (vect_bb_vectorization_profitable_p): Adjust. Compute
9042 actual scalar cost using the cost vector and the add_stmt_cost
9045 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
9047 PR rtl-optimization/85645
9048 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
9049 in the REG_CFA_REGISTER note for LR, don't leave it empty.
9051 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
9053 PR rtl-optimization/85645
9054 * shrink-wrap.c (spread_components): Return a boolean saying if
9055 anything was changed.
9056 (try_shrink_wrapping_separate): Iterate spread_components until
9057 nothing changes anymore.
9059 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
9061 PR rtl-optimization/85645
9062 * regrename.c (build_def_use): Also kill the chains that include the
9063 destination of a REG_CFA_REGISTER note.
9065 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
9067 PR rtl-optimization/85645
9068 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
9069 insn that has a REG_CFA_REGISTER note.
9071 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
9073 * cfgexpand.c (expand_clobber): New function.
9074 (expand_gimple_stmt_1): Use it.
9075 * tree-vect-stmts.c (vect_clobber_variable): New function,
9077 (vectorizable_simd_clone_call): ...here.
9078 (vectorizable_store): Emit a clobber either side of an
9079 IFN_STORE_LANES sequence.
9080 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
9082 2018-05-09 Tom de Vries <tom@codesourcery.com>
9085 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
9086 (define_insn "trap_if_false"): Add exit after trap.
9088 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
9090 PR rtl-optimization/85638
9091 * bb-reorder.c: Include common/common-target.h.
9092 (create_forwarder_block): New function extracted from...
9093 (fix_up_crossing_landing_pad): ...here. Rename into...
9094 (dw2_fix_up_crossing_landing_pad): ...this.
9095 (sjlj_fix_up_crossing_landing_pad): New function.
9096 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
9097 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
9098 from both partitions and exit the loop after one iteration.
9100 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
9103 * doc/extend.texi (PowerPC Built-in Functions): Rename this
9105 (Basic PowerPC Built-in Functions): The new name of the
9106 subsection previously known as "PowerPC Built-in Functions".
9107 (Basic PowerPC Built-in Functions Available on all Configurations):
9109 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
9111 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
9113 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
9115 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
9118 2018-05-08 Jim Wilson <jimw@sifive.com>
9120 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
9121 (LD_EMUL_SUFFIX): New.
9122 (LINK_SPEC): Use it.
9124 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
9126 * doc/extend.texi (PowerPC Built-in Functions): Rename this
9128 (Basic PowerPC Built-in Functions): The new name of the
9129 subsection previously known as "PowerPC Built-in Functions".
9130 (Basic PowerPC Built-in Functions Available on all Configurations):
9132 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
9134 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
9136 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
9138 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
9141 2018-05-08 Jakub Jelinek <jakub@redhat.com>
9144 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
9145 after cmpelim optimization.
9147 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
9149 * config.gcc: Support "goldmont".
9150 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
9151 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
9153 * config/i386/i386.c (m_GOLDMONT): Define.
9154 (processor_target_table): Add "goldmont".
9155 (PTA_GOLDMONT): Define.
9156 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
9157 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
9158 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
9159 (fold_builtin_cpu): Add "goldmont".
9160 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
9161 (ix86_option_override_internal): Add "goldmont".
9162 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
9163 (processor_type): Add PROCESSOR_GOLDMONT.
9164 * config/i386/i386.md: Add CPU "glm".
9165 * config/i386/glm.md: New file.
9166 * config/i386/x86-tune.def: Add m_GOLDMONT.
9167 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
9169 2018-05-08 Jakub Jelinek <jakub@redhat.com>
9172 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
9174 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
9175 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
9176 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
9179 * config/i386/i386.c (ix86_fold_builtin): Handle
9180 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
9183 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
9184 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
9186 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
9189 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
9190 (check_arch): Likewise.
9191 (check_fpu): Return the result rather than printing it.
9192 (end arch): Fix operator precedence.
9193 (end cpu): Likewise.
9194 (END): Print the result from check_fpu.
9196 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
9197 Alan Hayward <alan.hayward@arm.com>
9198 David Sherwood <david.sherwood@arm.com>
9200 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
9201 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
9202 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
9203 (*fcmuo<mode>_and): New patterns.
9205 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
9207 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
9208 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
9209 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
9210 (cmp_op, sve_imm_con): New code attributes.
9211 (SVE_COND_INT_CMP, imm_con): Delete.
9212 (cmp_op): Remove above unspecs from int attribute.
9213 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
9215 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
9216 comparison-specific unspecs.
9217 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
9218 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
9219 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
9220 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
9221 (*vec_fcm<cmp_op><mode>): Rename to...
9222 (*fcm<cmp_op><mode>): ...this and adjust likewise.
9223 (*vec_fcmuo<mode>): Rename to...
9224 (*fcmuo<mode>): ...this and adjust likewise.
9225 (*pred_fcm<cmp_op><mode>): New pattern.
9226 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
9227 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
9229 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
9231 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
9232 (aarch64_emit_sve_predicated_cond): New function.
9233 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
9234 (aarch64_emit_unspec_cond_or): Replace with...
9235 (aarch64_emit_sve_or_conds): ...this new function. Use
9236 aarch64_emit_sve_ptrue_op for the individual comparisons and
9237 aarch64_emit_binop to OR them together.
9238 (aarch64_emit_inverted_unspec_cond): Replace with...
9239 (aarch64_emit_sve_inverted_cond): ...this new function. Use
9240 aarch64_emit_sve_ptrue_op for the comparison and
9241 aarch64_emit_unop to invert the result.
9242 (aarch64_expand_sve_vec_cmp_float): Update after the above
9243 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
9245 2018-05-07 Nathan Sidwell <nathan@acm.org>
9247 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
9248 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
9249 (Backwards Compatibility): Likewise.
9251 2018-05-07 Luis Machado <luis.machado@linaro.org>
9255 2018-05-07 Luis Machado <luis.machado@linaro.org>
9257 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
9258 <prefetch_dynamic_strides>: New const bool field.
9259 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
9260 prefetch_dynamic_strides.
9261 (exynosm1_prefetch_tune): Likewise.
9262 (thunderxt88_prefetch_tune): Likewise.
9263 (thunderx_prefetch_tune): Likewise.
9264 (thunderx2t99_prefetch_tune): Likewise.
9265 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
9267 (aarch64_override_options_internal): Update to set
9268 PARAM_PREFETCH_DYNAMIC_STRIDES.
9269 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
9270 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
9271 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
9272 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
9273 prefetch-dynamic-strides setting.
9275 2018-05-07 Luis Machado <luis.machado@linaro.org>
9277 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
9278 <minimum_stride>: New const int field.
9279 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
9280 minimum_stride field.
9281 (exynosm1_prefetch_tune): Likewise.
9282 (thunderxt88_prefetch_tune): Likewise.
9283 (thunderx_prefetch_tune): Likewise.
9284 (thunderx2t99_prefetch_tune): Likewise.
9285 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
9286 (aarch64_override_options_internal): Update to set
9287 PARAM_PREFETCH_MINIMUM_STRIDE.
9288 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
9289 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
9290 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
9291 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
9292 stride is constant and is below the minimum stride threshold.
9294 2018-05-07 Luis Machado <luis.machado@linaro.org>
9296 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
9299 2018-05-07 Luis Machado <luis.machado@linaro.org>
9301 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
9302 <prefetch_dynamic_strides>: New const bool field.
9303 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
9304 prefetch_dynamic_strides.
9305 (exynosm1_prefetch_tune): Likewise.
9306 (thunderxt88_prefetch_tune): Likewise.
9307 (thunderx_prefetch_tune): Likewise.
9308 (thunderx2t99_prefetch_tune): Likewise.
9309 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
9311 (aarch64_override_options_internal): Update to set
9312 PARAM_PREFETCH_DYNAMIC_STRIDES.
9313 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
9314 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
9315 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
9316 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
9317 prefetch-dynamic-strides setting.
9319 2018-05-07 Luis Machado <luis.machado@linaro.org>
9321 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
9322 <minimum_stride>: New const int field.
9323 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
9324 minimum_stride field.
9325 (exynosm1_prefetch_tune): Likewise.
9326 (thunderxt88_prefetch_tune): Likewise.
9327 (thunderx_prefetch_tune): Likewise.
9328 (thunderx2t99_prefetch_tune): Likewise.
9329 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
9330 (aarch64_override_options_internal): Update to set
9331 PARAM_PREFETCH_MINIMUM_STRIDE.
9332 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
9333 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
9334 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
9335 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
9336 stride is constant and is below the minimum stride threshold.
9338 2018-05-06 Jakub Jelinek <jakub@redhat.com>
9341 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
9342 the type is addressable. Don't force op into register if it has
9345 2018-05-05 Roland McGrath <mcgrathr@google.com>
9348 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
9349 any section for which we don't know a specific type it should have,
9350 regardless of name. Previously this was done only for the exact
9351 names ".init_array", ".fini_array", and ".preinit_array".
9352 (default_elf_asm_named_section): Add comment about
9353 relationship with default_section_type_flags and SECTION_NOTYPE.
9354 (get_section): Don't consider it a type conflict if one side has
9355 SECTION_NOTYPE and the other doesn't, as long as neither has the
9356 SECTION_BSS et al used in the default_section_type_flags logic.
9358 2018-05-05 Tom de Vries <tom@codesourcery.com>
9361 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
9362 (workaround_barsyncs): New function.
9363 (nvptx_reorg): Use workaround_barsyncs.
9364 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
9365 (define_expand "nvptx_membar_cta"): New define_expand.
9366 (define_insn "*nvptx_membar_cta"): New insn.
9368 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
9370 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
9371 To improve optimization opportunities.
9372 * builtin-types.def: The new needed builtin types for the above.
9374 2018-05-04 Richard Biener <rguenther@suse.de>
9376 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
9377 * gimple-ssa-store-merging.c
9378 (imm_store_chain_info::output_merged_store): Remove redundant create,
9379 release split_store vector contents on failure.
9380 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
9381 scalar stmt vector on cache hit.
9383 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
9385 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
9387 * config.gcc (powerpc-xilinx-eabi*): Remove.
9388 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
9390 (fusion_addis_mem_combo_load): Ditto.
9391 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
9393 (rs6000_cpu_cpp_builtins): Ditto.
9394 * config/rs6000/rs6000-linux.c
9395 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
9396 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
9397 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
9399 (rs6000_setup_reg_addr_masks): Ditto.
9400 (rs6000_init_hard_regno_mode_ok): Ditto.
9401 (rs6000_option_override_internal): Ditto.
9402 (legitimate_lo_sum_address_p): Ditto.
9403 (rs6000_legitimize_address): Ditto.
9404 (rs6000_legitimize_reload_address): Ditto.
9405 (rs6000_legitimate_address_p): Ditto.
9406 (abi_v4_pass_in_fpr): Ditto.
9407 (setup_incoming_varargs): Ditto.
9408 (rs6000_gimplify_va_arg): Ditto.
9409 (rs6000_split_multireg_move): Ditto.
9410 (rs6000_savres_strategy): Ditto.
9411 (rs6000_emit_prologue_components): Ditto.
9412 (rs6000_emit_epilogue_components): Ditto.
9413 (rs6000_emit_prologue): Ditto.
9414 (rs6000_emit_epilogue): Ditto.
9415 (rs6000_elf_file_end): Ditto.
9416 (rs6000_function_value): Ditto.
9417 (rs6000_libcall_value): Ditto.
9418 * config/rs6000/rs6000.h: Ditto.
9419 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
9420 (TARGET_MINMAX): ... this. New.
9421 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
9422 * config/rs6000/rs6000.md: Remove Xilinx FP support.
9423 (*movsi_internal1_single): Delete.
9424 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
9425 mfpu=, mxilinx-fpu): Delete.
9426 * config/rs6000/singlefp.h: Delete.
9427 * config/rs6000/sysv4.h: Remove Xilinx FP support.
9428 * config/rs6000/t-rs6000: Ditto.
9429 * config/rs6000/t-xilinx: Delete.
9430 * config/rs6000/titan.md: Adjust for fp_type removal.
9431 * config/rs6000/vsx.md: Remove Xilinx FP support.
9432 (VStype_simple): Delete.
9433 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
9434 * config/rs6000/xfpu.h: Delete.
9435 * config/rs6000/xfpu.md: Delete.
9436 * config/rs6000/xilinx.h: Delete.
9437 * config/rs6000/xilinx.opt: Delete.
9438 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
9439 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
9441 2018-05-04 Tom de Vries <tom@codesourcery.com>
9444 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
9447 2018-05-04 Richard Biener <rguenther@suse.de>
9450 * tree-complex.c (update_complex_assignment): We are always in SSA form.
9451 (expand_complex_div_wide): Likewise.
9452 (expand_complex_operations_1): Likewise.
9453 (expand_complex_libcall): Preserve EH info of the original stmt.
9454 (tree_lower_complex): Handle removed blocks.
9455 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
9456 on complex multiplication and division libcall builtins.
9458 2018-05-04 Richard Biener <rguenther@suse.de>
9461 * fold-const.c (negate_expr_p): Restrict negation of operand
9462 zero of a division to when we know that can happen without
9464 (fold_negate_expr_1): Likewise.
9466 2018-05-04 Jakub Jelinek <jakub@redhat.com>
9469 * real.h (real_nextafter): Declare.
9470 * real.c (real_nextafter): New function.
9471 * fold-const-call.c (fold_const_nextafter): New function.
9472 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
9473 CASE_CFN_NEXTTOWARD.
9474 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
9475 even when arg1_mode is different from arg0_mode.
9477 2018-05-03 Nathan Sidwell <nathan@acm.org>
9479 * doc/extend.texi (Deprecated Features): Remove
9481 (Backwards Compatibility): Likewise.
9482 * doc/invoke.texi (C++ Language Options): Likewise.
9483 (C++ Dialect Options): Likewise.
9485 2018-05-03 Jakub Jelinek <jakub@redhat.com>
9488 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
9489 _mm512_mask_mullox_epi64): New intrinsics.
9491 2018-05-03 Tom de Vries <tom@codesourcery.com>
9494 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
9495 dump files): Add offload-tree.
9497 2018-05-03 Richard Biener <rguenther@suse.de>
9499 PR tree-optimization/85615
9500 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
9501 to loops not nested in BBs loop father to avoid creating multi-entry
9504 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9506 PR tree-optimization/70291
9507 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
9508 arguments. Change return type to tree. Emit libcall as a new
9509 statement rather than replacing existing one when inplace_p is true.
9510 (expand_complex_multiplication_components): New function.
9511 (expand_complex_multiplication): Expand floating-point complex
9512 multiplication using the above.
9513 (expand_complex_division): Rename inner_type parameter to type.
9514 Update expand_complex_libcall call-site.
9515 (expand_complex_operations_1): Update expand_complex_multiplication
9516 and expand_complex_division call-sites.
9518 2018-05-02 Jakub Jelinek <jakub@redhat.com>
9521 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
9522 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
9523 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
9524 the highest significant bit of the shift count mask is clear. In
9525 check whether and[sq]i3 is needed verify that all significant bits
9526 of the shift count other than the highest are set.
9528 2018-05-02 Tom de Vries <tom@codesourcery.com>
9531 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
9532 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
9533 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
9534 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
9535 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
9536 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
9537 __builtin_goacc_parlevel_size.
9539 2018-05-02 Richard Biener <rguenther@suse.de>
9541 PR tree-optimization/85597
9542 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
9543 do not use split vect_get_vec_defs call but call vect_get_slp_defs
9546 2018-05-02 Tom de Vries <tom@codesourcery.com>
9549 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
9550 dump files): Add ltrans-tree.
9552 2018-05-02 Tom de Vries <tom@codesourcery.com>
9555 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
9556 dump files): Add wpa-ipa.
9558 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
9560 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
9561 powerpc*-*-linux*paired* target.
9562 * config/rs6000/750cl.h: Delete.
9563 * config/rs6000/paired.h: Delete.
9564 * config/rs6000/paired.md: Delete.
9565 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
9567 * config/rs6000/rs6000-builtin.def: Remove paired float support.
9568 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
9569 comment. Remove paired float support.
9570 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
9571 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
9573 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
9574 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
9576 * config/rs6000/rs6000.c: Remove paired float support.
9577 (paired_expand_vector_init, paired_expand_vector_move,
9578 paired_emit_vector_compare, paired_emit_vector_cond_expr,
9579 (paired_expand_lv_builtin, paired_expand_stv_builtin,
9580 paired_expand_builtin, paired_expand_predicate_builtin,
9581 paired_init_builtins): Delete.
9582 * config/rs6000/rs6000.h: Remove paired float support.
9583 * config/rs6000/rs6000.md: Remove paired float support.
9584 (move_from_CR_ov_bit): Delete.
9585 * config/rs6000/rs6000.opt (mpaired): Delete.
9586 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
9587 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
9589 2018-05-02 Richard Biener <rguenther@suse.de>
9592 * gimplify.c (gimplify_save_expr): When in SSA form allow
9593 SAVE_EXPRs to compute to SSA vars.
9595 2018-05-02 Jakub Jelinek <jakub@redhat.com>
9598 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
9599 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
9600 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
9601 clobber operands[2], instead use a new pseudo. Formatting fixes.
9603 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
9605 PR tree-optimization/85586
9606 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
9607 exit early for statements in the same group if the accesses are
9610 2018-05-02 Tom de Vries <tom@codesourcery.com>
9613 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
9616 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
9618 PR tree-optimization/85143
9619 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
9621 2018-05-01 Tom de Vries <tom@codesourcery.com>
9624 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
9625 not found" error message.
9627 2018-05-01 Tom de Vries <tom@codesourcery.com>
9630 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
9631 * vec.c (test_ordered_remove_if): New function.
9632 (vec_c_tests): Call test_ordered_remove_if.
9633 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
9634 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
9635 * tree-vect-patterns.c (vect_pattern_recog_1): Use
9636 VEC_ORDERED_REMOVE_IF.
9638 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9640 PR tree-optimization/82665
9641 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
9642 pointer subtraction where arguments come from a memchr call.
9644 2018-05-01 Jakub Jelinek <jakub@redhat.com>
9646 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
9647 --push-state --as-needed and --pop-state instead of --as-needed and
9648 --no-as-needed if ld supports it.
9649 * configure: Regenerated.
9652 * doc/install.texi2html: Replace _002d with - and _002a with * in
9653 generated html files using sed.
9655 2018-04-30 David Malcolm <dmalcolm@redhat.com>
9658 * gcc-rich-location.c (blank_line_before_p): New function.
9659 (use_new_line): New function.
9660 (gcc_rich_location::add_fixit_insert_formatted): New function.
9661 * gcc-rich-location.h
9662 (gcc_rich_location::add_fixit_insert_formatted): New function.
9664 2018-04-30 David Malcolm <dmalcolm@redhat.com>
9666 * selftest.c (assert_streq): Rename "expected" and "actual" to
9667 "val1" and "val2". Extend NULL-handling to cover both inputs
9668 symmetrically, while still requiring both to be non-NULL for a pass.
9669 * selftest.h (assert_streq): Rename "expected" and "actual" to
9671 (ASSERT_EQ): Likewise.
9672 (ASSERT_EQ_AT): Likewise.
9673 (ASSERT_KNOWN_EQ): Likewise.
9674 (ASSERT_KNOWN_EQ_AT): Likewise.
9675 (ASSERT_NE): Likewise.
9676 (ASSERT_MAYBE_NE): Likewise.
9677 (ASSERT_MAYBE_NE_AT): Likewise.
9678 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
9679 the assertion to pass.
9680 (ASSERT_STREQ_AT): Likewise.
9682 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
9684 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
9685 interaction with -pie.
9687 2018-04-30 David Malcolm <dmalcolm@redhat.com>
9689 * selftest.h: Fix alphabetization of per-source-file selftest
9692 2018-04-30 Jason Merrill <jason@redhat.com>
9694 PR c++/61982 - dead stores to destroyed objects.
9695 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
9698 2018-04-30 Jason Merrill <jason@redhat.com>
9700 * tree.c (build_clobber): New.
9701 * tree.h: Declare it.
9702 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
9704 2018-04-30 David Malcolm <dmalcolm@redhat.com>
9706 * diagnostic-show-locus.c (layout::layout): Update for
9707 location_get_source_line returning a char_span.
9708 (struct char_span): Move to input.h.
9709 (struct correction): Update for fields in char_span becoming
9711 (struct source_line): Update for location_get_source_line
9712 returning a char_span.
9713 (layout::print_line): Likewise.
9714 * edit-context.c (edited_file::print_content): Likewise.
9715 (edited_file::print_diff_hunk): Likewise.
9716 (edited_file::print_run_of_changed_lines): Likewise.
9717 (edited_file::get_num_lines): Likewise.
9718 (edited_line::edited_line): Likewise.
9719 * final.c (asm_show_source): Likewise.
9720 * input.c (location_get_source_line): Convert return type
9721 from const char * to char_span, losing the final "line_len"
9723 (dump_location_info): Update for the above.
9724 (get_substring_ranges_for_loc): Likewise. Use a char_span
9725 when handling the literal within the line.
9726 (test_reading_source_line): Update for location_get_source_line
9727 returning a char_span.
9728 * input.h (class char_span): Move here from
9729 diagnostic-show-locus.c, converting from a struct to a class.
9730 Make data members private.
9731 (char_span::operator bool): New.
9732 (char_span::length): New.
9733 (char_span::get_buffer): New.
9734 (char_span::operator[]): New.
9735 (char_span::subspan): Make const.
9736 (char_span::xstrdup): New.
9737 (location_get_source_line): Convert return type from const char *
9738 to char_span, losing the final "line_size" param.
9740 2018-04-30 Jan Hubicka <jh@suse.cz>
9742 * lto-wrapper.c (ltrans_priorities): New static var.
9743 (cmp_priority): New.
9744 (run_gcc): Read priorities and if doing parallel build order
9745 the Makefile by them.
9747 2018-04-30 David Malcolm <dmalcolm@redhat.com>
9749 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
9751 2018-04-30 Richard Biener <rguenther@suse.de>
9753 * tree-cfg.c (verify_address): Remove base argument, add
9754 flag whether to check TREE_ADDRESSABLE and do that.
9755 (verify_expr): Remove.
9756 (verify_types_in_gimple_reference): Add pieces from verify_expr.
9757 (verify_gimple_assign_single): Likewise.
9758 (verify_gimple_switch): Likewise.
9759 (verify_expr_location_1): Dereference tp once. Add (disabled)
9760 piece from verify_expr.
9761 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
9763 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
9765 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
9767 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
9769 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
9770 (small_data_pattern): Likewise.
9771 (arc_rewrite_small_data): Likewise.
9772 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
9773 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
9774 (get_symbol_alignment): New function.
9775 (legitimate_small_data_address_p): Likewise.
9776 (legitimate_scaled_address): Update, call
9777 legitimate_small_data_address_p.
9778 (output_sdata): New static variable.
9779 (arc_print_operand): Update how we handle small data operands.
9780 (arc_print_operand_address): Likewise.
9781 (arc_legitimate_address_p): Update, use
9782 legitimate_small_data_address_p.
9783 (arc_rewrite_small_data_p): Remove.
9784 (arc_rewrite_small_data_1): Likewise.
9785 (arc_rewrite_small_data): Likewise.
9786 (small_data_pattern): Likewise.
9787 (compact_sda_memory_operand): Update to use
9788 legitimate_small_data_address_p and get_symbol_alignment.
9789 (prepare_move_operands): Don't rewite sdata pattern.
9790 (prepare_extend_operands): Remove.
9791 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
9793 (zero_extendqisi2): Likewise.
9794 (zero_extendhisi2): Likewise.
9795 (extendqihi2): Likewise.
9796 (extendqisi2): Likewise.
9797 (extendhisi2): Likewise.
9801 * config/arc/constraints.md (Usd): Change it to memory constraint.
9803 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
9805 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
9806 as source of std instructions.
9807 * config/arc/arc.md (movsi_insn): Update pattern predicate to
9808 allow 6-bit constants as source for store instructions.
9809 (movdi_insn): Update instruction pattern to allow 6-bit constants
9810 as source for store instructions.
9812 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
9814 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
9816 2018-04-30 Nathan Sidwell <nathan@acm.org>
9817 Sandra Loosemore <sandra@codesourcery.com>
9819 * dumpfile.c (dump_open): Allow '-' for stdout.
9820 * doc/invoke.texi (Developer Options): Document dump filename
9821 determination early. Document stdin/stdout selection.
9823 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
9825 Microblaze Target: PIC data text relative
9827 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
9828 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
9830 * config/microblaze/microblaze.h (microblaze_constant_address_p):
9831 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
9832 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
9833 New addressing mode for data-text relative position indepenedent code.
9834 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
9835 'ADDRESS_SYMBOLIC_TXT_REL'.
9836 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
9837 (microblaze_legitimate_pic_operand): Exclude function calls from
9838 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
9839 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
9841 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
9842 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
9843 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
9844 for 'address + offset'.
9845 (microblaze_expand_prologue): Add new function prologue call for
9847 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
9848 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
9849 table in case of TARGET_PIC_DATA_TEXT_REL.
9850 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
9851 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
9852 Add new macros 'UNSPEC_TEXT',
9853 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
9854 + exclude function calls from 'UNSPEC_PLT' in case of data text
9856 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
9857 new target hook for generating address diff vector tables in case of
9859 * doc/tm.texi : Regenerate.
9860 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
9861 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
9862 of addr diff vector generation.
9863 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
9864 target hook definition.
9865 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
9866 Add default function for generate_pic_addr_diff_vec -> flag_pic.
9867 * doc/invoke.texi (Add new pic option): Add new microblaze pic
9868 option for data text relative.
9870 2018-04-30 Richard Biener <rguenther@suse.de>
9872 * tree-chrec.h (evolution_function_is_constant_p): Remove
9874 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
9876 2018-04-30 Richard Biener <rguenther@suse.de>
9879 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
9881 2018-04-30 Richard Biener <rguenther@suse.de>
9883 PR tree-optimization/28364
9884 PR tree-optimization/85275
9885 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
9886 copying first exit test.
9888 2018-04-28 Mark Wielaard <mark@klomp.org>
9890 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
9892 (dwarf_AT): Handle DW_AT_addr_base.
9893 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
9895 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
9898 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
9899 (*ashl<dwi>3_doubleword_mask_1): Ditto.
9900 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
9901 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
9903 2018-04-28 Richard Biener <rguenther@suse.de>
9905 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
9906 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
9907 to reflect use. Only add interesting stmts.
9909 2018-04-27 Martin Jambor <mjambor@suse.cz>
9912 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
9913 the jump function allows for passing through aggregate values.
9915 2018-04-27 David Malcolm <dmalcolm@redhat.com>
9917 * input.h (in_system_header_at): Convert from macro to inline
9919 (from_macro_expansion_at): Likewise.
9920 (from_macro_definition_at): Likewise.
9922 2018-04-27 Jeff Law <law@redhat.com>
9924 * config.gcc: Mark tile* targets as deprecated/obsolete.
9926 2018-04-27 Richard Biener <rguenther@suse.de>
9928 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
9931 2018-04-27 Richard Biener <rguenther@suse.de>
9933 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
9935 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
9937 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
9938 with Yd constraint. Set "preferred_for_speed" attribute from
9939 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
9941 (*movdi_internal): Ditto.
9942 (movti_interunit splitters): Remove
9943 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
9944 (movdi_interunit splitters): Ditto.
9945 * config/i386/constraints.md (Ye): Remove.
9946 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
9948 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9951 * config/aarch64/constraints.md (Usg): Limit to 31.
9954 2018-04-27 Jakub Jelinek <jakub@redhat.com>
9956 PR tree-optimization/85529
9957 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
9958 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
9959 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
9960 zero extension or masking of the MSB bit.
9961 (optimize_range_tests): Add FIRST_BB argument, pass it through
9962 to optimize_range_tests_var_bound.
9963 (maybe_optimize_range_tests, reassociate_bb): Adjust
9964 optimize_range_tests callers.
9966 2018-04-26 Richard Biener <rguenther@suse.de>
9967 Jakub Jelinek <jakub@redhat.com>
9969 * cgraph.h (symbol_table): Just declare debug method here.
9970 * symtab.c (symbol_table::debug): Define.
9972 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
9974 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
9976 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
9978 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
9979 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
9980 (*movdi_internal): Substitute Yi and Yj constraint with x
9981 and Ym and Yn constraint with y constraint. Update "isa"
9982 attribute and set "preferred_for_speed" attribute from
9983 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
9984 (*movsi_internal): Ditto.
9985 (*movdf_internal): Ditto.
9986 (*movsf_internal): Ditto.
9987 (*zero_extendsidi2): Ditto.
9988 * config/i386/sse.md (vec_set<mode>_0): Ditto.
9989 (sse2_loadld): Ditto.
9990 (*vec_extract<ssevecmodelower>_0): Ditto.
9991 (*vec_extractv4si_0_zext_sse4): Ditto.
9992 (vec_concatv2di): Ditto.
9993 (*vec_dup<mode>): Ditto.
9994 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
9995 * config/i386/constraints.md (Yi): Remove.
10000 2018-04-26 Nathan Sidwell <nathan@acm.org>
10002 * dumpfile.c (dump_open): New.
10003 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
10004 (dump_finish): Detect stdio/stderr by value not name.
10006 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
10008 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
10010 2018-04-26 Tom de Vries <tom@codesourcery.com>
10013 * config/nvptx/nvptx.c (verify_neutering_jumps)
10014 (verify_neutering_labels): New function
10015 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
10017 2018-04-26 Tom de Vries <tom@codesourcery.com>
10020 * config/nvptx/nvptx.c (needs_neutering_p): New function.
10021 (nvptx_single): Use needs_neutering_p to skip over insns that do not
10024 2018-04-26 Richard Biener <rguenther@suse.de>
10025 Tom de Vries <tom@codesourcery.com>
10028 * lto-streamer-out.c (output_function): Fixup loops if required to match
10029 discovery done in the reader.
10031 2018-04-26 Richard Biener <rguenther@suse.de>
10033 PR tree-optimization/85116
10034 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
10035 have a loop exit from the single latch predecessor. Remove
10036 case of header with just condition.
10037 (ch_base::copy_headers): Exclude infinite loops from any
10039 (pass_ch::execute): Record exits.
10041 2018-04-26 Richard Biener <rguenther@suse.de>
10043 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
10044 prologue cost vector and pass it to vect_get_load_cost.
10045 (vect_get_peeling_costs_all_drs): Likewise.
10046 (vect_peeling_hash_get_lowest_cost): Likewise.
10047 (vect_enhance_data_refs_alignment): Likewise.
10049 2018-04-26 Richard Biener <rguenther@suse.de>
10051 PR middle-end/85450
10052 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
10053 checking of integer<->pointer conversions.
10054 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
10055 sign-/zero-extending pointer types.
10056 (expand_omp_for_static_chunk): Likewise.
10058 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
10059 Jean Lee <xiaoyur347@gmail.com>
10061 * config/mips/mips.c (mips_asan_shadow_offset): New function.
10062 (TARGET_ASAN_SHADOW_OFFSET): Define.
10063 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
10064 true for -fsanitize=address.
10066 2018-04-25 Mark Wielaard <mark@klomp.org>
10068 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
10071 2018-04-25 Jakub Jelinek <jakub@redhat.com>
10073 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
10074 than "alu", remove explicit "memory" and "imm_disp" attributes.
10075 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
10077 PR middle-end/85414
10078 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
10079 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
10080 gen_lowpart_no_emit.
10082 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
10085 * config/i386/i386.c (ix86_expand_builtin): Change memory
10086 operand to XI, extend p0 to Pmode.
10087 * config/i386/i386.md: Change unspec volatile and operand
10088 1 mode to XI, change operand 0 mode to P.
10090 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
10092 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
10093 GET_MODE_MASK before any checking.
10094 (nds32_can_use_bset_p): Likewise.
10095 (nds32_can_use_btgl_p): Likewise.
10097 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
10099 * config/nds32/nds32-doubleword.md: New define_split pattern for
10100 illegal register number.
10102 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
10104 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
10106 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
10108 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
10110 2018-04-25 Richard Biener <rguenther@suse.de>
10112 * lto-streamer.h (LTO_major_version): Bump to 8.
10114 2018-04-25 Jakub Jelinek <jakub@redhat.com>
10116 * BASE-VER: Set to 9.0.0.
10118 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
10120 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
10121 in __abskf2 and __powikf2.
10123 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10126 * config/aarch64/constraints.md (Usg, Usj): New constraints.
10127 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
10128 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
10129 Use the above on operand 2. Reindent.
10130 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
10132 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
10135 * common/config/i386/i386-common.c (ix86_handle_option): Don't
10137 * config/i386/i386.opt (mcet): Removed.
10138 * doc/install.texi: Remove -mcet documentation.
10139 * doc/invoke.texi: Likewise.
10141 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
10144 * doc/install.texi: Remove -mcet from bootstrap-cet.
10146 2018-04-24 Jakub Jelinek <jakub@redhat.com>
10149 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
10150 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
10154 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
10155 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
10156 containing a CONST_VECTOR.
10158 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
10160 * doc/install.texi: Update newlib dependency for nvptx.
10162 2018-04-24 Jakub Jelinek <jakub@redhat.com>
10165 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
10166 instead of INTVAL when shifting x left.
10168 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
10170 PR tree-optimization/85478
10171 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
10172 vect_grouped_store_supported for single element vectors.
10174 2018-04-24 Richard Biener <rguenther@suse.de>
10177 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
10178 load cost increase to the case of non-constant step.
10180 2018-04-24 Jakub Jelinek <jakub@redhat.com>
10183 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
10184 destination if any_malformed_asm.
10186 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
10188 PR middle-end/85496
10189 * expr.c (store_field): In the bitfield case, if the value comes from
10190 a function call and is returned in registers by means of a PARALLEL,
10191 do not change the mode of the temporary unless BLKmode and VOIDmode.
10193 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
10195 PR rtl-optimization/85423
10196 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
10197 dependencies to debug insns when the previous insn is non-debug.
10199 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
10201 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
10202 enums into a single definition.
10203 (fls): Fix predicates and printing.
10206 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
10208 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
10209 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
10210 and short u6 immediate.
10211 (check_if_valid_sleep_operand): Remove.
10212 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
10214 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
10216 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
10217 flag_always_save_lp condition.
10218 * config/nds32/nds32.opt (malways-save-lp): New option.
10220 2018-04-22 Shiva Chen <shiva0217@gmail.com>
10222 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
10223 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
10224 * config/nds32/nds32.h
10225 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
10226 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
10228 2018-04-22 Shiva Chen <shiva0217@gmail.com>
10230 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
10231 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
10233 2018-04-22 Shiva Chen <shiva0217@gmail.com>
10234 Chung-Ju Wu <jasonwucj@gmail.com>
10236 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
10238 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
10239 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
10241 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
10243 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
10245 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
10247 * config/nds32/nds32-protos.h (nds32_data_alignment,
10248 nds32_local_alignment): Declare.
10249 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
10250 nds32_local_alignment): New functions.
10251 (TARGET_CONSTANT_ALIGNMENT): Define.
10252 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
10254 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
10256 * config/nds32/nds32.c
10257 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
10258 (TARGET_MODES_TIEABLE_P): Likewise.
10260 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
10262 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
10263 level Ofast and Og.
10265 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
10266 Chung-Ju Wu <jasonwucj@gmail.com>
10268 * config/nds32/constants.md (unspec_volatile_element): Add enum values
10269 for unaligned access.
10270 * config/nds32/nds32-intrinsic.c: Implementation of expanding
10272 * config/nds32/nds32-intrinsic.md: Likewise.
10273 * config/nds32/nds32_intrinsic.h: Likewise.
10274 * config/nds32/nds32.h (nds32_builtins): Likewise.
10275 * config/nds32/nds32.opt (munaligned-access): New option.
10276 * config/nds32/nds32.c (nds32_asm_file_start): Display
10277 flag_unaligned_access status.
10279 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
10281 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
10282 -mno-relax is present.
10283 * config/riscv/linux.h (LINK_SPEC): Ditto.
10285 2018-04-20 Martin Sebor <msebor@redhat.com>
10288 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
10290 (gimple_fold_builtin_stxcpy_chk): Same.
10291 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
10293 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
10296 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
10297 __powikf2 when long double is IEEE 128-bit.
10299 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
10301 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
10302 step to make sure stack always aligned.
10304 2018-04-20 Carl Love <cel@us.ibm.com>
10307 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
10308 size check for arg0.
10310 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
10311 Tom de Vries <tom@codesourcery.com>
10314 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
10315 Emit insns for calls too.
10316 (nvptx_find_par): Always look for worker-level predecessor insn.
10317 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
10319 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
10320 (nvptx_process_pars): Propagate frames for calls.
10322 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
10325 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
10327 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
10328 (ix86_handle_option): Don't handle OPT_mibt.
10329 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
10331 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
10333 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
10334 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
10335 (ix86_target_macros): Define __CET__ with flag_cf_protection
10336 for -fcf-protection.
10337 * config/i386/i386.c (isa2_opts): Remove -mibt.
10338 * config/i386/i386.h (TARGET_IBT): Removed.
10339 (TARGET_IBT_P): Likewise.
10340 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
10341 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
10342 * config/i386/i386.opt (mcet): Update help message.
10343 (mshstk): Likewise.
10345 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
10346 -mcet as an alias for -mshstk.
10348 2018-04-20 Richard Biener <rguenther@suse.de>
10350 PR middle-end/85475
10351 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
10352 complexity by forcing a single use of the multiply operand.
10354 2018-04-20 Martin Jambor <mjambor@suse.cz>
10357 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
10358 recursion dependency to only apply to non-clones.
10360 2018-04-20 Martin Jambor <mjambor@suse.cz>
10363 * ipa-cp.c (create_specialized_node): Check that clones of
10364 self-recursive edges exist during IPA-CP.
10366 2018-04-19 Toon Moene <toon@moene.org>
10368 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
10371 2018-04-19 Jakub Jelinek <jakub@redhat.com>
10373 PR tree-optimization/85467
10374 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
10375 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
10376 VECTOR_CST element to type.
10378 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
10381 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
10382 * config/i386/i386.md (builtin_setjmp_setup): Removed.
10383 (builtin_longjmp): Likewise.
10384 (save_stack_nonlocal): New pattern.
10385 (restore_stack_nonlocal): Likewise.
10387 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
10390 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
10391 Replace ASM_OUTPUT_LABEL with fprintf.
10393 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
10396 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
10397 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
10398 * config/i386/i386-c.c (ix86_target_macros_internal): Also
10399 define __IBT__ and __SHSTK__ for -fcf-protection.
10400 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
10402 (ix86_trampoline_init): Likewise.
10403 (x86_output_mi_thunk): Likewise.
10404 (ix86_notrack_prefixed_insn_p): Likewise.
10405 (ix86_option_override_internal): Don't disallow -fcf-protection.
10406 * config/i386/i386.md (rdssp<mode>): Also enable for
10408 (incssp<mode>): Likewise.
10409 (nop_endbr): Likewise.
10410 * config/i386/i386.opt (mcet): Change help message to built-in
10413 (mshstk): Likewise.
10414 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
10415 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
10416 enable CET built-in functions.
10418 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
10420 * common/config/i386/i386-common.c
10421 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
10422 OPTION_MASK_ISA_MOVDIRI_UNSET,
10423 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
10424 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
10425 * config.gcc (movdirintrin.h): New header.
10426 * config/i386/cpuid.h (bit_MOVDIRI,
10427 bit_MOVDIR64B): New bits.
10428 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
10430 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
10431 (VOID, PVOID, PCVOID)): New function types.
10432 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
10433 __builtin_ia32_directstoreu_u64,
10434 __builtin_ia32_movdir64b): New builtins.
10435 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
10436 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
10438 (ix86_valid_target_attribute_inner_p): Ditto.
10439 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
10440 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
10441 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
10442 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
10443 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
10444 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
10445 (movdiri<mode>, movdir64b_<mode>): New.
10446 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
10447 * config/i386/immintrin.h: Include movdirintrin.h.
10448 * config/i386/movdirintrin.h: New file.
10449 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
10451 2018-04-19 Richard Biener <rguenther@suse.de>
10453 PR middle-end/85455
10454 * cfg.c (clear_bb_flags): When loop state says we have
10455 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
10457 2018-04-19 Richard Biener <rguenther@suse.de>
10459 PR tree-optimization/84737
10460 * tree-vect-data-refs.c (vect_copy_ref_info): New function
10461 copying restrict info.
10462 (vect_setup_realignment): Use it.
10463 * tree-vectorizer.h (vect_copy_ref_info): Declare.
10464 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
10465 the first DR to all generated stores.
10466 (vectorizable_load): Likewise for loads.
10468 2018-04-19 Jakub Jelinek <jakub@redhat.com>
10470 PR tree-optimization/85446
10471 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
10472 the integral and pointer types to have the same precision.
10474 * doc/install.texi: Document --disable-cet being the default and
10477 2018-04-18 Martin Liska <mliska@suse.cz>
10479 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
10482 2018-04-18 Martin Liska <mliska@suse.cz>
10485 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
10488 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
10489 arguments if they are comparable.
10491 2018-04-18 Martin Liska <mliska@suse.cz>
10494 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
10497 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
10500 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
10503 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
10504 ENDBR after calling __morestack.
10506 2018-04-18 David Malcolm <dmalcolm@redhat.com>
10509 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
10510 by using gcc_base_ver to generate a gcc_driver_version, and use
10511 it when generating GCC_DRIVER_NAME.
10512 * configure: Regenerate.
10514 2018-04-18 Jakub Jelinek <jakub@redhat.com>
10517 * config.gcc: Obsolete powerpc*-*-*spe*.
10519 2018-04-17 Jakub Jelinek <jakub@redhat.com>
10522 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
10523 (stabstr_D): Change type of unum from unsigned int to
10524 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
10527 2018-04-17 Jim Wilson <jimw@sifive.com>
10530 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
10531 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
10532 Set arg_pointer_offset after using pretend_args_size.
10534 2018-04-17 Jakub Jelinek <jakub@redhat.com>
10536 PR rtl-optimization/85431
10537 * dse.c (record_store): Ignore zero width stores.
10540 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
10541 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
10542 __builtin_stack_restore rather than after it.
10543 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
10544 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
10545 argument instead of virtual_dynamic_stack_rtx.
10547 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
10549 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
10551 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
10552 Add note to error message to explain internal mapping of overloaded
10553 built-in function name to non-overloaded built-in function name.
10554 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
10557 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
10560 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
10561 where the inputs overlap with the output.
10563 2018-04-17 Jakub Jelinek <jakub@redhat.com>
10565 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
10566 (=v, v) alternative and explicit "memory" attribute.
10567 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
10568 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
10570 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
10571 "sselog1" type instead of "sselog".
10572 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
10573 "sselog". Remove explicit "memory" attribute.
10574 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
10575 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
10577 (vec_extract_hi_v32hi): Merge all alternatives into one, use
10578 "sselog1" type instead of "sselog". Remove explicit "memory"
10580 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
10581 use "sselog1" type instead of "sselog". Remove explicit "memory"
10583 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
10584 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
10586 (vec_extract_hi_v64qi): Merge all alternatives into one, use
10587 "sselog1" type instead of "sselog". Remove explicit "memory"
10589 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
10590 use "sselog1" type instead of "sselog". Remove explicit "memory"
10594 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
10596 PR middle-end/85414
10597 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
10600 2018-04-17 Martin Jambor <mjambor@suse.cz>
10603 * ipa-cp.c (create_specialized_node): Call
10604 expand_all_artificial_thunks if necessary.
10606 2018-04-17 Martin Liska <mliska@suse.cz>
10609 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
10610 in message, remote space in between '_G' and '('.
10612 2018-04-17 Jakub Jelinek <jakub@redhat.com>
10615 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
10616 avx512f_vmcmp<mode>3<round_saeonly_name>,
10617 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
10618 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
10619 avx512f_rndscale<mode><round_saeonly_name>,
10620 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
10621 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
10622 Use %<iptr>2 instead of %2 for -masm=intel.
10623 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
10624 avx512f_vcvttss2usi<round_saeonly_name>,
10625 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
10627 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
10628 avx512f_vcvttsd2usi<round_saeonly_name>,
10629 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
10630 Use %q1 instead of %1 for -masm=intel.
10631 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
10632 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
10633 of %3 for -masm=intel.
10634 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
10636 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
10638 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
10640 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
10641 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
10643 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
10645 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
10646 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
10647 %g1 and one with %0 and %1.
10648 (avx512er_vmrcp28<mode><round_saeonly_name>,
10649 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
10650 %1 for -masm=intel.
10651 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
10652 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
10653 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
10654 of %0 and %{%4%} for -masm=intel.
10655 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
10656 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
10657 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
10658 order of %0 and %{%5%}%{z%} for -masm=intel.
10660 2018-04-17 Jan Hubicka <jh@suse.cz>
10663 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
10665 2018-04-17 Martin Liska <mliska@suse.cz>
10668 * multiple_target.c (create_dispatcher_calls): Set apostrophes
10669 for target_clone error message. Make default implementation
10670 clone to be a local declaration.
10671 (separate_attrs): Add new argument and check for an empty
10673 (expand_target_clones): Handle it.
10674 (ipa_target_clone): Make redirection just for target_clones
10677 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
10678 Tom de Vries <tom@codesourcery.com>
10680 PR middle-end/84955
10681 * omp-expand.c (expand_oacc_for): Add dummy false branch for
10682 tiled basic blocks without omp continue statements.
10684 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
10687 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
10688 vec_extract expression as having side effects to make sure it gets
10691 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
10694 * config/i386/i386.c (get_builtin_code_for_version): Check
10697 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
10700 * config.gcc: Support "skylake".
10701 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10703 * config/i386/i386.c (m_SKYLAKE): Define.
10704 (processor_target_table): Add "skylake".
10705 (ix86_option_override_internal): Add "skylake".
10706 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
10707 PROCESSOR_CANNONLAKE.
10708 (get_builtin_code_for_version): Fix priority for
10709 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
10710 PROCESSOR_SKYLAKE-AVX512.
10711 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
10712 (processor_type): Add PROCESSOR_SKYLAKE.
10714 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
10715 Jason Merrill <jason@redhat.com>
10718 * convert.c (convert_to_integer_1): Use direct recursion for
10719 enumeral types and types with a precision less than the number
10720 of bits in their mode.
10722 2018-04-16 Julia Koval <julia.koval@intel.com>
10725 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
10726 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
10728 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
10731 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
10732 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
10733 and -mno-direct-move.
10735 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
10738 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
10739 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
10740 Ensure negative shifts result in {0}.
10742 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
10744 PR rtl-optimization/79916
10745 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
10746 regs (if any) to define how to gnerate SD moves when LRA is in
10749 2018-04-13 Jakub Jelinek <jakub@redhat.com>
10751 PR rtl-optimization/85393
10752 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
10753 * except.c (expand_dw2_landing_pad_for_region): Make static.
10754 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
10755 a label and unconditional jump to old_bb, rather than
10756 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
10759 PR rtl-optimization/85376
10760 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
10761 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
10762 instead of a specific value.
10764 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
10765 Bin Cheng <bin.cheng@arm.com>
10767 PR tree-optimization/82965
10768 PR tree-optimization/83991
10769 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
10770 by_profile_only parameter.
10771 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
10772 information if the loop was predicted to iterate too many times.
10773 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
10775 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
10778 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
10781 2018-04-13 Martin Liska <mliska@suse.cz>
10782 Jakub Jelinek <jakub@redhat.com>
10784 PR middle-end/81657
10785 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
10786 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
10787 * builtins.c (expand_builtin_memory_copy_args): Use
10788 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
10789 handle dest_addr == pc_rtx.
10791 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
10794 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
10795 asked to not generate direct moves.
10796 (fix_trunc<mode>si2_stfiwx): Similar.
10797 (fix_trunc<mode>si2_internal): Similar.
10799 2018-04-12 Jakub Jelinek <jakub@redhat.com>
10802 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
10803 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
10804 lookup if dest in some wider mode is known to be const0_rtx and
10805 if so, record permanent equivalence for it to be ZERO_EXTEND of
10806 the narrower mode destination.
10808 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
10810 * lto-streamer-out.c (output_function): Revert 259346.
10811 * omp-expand.c (expand_oacc_for): Likewise.
10813 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
10815 PR rtl-optimization/85354
10816 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
10817 * sel-sched.c (sel_global_init): ... here.
10819 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
10822 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
10823 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
10824 mode for PE-COFF targets.
10825 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
10826 (i386_pe_asm_lto_end): Likewise.
10827 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
10828 (TARGET_ASM_LTO_END): Likewise.
10829 * config/i386/winnt.c (saved_debug_info_level): New static variable.
10830 (i386_pe_asm_lto_start): New function.
10831 (i386_pe_asm_lto_end): Likewise.
10833 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
10834 Richard Biener <rguenther@suse.de>
10836 PR middle-end/84955
10837 * lto-streamer-out.c (output_function): Fix CFG loop state before
10839 * omp-expand.c (expand_oacc_for): Handle calls to internal
10840 functions like regular functions.
10842 2018-04-12 Richard Biener <rguenther@suse.de>
10845 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
10846 for the early LTO debug to properly generate references to it
10847 during DIE emission. Do not re-use that for the skeleton for
10849 (dwarf2out_early_finish): Likewise.
10851 2018-04-12 Jakub Jelinek <jakub@redhat.com>
10854 * config/i386/sse.md
10855 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
10856 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
10857 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
10858 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
10859 and output is a reg, avoid creating invalid lowpart subreg, but
10860 instead split into a 512-bit move. Don't split if not AVX512VL,
10861 input is xmm16+ reg and output is a mem.
10862 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
10863 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
10864 xmm16+ reg and output is a mem.
10866 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10868 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
10869 also for flag_dwarf2_cfi_asm.
10871 2018-04-12 Jakub Jelinek <jakub@redhat.com>
10873 PR rtl-optimization/85342
10874 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
10875 a bool scalar var inside of the loop instead. Don't try to update
10876 recog_data.operand after failed apply_change_group.
10878 2018-04-12 Tom de Vries <tom@codesourcery.com>
10881 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
10882 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
10883 array with flexible array member as array without given dimension.
10884 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
10885 argument for undefined param to true.
10887 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
10890 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
10891 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
10892 from PowerPC section.
10893 * config/rs6000/sysv4.opt (mcall-): Improve help text.
10894 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
10895 help text that is too long.
10896 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
10897 help text that is too long.
10898 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
10899 help text that is too long.
10901 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
10903 * config/alpha/alpha.md (stack_probe_internal): Rename
10904 from "probe_stack". Update all callers.
10906 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
10908 PR rtl-optimization/84566
10909 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
10910 sched_macro_fuse_insns.
10912 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
10915 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
10916 (compute_block_dependences): ... from here.
10918 2018-04-11 Jakub Jelinek <jakub@redhat.com>
10920 PR tree-optimization/85331
10921 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
10922 from int to HOST_WIDE_INT.
10924 2018-04-11 Martin Jambor <mjambor@suse.cz>
10927 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
10928 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
10929 not the same as the source val.
10930 (cgraph_edge_brings_value_p): New parameter.
10931 (gather_edges_for_value): Pass destination value to
10932 cgraph_edge_brings_value_p.
10933 (perhaps_add_new_callers): Likewise.
10934 (get_info_about_necessary_edges): Likewise and exclude values brought
10935 only by self-recursive edges.
10936 (create_specialized_node): Redirect only clones of self-calling edges.
10937 (+self_recursive_pass_through_p): New function.
10938 (find_more_scalar_values_for_callers_subset): Use it.
10939 (find_aggregate_values_for_callers_subset): Likewise.
10940 (known_aggs_to_agg_replacement_list): Removed.
10941 (decide_whether_version_node): Re-calculate known constants for all
10942 remaining context clones.
10944 2018-04-11 Richard Biener <rguenther@suse.de>
10947 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
10948 from early DWARF output.
10949 (dwarf2out_early_finish): Output line info unconditionally into
10950 early DWARF and add reference to it.
10952 2018-04-11 Jakub Jelinek <jakub@redhat.com>
10955 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
10956 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
10957 other than V2DFmode using iptr mode attribute.
10958 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
10960 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
10962 PR rtl-optimization/84659
10963 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
10965 2018-04-11 Jakub Jelinek <jakub@redhat.com>
10968 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
10970 (output_loc_list): Pass address of a dummy size variable even in the
10971 locview handling loop.
10972 (index_location_lists): Add comment on why skip_loc_list_entry can't
10975 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
10978 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
10981 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
10984 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
10985 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
10986 and -mstring-compare-inline-limit.
10988 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
10991 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
10992 for stack clash protection in a register whenever we need it to be in
10995 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
10997 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
10998 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
11000 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
11003 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
11005 (mlong-double-): Ditto.
11006 * config/rs6000/sysv4.opt (msdata=): Ditto.
11007 (mtls-size=): Ditto.
11009 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
11011 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
11012 erroneous entries for
11013 "vector int vec_ldl (int, long int *)", and
11014 "vector unsigned int vec_ldl (int, unsigned long int *)".
11015 Add comments and entries for
11016 "vector bool char vec_ldl (int, bool char *)",
11017 "vector bool short vec_ldl (int, bool short *)",
11018 "vector bool int vec_ldl (int, bool int *)",
11019 "vector bool long long vec_ldl (int, bool long long *)",
11020 "vector pixel vec_ldl (int, pixel *)",
11021 "vector long long vec_ldl (int, long long *)",
11022 "vector unsigned long long vec_ldl (int, unsigned long long *)".
11023 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
11024 type tree bool_long_long_type_node and correct definition of
11025 bool_V2DI_type_node to make reference to this new type tree.
11026 (rs6000_mangle_type): Replace erroneous reference to
11027 bool_long_type_node with bool_long_long_type_node.
11028 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
11029 comments to emphasize sign distinctions for char and int types and
11030 replace RS6000_BTI_bool_long constant with
11031 RS6000_BTI_bool_long_long constant. Also add comment to restrict
11032 use of RS6000_BTI_pixel.
11033 (bool_long_type_node): Remove this macro definition.
11034 (bool_long_long_type_node): New macro definition
11036 2018-04-10 Jakub Jelinek <jakub@redhat.com>
11038 PR rtl-optimization/85300
11039 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
11040 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
11041 simplify_unary_operation fails.
11043 2018-04-10 Martin Liska <mliska@suse.cz>
11045 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
11046 cgraph_edge and ipa_ref.
11048 2018-04-10 Jakub Jelinek <jakub@redhat.com>
11052 * config/i386/sse.md
11053 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
11054 computation of the VEC_MERGE selector from mask.
11055 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
11056 Fix decoding of the VEC_MERGE selector into mask.
11058 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
11060 PR tree-optimization/85286
11061 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
11063 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
11065 * final.c (final_1): Set insn_last_address as well as
11066 insn_current_address.
11068 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11071 * explow.c (emit_stack_probe): Call validize_mem on memory location
11072 before passing it to gen_probe_stack. Create address operand and
11073 legitimize it for the probe_stack_address case.
11075 2018-04-09 Jan Hubicka <jh@suse.cz>
11078 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
11079 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
11080 * tree.c (free_lang_data_in_type): Fix handling of binfos;
11082 (free_lang_data): Rebuild type inheritance graph.
11084 2018-04-09 Martin Sebor <msebor@redhat.com>
11086 * invoke.texi (-finline-small-functions): Mention other optimization
11088 (-findirect-inlining, -fpartial-inlining): Same.
11089 (-finline-functions-called-once): Same.
11090 (-freorder-blocks-and-partition): Same.
11092 2018-04-09 Jan Hubicka <jh@suse.cz>
11095 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
11096 jumps; choose last target that matches the criteria (i.e.
11097 no partition changes for non-crossing jumps).
11098 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
11099 support for redirecting crossing jumps to non-crossing.
11101 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
11103 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
11104 also for naked functions.
11106 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
11108 * config/arc/arc.md (add_shift): New pattern.
11109 (add_shift2): Likewise.
11110 (sub_shift): Likewise.
11111 (sub_shift_cmp0_noout): Likewise.
11112 (compare_si_ashiftsi): Likewise.
11113 (xbfu_cmp0_noout): New combine pattern.
11114 (xbfu_cmp0"): Likewise.
11115 (movsi_set_cc_insn): Place the predicable variant first.
11116 (commutative_binary_cmp0_noout): Remove clobber.
11117 (commutative_binary_cmp0): New pattern.
11118 (noncommutative_binary_cmp0): Likewise.
11119 (noncommutative_binary_cmp0_noout): Likewise.
11120 (noncommutative_binary_comparison_result_used): Removed.
11121 (rsub_cmp0): New pattern.
11122 (rsub_cmp0_noout): Likewise.
11123 (extzvsi): Changed, keep only meaningful variants.
11124 (SQH, SEZ): New iterators.
11125 (SQH_postfix): New mode attribute.
11126 (SEZ_prefix): New code attribute.
11127 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
11128 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
11129 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
11130 of numerical value.
11131 (noncommutative_operator): Check the availability of barrel
11134 2018-04-09 Richard Biener <rguenther@suse.de>
11136 PR tree-optimization/85284
11137 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
11138 Only use the niter constraining form of simple_iv when the exit
11139 is always executed.
11141 2018-04-09 Tom de Vries <tom@codesourcery.com>
11144 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
11145 (define_expand "*memory_barrier"): New define_expand.
11146 (define_insn "memory_barrier"): New insn.
11148 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
11150 PR rtl-optimization/80463
11151 PR rtl-optimization/83972
11152 PR rtl-optimization/83480
11154 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
11155 correct producer for the insn.
11156 (tidy_control_flow): Fixup seqnos in case of debug insns.
11158 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
11160 PR rtl-optimization/83913
11162 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
11163 different sched-times when merging exprs.
11165 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
11167 PR rtl-optimization/83962
11169 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
11170 tidy_fallthru_edge and tidy_control_flow.
11172 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
11174 PR rtl-optimization/83530
11176 * sel-sched.c (force_next_insn): New global variable.
11177 (remove_insn_for_debug): When force_next_insn is true, also leave only
11178 next insn in the ready list.
11179 (sel_sched_region): When the region wasn't scheduled, make another pass
11180 over it with force_next_insn set to 1.
11182 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
11184 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
11186 * config/nds32/constants.md (unspec_volatile_element): Add enum values
11187 for interrupt control.
11188 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
11189 functions for interrupt control.
11190 * config/nds32/nds32-intrinsic.md: Likewise.
11191 * config/nds32/nds32_intrinsic.h: Likewise.
11192 * config/nds32/nds32.h (nds32_builtins): Likewise.
11194 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
11196 * config/nds32/nds32.c (nds32_init_machine_status,
11197 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
11198 strict_aligned_p field.
11199 (nds32_expand_to_rtl_hook): New function.
11200 (TARGET_EXPAND_TO_RTL_HOOK): Define.
11201 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
11203 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
11204 Chung-Ju Wu <jasonwucj@gmail.com>
11206 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
11207 * config/nds32/nds32-n7.md: New file.
11208 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
11209 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
11211 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
11212 * config/nds32/nds32.md (pipeline_model): Add n7.
11213 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
11214 * config/nds32/pipelines.md: Include n7 settings.
11216 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
11217 Chung-Ju Wu <jasonwucj@gmail.com>
11219 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
11220 * config/nds32/nds32-e8.md: New file.
11221 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
11222 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
11224 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
11225 * config/nds32/nds32.md (pipeline_model): Add e8.
11226 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
11227 * config/nds32/pipelines.md: Include e8 settings.
11229 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
11230 Chung-Ju Wu <jasonwucj@gmail.com>
11232 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
11233 * config/nds32/nds32-n8.md: New file.
11234 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
11235 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
11237 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
11238 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
11239 * config/nds32/nds32.md (pipeline_model): Add n8.
11240 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
11241 * config/nds32/pipelines.md: Include n8 settings.
11243 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
11244 Chung-Ju Wu <jasonwucj@gmail.com>
11246 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
11247 * config/nds32/nds32-n9-2r1w.md: New file.
11248 * config/nds32/nds32-n9-3r2w.md: New file.
11249 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
11250 nds32_register_ports): New or modify for cpu n9.
11251 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
11253 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
11254 * config/nds32/nds32-utils.c: New file.
11255 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
11256 TARGET_MUL_SLOW): Define.
11257 * config/nds32/nds32.md (pipeline_model): New attribute.
11258 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
11259 New options that support cpu n9.
11260 * config/nds32/pipelines.md: Include n9 settings.
11261 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
11263 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
11265 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
11266 information if necessary.
11267 (output_cond_branch_compare_zero): Likewise.
11268 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
11269 (nds32_target_alignment): Refine for alignment.
11270 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
11271 (FUNCTION_BOUNDARY): Modify.
11272 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
11274 * config/nds32/nds32.opt (malways-align, malign-functions): New.
11276 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
11278 * config/nds32/constants.md (unspec_volatile_element): Add values for
11279 TLB operation and data prefetch.
11280 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
11281 functions for TLB operation and data prefetch.
11282 * config/nds32/nds32-intrinsic.md: Likewise.
11283 * config/nds32/nds32_intrinsic.h: Likewise.
11284 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
11285 (nds32_print_operand): Likewise.
11286 * config/nds32/nds32.h (nds32_builtins): Likewise.
11288 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
11289 Andrew Pinski <pinsika@gcc.gnu.org>
11291 PR middle-end/82976
11292 * match.pd: Use constant_boolean_node of correct type instead of
11293 boolean_true_node or boolean_false_node for simplifying
11294 pointer comparisons to zero.
11296 2018-04-07 Jakub Jelinek <jakub@redhat.com>
11298 PR tree-optimization/80021
11299 * tree.c (verify_type_variant): Make error call in verify_variant_match
11300 translatable and remove final full stop.
11302 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
11304 * config/nds32/constants.md (unspec_volatile_element): Add
11305 UNSPEC_VOLATILE_EH_RETURN.
11306 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
11307 nds32_output_stack_pop): Support dwarf exception handling process.
11308 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
11309 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
11310 exception handling process.
11311 (nds32_compute_stack_frame): Likewise.
11312 (nds32_return_addr_rtx): Likewise.
11313 (nds32_initial_elimination_offset): Likewise.
11314 (nds32_expand_prologue): Likewise.
11315 (nds32_expand_epilogue): Likewise.
11316 (nds32_dynamic_chain_address): New function.
11317 * config/nds32/nds32.h (machine_function): Add fields for dwarf
11318 exception handling.
11319 (DYNAMIC_CHAIN_ADDRESS): Define.
11320 (EH_RETURN_DATA_REGNO): Define.
11321 (EH_RETURN_STACKADJ_RTX): Define.
11322 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
11323 patterns for dwarf exception handling.
11325 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
11327 * config/nds32/nds32.h: Clean up obsolete macros.
11329 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
11331 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
11332 Add enum values for particular instructions.
11333 * config/nds32/nds32-intrinsic.c: Implementation of expanding
11334 particular intrinsic functions.
11335 * config/nds32/nds32-intrinsic.md: Likewise.
11336 * config/nds32/nds32_intrinsic.h: Likewise.
11337 * config/nds32/nds32.h (nds32_builtins): Likewise.
11338 * config/nds32/nds32.md (type): Add pbsad and pbsada.
11339 (btst, ave): New patterns for particular instructions.
11341 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
11343 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
11344 Add enum values for atomic load/store and memory sync.
11345 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
11347 * config/nds32/nds32-intrinsic.md: Likewise.
11348 * config/nds32/nds32_intrinsic.h: Likewise.
11349 * config/nds32/nds32.h (nds32_builtins): Likewise.
11351 2018-04-07 Jakub Jelinek <jakub@redhat.com>
11353 PR tree-optimization/85257
11354 * fold-const.c (native_encode_vector): If not all elts could fit
11355 and off is -1, return 0 rather than offset.
11356 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
11357 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
11358 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
11359 adjust buffer in native_interpret_expr call.
11361 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
11363 * config/nds32/constants.md (unspec_volatile_element): Add cache
11364 control enum values.
11365 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
11366 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
11367 * config/nds32/nds32.c (nds32_cctl_names): New.
11368 (nds32_print_operand): Handle cache control register names.
11369 * config/nds32/nds32.h (nds32_builtins): New enum values.
11370 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
11372 * config/nds32/nds32.md (type): Add mmu.
11373 * config/nds32/pipelines.md (simple_insn): Add mmu.
11375 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
11377 * config/nds32/nds32.md (type): Remove call.
11378 * config/nds32/pipelines.md (simple_insn): Likewise.
11380 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
11382 * config/nds32/constants.md (unspec_volatile_element): Add
11383 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
11384 UNSPEC_VOLATILE_FMFCFG.
11385 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
11386 description for fmfcfg and fmfcsr.
11387 (bdesc_1arg): Add fmtcsr.
11388 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
11389 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
11390 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
11391 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
11392 unspec_fmfcfg): New patterns.
11393 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
11394 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
11395 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
11396 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
11397 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
11398 __nds32__fmfcfg): Define.
11400 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
11402 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
11403 intrinsic register names.
11404 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
11405 intrinsic register enum values and macros.
11407 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
11409 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
11410 for load/store addressing form.
11411 (nds32_print_operand_address): Likewise.
11413 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
11416 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
11417 based on LABEL_REF. Remove useless assertion.
11418 (pic_address_needs_scratch): Fix formatting.
11419 (sparc_legitimize_pic_address): Minor tweaks.
11420 (sparc_delegitimize_address): Adjust assertion accordingly.
11421 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
11422 into symbolic_operand.
11423 (movsi_high_pic_label_ref): Likewise.
11424 (movsi_lo_sum_pic_label_ref): Likewise.
11425 (movdi_pic_label_ref): Likewise.
11426 (movdi_high_pic_label_ref): Likewise.
11427 (movdi_lo_sum_pic_label_ref): Likewise.
11429 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
11431 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
11432 custom LIB_SPEC setup.
11434 2018-04-06 Ruslan Bukin <br@bsdpad.com>
11435 Kito Cheng <kito.cheng@gmail.com>
11437 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
11438 * config/riscv/freebsd.h: New.
11440 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
11442 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
11443 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
11446 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
11447 Kito Cheng <kito.cheng@gmail.com>
11449 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
11450 nds32_output_call, nds32_symbol_binds_local_p): New functions.
11451 * config/nds32/nds32-protos.h (nds32_output_call,
11452 nds32_output_return): Declare.
11453 * config/nds32/nds32.md: Refine all the call and return patterns.
11455 2018-04-06 Jakub Jelinek <jakub@redhat.com>
11458 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
11459 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
11461 PR rtl-optimization/84872
11462 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
11463 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
11464 EDGE_CROSSING edge.
11466 2018-04-06 Tamar Christina <tamar.christina@arm.com>
11468 * expr.c (copy_blkmode_to_reg): Revert 254862.
11469 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
11471 2018-04-06 Richard Biener <rguenther@suse.de>
11473 PR middle-end/85244
11474 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
11475 after seeing a component reference with an adjacent field. Treat
11476 refs to arrays at struct end of external decls similar to
11477 refs to unconstrained commons.
11479 2018-04-06 Jakub Jelinek <jakub@redhat.com>
11482 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
11483 look through SAVE_EXPRs with non-side-effects argument. Adjust
11485 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
11488 2018-04-06 Richard Biener <rguenther@suse.de>
11490 PR middle-end/85180
11491 * alias.c (find_base_term): New wrapper around find_base_term
11492 unwinding CSELIB_VAL_PTR changes.
11493 (find_base_term): Do not restore CSELIB_VAL_PTR during the
11496 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11498 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
11500 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
11501 constant definitions.
11502 ("nop"): lr 0,0 -> nopr r0
11503 ("nop_lr0", "nop_lr1"): New insn definitions.
11505 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
11507 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
11508 NDS32_V3PUSH_AVAILABLE_P macro.
11510 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
11511 Chung-Ju Wu <jasonwucj@gmail.com>
11513 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
11514 (nds32*-*-*): Add float and fpu_config into supported_defaults.
11515 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
11516 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
11517 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
11518 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
11519 * config/nds32/constraints.md: New constraints and checking for hard
11520 float configuration.
11521 * config/nds32/iterators.md: New mode iterator and attribute for hard
11522 float configuration.
11523 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
11525 * config/nds32/nds32-fpu.md: New file.
11526 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
11527 deal with hard float code generation.
11528 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
11530 (abi_type, float_reg_number): New enum type.
11531 * config/nds32/nds32-predicates.c: New predicates for hard float.
11532 * config/nds32/nds32-protos.h: Declare functions for hard float.
11533 * config/nds32/nds32.c: Implementation for hard float configuration.
11534 * config/nds32/nds32.h: Definitions for hard float configuration.
11535 * config/nds32/nds32.md: Include hard float machine description and
11536 modify patterns for hard float configuration.
11537 * config/nds32/nds32.opt: New options for hard float configuration.
11538 * config/nds32/predicates.md: New predicates for hard float
11541 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
11543 * common/config/nds32/nds32-common.c
11544 (nds32_option_optimization_table): Enable -mreleax-hint by default.
11546 2018-04-05 Jakub Jelinek <jakub@redhat.com>
11548 PR middle-end/85195
11549 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
11550 CONSTRUCTOR_ELT (ctor, ...)->value.
11552 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
11555 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
11557 2018-04-05 Tom de Vries <tom@codesourcery.com>
11560 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
11563 2018-04-05 Shiva Chen <shiva0217@gmail.com>
11564 Kito Cheng <kito.cheng@gmail.com>
11566 * config/nds32/constraints.md (U33): Fine-tune checking condition.
11567 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
11568 * config/nds32/nds32.h (nds32_16bit_address_type): Add
11569 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
11571 2018-04-05 Shiva Chen <shiva0217@gmail.com>
11572 Kito Cheng <kito.cheng@gmail.com>
11574 * config/nds32/constraints.md (Ufe): New memory constraint.
11575 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
11576 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
11577 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
11579 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
11580 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
11582 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11584 * config/nds32/nds32.md: Use optimize_size in the condition for
11585 alu-shift instructions.
11587 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11589 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
11591 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11593 * config/nds32/nds32.md (negsi2): Refine pattern.
11595 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
11596 Chung-Ju Wu <jasonwucj@gmail.com>
11598 * config/nds32/iterators.md (shift_rotate): New code iterator.
11599 (shift): New code attribute.
11600 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
11601 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
11602 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
11603 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
11604 bit-wise operations.
11605 (andsi3, *andsi3): Ditto.
11606 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
11607 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
11608 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
11609 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
11610 nds32_ior_operand, nds32_xor_operand): New predicates.
11612 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11614 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
11615 (addsi3, subsi3): ... this.
11617 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11619 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
11621 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11623 * config/nds32/nds32.md: Adjust indention.
11625 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
11627 * config/nds32/nds32.md (feature): New attribute.
11629 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
11631 * config/nds32/nds32.md (subtype): New attribute.
11633 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
11636 * config/arm/arm-builtins.c (arm_expand_builtin): Change
11637 expansion to perform a bitwise AND of the argument followed by a
11638 boolean negation of the result.
11640 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
11642 PR rtl-optimization/84878
11643 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
11644 the basic block. Assert the use reference is not artificial and that
11645 it has an associated insn.
11647 2018-04-04 Michael Matz <matz@suse.de>
11649 * builtins.c (compute_objsize): Pass correct operand
11650 to array_at_struct_end_p.
11652 2018-04-04 Richard Biener <rguenther@suse.de>
11655 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
11656 from contexts for DINFO_LEVEL_TERSE and below.
11658 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
11660 * config/nds32/nds32-doubleword.md (move_<mode>): Require
11661 resiter_operand condition.
11662 * config/nds32/nds32.md (*move<mode>): Ditto.
11664 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
11665 Monk Chiang <sh.chiang04@gmail.com>
11667 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
11669 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
11671 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
11673 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
11674 Kito Cheng <kito.cheng@gmail.com>
11676 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
11677 nds32_cond_code_str, output_cond_branch,
11678 output_cond_branch_compare_zero, nds32_expand_cbranch,
11679 nds32_expand_cstore, nds32_expand_movcc,
11680 nds32_output_cbranchsi4_equality_zero,
11681 nds32_output_cbranchsi4_equality_reg,
11682 nds32_output_cbranchsi4_equality_reg_or_const_int,
11683 nds32_output_cbranchsi4_greater_less_zero: New functions.
11684 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
11685 nds32_expand_cstore, nds32_expand_movcc,
11686 nds32_output_cbranchsi4_equality_zero,
11687 nds32_output_cbranchsi4_equality_reg,
11688 nds32_output_cbranchsi4_equality_reg_or_const_int,
11689 nds32_output_cbranchsi4_greater_less_zero): Declare.
11690 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
11691 nds32_rimm11s_operand): New predicates.
11692 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
11693 * config/nds32/nds32.md: Rewrite all the branch and conditional move
11696 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
11698 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
11699 * config/nds32/nds32.md: Ditto.
11700 * config/nds32/pipelines.md: Ditto.
11702 2018-04-04 Richard Biener <rguenther@suse.de>
11704 PR tree-optimization/85168
11705 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
11706 propagating abnormals.
11708 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
11710 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
11712 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
11713 Kito Cheng <kito.cheng@gmail.com>
11715 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
11716 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
11717 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
11718 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
11719 * config/nds32/nds32.md (sibcall_internal): New.
11720 (sibcall_register): Remove.
11721 (sibcall_immediate): Remove.
11722 (sibcall_value_internal): New.
11723 (sibcall_value_register): Remove.
11724 (sibcall_value_immediate): Remove.
11725 * config/nds32/predicates.md (nds32_general_register_operand): New.
11726 (nds32_call_address_operand): New.
11728 2018-04-03 Jakub Jelinek <jakub@redhat.com>
11730 PR rtl-optimization/85167
11731 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
11732 bb_defs if *split_p, instead preinitialize it to NULL.
11734 PR tree-optimization/85156
11735 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
11736 evaluating the argument multiple times.
11738 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
11740 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
11742 (_mm_cvtpd_ps): Likewise.
11743 (_mm_cvttpd_epi32): Likewise.
11744 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
11745 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
11746 vector, pixel, and bool following altivec.h include.
11748 2018-04-03 Martin Sebor <msebor@redhat.com>
11750 * doc/extend.texi (Common Function Attributes): Clarify.
11751 (const attribute): Likewise.
11752 (pure attribute): Likewise.
11754 2018-04-03 Jakub Jelinek <jakub@redhat.com>
11757 * config/i386/i386.c (ix86_expand_vector_set): Use
11758 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
11760 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
11762 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
11763 instructions when changing rounding bits to preserve precision bits
11764 in the x87 control word.
11766 2018-04-03 Martin Liska <mliska@suse.cz>
11768 PR tree-optimization/82491
11769 * rtl.h (strip_offset_and_add): Replace += suboffset with
11770 poly_uint64 () + suboffset.
11772 2018-03-29 Martin Liska <mliska@suse.cz>
11773 Martin Jambor <mjambor@suse.cz>
11776 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
11777 param_type is not an integral or pointer type.
11779 2018-04-03 Richard Biener <rguenther@suse.de>
11781 * sese.h (recompute_all_dominators): Remove.
11783 2018-04-02 Martin Sebor <msebor@redhat.com>
11785 * doc/invoke.texi (-Wrestrict): Fix typos.
11787 2018-04-02 Jim Wilson <jimw@sifive.com>
11789 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
11790 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
11791 (<optab>di3, <optab>si3_extend): Likewise.
11792 (<optab>si3_mask, <optab>si3_mask_1): New.
11793 (<optab>di3_mask, <optab>di3_mask_1): New.
11794 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
11795 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
11796 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
11798 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
11800 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
11803 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
11805 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
11806 (nds32_canonicalize_comparison): New function.
11808 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
11809 Kito Cheng <kito.cheng@gmail.com>
11810 Kuan-Lin Chen <kuanlinchentw@gmail.com>
11812 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
11813 * config/nds32/constants.md (unspec_volatile_element): Add
11814 UNSPEC_VOLATILE_RELAX_GROUP.
11815 * config/nds32/nds32-relax-opt.c: New file.
11816 * config/nds32/nds32-predicates.c
11817 (nds32_symbol_load_store_p): New function.
11818 * config/nds32/nds32-protos.h
11819 (nds32_symbol_load_store_p): Declare function.
11820 (make_pass_nds32_relax_opt): Declare new rtl pass function.
11821 * config/nds32/nds32.c
11822 (nds32_register_pass): New function to register pass.
11823 (nds32_register_passes): New function to register passes.
11824 * config/nds32/nds32.md (relax_group): New pattern.
11825 * config/nds32/nds32.opt (mrelax-hint): New option.
11826 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
11828 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
11830 * config/nds32/t-nds32: Modify files dependency.
11832 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
11834 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
11835 (PROFILE_HOOK): Define its implementation.
11837 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
11839 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
11840 type and 32-bit size.
11842 2018-04-01 Jakub Jelinek <jakub@redhat.com>
11844 PR middle-end/85090
11845 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
11846 (V_128_256): New mode iterator.
11847 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
11848 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
11849 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
11851 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
11852 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
11854 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
11857 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
11858 NaN inputs correctly.
11860 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
11863 * config/rs6000/vsx.md (??r): New mode attribute.
11864 (*vsx_mov<mode>_64bit): Use it.
11865 (*vsx_mov<mode>_32bit): Likewise.
11867 2018-03-30 Martin Sebor <msebor@redhat.com>
11869 PR tree-optimization/84818
11870 * builtins.c (check_access): Use warning_n.
11872 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11875 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
11877 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
11880 2018-03-30 Julia Koval <julia.koval@intel.com>
11883 * x86-tune.def (movx, partial_reg_dependency): Enable for
11886 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
11888 PR inline-asm/84985
11889 * lra-constraints.c (process_alt_operands): Move setting
11890 this_alternative_matches below.
11892 2018-03-29 Martin Liska <mliska@suse.cz>
11895 * doc/invoke.texi: Document how LTO works with debug info.
11896 Describe auto-load support of binutils. Mention 'x86-64'
11897 as valid option value of -march option.
11899 2018-03-29 Jakub Jelinek <jakub@redhat.com>
11901 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
11904 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
11905 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
11906 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
11909 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
11912 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
11913 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
11914 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
11915 for RS6000_BTM_POWERPC64.
11916 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
11917 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
11918 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
11921 (DIVDEU): Likewise.
11923 2018-03-28 Carl Love <cel@us.ibm.com>
11926 2017-09-27 Carl Love <cel@us.ibm.com>
11928 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
11929 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
11930 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
11933 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11935 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
11936 instead of __vector bool.
11937 (_mm_max_pu8): Likewise.
11938 (_mm_min_pi16): Likewise.
11940 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
11943 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
11944 (DIVWEUO): Likewise.
11945 (DIVDEO): Likewise.
11946 (DIVDEUO): Likewise.
11947 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
11948 DIVWEUO and DIVDEUO.
11949 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
11950 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
11951 (div_extend): Likewise.
11952 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
11954 (__builtin_divweuo): Likewise.
11955 (__builtin_divdeo): Likewise.
11956 (__builtin_divdeuo): Likewise.
11958 2018-03-28 Jakub Jelinek <jakub@redhat.com>
11961 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
11962 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
11964 PR tree-optimization/82004
11965 * gimple-match-head.c (optimize_pow_to_exp): New function.
11966 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
11967 Don't fold to exp if optimize_pow_to_exp is false.
11969 2018-03-28 Martin Liska <mliska@suse.cz>
11972 * calls.c (initialize_argument_information): Fix trailing space.
11973 * common.opt: Fix typo and provide better explanation for
11974 -fsanitize-coverage option.
11975 * config/i386/i386.opt: Fix typo.
11977 2018-03-28 Jakub Jelinek <jakub@redhat.com>
11978 Martin Liska <mliska@suse.cz>
11981 * gimplify.c (asan_poison_variable): Don't do the check for
11982 gimplify_omp_ctxp here.
11983 (gimplify_decl_expr): Do it here.
11984 (gimplify_target_expr): Likewise.
11986 2018-03-28 Martin Liska <mliska@suse.cz>
11989 * config/i386/i386.c (ix86_function_arg_advance): Do not call
11990 chkp_type_bounds_count if MPX is not enabled.
11992 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
11994 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
11996 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
11999 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
12000 function to create the function decl for complex long double
12001 multiply and divide for -mabi=ieeelongdouble.
12002 (init_float128_ieee): Call it.
12004 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
12007 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
12008 -fcf-protection=branch -mibt.
12009 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
12011 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
12014 * config/arm/arm.c (arm_valid_symbolic_address): Handle
12015 arm_word_relocations.
12017 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
12020 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
12021 extern array declarations.
12023 2018-03-27 Richard Biener <rguenther@suse.de>
12025 PR middle-end/84067
12026 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
12027 explicit single_use checks.
12029 2018-03-27 Richard Biener <rguenther@suse.de>
12031 PR tree-optimization/85082
12032 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
12035 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
12037 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
12038 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
12039 Turn on fasynchronous-unwind-tables and funwind-tables.
12041 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
12044 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
12045 (*bmi_blsr_<mode>_ccz): Ditto.
12047 2018-03-26 Tom de Vries <tom@codesourcery.com>
12049 PR tree-optimization/85063
12050 * omp-general.c (offloading_function_p): New function. Factor out
12052 * omp-offload.c (pass_omp_target_link::gate): ... here.
12053 * omp-general.h (offloading_function_p): Declare.
12054 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
12055 with attribute omp declare target for offloading functions.
12057 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
12059 PR tree-optimization/84005
12060 * tree-data-ref.h (get_base_for_alignment): Declare.
12061 * tree-data-ref.c (get_base_for_alignment_1): New function.
12062 (get_base_for_alignment): Likewise.
12063 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
12064 get_base_for_alignment to find a suitable base object, instead
12065 of always using drb->base_address.
12067 2018-03-23 Jakub Jelinek <jakub@redhat.com>
12069 PR inline-asm/85022
12070 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
12071 known size by default.
12073 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
12075 PR inline-asm/85030
12076 * lra-constraints.c (process_alt_operands): Don't match BLKmode
12077 and non BLKmode operands.
12079 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12082 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
12083 Clean up attributes.
12085 2018-03-23 Richard Biener <rguenther@suse.de>
12088 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
12089 we are going to emit early debug for LTO.
12091 2018-03-23 Jakub Jelinek <jakub@redhat.com>
12093 PR inline-asm/85034
12094 * function.c (match_asm_constraints_1): Don't optimize if input
12095 doesn't satisfy general_operand predicate for output's mode.
12097 PR inline-asm/85022
12098 * alias.c (write_dependence_p): Don't require for x_canonicalized
12099 non-VOIDmode if x has VOIDmode.
12102 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
12103 just don't try to optimize it rather than assert it never happens.
12105 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
12107 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
12108 macro expansions for definition of ST_INTERNAL_<mode> and
12109 LD_INTERNAL_<mode> builtins.
12110 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
12112 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
12114 (altivec_expand_st_builtin): Likewise.
12115 (altivec_expand_builtin): Remove calls to deleted functions.
12116 (rs6000_address_for_altivec): Delete this function.
12117 * config/rs6000/vector.md: Remove expands for
12118 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
12120 2018-03-22 Sudakshina Das <sudi.das@arm.com>
12123 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
12124 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
12125 re-computing once computed.
12126 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
12127 (arm_init_machine_status): Initialize
12128 machine->static_chain_stack_bytes.
12130 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
12133 * doc/extend.texi: Add four new prototypes for vec_ld.
12134 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
12135 definitions for more logical presentation.
12136 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
12137 entries for V1TI variants of __builtin_altivec_ld builtin.
12138 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
12139 handling of V1TI variant of LVX icode pattern.
12140 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
12141 (rs6000_gimple_fold_builtin): Likewise.
12142 (altivec_init_builtins): Add code to define
12143 __builtin_altivec_lvx_v1ti function.
12145 2018-03-22 Jakub Jelinek <jakub@redhat.com>
12147 PR inline-asm/84941
12148 * function.c (match_asm_constraints_1): Don't do the optimization
12149 if input isn't a REG, SUBREG, MEM or constant.
12151 2018-03-22 Tom de Vries <tom@codesourcery.com>
12153 PR tree-optimization/84956
12154 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
12155 bb_has_abnormal_pred.
12157 2018-03-22 Jakub Jelinek <jakub@redhat.com>
12160 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
12161 DECL_INITIAL (decl) to decl at the end.
12162 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
12163 adjust the comment.
12165 2018-03-21 Joseph Myers <joseph@codesourcery.com>
12167 * doc/extend.texi (__builtin_tgmath): Document when complex
12168 integer types are treated as _Complex _Float64.
12170 2018-03-21 Tom de Vries <tom@codesourcery.com>
12172 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
12174 2018-03-21 Jakub Jelinek <jakub@redhat.com>
12176 PR tree-optimization/84960
12177 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
12178 if it is ENTRY block, move them into single succ of ENTRY in that case.
12180 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
12182 PR tree-optimization/84811
12183 * poly-int.h (poly_span_traits): Remove the T3 parameter and
12184 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
12185 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
12186 (known_subrange_p): Update accordingly. Cast each value involved
12187 in the size comparison, rather than casting the result of the
12190 2018-03-21 Jakub Jelinek <jakub@redhat.com>
12192 PR tree-optimization/84982
12193 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
12194 by flipping the least significant bit rather than all bits from
12195 bitpos to bitpos + bitsize - 1.
12197 2018-03-21 Nathan Sidwell <nathan@acm.org>
12199 * doc/extend.texi (Deprecated Features): Remove mention of
12200 long-deleted deprecations.
12202 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12205 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
12206 * configure: Regenerate.
12208 2018-03-21 Tom de Vries <tom@codesourcery.com>
12210 PR tree-optimization/83126
12211 * tree-parloops.c (num_phis): New function.
12212 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
12214 2018-03-21 Nathan Sidwell <nathan@acm.org>
12216 * doc/extend.texi (Deprecated Features): Update deprecated flags,
12217 mention anon-struct/union members and trailing attributes.
12219 2018-03-21 Bin Cheng <bin.cheng@arm.com>
12221 PR tree-optimization/84969
12222 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
12223 builtin memset partitions if they set different rhs values.
12225 2018-03-21 Jakub Jelinek <jakub@redhat.com>
12227 PR rtl-optimization/84989
12228 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
12229 VEC_DUPLICATE with scalar result mode.
12231 2018-03-21 Martin Liska <mliska@suse.cz>
12234 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
12235 not intended return statement.
12237 2018-03-21 Martin Liska <mliska@suse.cz>
12240 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
12241 (chkp_find_bound_slots_1): Limit number of iterations.
12243 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
12246 * Minor grammar fixes for x86 options.
12248 2018-03-20 Jakub Jelinek <jakub@redhat.com>
12251 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
12252 holding REG_CFA_RESTORE notes, instead turn them into a USE.
12254 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
12257 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
12258 (altivec_lvx_<mode>_1op): Likewise.
12259 (altivec_stvx_<mode>_2op): Likewise.
12260 (altivec_stvx_<mode>_1op): Likewise.
12261 (altivec_lvx_<VM2:mode>): New define_expand.
12262 (altivec_stvx_<VM2:mode>): Likewise.
12263 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
12264 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
12265 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
12266 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
12267 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
12268 (rs6000_gen_lvx): Likewise.
12269 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
12270 (altivec_expand_stv_builtin): Likewise.
12271 (altivec_expand_builtin): Likewise.
12272 * config/rs6000/vector.md: Likewise.
12274 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12277 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
12280 2018-03-20 Richard Biener <rguenther@suse.de>
12283 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
12284 sign-conversions as zero, fall back to standard scalar_stmt
12287 2018-03-20 Martin Liska <mliska@suse.cz>
12290 * predict.c (rebuild_frequencies): Handle case when we have
12291 PROFILE_ABSENT, but flag_guess_branch_prob is false.
12293 2018-03-20 Jakub Jelinek <jakub@redhat.com>
12296 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
12297 flag_section_anchors.
12298 * varasm.c (use_blocks_for_decl_p): Remove hack for
12299 dw2_force_const_mem.
12302 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
12304 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
12305 be created, use lowpart_subreg of operands[0] rather than operands[0]
12307 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
12308 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
12309 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
12310 and n constraint instead of aarch64_shift_imm_di and Usd.
12311 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
12312 (*aarch64_<optab>_reg_minus<mode>3): ... this.
12314 2018-03-20 Sudakshina Das <sudi.das@arm.com>
12317 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
12318 to favor GPR over NEON registers.
12319 (<shift>di3_neon): Likewise.
12321 2018-03-20 Tom de Vries <tom@codesourcery.com>
12324 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
12325 (nvptx_process_pars): Emit bar.sync asap and alap.
12327 2018-03-20 Tom de Vries <tom@codesourcery.com>
12330 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
12331 seen_label if seen_label is already set.
12333 2018-03-20 Jakub Jelinek <jakub@redhat.com>
12336 * config/i386/i386.c (fold_builtin_cpu): For features above 31
12337 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
12338 Use 1U instead of 1. Formatting fixes.
12341 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
12342 instead of TREE_TYPE (s1) for the return value.
12344 2018-03-19 Jakub Jelinek <jakub@redhat.com>
12346 PR tree-optimization/84946
12347 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
12348 bitsize + bitsize in poly_uint64 rather than poly_int64.
12351 * dwarf2asm.c: Include fold-const.c.
12352 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
12353 of decl rather than decl itself.
12355 PR rtl-optimization/84643
12356 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
12358 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
12361 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
12362 calling assemble_variable.
12364 2018-03-19 Sudakshina Das <sudi.das@arm.com>
12367 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
12368 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
12370 2018-03-19 Jim Wilson <jimw@sifive.com>
12373 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
12374 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
12375 (riscv_first_stack_step): Likewise.
12376 (riscv_option_override): Use STACK_BOUNDARY instead of
12377 MIN_STACK_BOUNDARY.
12378 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
12379 MIN_STACK_BOUNDARY.
12380 (BIGGEST_ALIGNMENT): Set to 128.
12381 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
12382 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
12385 2018-03-19 Richard Biener <rguenther@suse.de>
12387 PR tree-optimization/84933
12388 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
12389 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
12391 2018-03-19 Richard Biener <rguenther@suse.de>
12393 PR tree-optimization/84859
12394 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
12395 (cond_if_else_store_replacement): Perform sinking operation on
12396 single-store BBs regardless of MAX_STORES_TO_SINK setting.
12397 Generalize what a BB with a single eligible store is.
12399 2018-03-19 Richard Biener <rguenther@suse.de>
12401 PR tree-optimization/84929
12402 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
12403 chrec_is_positive against non-chrec arg.
12405 2018-03-19 Tamar Christina <tamar.christina@arm.com>
12408 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
12410 2018-03-18 Martin Liska <mliska@suse.cz>
12412 PR rtl-optimization/84635
12413 * regrename.c (build_def_use): Use matches_mode only when
12416 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
12418 PR tree-optimization/84913
12419 * tree-vect-loop.c (vectorizable_reduction): Don't try to
12420 vectorize chains of COND_EXPRs.
12422 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
12424 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
12426 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
12428 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
12430 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
12432 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
12434 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
12435 Kito Cheng <kito.cheng@gmail.com>
12437 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
12438 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
12439 (nds32_adjust_reg_alloc_order): New function.
12440 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
12442 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
12444 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
12445 nds32_print_operand, nds32_print_operand_address): Use
12446 HOST_WIDE_INT_PRINT_DEC instead.
12448 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
12450 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
12452 2018-03-17 Jakub Jelinek <jakub@redhat.com>
12455 * config/i386/i386.c (initial_ix86_tune_features,
12456 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
12457 unsigned long long.
12458 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
12459 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
12460 rather than 1u << ix86_tune. Formatting fix.
12461 (ix86_option_override_internal): Change ix86_arch_mask from
12462 unsigned int to unsigned HOST_WIDE_INT, initialize to
12463 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
12464 (ix86_function_specific_restore): Likewise.
12466 2018-03-16 Jakub Jelinek <jakub@redhat.com>
12469 * postreload.c (reload_combine_recognize_pattern): Perform
12470 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
12471 truncate_int_for_mode the result for the destination's mode.
12474 * hsa-gen.c (mem_type_for_type): Fix comment typo.
12475 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
12477 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
12480 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
12483 * lra-assigns.c (lra_split_hard_reg_for): Don't use
12484 regno_allocno_class_array and sorted_pseudos.
12485 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
12486 insns where regno is used.
12488 2018-03-16 Martin Liska <mliska@suse.cz>
12491 * multiple_target.c (create_dispatcher_calls): Redirect
12492 reference in the symbol table.
12494 2018-03-16 Martin Liska <mliska@suse.cz>
12497 * multiple_target.c (create_dispatcher_calls): Redirect also
12500 2018-03-16 Jakub Jelinek <jakub@redhat.com>
12504 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
12505 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
12506 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
12508 2018-03-16 Julia Koval <julia.koval@intel.com>
12510 * doc/invoke.texi (Skylake Server): Add CLWB.
12511 Cannonlake): Remove CLWB.
12513 2018-03-16 Jakub Jelinek <jakub@redhat.com>
12515 PR tree-optimization/84841
12516 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
12518 (FLOAT_ONE_CONST_TYPE): Define.
12519 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
12520 (sort_by_operand_rank): Put entries with higher constant_type last
12521 rather than first to match comments.
12523 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
12525 * config/nios2/nios2.md (movsi_internal): Fix thinko in
12528 2018-03-15 Jakub Jelinek <jakub@redhat.com>
12531 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
12532 check and use address of target always.
12534 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
12537 * config/i386/i386.c (indirect_thunk_needed): Update comments.
12538 (indirect_thunk_bnd_needed): Likewise.
12539 (indirect_thunks_used): Likewise.
12540 (indirect_thunks_bnd_used): Likewise.
12541 (indirect_return_needed): New.
12542 (indirect_return_bnd_needed): Likewise.
12543 (output_indirect_thunk_function): Add a bool argument for
12545 (output_indirect_thunk_function): Don't generate alias for
12546 function return thunk.
12547 (ix86_code_end): Call output_indirect_thunk_function to generate
12548 function return thunks.
12549 (ix86_output_function_return): Set indirect_return_bnd_needed
12550 and indirect_return_needed instead of indirect_thunk_bnd_needed
12551 and indirect_thunk_needed.
12553 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
12555 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
12556 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
12557 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
12559 2018-03-15 David Malcolm <dmalcolm@redhat.com>
12560 Paul Hua <paul.hua.gm@gmail.com>
12563 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
12565 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
12567 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
12568 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
12569 resp. SFmode cases.
12571 2018-03-15 Tamar Christina <tamar.christina@arm.com>
12574 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
12575 instead of GET_MODE_SIZE when comparing Units.
12577 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
12580 * varasm.c (hash_section): Return an unchangeble hash value
12581 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
12582 Return !aarch64_can_use_per_function_literal_pools_p ().
12584 2018-03-15 Jakub Jelinek <jakub@redhat.com>
12587 * optabs.c (emit_conditional_move): Pass address of cmode's copy
12588 rather than address of cmode as last argument to prepare_cmp_insn.
12590 2018-03-15 Julia Koval <julia.koval@intel.com>
12592 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
12593 F_AVX512VNNI, F_AVX512BITALG): New.
12595 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
12598 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
12599 insn for floating-point loads and stores.
12601 2018-03-14 Carl Love <cel@us.ibm.com>
12603 * config/rs6000/rs6000-c.c: Add macro definitions for
12604 ALTIVEC_BUILTIN_VEC_PERMXOR.
12605 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
12606 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
12607 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
12608 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
12610 * config/doc/extend.texi: Add prototypes for vec_permxor.
12612 2018-03-14 David Malcolm <dmalcolm@redhat.com>
12615 * diagnostic-show-locus.c (class layout_point): Convert m_line
12616 from int to linenum_type.
12617 (line_span::comparator): Use linenum "compare" function when
12618 comparing line numbers.
12619 (test_line_span): New function.
12620 (layout_range::contains_point): Convert param "row" from int to
12622 (layout_range::intersects_line_p): Likewise.
12623 (layout::will_show_line_p): Likewise.
12624 (layout::print_source_line): Likewise.
12625 (layout::should_print_annotation_line_p): Likewise.
12626 (layout::print_annotation_line): Likewise.
12627 (layout::print_leading_fixits): Likewise.
12628 (layout::annotation_line_showed_range_p): Likewise.
12629 (struct line_corrections): Likewise for field m_row.
12630 (line_corrections::line_corrections): Likewise for param "row".
12631 (layout::print_trailing_fixits): Likewise.
12632 (layout::get_state_at_point): Likewise.
12633 (layout::get_x_bound_for_row): Likewise.
12634 (layout::print_line): Likewise.
12635 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
12636 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
12637 * input.c (selftest::test_linenum_comparisons): New function.
12638 (selftest::input_c_tests): Call it.
12639 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
12640 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
12641 * selftest.h (ASSERT_GT): New macro.
12642 (ASSERT_GT_AT): New macro.
12643 (ASSERT_LT): New macro.
12644 (ASSERT_LT_AT): New macro.
12646 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
12648 PR rtl-optimization/84780
12649 * combine.c (distribute_links): Don't make a link based on pc_rtx.
12651 2018-03-14 Martin Liska <mliska@suse.cz>
12653 * tree.c (record_node_allocation_statistics): Use
12654 get_stats_node_kind.
12655 (get_stats_node_kind): New function extracted from
12656 record_node_allocation_statistics.
12657 (free_node): Use get_stats_node_kind.
12659 2018-03-14 Richard Biener <rguenther@suse.de>
12661 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
12662 that the value-set of ANTIC_IN doesn't grow.
12665 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
12667 (BB_VISITED_WITH_VISITED_SUCCS): New define.
12668 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
12670 2018-03-14 Julia Koval <julia.koval@intel.com>
12672 * config.gcc (icelake-client, icelake-server): New.
12674 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
12675 (initial_ix86_arch_features): Ditto.
12676 (PTA_SKYLAKE): Add SGX.
12677 (PTA_ICELAKE): Remove.
12678 (PTA_ICELAKE_CLIENT): New.
12679 (PTA_ICELAKE_SERVER): New.
12680 (ix86_option_override_internal): Split up icelake on icelake client and
12682 (get_builtin_code_for_version): Ditto.
12683 (fold_builtin_cpu): Ditto.
12684 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
12685 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
12686 * config/i386/i386.h (processor_type): Ditto.
12687 * doc/invoke.texi: Ditto.
12689 2018-03-14 Jakub Jelinek <jakub@redhat.com>
12692 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
12693 INTEGER_CST offset, add it together with bitpos / 8 and
12694 sign extend based on POINTER_SIZE.
12698 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
12701 * config/i386/constraints.md (Yc): New register constraint.
12702 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
12703 Use Yc constraint for alternative 2 of operand 0. Remove
12704 preferred_for_speed attribute.
12706 2018-03-14 Richard Biener <rguenther@suse.de>
12708 PR tree-optimization/84830
12709 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
12710 with the old one to avoid oscillations.
12712 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
12715 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
12717 (assign_by_spills): Return a flag of reload assignment failure.
12718 Do not process the reload assignment failures. Do not spill other
12719 reload pseudos if they has the same reg class. Update n if
12721 (lra_assign): Add a return arg. Set up from the result of
12722 assign_by_spills call.
12723 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
12724 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
12725 usage_insns if it is not NULL.
12726 (spill_hard_reg_in_range): New function.
12727 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
12728 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
12729 function prototypes.
12730 (lra_assign): Change prototype.
12731 * lra.c (lra): Add code to deal with fails by splitting hard reg
12734 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
12736 * config/riscv/riscv.opt (mrelax): New option.
12737 * config/riscv/riscv.c (riscv_file_start): Emit ".option
12738 "norelax" when riscv_mrelax is disabled.
12739 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
12741 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12744 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
12745 reassociation for int modes.
12747 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
12749 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
12750 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
12752 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
12753 * config/aarch64/aarch64-sve.md
12754 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
12755 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
12756 (*extend<mode><Vwide>2): Rename to...
12757 (aarch64_sve_extend<mode><Vwide>2): ...this.
12758 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
12759 renaming the old pattern to...
12760 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
12762 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
12763 define_expand, renaming the old pattern to...
12764 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
12765 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
12766 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
12767 account when deciding which SVE instruction the optab should use.
12768 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
12770 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
12772 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
12773 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
12774 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
12775 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
12776 (tlsdesc_small_<mode>): Turn a define_expand and use
12777 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
12778 (tlsdesc_small_advsimd_<mode>): ...this.
12779 (tlsdesc_small_sve_<mode>): New pattern.
12781 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
12783 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
12784 (UNSPEC_UMUL_HIGHPART): New constants.
12785 (MUL_HIGHPART): New int iteraor.
12786 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
12787 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
12789 (*<su>mul<mode>3_highpart): New define_insn.
12791 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
12794 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
12797 2018-03-13 Martin Liska <mliska@suse.cz>
12800 * (sem_item_optimizer::sem_item_optimizer): Initialize new
12802 (sem_item_optimizer::~sem_item_optimizer): Release it.
12803 (sem_item_optimizer::merge_classes): Register variable aliases.
12804 (sem_item_optimizer::fixup_pt_set): New function.
12805 (sem_item_optimizer::fixup_points_to_sets): Likewise.
12806 * ipa-icf.h: Declare new variables and functions.
12808 2018-03-13 Jakub Jelinek <jakub@redhat.com>
12810 PR middle-end/84834
12811 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
12812 integer_pow2p@2 and test integer_pow2p in condition.
12813 (A < 0 ? C : 0): Similarly for @1.
12815 PR middle-end/84831
12816 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
12817 characters starting at p contain '\0' character, don't look beyond
12821 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
12822 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
12825 * reg-stack.c (change_stack): Change update_end var from int to
12826 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
12827 also call set_block_for_insn on the newly added insns and rescan.
12830 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
12831 on the last operand.
12834 * tree.c (stabilize_reference_1): Return save_expr (e) for
12835 STATEMENT_LIST even if it doesn't have side-effects.
12837 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
12839 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
12841 2018-03-12 Renlin Li <renlin.li@arm.com>
12843 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
12844 aarch64_output_scalar_simd_mov_immediate.
12846 2018-03-12 Martin Sebor <msebor@redhat.com>
12848 PR tree-optimization/83456
12849 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
12850 for perfectly overlapping calls to memcpy.
12851 (gimple_fold_builtin_memory_chk): Same.
12852 (gimple_fold_builtin_strcpy): Handle no-warning.
12853 (gimple_fold_builtin_stxcpy_chk): Same.
12854 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
12856 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
12858 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
12859 parameter. Use it for SFmode.
12860 (rs6000_function_arg_advance_1): Adjust.
12861 (rs6000_function_arg): Adjust.
12862 (rs6000_gimplify_va_arg): Pass false for that new parameter.
12864 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
12866 PR rtl-optimization/84169
12867 PR rtl-optimization/84780
12868 * combine.c (can_combine_p): Check for a 2-insn combination whether
12869 the destination register is used between the two insns, too.
12871 2018-03-12 Richard Biener <rguenther@suse.de>
12873 PR tree-optimization/84803
12874 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
12875 for refs DR analysis didn't process.
12877 2018-03-12 Richard Biener <rguenther@suse.de>
12879 PR tree-optimization/84777
12880 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
12881 force-vectorize loops ignore whether we are optimizing for size.
12883 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
12885 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
12886 (TARGET_MD_ASM_ADJUST): Define.
12888 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
12889 Kito Cheng <kito.cheng@gmail.com>
12890 Chung-Ju Wu <jasonwucj@gmail.com>
12892 * config/nds32/nds32.c (nds32_compute_stack_frame,
12893 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
12894 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
12895 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
12896 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
12897 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
12898 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
12899 * config/nds32/nds32.md (prologue, epilogue): Use macro
12900 NDS32_V3PUSH_AVAILABLE_P to do checking.
12902 2018-03-11 Jakub Jelinek <jakub@redhat.com>
12905 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
12906 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
12907 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
12908 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
12909 addition of most attributes on !orig_type_die or the attribute not
12910 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
12912 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
12913 Chung-Ju Wu <jasonwucj@gmail.com>
12915 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
12916 __NDS32_VH__ macro.
12917 * config/nds32/nds32.opt (mvh): New option.
12919 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
12920 Chung-Ju Wu <jasonwucj@gmail.com>
12922 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
12924 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
12925 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
12928 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
12929 Chung-Ju Wu <jasonwucj@gmail.com>
12931 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
12933 * config/nds32/nds32-multiple.md (strlensi): New pattern.
12934 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
12936 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
12937 Kito Cheng <kito.cheng@gmail.com>
12938 Chung-Ju Wu <jasonwucj@gmail.com>
12940 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
12941 UNSPEC_FFMISM and UNSPEC_FLMISM.
12942 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
12943 for ffb, ffmism and flmism.
12944 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
12945 (unspec_ffmism): Ditto.
12946 (unspec_flmism): Ditto.
12947 (nds32_expand_builtin_impl): Check if string extension is available.
12948 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
12949 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
12951 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
12954 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
12957 * lra-assigns.c (assign_by_spills): Return a flag of reload
12958 assignment failure. Do not process the reload assignment
12959 failures. Do not spill other reload pseudos if they has the same
12961 (lra_assign): Add a return arg. Set up from the result of
12962 assign_by_spills call.
12963 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
12964 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
12965 usage_insns if it is not NULL.
12966 (spill_hard_reg_in_range): New function.
12967 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
12968 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
12969 function prototypes.
12970 (lra_assign): Change prototype.
12971 * lra.c (lra): Add code to deal with fails by splitting hard reg
12974 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
12977 * config/i386/i386.opt: Replace Enforcment with Enforcement.
12979 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
12982 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
12983 (dw_val_node): Add val_symbolic_view.
12984 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
12985 (symview_upper_bound): New.
12986 (new_line_info_table): Initialize symviews_since_reset.
12987 (dwarf2out_source_line): Count symviews_since_reset and set
12988 symview_upper_bound.
12989 (dw_val_equal_p): Handle symview.
12990 (add_AT_symview): New.
12991 (print_dw_val): Handle symview.
12992 (attr_checksum, attr_checksum_ordered): Likewise.
12993 (same_dw_val_p, size_of_die): Likewise.
12994 (value_format, output_die): Likewise.
12995 (add_high_low_attributes): Use add_AT_symview for entry_view.
12996 (dwarf2out_finish): Reset symview_upper_bound, clear
12999 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
13002 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
13003 Add strict argument and use it.
13004 (rs6000_split_multireg_move): Update for new strict argument.
13005 (mem_operand_gpr): Disallow all non-offsettable addresses.
13006 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
13008 2018-03-09 Jakub Jelinek <jakub@redhat.com>
13011 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
13012 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
13013 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
13016 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
13017 decl, use remap_type if we want to use the type.
13019 2018-03-09 Martin Sebor <msebor@redhat.com>
13021 PR tree-optimization/84526
13022 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
13024 (builtin_access::generic_overlap): Be prepared to handle non-array
13027 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
13029 PR rtl-optimization/84682
13030 * lra-constraints.c (process_address_1): Check is_address flag
13031 for address constraints.
13032 (process_alt_operands): Likewise.
13033 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
13034 preprocess_constraints.
13035 * recog.h (preprocess_constraints): Add oploc parameter.
13037 * recog.c (preprocess_constraints): Test address_operand for
13038 CT_ADDRESS constraints.
13040 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
13043 * lra-assigns.c (assign_by_spills): Return a flag of reload
13044 assignment failure. Do not process the reload assignment
13045 failures. Do not spill other reload pseudos if they has the same
13047 (lra_assign): Add a return arg. Set up from the result of
13048 assign_by_spills call.
13049 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
13050 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
13051 usage_insns if it is not NULL.
13052 (spill_hard_reg_in_range): New function.
13053 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
13054 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
13055 function prototypes.
13056 (lra_assign): Change prototype.
13057 * lra.c (lra): Add code to deal with fails by splitting hard reg
13060 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13063 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
13064 Accept complain bool parameter. Only emit errors if it is true.
13065 (arm_parse_cpu_option_name): Likewise.
13066 (arm_target_thumb_only): Adjust callers of the above.
13067 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
13068 prototype to take a default true bool parameter.
13069 (arm_parse_arch_option_name): Likewise.
13071 2018-03-09 David Malcolm <dmalcolm@redhat.com>
13072 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
13076 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
13077 * configure: Regenerate.
13078 * configure.ac ("linker --version-script option"): New.
13079 ("linker soname option"): New.
13081 2018-03-09 Richard Biener <rguenther@suse.de>
13083 PR tree-optimization/84775
13084 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
13085 immediate uses of predicate stmts and mark them modified.
13088 PR tree-optimization/84178
13089 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
13091 (version_loop_for_if_conversion): Delay update_ssa call.
13092 (tree_if_conversion): Delay update_ssa until after predicate
13095 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
13098 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
13099 when the function accesses prior frames.
13101 2018-03-08 Jakub Jelinek <jakub@redhat.com>
13104 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
13105 gen_llsym, otherwise call maybe_gen_llsym.
13107 PR inline-asm/84742
13108 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
13109 has ',' character inside of it.
13111 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13114 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
13115 as clobbering CC_REGNUM.
13117 2018-03-08 Richard Biener <rguenther@suse.de>
13119 PR middle-end/84552
13120 * tree-scalar-evolution.c: Include tree-into-ssa.h.
13121 (follow_copies_to_constant): Do not follow SSA names registered
13124 2018-03-08 Richard Biener <rguenther@suse.de>
13126 PR tree-optimization/84178
13127 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
13129 (version_loop_for_if_conversion): Delay update_ssa call.
13130 (tree_if_conversion): Delay update_ssa until after predicate
13133 2018-03-08 David Malcolm <dmalcolm@redhat.com>
13135 PR tree-optimization/84178
13136 * tree-if-conv.c (release_bb_predicate): Remove the
13137 the assertion that the stmts have NULL use_ops.
13138 Discard the statements, asserting that they haven't
13139 yet been added to a BB.
13141 2018-03-08 Richard Biener <rguenther@suse.de>
13143 PR tree-optimization/84746
13144 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
13145 (phi_translate): Pass in destination ANTIC_OUT set.
13146 (phi_translate_1): Likewise. For a simplified result lookup
13147 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
13148 (phi_translate_set): Adjust.
13149 (do_pre_regular_insertion): Likewise.
13150 (do_pre_partial_partial_insertion): Likewise.
13152 2018-03-08 Martin Liska <mliska@suse.cz>
13154 PR gcov-profile/84735
13155 * doc/gcov.texi: Document usage of profile files.
13156 * gcov-io.h: Document changes in the format.
13158 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
13162 * dwarf2out.c (struct dw_line_info_table): Update comments for
13164 (FORCE_RESET_NEXT_VIEW): New.
13165 (FORCE_RESETTING_VIEW_P): New.
13166 (RESETTING_VIEW_P): Check for -1 too.
13167 (ZERO_VIEW_P): Likewise.
13168 (new_line_info_table): Force-reset next view.
13169 (dwarf2out_begin_function): Likewise.
13170 (dwarf2out_source_line): Simplify zero_view_p initialization.
13171 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
13172 view directly. Omit view when omitting .loc at line 0.
13174 2018-03-08 Jakub Jelinek <jakub@redhat.com>
13176 PR tree-optimization/84740
13177 * tree-switch-conversion.c (process_switch): Call build_constructors
13178 only if info.phi_count is non-zero.
13180 PR tree-optimization/84739
13181 * tree-tailcall.c (find_tail_calls): Check call arguments against
13182 DECL_ARGUMENTS (current_function_decl) rather than
13183 DECL_ARGUMENTS (func) when checking for tail recursion.
13185 2018-03-07 Jakub Jelinek <jakub@redhat.com>
13187 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
13188 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
13189 Volker Reichelt's entry and add entries for people that perform
13190 GCC fuzzy testing and report numerous bugs.
13192 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
13195 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
13196 readonly data in sdata, if that is disabled.
13197 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
13198 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
13199 -mreadonly-in-sdata option.
13201 2018-03-07 Martin Sebor <msebor@redhat.com>
13203 PR tree-optimization/84468
13204 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
13205 basic block when looking for nul assignment.
13207 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
13210 * except.h (output_function_exception_table): Adjust prototype.
13211 * except.c (output_function_exception_table): Remove FNNAME parameter
13212 and add SECTION parameter. Ouput one part of the table at a time.
13213 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
13214 the first part of the exception table and emit unwind directives.
13215 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
13216 (i386_pe_seh_cold_init): Likewise.
13217 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
13218 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
13219 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
13220 (ix86_output_call_insn): Emit a nop in one more case for SEH.
13221 * config/i386/winnt.c: Include except.h.
13222 (struct seh_frame_state): Add reg_offset, after_prologue and
13223 in_cold_section fields.
13224 (i386_pe_seh_end_prologue): Set seh->after_prologue.
13225 (i386_pe_seh_cold_init): New function.
13226 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
13227 to seh->in_cold_section.
13228 (seh_emit_push): Record the offset of the push.
13229 (seh_emit_save): Record the offet of the save.
13230 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
13231 Test seh->after_prologue to disregard the epilogue.
13232 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
13233 (i386_pe_end_cold_function): New function.
13235 2018-03-07 Jakub Jelinek <jakub@redhat.com>
13238 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
13239 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
13242 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
13244 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
13245 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
13247 PR middle-end/84723
13248 * multiple_target.c: Include tree-inline.h and intl.h.
13249 (expand_target_clones): Diagnose and fail if node->definition and
13250 !tree_versionable_function_p (node->decl).
13252 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
13254 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
13256 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
13257 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
13258 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
13260 2018-03-06 Jakub Jelinek <jakub@redhat.com>
13263 * combine.c (try_combine): Use reg_or_subregno instead of handling
13264 just paradoxical SUBREGs and REGs.
13266 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
13268 * config/arc/arc.c (arc_finalize_pic): Remove function.
13269 (arc_must_save_register): We use single base PIC register, remove
13270 checks to save/restore the PIC register.
13271 (arc_expand_prologue): Likewise.
13272 * config/arc/arc-protos.h (arc_set_default_type_attributes):
13274 (arc_verify_short): Likewise.
13275 (arc_attr_type): Likewise.
13276 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
13277 (walk_stores): Likewise.
13278 (arc_address_cost): Make it static.
13279 (arc_verify_short): Likewise.
13280 (branch_dest): Likewise.
13281 (arc_attr_type): Likewise.
13282 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
13283 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
13284 (arc_final_prescan_insn): Remove inserting the nops due to
13285 hardware hazards. It is done in reorg step.
13286 (insn_length_variant_t): Remove.
13287 (insn_length_parameters_t): Likewise.
13288 (arc_insn_length_parameters): Likewise.
13289 (arc_get_insn_variants): Likewise.
13290 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
13292 2018-03-06 Jakub Jelinek <jakub@redhat.com>
13294 PR inline-asm/84683
13295 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
13298 PR tree-optimization/84687
13299 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
13301 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
13303 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13305 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
13306 Rename to ppc_speculation_barrier.
13307 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
13308 __builtin_ppc_speculation_barrier.
13310 2018-03-05 Jakub Jelinek <jakub@redhat.com>
13313 * combine.c (combine_simplify_rtx): Don't try to simplify if
13314 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
13317 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
13319 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
13320 to 32 bytes when compiling for POWER9.
13322 2018-03-05 Jakub Jelinek <jakub@redhat.com>
13325 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
13326 regparm >= 3 with no arg reg available also for calls with
13327 flag_force_indirect_call. Pass decl to ix86_function_regparm.
13330 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
13332 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
13334 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
13337 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
13339 2018-03-05 Richard Biener <rguenther@suse.de>
13341 PR tree-optimization/84486
13342 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
13343 When inserting a __builtin_assume_aligned call set the LHS
13344 SSA name alignment info accordingly.
13346 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
13348 PR tree-optimization/84114
13349 * config/aarch64/aarch64.c (aarch64_reassociation_width)
13350 Avoid reassociation of FLOAT_MODE addition.
13352 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
13354 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
13355 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
13356 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
13357 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
13358 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
13359 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
13360 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
13362 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
13363 __builtin_ia32_wbinvd): New builtins.
13364 (SPECIAL_ARGS2): New.
13365 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
13366 (SPECIAL_ARGS2): New.
13367 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
13368 (ix86_valid_target_attribute_inner_p): Ditto.
13369 (ix86_init_mmx_sse_builtins): Add special_args2.
13370 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
13371 TARGET_WBNOINVD_P): New.
13372 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
13373 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
13374 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
13375 * config/i386/immintrin.h (_wbinvd): New intrinsic.
13376 * config/i386/pconfigintrin.h: New file.
13377 * config/i386/wbnoinvdintrin.h: Ditto.
13378 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
13380 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
13382 2018-03-05 Richard Biener <rguenther@suse.de>
13384 PR tree-optimization/84670
13385 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
13387 (BB_VISITED_WITH_VISITED_SUCCS): New define.
13388 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
13389 (compute_antic_aux): Only assert the number of values in ANTIC_IN
13390 doesn't grow if all successors (recursively) were visited at least
13393 2018-03-05 Richard Biener <rguenther@suse.de>
13395 PR tree-optimization/84650
13396 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
13397 if executed in the loop pipeline.
13399 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
13401 * doc/configfiles.texi (Configuration Files): Move info about
13402 conditionalizing $target-protos.h to...
13403 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
13404 differs from $target-protos.h.
13406 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
13407 Chung-Ju Wu <jasonwucj@gmail.com>
13409 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
13410 * config/nds32/nds32-multiple.md (setmemsi): Define.
13411 * config/nds32/nds32-memory-manipulation.c
13412 (nds32_gen_dup_4_byte_to_word_value): New.
13413 (emit_setmem_word_loop): New.
13414 (emit_setmem_byte_loop): New.
13415 (nds32_expand_setmem_loop): New.
13416 (nds32_expand_setmem_loop_v3m): New.
13417 (nds32_expand_setmem_unroll): New.
13418 (nds32_expand_setmem): New.
13420 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
13421 Chung-Ju Wu <jasonwucj@gmail.com>
13423 * config/nds32/nds32-memory-manipulation.c
13424 (nds32_emit_load_store): New.
13425 (nds32_emit_post_inc_load_store): New.
13426 (nds32_emit_mem_move): New.
13427 (nds32_emit_mem_move_block): New.
13428 (nds32_expand_movmemsi_loop_unknown_size): New.
13429 (nds32_expand_movmemsi_loop_known_size): New.
13430 (nds32_expand_movmemsi_loop): New.
13431 (nds32_expand_movmemsi_unroll): New.
13432 (nds32_expand_movmemqi): Rename ...
13433 (nds32_expand_movmemsi): ... to this.
13434 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
13435 (movmemsi): ... to this.
13436 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
13437 (nds32_expand_movmemsi): ... to this.
13439 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
13440 Monk Chiang <sh.chiang04@gmail.com>
13441 Chung-Ju Wu <jasonwucj@gmail.com>
13443 * config/nds32/nds32-protos.h
13444 (nds32_expand_load_multiple): New arguments.
13445 (nds32_expand_store_multiple): Ditto.
13446 (nds32_valid_multiple_load_store): Rename ...
13447 (nds32_valid_multiple_load_store_p): ... to this.
13448 * config/nds32/nds32-memory-manipulation.c
13449 (nds32_expand_load_multiple): Refine implementation.
13450 (nds32_expand_store_multiple): Ditto.
13451 * config/nds32/nds32-multiple.md
13452 (load_multiple): Update nds32_expand_load_multiple interface.
13453 (store_multiple): Update nds32_expand_store_multiple interface.
13454 * config/nds32/nds32-predicates.c
13455 (nds32_valid_multiple_load_store): Rename ...
13456 (nds32_valid_multiple_load_store_p): ... to this and refine
13458 * config/nds32/predicates.md
13459 (nds32_load_multiple_and_update_address_operation): New predicate.
13460 (nds32_store_multiple_and_update_address_operation): New predicate.
13462 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
13463 Chung-Ju Wu <jasonwucj@gmail.com>
13465 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
13466 (combo): New attribute.
13467 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
13469 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
13471 * config/nds32/nds32.opt: Change -mcmodel= default value.
13473 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
13474 Monk Chiang <sh.chiang04@gmail.com>
13475 Chung-Ju Wu <jasonwucj@gmail.com>
13477 * config/nds32/constants.md (unspec_element): New enum.
13478 * config/nds32/constraints.md (Umw): New constraint.
13479 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
13480 * config/nds32/nds32-intrinsic.md: Likewise.
13481 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
13482 (nds32_valid_smw_lwm_base_p): New.
13483 (nds32_output_smw_single_word): New.
13484 (nds32_output_lmw_single_word): New.
13485 (nds32_expand_unaligned_load): New.
13486 (nds32_expand_unaligned_store): New.
13487 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
13488 (nds32_output_smw_single_word): Declare.
13489 (nds32_output_lmw_single_word): Declare.
13490 (nds32_expand_unaligned_load): Declare.
13491 (nds32_expand_unaligned_store): Declare.
13492 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
13493 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
13494 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
13495 NDS32_BUILTIN_UASTORE_DW.
13496 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
13499 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
13500 Kito Cheng <kito.cheng@gmail.com>
13501 Chung-Ju Wu <jasonwucj@gmail.com>
13503 * config/nds32/nds32-intrinsic.c
13504 (nds32_expand_builtin_null_ftype_reg): Delete.
13505 (nds32_expand_builtin_reg_ftype_imm): Ditto.
13506 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
13507 (nds32_read_argument): New.
13508 (nds32_legitimize_target): Ditto.
13509 (nds32_legitimize_argument): Ditto.
13510 (nds32_check_constant_argument): Ditto.
13511 (nds32_expand_unop_builtin): Ditto.
13512 (nds32_expand_unopimm_builtin): Ditto.
13513 (nds32_expand_binop_builtin): Ditto.
13514 (nds32_builtin_decl_impl): Ditto.
13515 (builtin_description): Ditto.
13516 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
13517 (nds32_init_builtins_impl): Ditto.
13518 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
13519 (nds32_builtin_decl): New.
13520 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
13521 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
13523 2018-03-02 Jeff Law <law@redhat.com>
13525 * reorg.c (stop_search_p): Handle DEBUG_INSN.
13526 (redundant_insn, fill_simple_delay_slots): Likewise.
13527 (fill_slots_from_thread): Likewise.
13528 * resource.c (mark_referenced_resources): Likewise.
13529 (mark_set_resources, find_dead_or_set_registers): Likewise.
13531 2018-03-02 Jakub Jelinek <jakub@redhat.com>
13533 * substring-locations.h (format_warning_va): Formatting fix for
13534 ATTRIBUTE_GCC_DIAG.
13535 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
13537 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
13538 * substring-locations.c: Include intl.h.
13539 (format_warning_va): Turned into small wrapper around
13540 format_warning_n_va, renamed to ...
13541 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
13542 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
13544 (format_warning_at_substring_n): New function.
13545 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
13546 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
13547 format_warning_at_substring with just a shorter name instead of
13548 const function pointer.
13549 (fmtwarn_n): New function.
13550 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
13551 appropriate, get rid of all the fmtstr temporaries, move conditionals
13552 with G_() wrapped string literals directly into fmtwarn arguments,
13553 cast dir.len to (int), formatting fixes.
13555 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
13557 * doc/invoke.texi: Remove "Cilk Plus" references.
13559 2018-03-02 Jakub Jelinek <jakub@redhat.com>
13560 Richard Biener <rguenther@suse.de>
13563 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
13564 for error or warning attributes if CALL_FROM_THUNK_P is set.
13567 2018-03-02 Jakub Jelinek <jakub@redhat.com>
13570 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
13571 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
13574 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
13575 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
13577 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
13578 instead of -1U in last predictors element's probability member.
13580 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
13583 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
13584 arguments if they are comparable.
13586 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
13588 PR tree-optimization/84634
13589 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
13590 masks and masked_loop_p with a single loop_masks, making sure it's
13591 null for bb vectorization.
13593 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
13595 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
13596 (vect_analyze_data_ref_access): Use loop->safe_len rather than
13597 loop->force_vectorize to check whether there is no alias.
13599 2018-03-02 Jakub Jelinek <jakub@redhat.com>
13602 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
13604 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
13606 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
13607 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
13608 instead of a loop around prev_real_insn.
13609 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
13612 PR inline-asm/84625
13613 * config/i386/i386.c (ix86_print_operand): Use conditional
13614 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
13617 2018-03-02 Richard Biener <rguenther@suse.de>
13619 PR tree-optimization/84427
13620 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
13621 (bitmap_set_subtract_values): Rewrite to handle multiple
13624 (prune_clobbered_mems): Likewise.
13625 (phi_translate): Take edge instead of pred/phiblock.
13626 (phi_translate_1): Likewise.
13627 (phi_translate_set): Likewise. Insert all translated
13628 exprs for a value into the set, keeping possibly multiple
13629 expressions per value.
13630 (compute_antic_aux): Adjust for phi_translate changes.
13631 When intersecting union the expressions and prune those
13632 not in the final value set, keeping possibly multiple
13633 expressions per value. Do not use value-insertion
13634 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
13635 all expressions. Add verification that the value-sets
13636 only shrink during iteration.
13637 (compute_partial_antic_aux): Adjust for the phi_translate changes.
13638 (do_pre_regular_insertion): Likewise.
13639 (do_pre_partial_partial_insertion): Likewise.
13641 2018-03-02 Richard Biener <rguenther@suse.de>
13644 * config/darwin.c (saved_debug_info_level): New static global.
13645 (darwin_asm_lto_start): Disable debug info generation for LTO out.
13646 (darwin_asm_lto_end): Restore debug info generation settings.
13648 2018-03-01 Martin Liska <mliska@suse.cz>
13651 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
13652 volatile arguments.
13654 2018-03-01 Richard Biener <rguenther@suse.de>
13657 * dwarf2out.c (gen_variable_die): Properly handle late VLA
13658 type annotation with LTO when debug was disabled at compile-time.
13660 2018-03-01 Matthew Fortune <mfortune@gmail.com>
13662 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
13664 (mips_final_postscan_insn): Likewise.
13666 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
13668 PR rtl-optimization/84528
13669 * alias.c (init_alias_target): Add commentary.
13670 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
13671 a unique base value if the frame pointer is not eliminated
13672 to the stack pointer.
13674 2018-03-01 Tom de Vries <tom@codesourcery.com>
13676 PR rtl-optimization/83327
13677 * lra-int.h (hard_regs_spilled_into): Declare.
13678 * lra.c (hard_regs_spilled_into): Define.
13679 (init_reg_info): Init hard_regs_spilled_into.
13680 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
13681 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
13682 (process_bb_lives): Handle hard_regs_spilled_into.
13683 (lra_create_live_ranges_1): Before doing liveness propagation, clear
13684 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
13686 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
13688 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
13689 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
13690 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
13691 * config/rs6000/aix72.h: New file.
13693 2018-02-28 Jakub Jelinek <jakub@redhat.com>
13695 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
13696 instead of warning_at with conditional singular and plural messages
13700 * stor-layout.c (update_alignment_for_field): For
13701 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
13702 && !DECL_PACKED (field), do the alignment update, just use
13703 only desired_align instead of MAX (type_align, desired_align)
13705 (place_field): Don't do known_align < desired_align handling
13706 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
13707 is non-NULL, instead do it after rli->prev_field handling and
13708 only if not within a bitfield word. For DECL_PACKED (field)
13709 use type_align of BITS_PER_UNIT.
13711 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
13713 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
13714 superfluous parentheses and trailing spaces.
13716 2018-02-28 Richard Biener <rguenther@suse.de>
13718 PR tree-optimization/84584
13719 * graphite-scop-detection.c (scop_detection::add_scop): Discard
13720 SCoPs with fake exit edge.
13722 2018-02-28 Martin Liska <mliska@suse.cz>
13725 * timevar.c (timer::print): Fix format to properly print 100%
13728 2018-02-28 Richard Biener <rguenther@suse.de>
13730 PR middle-end/84607
13731 * genmatch.c (capture_info::walk_match): Do not mark
13732 captured expressions without operands as expr_p given
13733 they act more like predicates and should be subject to
13734 "lost tail" side-effect preserving.
13736 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
13738 PR rtl-optimization/81611
13739 * auto-inc-dec.c (attempt_change): Move dead note from
13740 mem_insn if it's the next use of regno
13741 (find_address): Take address use of reg holding
13742 non-incremented value. Add parm to limit search to the named
13744 (merge_in_block): Attempt to use a mem insn that is the next
13745 use of the original regno.
13747 2018-02-27 Martin Sebor <msebor@redhat.com>
13750 * doc/invoke.texi (-Wmissing-attributes): New option.
13751 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
13753 2018-02-27 Martin Sebor <msebor@redhat.com>
13755 PR translation/84207
13756 * diagnostic-core.h (warning_n, error_n, inform_n): Change
13757 n argument to unsigned HOST_WIDE_INT.
13758 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
13759 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
13760 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
13761 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
13763 2018-02-27 Richard Biener <rguenther@suse.de>
13765 PR tree-optimization/84512
13766 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
13767 Do not use the estimate returned from record_stmt_cost for
13768 the scalar iteration cost but sum properly using add_stmt_cost.
13770 2018-02-27 Richard Biener <rguenther@suse.de>
13772 PR tree-optimization/84466
13773 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
13774 Adjust last change to less strictly validate use operands.
13776 2018-02-27 Martin Liska <mliska@suse.cz>
13778 PR gcov-profile/84548
13779 * gcov.c (process_file): Allow partial overlap and consider it
13780 also as group functions.
13781 (output_lines): Properly calculate range of lines for a group.
13783 2018-02-27 Martin Liska <mliska@suse.cz>
13785 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
13786 'ggc' suffixes. Change first column width.
13787 (timer::print): Fix formatting of the column.
13789 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
13791 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
13792 preserve inline entry blocks for the sake of debug inline
13793 entry point markers alone.
13794 (remove_unused_locals): Suggest in comments a better place to
13795 force the preservation of inline entry blocks that are
13796 otherwise unused, but do not preserve them.
13798 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
13800 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
13802 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
13805 * config/i386/constraints.md (Bs): Replace
13806 ix86_indirect_branch_register with
13807 TARGET_INDIRECT_BRANCH_REGISTER.
13809 * config/i386/i386.md (indirect_jump): Likewise.
13810 (tablejump): Likewise.
13811 (*sibcall_memory): Likewise.
13812 (*sibcall_value_memory): Likewise.
13813 Peepholes of indirect call and jump via memory: Likewise.
13814 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
13815 (*sibcall_value_GOT_32): Likewise.
13816 * config/i386/predicates.md (indirect_branch_operand): Likewise.
13817 (GOT_memory_operand): Likewise.
13818 (call_insn_operand): Likewise.
13819 (sibcall_insn_operand): Likewise.
13820 (GOT32_symbol_operand): Likewise.
13821 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
13823 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
13825 PR rtl-optimization/83496
13826 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
13827 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
13828 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
13829 redundant insn, if any.
13830 (relax_delay_slots): Likewise.
13831 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
13833 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
13835 PR tree-optimization/83965
13836 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
13837 that grouped statements are part of a reduction chain. Return
13838 true if the statement is not marked as a reduction itself but
13839 is part of a group.
13840 (vect_recog_dot_prod_pattern): Don't check whether the statement
13841 is part of a group here.
13842 (vect_recog_sad_pattern): Likewise.
13843 (vect_recog_widen_sum_pattern): Likewise.
13845 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
13848 * final.c (rest_of_clean_state): Also look for calls inside sequences.
13850 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
13853 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
13855 (ix86_output_indirect_function_return): New prototype.
13856 (ix86_split_simple_return_pop_internal): Likewise.
13857 * config/i386/i386.c (indirect_return_via_cx): New.
13858 (indirect_return_via_cx_bnd): Likewise.
13859 (indirect_thunk_name): Handle return va CX_REG.
13860 (output_indirect_thunk_function): Create alias for
13861 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
13862 (ix86_output_indirect_jmp): Remove the bool argument.
13863 (ix86_output_indirect_function_return): New function.
13864 (ix86_split_simple_return_pop_internal): Likewise.
13865 * config/i386/i386.md (*indirect_jump): Don't pass false
13866 to ix86_output_indirect_jmp.
13867 (*tablejump_1): Likewise.
13868 (simple_return_pop_internal): Change it to define_insn_and_split.
13869 Call ix86_split_simple_return_pop_internal to split it for
13870 -mfunction-return=.
13871 (simple_return_indirect_internal): Call
13872 ix86_output_indirect_function_return instead of
13873 ix86_output_indirect_jmp.
13875 2018-02-26 Jakub Jelinek <jakub@redhat.com>
13878 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
13879 memset and value initialization afterwards.
13881 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
13883 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
13885 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
13888 * common/config/aarch64/aarch64-common.c
13889 (aarch_option_optimization_table[]): Switch
13890 off fomit-frame-pointer
13892 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
13893 Chung-Ju Wu <jasonwucj@gmail.com>
13895 * config/nds32/nds32-multiple.md (load_multiple): Disallow
13897 (store_multiple): Ditto.
13899 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
13901 * config.gcc: Add --with-cpu support for nds32 target.
13902 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
13903 * config/nds32/nds32.opt: Add -mcpu= option.
13905 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
13907 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
13908 isel=yes): Warn for these deprecated options.
13910 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
13912 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
13913 ISA_2_5_MASKS_EMBEDDED.
13915 2018-02-23 Jakub Jelinek <jakub@redhat.com>
13917 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
13918 p->max as pointers rather than using iterative_hash_expr.
13920 2018-02-23 Carl Love <cel@us.ibm.com>
13922 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
13923 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
13925 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
13926 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
13927 P8V_BUILTIN_VEC_VUNSIGNED2.
13929 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
13932 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
13933 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
13935 * lra-constraints.c (curr_insn_transform): Set up
13936 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
13938 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
13939 * lra-eliminations.c (spill_pseudos): Ditto.
13940 (process_insn_for_elimination): Ditto.
13941 * lra-lives.c (reg_early_clobber_p): Use the new macros.
13942 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
13943 LRA_NON_CLOBBERED_ALT.
13945 2018-02-22 Martin Sebor <msebor@redhat.com>
13947 PR tree-optimization/84480
13948 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
13949 to maybe_diag_stxncpy_trunc. Call it.
13950 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
13951 from gimple_fold_builtin_strcpy. Print inlining stack.
13952 (handle_builtin_stxncpy): Print inlining stack.
13953 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
13955 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
13958 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
13959 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
13960 and -fcheck-pointer-bounds are used together.
13961 (indirect_thunk_prefix): New enum.
13962 (indirect_thunk_need_prefix): New function.
13963 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
13964 "_nt" instead of "_bnd" for NOTRACK prefix.
13965 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
13966 (output_indirect_thunk_function): Likewise.
13968 (ix86_code_end): Update output_indirect_thunk_function calls.
13969 (ix86_output_indirect_branch_via_reg): Replace
13970 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
13971 (ix86_output_indirect_branch_via_push): Likewise.
13972 (ix86_output_function_return): Likewise.
13973 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
13974 incompatible with -fcf-protection=branch and
13975 -fcheck-pointer-bounds.
13977 2018-02-22 Steve Ellcey <sellcey@cavium.com>
13980 * config/aarch64/aarch64.c (aarch64_print_address_internal):
13981 Change gcc_assert call to output_operand_lossage.
13983 2018-02-22 Steve Ellcey <sellcey@cavium.com>
13985 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
13987 2018-02-22 DJ Delorie <dj@redhat.com>
13988 Sebastian Perta <sebastian.perta@renesas.com>
13989 Oleg Endo <olegendo@gcc.gnu.org>
13991 * config/rx/rx.c (rx_rtx_costs): New function.
13992 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
13994 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
13996 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
13998 2018-02-22 Martin Liska <mliska@suse.cz>
14001 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
14002 Add "native" as a possible value.
14004 2018-02-22 Martin Liska <mliska@suse.cz>
14007 * config/i386/i386.c (ix86_option_override_internal):
14008 Add "native" as a possible value for -march and -mtune.
14010 2018-02-22 Jakub Jelinek <jakub@redhat.com>
14013 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
14014 to all type variants.
14016 PR tree-optimization/84503
14017 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
14018 width as info->bitpos + info->bitsize - start.
14019 (merged_store_group::merge_overlapping): Simplify width computation.
14020 (check_no_overlap): New function.
14021 (imm_store_chain_info::try_coalesce_bswap): Compute expected
14022 start + width and last_order of the group, fail if check_no_overlap
14024 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
14025 to group if check_no_overlap fails.
14027 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
14029 * config/rs6000/altivec.md: Delete contraint arguments to
14030 define_expand, define_split, and define_peephole2, and in
14031 define_insn_and_split if always unused.
14032 * config/rs6000/darwin.md: Ditto.
14033 * config/rs6000/dfp.md: Ditto.
14034 * config/rs6000/rs6000.md: Ditto.
14035 * config/rs6000/sync.md: Ditto.
14036 * config/rs6000/vector.md: Ditto.
14037 * config/rs6000/vsx.md: Ditto.
14039 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
14041 * config/rs6000/altivec.md: Write output control strings as braced
14042 blocks instead of double-quoted strings.
14043 * config/rs6000/darwin.md: Ditto.
14044 * config/rs6000/rs6000.md: Ditto.
14045 * config/rs6000/vector.md: Ditto.
14046 * config/rs6000/vsx.md: Ditto.
14048 2018-02-21 Jason Merrill <jason@redhat.com>
14050 PR c++/84314 - ICE with templates and fastcall attribute.
14051 * attribs.c (build_type_attribute_qual_variant): Remove assert.
14053 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
14055 * ipa-cp.c (determine_versionability): Fix comment typos.
14057 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
14060 * ipa-cp.c (determine_versionability): Do not version functions caling
14063 2018-02-21 Martin Liska <mliska@suse.cz>
14066 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
14067 Add "native" as a possible value.
14068 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
14069 the macro when native cpu detection is available.
14071 2018-02-21 Martin Liska <mliska@suse.cz>
14074 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
14075 Add "native" as a possible value.
14076 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
14077 when native cpu detection is available.
14079 2018-02-21 Jakub Jelinek <jakub@redhat.com>
14080 Martin Sebor <msebor@redhat.com>
14082 PR tree-optimization/84478
14083 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
14085 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
14086 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
14087 support which is conservatively correct, for 2 only stay conservative
14088 for maxlen. Formatting and comment capitalization fixes. Add STRICT
14089 argument to the 2 argument get_range_strlen, adjust 6 arg
14090 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
14092 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
14093 (gimple_fold_builtin_strlen): Pass true as last argument to
14096 2018-02-20 Martin Sebor <msebor@redhat.com>
14098 PR middle-end/84095
14099 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
14100 (builtin_memref::set_base_and_offset): Same. Handle inner references.
14101 (builtin_memref::builtin_memref): Factor out parts into
14102 set_base_and_offset and call it.
14104 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
14106 PR middle-end/84406
14107 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
14108 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
14109 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
14110 search at the associated MODE_INT.
14112 2018-02-20 Jeff Law <law@redhat.com>
14114 PR middle-end/82123
14115 PR tree-optimization/81592
14116 PR middle-end/79257
14117 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
14118 for range data rather than using global data.
14119 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
14120 range data rather than using global data.
14121 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
14122 pass it to children as needed.
14123 (struct directive::fmtresult): Similarly.
14124 (struct directive::set_width): Similarly.
14125 (struct directive::set_precision): Similarly.
14126 (format_integer, format_directive, parse_directive): Similarly.
14127 (format_none): Accept unnamed vr_values parameter.
14128 (format_percent, format_floating, format_character): Similarly.
14129 (format_string, format_plain): Similarly.
14130 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
14131 the EVRP range analyzer for range data rather than using global data.
14132 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
14133 gimple-ssa-evrp-analyze.h
14134 (class sprintf_dom_walker): Add after_dom_children member function.
14135 Add evrp_range_analyzer member.
14136 (sprintf_dom_walker::before_dom_children): Call into the EVRP
14137 range analyzer as needed.
14138 (sprintf_dom_walker::after_dom_children): New member function.
14139 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
14141 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
14142 (evrp_range_analyzer::pop_to_marker): Likewise.
14144 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
14146 PR tree-optimization/84419
14147 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
14148 with the required type if its current type is compatible but
14151 2018-02-20 Jakub Jelinek <jakub@redhat.com>
14153 PR middle-end/82004
14154 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
14155 after vectorization.
14157 2018-02-20 Martin Liska <mliska@suse.cz>
14160 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
14161 possible values if we don't have a hint.
14163 2018-02-20 Martin Liska <mliska@suse.cz>
14167 * final.c (shorten_branches): Build align_tab array with one
14169 * opts.c (finish_options): Add alignment option limit check.
14170 (MAX_CODE_ALIGN): Likewise.
14171 (MAX_CODE_ALIGN_VALUE): Likewise.
14172 * doc/invoke.texi: Document maximum allowed option value for
14173 all -falign-* options.
14175 2018-02-19 Jakub Jelinek <jakub@redhat.com>
14178 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
14179 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
14180 * var-tracking.c (emit_note_insn_var_location): Remove all references
14181 to NOTE_INSN_CALL_ARG_LOCATION.
14182 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
14183 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
14184 Use copy_rtx_if_shared.
14185 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
14186 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
14187 (dwarf2out_var_location): Remove handling of
14188 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
14190 * final.c (final_scan_insn): Remove all references to
14191 NOTE_INSN_CALL_ARG_LOCATION.
14192 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
14193 before dumping final insns.
14194 * except.c (emit_note_eh_region_end): Remove all references to
14195 NOTE_INSN_CALL_ARG_LOCATION.
14196 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
14197 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
14198 * config/arc/arc.c (hwloop_optimize): Likewise.
14199 * config/arm/arm.c (create_fix_barrier): Likewise.
14200 * config/s390/s390.c (s390_chunkify_start): Likewise.
14201 * config/sh/sh.c (find_barrier): Likewise.
14202 * config/i386/i386.c (rest_of_insert_endbranch,
14203 ix86_seh_fixup_eh_fallthru): Likewise.
14204 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
14205 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
14206 * config/frv/frv.c (frv_function_prologue): Likewise.
14207 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
14209 (note_outside_basic_block_p): Remove all references to
14210 NOTE_INSN_CALL_ARG_LOCATION.
14211 * gengtype.c (adjust_field_rtx_def): Likewise.
14212 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
14214 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
14215 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
14218 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
14221 PR tree-optimization/84452
14222 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
14223 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
14226 2018-02-19 Martin Liska <mliska@suse.cz>
14229 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
14231 2018-02-19 Martin Liska <mliska@suse.cz>
14232 Richard Sandiford <richard.sandiford@linaro.org>
14234 PR tree-optimization/82491
14235 * gimple-fold.c (get_base_constructor): Make earlier bail out
14238 2018-02-19 Carl Love <cel@us.ibm.com>
14240 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
14241 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
14243 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
14244 P8V_BUILTIN_VEC_NEG.
14246 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
14248 * config/rl78/rl78.md (movdf): New define expand.
14250 2018-02-19 Martin Liska <mliska@suse.cz>
14253 * doc/invoke.texi: Fix typo.
14254 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
14256 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
14258 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
14259 handle rs6000_single_float and rs6000_double_float specially for
14262 2018-02-16 Jeff Law <law@redhat.com>
14264 * config/rx/rx.c (add_pop_cfi_notes): New function.;
14265 (pop_regs): Use it.
14267 2018-02-16 Jakub Jelinek <jakub@redhat.com>
14270 * ipa-inline.c (inline_small_functions): Fix a typo.
14272 2018-02-16 Nathan Sidwell <nathan@acm.org>
14274 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
14276 2018-02-16 Carl Love <cel@us.ibm.com>
14278 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
14279 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
14280 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
14281 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
14282 expansion to P8V_BUILTIN_VEC_FLOAT2.
14284 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
14286 PR rtl-optimization/70023
14287 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
14288 src_regno into account.
14290 2018-02-16 Carl Love <cel@us.ibm.com>
14292 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
14293 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
14294 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
14295 * config/rs6000/rs6000.c: Remove case statements for
14296 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
14297 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
14298 and P9V_BUILTIN_VEC_VINSERT4B.
14299 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
14300 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
14301 * config/rs6000/vsx.md:
14302 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
14305 2018-02-16 Carl Love <cel@us.ibm.com>
14307 * config/rs6000/altivec.h: Add builtin names vec_extract4b
14309 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
14311 * config/rs6000/rs6000-c.c: Add the definitions for
14312 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
14313 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
14314 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
14315 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
14316 definition for insert4b and define insn *insert3b_internal.
14317 * doc/extend.texi: Add documentation for vec_extract4b.
14319 2018-02-16 Nathan Sidwell <nathan@acm.org>
14321 * doc/extend.texi (Backwards Compatibility): Mention friend
14322 injection. Note for-scope is deprecated.
14323 * doc/invoke.texi (-ffriend-injection): Deprecate.
14325 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
14327 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
14328 that moved to I2, also allow destinations that are a paradoxical
14329 subreg (instead of a normal reg).
14331 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
14334 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
14337 2018-02-16 Richard Biener <rguenther@suse.de>
14339 PR tree-optimization/84037
14340 PR tree-optimization/84016
14342 * config/i386/i386.c (ix86_builtin_vectorization_cost):
14343 Adjust vec_construct for the fact we need additional higher latency
14344 128bit inserts for AVX256 and AVX512 vector builds.
14345 (ix86_add_stmt_cost): Scale vector construction cost for
14348 2018-02-16 Richard Biener <rguenther@suse.de>
14350 PR tree-optimization/84417
14351 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
14352 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
14353 (non_rewritable_lvalue_p): Likewise, use poly-ints.
14355 2018-02-16 Martin Liska <mliska@suse.cz>
14358 * internal-fn.def (ASAN_CHECK): Set proper flags.
14359 (ASAN_MARK): Likewise.
14361 2018-02-16 Julia Koval <julia.koval@intel.com>
14363 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
14364 from PTA_CANNONLAKE.
14366 2018-02-16 Jakub Jelinek <jakub@redhat.com>
14369 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
14370 Use ++iter rather than iter++ for std::list iterators.
14371 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
14372 defer deleting them until all nodes in the forest are processed. Do
14373 free even leaf nodes. Change to_process into auto_vec.
14376 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
14377 * vec.h (vec_default_construct): Use memset instead of placement new
14378 if BROKEN_VALUE_INITIALIZATION is defined.
14379 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
14380 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
14383 PR rtl-optimization/83723
14384 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
14385 * lra.c (lra_substitute_pseudo): Likewise. If true, use
14386 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
14388 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
14390 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
14392 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
14394 PR rtl-optimization/81443
14395 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
14396 from inner REGs to paradoxical SUBREGs.
14398 2018-02-16 Richard Biener <rguenther@suse.de>
14400 PR tree-optimization/84399
14401 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
14402 For operands we can analyze at their definition make sure we can
14403 analyze them at each use as well.
14405 2018-02-16 Richard Biener <rguenther@suse.de>
14407 PR tree-optimization/84190
14408 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
14409 volatile accesses if the decl isn't volatile.
14411 2018-02-15 Jason Merrill <jason@redhat.com>
14413 PR c++/84314 - ICE with templates and fastcall attribute.
14414 * attribs.c (build_type_attribute_qual_variant): Don't clobber
14415 TYPE_CANONICAL on an existing type.
14417 2018-02-15 Jakub Jelinek <jakub@redhat.com>
14419 PR tree-optimization/84383
14420 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
14421 dstoff nor call operand_equal_p if dstbase is NULL.
14423 PR tree-optimization/84334
14424 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
14425 also a CONSTANT_CLASS_P, punt.
14427 2018-02-14 Jim Wilson <jimw@sifive.com>
14429 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
14430 first SMALL_OPERAND check. New local min_second_step. Move assert
14431 to where locals are set. Add TARGET_RVC support.
14432 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
14434 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
14436 * doc/invoke.texi: Correct -Wformat-overflow code sample.
14438 2018-02-14 Martin Sebor <msebor@redhat.com>
14440 PR tree-optimization/83698
14441 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
14442 arrays constrain the offset range to their bounds.
14443 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
14444 (builtin_access::overlap): Avoid setting the size of overlap if it's
14446 (maybe_diag_overlap): Also consider arrays when deciding what values
14447 of offsets to include in diagnostics.
14449 2018-02-14 Martin Sebor <msebor@redhat.com>
14452 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
14453 that correspond to the kind of a declaration.
14455 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
14458 * config/pa/pa.md: Load address of PIC label using the linkage table
14459 if the label is nonlocal.
14461 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
14463 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
14464 warning message if user requests -maltivec=be.
14465 * doc/invoke.texi: Document deprecation of -maltivec=be.
14467 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
14470 * config/rs6000/rs6000-c.c: Update definitions for
14471 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
14472 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
14474 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
14477 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
14478 add _get_ssp intrinsics. Remove argument from
14479 __builtin_ia32_rdssp[d|q].
14480 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
14481 * config/i386/i386-builtin.def: Remove argument from
14482 __builtin_ia32_rdssp[d|q].
14483 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
14484 ix86_expand_special_args_builtin for _rdssp[d|q].
14485 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
14486 Clear register before usage.
14487 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
14488 Add documentation for new _get_ssp and _inc_ssp intrinsics.
14490 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
14492 PR tree-optimization/84357
14493 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
14494 operand 1 of an ARRAY_REF too.
14496 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
14499 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
14500 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
14502 (set_of_reg): New struct.
14503 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
14504 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
14505 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
14507 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
14508 Split into bitclr, bitset, bitinvert patterns if appropriate.
14509 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
14510 use rx_fuse_in_memory_bitop.
14511 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
14512 to named insn, correct maximum insn length.
14514 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
14517 * machmode.def: Define a complex mode for PARTIAL_INT.
14518 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
14520 * doc/rtl.texi: Document CSPImode.
14521 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
14523 (msp430_hard_regno_nregs_with_padding): Likewise.
14525 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
14528 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
14530 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
14532 PR rtl-optimization/84169
14533 * combine.c (try_combine): New variable split_i2i3. Set it to true if
14534 we generated a parallel as new i3 and we split that to new i2 and i3
14535 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
14536 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
14537 those to i2, not i1. Partially rewrite this scan code.
14539 2018-02-13 Jakub Jelinek <jakub@redhat.com>
14542 * stor-layout.c (place_field): For variable length fields, adjust
14543 offset_align afterwards not just based on the field's alignment,
14544 but also on the size.
14546 PR middle-end/84309
14547 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
14548 of exps and logs in the use_exp2 case.
14550 2018-02-13 Jeff Law <law@redhat.com>
14552 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
14553 entry for "vector".
14555 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
14558 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
14562 * common.opt (gas-loc-support, gas-locview-support): New.
14563 (ginline-points, ginternal-reset-location-views): New.
14564 * doc/invoke.texi: Document them. Use @itemx where intended.
14565 (gvariable-location-views): Adjust.
14566 * target.def (reset_location_view): New.
14567 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
14568 (TARGET_RESET_LOCATION_VIEW): New.
14569 * doc/tm.texi: Rebuilt.
14570 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
14571 (dwarf2out_default_as_locview_support): New.
14572 (output_asm_line_debug_info): Use option variables.
14573 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
14574 (output_loc_list): Likewise.
14575 (add_high_low_attributes): Check option variables.
14576 Don't output entry view attribute in strict mode.
14577 (gen_inlined_subroutine_die): Check option variables.
14578 (dwarf2out_inline_entry): Likewise.
14579 (init_sections_and_labels): Likewise.
14580 (dwarf2out_early_finish): Likewise.
14581 (maybe_reset_location_view): New, from...
14582 (dwarf2out_var_location): ... here. Call it.
14583 * debug.h (dwarf2out_default_as_loc_support): Declare.
14584 (dwarf2out_default_as_locview_support): Declare.
14585 * hooks.c (hook_int_rtx_insn_0): New.
14586 * hooks.h (hook_int_rtx_insn_0): Declare.
14587 * toplev.c (process_options): Take -gas-loc-support and
14588 -gas-locview-support from dwarf2out. Enable
14589 -gvariable-location-views by default only with locview
14590 assembler support. Enable -ginternal-reset-location-views by
14591 default only if the target defines the corresponding hook.
14592 Enable -ginline-points by default if location views are
14593 enabled; force it disabled if statement frontiers are
14595 * tree-inline.c (expand_call_inline): Check option variables.
14596 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
14598 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
14600 PR tree-optimization/84321
14601 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
14602 handling. Also check whether the anti-range contains any values
14603 that satisfy the mask; switch to a VR_RANGE if not.
14605 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
14608 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
14610 2018-02-13 Martin Jambor <mjambor@suse.cz>
14613 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
14614 of call statements, also set location of a load to a temporary.
14616 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
14618 * config/rl78/rl78.c (add_vector_labels): New function.
14619 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
14620 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
14621 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
14622 which checks that no arguments are passed.
14623 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
14624 * doc/extend.texi: Documentation for the new attribute.
14626 2018-02-13 Andreas Schwab <schwab@suse.de>
14628 * config/riscv/linux.h (CPP_SPEC): Define.
14630 2018-02-13 Jakub Jelinek <jakub@redhat.com>
14633 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
14634 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
14635 OPTION_MASK_ISA_AES as first argument to def_builtin_const
14636 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
14637 instead of OPTION_MASK_ISA_PCLMUL as first argument to
14638 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
14639 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
14640 temporarily for AES and PCLMUL builtins.
14642 PR tree-optimization/84339
14643 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
14644 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
14647 PR middle-end/84309
14648 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
14649 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
14650 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
14652 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
14654 * omp-simd-clone.h: New file.
14655 * omp-simd-clone.c: Include omp-simd-clone.h.
14656 (expand_simd_clones): No longer static.
14657 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
14658 cgraph.h and omp-simd-clone.h.
14659 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
14660 (vect_recog_widen_shift_pattern): Formatting fix.
14661 (vect_pattern_recog_1): Don't check optab for calls.
14664 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
14665 operands[2] into a REG before using gen_lowpart on it.
14667 2018-02-12 Jeff Law <law@redhat.com>
14670 * config/sh/sh.c (find_barrier): Consider a sibling call
14673 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
14674 successfully back substituting a reg.
14676 2018-02-12 Richard Biener <rguenther@suse.de>
14678 PR tree-optimization/84037
14679 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
14680 parameter, move visited init to caller.
14681 (vect_slp_analyze_operations): Separate cost from validity
14682 check, initialize visited once for all instances.
14683 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
14685 * tree-vect-stmts.c (vect_model_simple_cost): Make early
14687 (vect_model_promotion_demotion_cost): Likewise.
14688 (vectorizable_bswap): Guard cost modeling with !slp_node
14689 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
14691 (vectorizable_call): Likewise.
14692 (vectorizable_conversion): Likewise.
14693 (vectorizable_assignment): Likewise.
14694 (vectorizable_shift): Likewise.
14695 (vectorizable_operation): Likewise.
14696 (vectorizable_store): Likewise.
14697 (vectorizable_load): Likewise.
14698 (vectorizable_condition): Likewise.
14699 (vectorizable_comparison): Likewise.
14701 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
14704 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
14705 (ASAN_MARK): Fix fnspec to account for return value, change pointer
14706 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
14708 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
14710 PR middle-end/83665
14711 * params.def (inline-min-speedup): Increase from 8 to 15.
14712 (max-inline-insns-auto): Decrease from 40 to 30.
14713 * ipa-split.c (consider_split): Add some buffer for function to
14714 be considered inlining candidate.
14715 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
14718 2018-02-12 Richard Biener <rguenther@suse.de>
14720 PR tree-optimization/84037
14721 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
14722 matched stmts if we cannot swap the non-matched ones.
14724 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
14726 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
14727 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
14728 _mm_maskz_scalef_round_ss): New intrinsics.
14729 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
14730 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
14731 __builtin_ia32_scalefss_round): Remove.
14732 (__builtin_ia32_scalefsd_mask_round,
14733 __builtin_ia32_scalefss_mask_round): New intrinsics.
14734 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
14735 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
14736 ((match_operand:VF_128 2 "<round_nimm_predicate>"
14737 "<round_constraint>")): Changed to ...
14738 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
14739 "<round_scalar_constraint>")): ... this.
14740 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
14741 %0, %1, %2<round_op3>}"): Changed to ...
14742 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
14743 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
14744 %2<round_scalar_mask_op3>}"): ... this.
14745 * config/i386/subst.md (round_scalar_nimm_predicate): New.
14747 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
14749 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
14750 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
14751 (_mm_maskz_sqrt_round_ss): New intrinsics.
14752 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
14753 (__builtin_ia32_sqrtsd_mask_round)
14754 (__builtin_ia32_sqrtss_mask_round): New builtins.
14755 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
14756 (__builtin_ia32_sqrtss_round): Remove.
14757 (__builtin_ia32_sqrtsd_mask_round)
14758 (__builtin_ia32_sqrtss_mask_round): New builtins.
14759 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
14760 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
14761 ((match_operand:VF_128 1 "vector_operand"
14762 "xBm,<round_constraint>")): Changed to ...
14763 ((match_operand:VF_128 1 "vector_operand"
14764 "xBm,<round_scalar_constraint>")): ... this.
14765 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
14766 %0, %2, %<iptr>1<round_op3>}): Changed to ...
14767 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
14768 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
14769 %<iptr>1<round_scalar_mask_op3>}): ... this.
14770 ((set_attr "prefix" "<round_prefix>")): Changed to ...
14771 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
14773 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
14776 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
14777 Cast vec_cmpeq result to correct type.
14778 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
14779 Cast vec_cmpgt result to correct type.
14781 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
14783 * final.c (final_scan_insn_1): Renamed from...
14784 (final_scan_insn): ... this. New wrapper, to recover
14785 seen from the outermost call in recursive ones.
14786 * config/sparc/sparc.c (output_return): Drop seen from call.
14787 (output_sibcall): Likewise.
14788 * config/visium/visium.c (output_branch): Likewise.
14790 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
14792 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
14795 2018-02-10 Alan Modra <amodra@gmail.com>
14798 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
14799 Specify LR as an input.
14801 2018-02-10 Jakub Jelinek <jakub@redhat.com>
14804 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
14805 remove_member_access_dummy_vars): New functions.
14806 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
14807 lower_omp_1, execute_lower_omp): Use them.
14809 PR rtl-optimization/84308
14810 * shrink-wrap.c (spread_components): Release todo vector.
14812 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
14814 PR rtl-optimization/57193
14815 * ira-color.c (struct allocno_color_data): Add member
14816 conflict_allocno_hard_prefs.
14817 (update_conflict_allocno_hard_prefs): New.
14818 (bucket_allocno_compare_func): Add a preference based on
14819 conflict_allocno_hard_prefs.
14820 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
14821 (color_allocnos): Remove a dead code. Initiate
14822 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
14824 2018-02-09 Jakub Jelinek <jakub@redhat.com>
14827 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
14828 constraint from =wa to wa. Avoid a subreg on the output operand,
14829 instead use a pseudo and subreg it in a move.
14830 (p9_xxbrd_<mode>): Changed to ...
14831 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
14832 (p9_xxbrd_v2df): New expander.
14833 (p9_xxbrw_<mode>): Changed to ...
14834 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
14835 (p9_xxbrw_v4sf): New expander.
14837 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
14839 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
14841 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
14844 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
14845 multiply in 32-bit mode.
14846 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
14847 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
14850 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
14852 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
14853 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
14854 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
14855 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
14857 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
14860 * dwarf2out.c (is_trivial_indirect_ref): New function.
14861 (dwarf2out_late_global_decl): Do not generate a location
14862 attribute for variables that have a non-trivial DECL_VALUE_EXPR
14863 and that are not defined in the current unit.
14865 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
14867 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
14868 instead of a libcall for UNORDERED.
14870 2018-02-09 Tamar Christina <tamar.christina@arm.com>
14873 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
14874 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
14876 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
14879 * config/s390/s390.c (s390_set_current_function): Invoke
14880 s390_indirect_branch_settings also if fndecl didn't change.
14882 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
14884 * config/rs6000/rs6000.md (blockage): Set length to zero.
14886 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
14888 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
14890 2018-02-09 Jakub Jelinek <jakub@redhat.com>
14893 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
14894 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
14898 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
14899 PARALLEL incoming that failed vt_get_decl_and_offset check.
14901 PR middle-end/84237
14902 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
14903 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
14905 (get_variable_section): For decls in named .bss* sections pass true as
14906 second argument to bss_initializer_p.
14908 2018-02-09 Marek Polacek <polacek@redhat.com>
14909 Jakub Jelinek <jakub@redhat.com>
14912 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
14913 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
14914 Sync some changes from cxx_fold_indirect_ref.
14916 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
14918 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
14920 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
14921 (BLOCK_INLINE_ENTRY_LABEL): New.
14922 (dwarf2out_var_location): Disregard inline entry markers.
14923 (inline_entry_data): New struct.
14924 (inline_entry_data_hasher): New hashtable type.
14925 (inline_entry_data_hasher::hash): New.
14926 (inline_entry_data_hasher::equal): New.
14927 (inline_entry_data_table): New variable.
14928 (add_high_low_attributes): Add DW_AT_entry_pc and
14929 DW_AT_GNU_entry_view attributes if a pending entry is found
14930 in inline_entry_data_table. Add old entry_pc attribute only
14931 if debug nonbinding markers are disabled.
14932 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
14933 markers are enabled.
14934 (block_within_block_p, dwarf2out_inline_entry): New.
14935 (dwarf2out_finish): Check that no entries remained in
14936 inline_entry_data_table.
14937 * final.c (reemit_insn_block_notes): Handle inline entry notes.
14938 (final_scan_insn, notice_source_line): Likewise.
14939 (rest_of_clean_state): Skip inline entry markers.
14940 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
14942 * gimple.c (gimple_build_debug_inline_entry): New.
14943 * gimple.h (enum gimple_debug_subcode): Add
14944 GIMPLE_DEBUG_INLINE_ENTRY.
14945 (gimple_build_debug_inline_entry): Declare.
14946 (gimple_debug_inline_entry_p): New.
14947 (gimple_debug_nonbind_marker_p): Adjust.
14948 * insn-notes.def (INLINE_ENTRY): New.
14949 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
14950 inline entry marker notes.
14951 (print_insn): Likewise.
14952 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
14953 (INSN_DEBUG_MARKER_KIND): Likewise.
14954 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
14955 * tree-inline.c (expand_call_inline): Build and insert
14956 debug_inline_entry stmt.
14957 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
14958 inline entry blocks early, if nonbind markers are enabled.
14959 (dump_scope_block): Dump fragment info.
14960 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
14961 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
14962 (gimple_build_debug_inline_entry): New.
14963 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
14964 Enable/disable inline entry points too.
14965 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
14966 (DEBUG_INSN): Describe inline entry markers.
14968 * common.opt (gvariable-location-views): New.
14969 (gvariable-location-views=incompat5): New.
14970 * config.in: Rebuilt.
14971 * configure: Rebuilt.
14972 * configure.ac: Test assembler for view support.
14973 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
14974 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
14975 * dwarf2out.c (var_loc_view): New typedef.
14976 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
14977 (dwarf2out_locviews_in_attribute): New.
14978 (dwarf2out_locviews_in_loclist): New.
14979 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
14980 (enum dw_line_info_opcode): Add LI_adv_address.
14981 (struct dw_line_info_table): Add view.
14982 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
14983 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
14984 (zero_view_p): New variable.
14985 (ZERO_VIEW_P): New macro.
14986 (output_asm_line_debug_info): New.
14987 (struct var_loc_node): Add view.
14988 (add_AT_view_list, AT_loc_list): New.
14989 (add_var_loc_to_decl): Add view param. Test it against last.
14990 (new_loc_list): Add view params. Record them.
14991 (AT_loc_list_ptr): Handle loc and view lists.
14992 (view_list_to_loc_list_val_node): New.
14993 (print_dw_val): Handle dw_val_class_view_list.
14994 (size_of_die): Likewise.
14995 (value_format): Likewise.
14996 (loc_list_has_views): New.
14997 (gen_llsym): Set vl_symbol too.
14998 (maybe_gen_llsym, skip_loc_list_entry): New.
14999 (dwarf2out_maybe_output_loclist_view_pair): New.
15000 (output_loc_list): Output view list or entries too.
15001 (output_view_list_offset): New.
15002 (output_die): Handle dw_val_class_view_list.
15003 (output_dwarf_version): New.
15004 (output_compilation_unit_header): Use it.
15005 (output_skeleton_debug_sections): Likewise.
15006 (output_rnglists, output_line_info): Likewise.
15007 (output_pubnames, output_aranges): Update version comments.
15008 (output_one_line_info_table): Output view numbers in asm comments.
15009 (dw_loc_list): Determine current endview, pass it to new_loc_list.
15010 Call maybe_gen_llsym.
15011 (loc_list_from_tree_1): Adjust.
15012 (add_AT_location_description): Create view list attribute if
15013 needed, check it's absent otherwise.
15014 (convert_cfa_to_fb_loc_list): Adjust.
15015 (maybe_emit_file): Call output_asm_line_debug_info for test.
15016 (dwarf2out_var_location): Reset views as needed. Precompute
15017 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
15018 attribute. Set view.
15019 (new_line_info_table): Reset next view.
15020 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
15021 (dwarf2out_source_line): Likewise. Output view resets and labels to
15022 the assembler, or select appropriate line info opcodes.
15023 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
15024 (optimize_string_length): Catch it. Adjust.
15025 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
15026 dw_val_class_view_list, and remove it if no longer needed.
15027 (hash_loc_list): Hash view numbers.
15028 (loc_list_hasher::equal): Compare them.
15029 (optimize_location_lists): Check whether a view list symbol is
15030 needed, and whether the locview attribute is present, and
15031 whether they match. Remove the locview attribute if no longer
15033 (index_location_lists): Call skip_loc_list_entry for test.
15034 (dwarf2out_finish): Call output_asm_line_debug_info for test.
15035 Use output_dwarf_version.
15036 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
15037 (struct dw_val_node): Add val_view_list.
15038 * final.c (SEEN_NEXT_VIEW): New.
15039 (set_next_view_needed): New.
15040 (clear_next_view_needed): New.
15041 (maybe_output_next_view): New.
15042 (final_start_function): Rename to...
15043 (final_start_function_1): ... this. Take pointer to FIRST,
15044 add SEEN parameter. Emit param bindings in the initial view.
15045 (final_start_function): Reintroduce SEEN-less interface.
15046 (final): Rename to...
15047 (final_1): ... this. Take SEEN parameter. Output final pending
15048 next view at the end.
15049 (final): Reintroduce seen-less interface.
15050 (final_scan_insn): Output pending next view before switching
15051 sections or ending a block. Mark the next view as needed when
15052 outputting variable locations. Notify debug backend of section
15053 changes, and of location view changes.
15054 (rest_of_handle_final): Adjust.
15055 * toplev.c (process_options): Autodetect value for debug variable
15056 location views option. Warn on incompat5 without -gdwarf-5.
15057 * doc/invoke.texi (gvariable-location-views): New.
15058 (gvariable-location-views=incompat5): New.
15059 (gno-variable-location-views): New.
15061 2018-02-08 David Malcolm <dmalcolm@redhat.com>
15063 PR tree-optimization/84136
15064 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
15065 that the result of find_edge is non-NULL.
15067 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
15070 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
15071 storing integer register in SImode. Fix cost of 256 and 512
15072 byte aligned SSE register store.
15074 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
15076 * config/i386/i386.c (ix86_multiplication_cost): Fix
15077 multiplication cost for TARGET_AVX512DQ.
15079 2018-02-08 Marek Polacek <polacek@redhat.com>
15081 PR tree-optimization/84238
15082 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
15085 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
15087 PR tree-optimization/84265
15088 * tree-vect-stmts.c (vectorizable_store): Don't treat
15089 VMAT_CONTIGUOUS accesses as grouped.
15090 (vectorizable_load): Likewise.
15092 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
15094 PR tree-optimization/81635
15095 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
15096 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
15097 (test_round_for_mask): New functions.
15098 (wide_int_cc_tests): Call test_round_for_mask.
15099 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
15100 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
15101 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
15102 range returned by get_range_info.
15104 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
15107 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
15108 * symtab.c: Include builtins.h
15109 (symtab_node::output_to_lto_symbol_table_p): Move here
15110 from lto-streamer-out.c:output_symbol_p.
15111 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
15112 (output_symbol_p): Move all logic to symtab.c
15113 (produce_symtab): Update.
15115 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
15117 * config/s390/s390-opts.h (enum indirect_branch): Define.
15118 * config/s390/s390-protos.h (s390_return_addr_from_memory)
15119 (s390_indirect_branch_via_thunk)
15120 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
15121 (enum s390_indirect_branch_type): Define.
15122 * config/s390/s390.c (struct s390_frame_layout, struct
15123 machine_function): Remove.
15124 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
15125 (indirect_branch_table_label_no, indirect_branch_table_name):
15127 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
15128 (enum s390_indirect_branch_option): Define.
15129 (s390_return_addr_from_memory): New function.
15130 (s390_handle_string_attribute): New function.
15131 (s390_attribute_table): Add new attribute handler.
15132 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
15133 (s390_indirect_branch_via_thunk): New function.
15134 (s390_indirect_branch_via_inline_thunk): New function.
15135 (s390_function_ok_for_sibcall): When jumping via thunk disallow
15136 sibling call optimization for non z10 compiles.
15137 (s390_emit_call): Force indirect branch target to be a single
15138 register. Add r1 clobber for non-z10 compiles.
15139 (s390_emit_epilogue): Emit return jump via return_use expander.
15140 (s390_reorg): Handle JUMP_INSNs as execute targets.
15141 (s390_option_override_internal): Perform validity checks for the
15142 new command line options.
15143 (s390_indirect_branch_attrvalue): New function.
15144 (s390_indirect_branch_settings): New function.
15145 (s390_set_current_function): Invoke s390_indirect_branch_settings.
15146 (s390_output_indirect_thunk_function): New function.
15147 (s390_code_end): Implement target hook.
15148 (s390_case_values_threshold): Implement target hook.
15149 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
15151 * config/s390/s390.h (struct s390_frame_layout)
15152 (struct machine_function): Move here from s390.c.
15153 (TARGET_INDIRECT_BRANCH_NOBP_RET)
15154 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
15155 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
15156 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
15157 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
15158 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
15159 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
15160 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
15161 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
15162 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
15163 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
15164 (mnemonic attribute): Add values which aren't recognized
15166 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
15167 pattern for branch conversion. Fix mnemonic attribute.
15168 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
15169 indirect branch via thunk if requested.
15170 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
15171 ("*indirect_jump"): Disable for branch conversion using out of
15173 ("indirect_jump_via_thunk<mode>_z10")
15174 ("indirect_jump_via_thunk<mode>")
15175 ("indirect_jump_via_inlinethunk<mode>_z10")
15176 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
15177 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
15178 ("casesi_jump_via_inlinethunk<mode>_z10")
15179 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
15180 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
15181 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
15182 ("*indirect2_jump"): Disable for branch conversion.
15183 ("casesi_jump"): Turn into expander and expand patterns for branch
15185 ("return_use"): New expander.
15186 ("*return"): Emit return via thunk and rename it to ...
15187 ("*return<mode>"): ... this one.
15188 * config/s390/s390.opt: Add new options and and enum for the
15191 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
15193 * lra-constraints.c (match_reload): Unconditionally use
15194 gen_lowpart_SUBREG, rather than selecting between that
15195 and equivalent gen_rtx_SUBREG code.
15197 2018-02-08 Richard Biener <rguenther@suse.de>
15199 PR tree-optimization/84233
15200 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
15201 changed flag instead of boguously re-using phi_inserted.
15203 2018-02-08 Martin Jambor <mjambor@suse.cz>
15205 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
15206 static local variables.
15208 2018-02-08 Richard Biener <rguenther@suse.de>
15210 PR tree-optimization/84278
15211 * tree-vect-stmts.c (vectorizable_store): When looking for
15212 smaller vector types to perform grouped strided loads/stores
15213 make sure the mode is supported by the target.
15214 (vectorizable_load): Likewise.
15216 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
15218 * config/aarch64/aarch64.c (aarch64_components_for_bb):
15219 Increase LDP/STP opportunities by adding adjacent callee-saves.
15221 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
15223 PR rtl-optimization/84068
15224 PR rtl-optimization/83459
15225 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
15227 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
15229 PR tree-optimization/84224
15230 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
15231 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
15232 non-zero arguments.
15234 2018-02-07 Iain Sandoe <iain@codesourcery.com>
15237 * config/rs6000/altivec.md (*restore_world): Remove LR use.
15238 * config/rs6000/predicates.md (restore_world_operation): Adjust op
15239 count, remove one USE.
15241 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
15243 * doc/install.texi (Configuration): Document the
15244 --with-long-double-format={ibm,ieee} PowerPC configuration
15248 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
15249 Convert from define_expand to be define_insn_and_split. Rework
15250 float/double/_Float128 conversions to QI/HI/SImode to work with
15251 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
15252 conversions to QI/HImode types did a store and then a load to
15253 truncate the value. For conversions to VSX registers, don't split
15254 the insn, instead emit the code directly. Use the code iterator
15255 any_fix to combine signed and unsigned conversions.
15256 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
15257 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
15258 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
15259 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
15260 (fix_<mode>di2_hw): Likewise.
15261 (fixuns_<mode>di2_hw): Likewise.
15262 (fix_<mode>si2_hw): Likewise.
15263 (fixuns_<mode>si2_hw): Likewise.
15264 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
15265 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
15266 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
15267 fix<uns>_trunc<SFDF:mode>si2_p8.
15268 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
15269 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
15270 (fix<uns>_<mode>_mem): Likewise.
15271 (fctiw<u>z_<mode>_mem): Likewise.
15272 (fix<uns>_<mode>_mem): Likewise.
15273 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
15274 the register allocator from doing a direct move to the GPRs to do
15275 a store, and instead use the ISA 3.0 store byte/half-word from
15276 vector register instruction. For IEEE 128-bit floating point,
15277 also optimize stores of 32-bit ints.
15278 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
15280 2018-02-07 Alan Hayward <alan.hayward@arm.com>
15282 * genextract.c (push_pathstr_operand): New function to support
15284 (walk_rtx): Call push_pathstr_operand.
15285 (print_path): Support [a-zA-Z].
15287 2018-02-07 Richard Biener <rguenther@suse.de>
15289 PR tree-optimization/84037
15290 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
15291 (cse_and_gimplify_to_preheader): Declare.
15292 (vect_get_place_in_interleaving_chain): Likewise.
15293 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15295 (_loop_vec_info::~_loop_vec_info): Delete it.
15296 (cse_and_gimplify_to_preheader): New function.
15297 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
15298 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
15299 (vectorizable_load): Likewise. For grouped stores always base
15300 the IV on the first element.
15301 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
15302 condition before gimplifying.
15304 2018-02-07 Jakub Jelinek <jakub@redhat.com>
15306 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
15307 *DIV_EXPR and *MOD_EXPR.
15309 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
15312 * config/i386/i386.c (ix86_option_override_internal): Mask out
15313 the CF_SET bit when checking -fcf-protection.
15315 2018-02-07 Tom de Vries <tom@codesourcery.com>
15318 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
15321 2018-02-07 Richard Biener <rguenther@suse.de>
15323 PR tree-optimization/84204
15324 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
15327 PR tree-optimization/84205
15328 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
15329 special-case isl_ast_op_zdiv_r.
15331 PR tree-optimization/84223
15332 * graphite-scop-detection.c (gather_bbs::before_dom_children):
15333 Only add conditions from within the region.
15334 (gather_bbs::after_dom_children): Adjust.
15336 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
15339 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
15340 * config/avr/avr.md: Only post-reload split REG-REG moves if
15341 either register is GENERAL_REG_P.
15343 2018-02-07 Jakub Jelinek <jakub@redhat.com>
15345 PR tree-optimization/84235
15346 * tree-ssa-scopedtables.c
15347 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
15348 if the subtraction is performed in floating point type where NaNs are
15349 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
15350 build 1. Formatting fix.
15352 2018-02-06 Jakub Jelinek <jakub@redhat.com>
15355 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
15356 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
15357 and skip it regardless of bb boundaries. Use CALL_P macro,
15358 don't test INSN_P (insn) together with CALL_P or JUMP_P check
15359 unnecessarily, formatting fix.
15361 2018-02-06 Michael Collison <michael.collison@arm.com>
15363 * config/arm/thumb2.md:
15364 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
15365 (*thumb_mov_notscc): Ditto.
15367 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
15370 * config/rs6000/rs6000.md (su code attribute): Use "u" for
15371 unsigned_fix, not "s".
15373 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15375 * configure.ac (gcc_fn_eh_frame_ro): New function.
15376 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
15377 correct .eh_frame permissions.
15378 * configure: Regenerate.
15380 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
15382 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
15383 irrelevant options.
15385 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15387 * config/rs6000/rs6000.c (rs6000_option_override_internal):
15388 Display warning message for -mno-speculate-indirect-jumps.
15390 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
15392 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
15394 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
15396 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
15398 PR tree-optimization/84225
15399 * tree-eh.c (find_trapping_overflow): Only call
15400 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
15402 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
15405 * config/i386/i386.c: Reimplement the check of possible options
15406 -mibt/-mshstk conbination. Change error messages.
15407 * doc/invoke.texi: Fix a typo: remove extra '='.
15409 2018-02-06 Marek Polacek <polacek@redhat.com>
15411 PR tree-optimization/84228
15412 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
15414 2018-02-06 Tamar Christina <tamar.christina@arm.com>
15417 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
15418 emitted arch directives.
15419 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
15420 __ARM_FEATURE_COPROC before changing architectures.
15422 2018-02-06 Richard Biener <rguenther@suse.de>
15424 * config/i386/i386.c (print_reg): Fix typo.
15425 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
15427 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
15429 * configure: Regenerate.
15431 2018-02-05 Martin Sebor <msebor@redhat.com>
15433 PR tree-optimization/83369
15434 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
15437 2018-02-05 Martin Liska <mliska@suse.cz>
15439 * doc/invoke.texi: Cherry-pick upstream r323995.
15441 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
15443 * ira.c (ira_init_register_move_cost): Adjust comment.
15445 2018-02-05 Martin Liska <mliska@suse.cz>
15447 PR gcov-profile/84137
15448 * doc/gcov.texi: Fix typo in documentation.
15450 2018-02-05 Martin Liska <mliska@suse.cz>
15452 PR gcov-profile/83879
15453 * doc/gcov.texi: Document necessity of --dynamic-list-data when
15454 using dlopen functionality.
15456 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
15458 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
15459 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
15460 _mm_maskz_range_ss, _mm_mask_range_round_ss,
15461 _mm_maskz_range_round_ss): New intrinsics.
15462 (__builtin_ia32_rangesd128_round)
15463 (__builtin_ia32_rangess128_round): Remove.
15464 (__builtin_ia32_rangesd128_mask_round,
15465 __builtin_ia32_rangess128_mask_round): New builtins.
15466 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
15467 __builtin_ia32_rangess128_round): Remove.
15468 (__builtin_ia32_rangesd128_mask_round,
15469 __builtin_ia32_rangess128_mask_round): New builtins.
15470 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
15471 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
15472 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
15473 "<round_saeonly_constraint>")): Changed to ...
15474 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
15475 "<round_saeonly_scalar_constraint>")): ... this.
15476 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
15477 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
15478 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
15479 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
15480 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
15482 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
15484 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
15486 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
15487 Remove all values except native, 8540 and 8548.
15489 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
15491 * config/i386/i386.c (ix86_output_function_return): Pass
15492 INVALID_REGNUM, instead of -1, as invalid register number to
15493 indirect_thunk_name and output_indirect_thunk.
15495 2018-02-02 Julia Koval <julia.koval@intel.com>
15497 * config.gcc: Add -march=icelake.
15498 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
15499 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
15500 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
15501 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
15502 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
15503 (processor_target_table): Add icelake.
15504 (ix86_option_override_internal): Handle new PTAs.
15505 (get_builtin_code_for_version): Handle icelake.
15506 (M_INTEL_COREI7_ICELAKE): New.
15507 (fold_builtin_cpu): Handle icelake.
15508 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
15509 * doc/invoke.texi: Add -march=icelake.
15511 2018-02-02 Julia Koval <julia.koval@intel.com>
15513 * config/i386/i386.c (ix86_option_override_internal): Change flags type
15514 to wide_int_bitmask.
15515 * wide-int-bitmask.h: New.
15517 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
15520 * config/i386/i386.md: Replace Pmode with word_mode in
15521 builtin_setjmp_setup and builtin_longjmp to support x32.
15523 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
15527 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
15529 (rs6000_supported_cpu_names): New static variable.
15530 (linux_cpu_translation_table): Likewise.
15531 (elf_platform) <cpu>: Define new static variable and use it.
15532 Translate kernel AT_PLATFORM name to canonical name if needed.
15533 Error if platform name is unknown.
15535 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
15538 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
15540 2018-02-01 Jeff Law <law@redhat.com>
15543 * config/i386/i386.c (release_scratch_register_on_entry): Add new
15544 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
15545 the scratch if RELEASE_VIA_POP is false.
15546 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
15547 If we have to save a temporary register, decrement SIZE appropriately.
15548 Pass new arguments to release_scratch_register_on_entry.
15549 (ix86_adjust_stack_and_probe): Likewise.
15550 (ix86_emit_probe_stack_range): Pass new arguments to
15551 release_scratch_register_on_entry.
15553 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
15555 PR rtl-optimization/84157
15556 * combine.c (change_zero_ext): Use REG_P predicate in
15557 front of HARD_REGISTER_P predicate.
15559 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
15561 * config/avr/avr.c (avr_option_override): Move disabling of
15562 -fdelete-null-pointer-checks to...
15563 * common/config/avr/avr-common.c (avr_option_optimization_table):
15566 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
15568 PR tree-optimization/81635
15569 * tree-data-ref.c (split_constant_offset_1): For types that
15570 wrap on overflow, try to use range info to prove that wrapping
15573 2018-02-01 Renlin Li <renlin.li@arm.com>
15576 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
15577 TAILCALL_ADDR_REGS.
15578 (aarch64_register_move_cost): Likewise.
15579 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
15580 TAILCALL_ADDR_REGS.
15581 (REG_CLASS_NAMES): Likewise.
15582 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
15583 TAILCALL_ADDR_REGS. Remove IP registers.
15584 * config/aarch64/aarch64.md (Ucs): Update register constraint.
15586 2018-02-01 Richard Biener <rguenther@suse.de>
15588 * domwalk.h (dom_walker::dom_walker): Add additional constructor
15589 for specifying RPO order and allow NULL for that.
15590 * domwalk.c (dom_walker::dom_walker): Likewise.
15591 (dom_walker::walk): Handle NULL RPO order.
15592 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
15594 (rewrite_update_dom_walker): Likewise.
15595 (mark_def_dom_walker): Likewise.
15597 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
15599 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
15600 (aarch64_maybe_expand_sve_subreg_move): Declare.
15601 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
15602 * config/aarch64/predicates.md (aarch64_any_register_operand): New
15604 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
15605 that are semantically a reverse operation.
15606 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
15607 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
15608 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
15610 (aarch64_can_change_mode_class): For big-endian, forbid changes
15611 between two SVE modes if they have different element sizes.
15613 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
15615 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
15616 the TImode handling for big-endian targets.
15618 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
15620 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
15621 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
15623 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
15624 Remove BSWAP handing for big-endian targets and use the form of
15625 LD1RQ appropariate for the mode.
15627 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
15629 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
15630 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
15631 duplicated element.
15633 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
15636 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
15637 check for operands that need to go through aarch64_sve_reload_be.
15639 2018-02-01 Jakub Jelinek <jakub@redhat.com>
15641 PR tree-optimization/81661
15642 PR tree-optimization/84117
15643 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
15644 * tree-eh.c: Include gimplify.h.
15645 (find_trapping_overflow, replace_trapping_overflow,
15646 rewrite_to_non_trapping_overflow): New functions.
15647 * tree-vect-loop.c: Include tree-eh.h.
15648 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
15649 * tree-data-ref.c: Include tree-eh.h.
15650 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
15652 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
15654 PR rtl-optimization/84123
15655 * combine.c (change_zero_ext): Check if hard register satisfies
15656 can_change_dest_mode before calling gen_lowpart_SUBREG.
15658 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
15661 * ira.c (ira_init_register_move_cost): Remove assert.
15663 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
15665 PR rtl-optimization/84071
15666 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
15667 * doc/tm.texi: Regenerate.
15669 2018-01-31 Richard Biener <rguenther@suse.de>
15671 PR tree-optimization/84132
15672 * tree-data-ref.c (analyze_miv_subscript): Properly
15673 check whether evolution_function_is_affine_multivariate_p
15674 before calling gcd_of_steps_may_divide_p.
15676 2018-01-31 Julia Koval <julia.koval@intel.com>
15679 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
15680 * config/i386/i386.md (rdpid_rex64) New.
15681 (rdpid): Make 32bit only.
15683 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
15686 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
15687 an IDENTIFIER_NODE for FUNCTION_TYPE's.
15689 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
15692 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
15694 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
15696 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
15698 PR rtl-optimization/84071
15699 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
15700 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
15702 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
15704 * config/arc/arc.c (arc_handle_aux_attribute): New function.
15705 (arc_attribute_table): Add 'aux' attribute.
15706 (arc_in_small_data_p): Consider aux like variables.
15707 (arc_is_aux_reg_p): New function.
15708 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
15709 (arc_get_aux_arg): New function.
15710 (prepare_move_operands): Handle aux-register access.
15711 (arc_handle_aux_attribute): New function.
15712 * doc/extend.texi (ARC Variable attributes): Add subsection.
15714 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
15716 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
15717 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
15718 (arc_attribute_table): Add 'uncached' attribute.
15719 (arc_print_operand): Print '.di' flag for uncached memory
15721 (arc_in_small_data_p): Do not consider for small data the uncached
15723 (arc_is_uncached_mem_p): New function.
15724 * config/arc/predicates.md (compact_store_memory_operand): Check
15725 for uncached memory accesses.
15726 (nonvol_nonimm_operand): Likewise.
15727 * doc/extend.texi (ARC Type Attribute): New subsection.
15729 2018-01-31 Jakub Jelinek <jakub@redhat.com>
15732 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
15733 falign-loops=): Add Optimization flag.
15735 2018-01-30 Jeff Law <law@redhat.com>
15738 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
15739 INT_REGISTERS_SAVED. Check it prior to calling
15740 get_scratch_register_on_entry.
15741 (ix86_adjust_stack_and_probe): Similarly.
15742 (ix86_emit_probe_stack_range): Similarly.
15743 (ix86_expand_prologue): Corresponding changes.
15745 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15748 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
15749 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
15751 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
15754 * lra-constraints.c (curr_insn_transform): Process AND in the
15757 2018-01-30 Jakub Jelinek <jakub@redhat.com>
15759 PR rtl-optimization/83986
15760 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
15761 dependence against last_pending_memory_flush in addition to
15762 pending_jump_insns.
15764 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
15766 PR tree-optimization/81611
15767 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
15770 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15773 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
15776 2018-01-30 Richard Biener <rguenther@suse.de>
15777 Jakub Jelinek <jakub@redhat.com>
15779 PR tree-optimization/84111
15780 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
15781 inner loops added during recursion, as they don't have up-to-date
15784 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
15787 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
15788 (can_inline_edge_by_limits_p): ... here.
15789 (can_early_inline_edge_p, check_callers,
15790 update_caller_keys, update_callee_keys, recursive_inlining,
15791 add_new_edges_to_heap, speculation_useful_p,
15792 inline_small_functions,
15793 inline_small_functions, flatten_function,
15794 inline_to_all_callers_1): Update.
15796 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
15798 * profile-count.c (profile_count::combine_with_ipa_count): Handle
15801 2018-01-30 Richard Biener <rguenther@suse.de>
15803 PR tree-optimization/83008
15804 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
15805 invariant and constant vector uses in stmts when they need
15806 more than one stmt.
15808 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15811 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
15812 * configure: Regenerate.
15814 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
15816 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
15818 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
15819 Use gen_rtx_REG rather than gen_lowpart.
15821 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
15823 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
15824 rather than 0 when creating partial subregs.
15826 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
15828 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
15831 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
15834 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
15835 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
15836 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
15837 flags. This restores the settings used before the 2017-07-24.
15838 Turning off pre increment/decrement/modify allows IVOPTS to
15839 optimize DF/SF loops where the index is an int.
15841 2018-01-29 Richard Biener <rguenther@suse.de>
15842 Kelvin Nilsen <kelvin@gcc.gnu.org>
15845 * tree-vect-stmts.c (vectorizable_call): Don't call
15846 targetm.vectorize_builtin_md_vectorized_function if callee is
15849 2018-01-22 Carl Love <cel@us.ibm.com>
15851 * doc/extend.tex: Fix typo in second arg in
15852 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
15854 2018-01-29 Richard Biener <rguenther@suse.de>
15856 PR tree-optimization/84086
15857 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
15858 (flush_ssaname_freelist): When SSA names were released reset
15859 the SCEV hash table.
15861 2018-01-29 Richard Biener <rguenther@suse.de>
15863 PR tree-optimization/84057
15864 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
15865 removed paths when removing edges.
15867 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
15869 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
15870 -mfunction-return=@var{choice}.
15872 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
15874 PR diagnostic/84034
15875 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
15876 Handle CR like TAB.
15877 (layout::print_source_line): Likewise.
15878 (test_get_line_width_without_trailing_whitespace): Add test cases.
15880 2018-01-27 Jakub Jelinek <jakub@redhat.com>
15882 PR middle-end/84040
15883 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
15886 2018-01-26 Jim Wilson <jimw@sifive.com>
15888 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
15890 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
15893 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15895 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
15896 and CMP + SUB-immediate -> SUBS.
15898 2018-01-26 Martin Sebor <msebor@redhat.com>
15900 PR tree-optimization/83896
15901 * tree-ssa-strlen.c (get_string_len): Rename...
15902 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
15903 Avoid assuming length is constant.
15904 (handle_char_store): Use HOST_WIDE_INT for string length.
15906 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
15909 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
15910 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
15912 2018-01-26 Richard Biener <rguenther@suse.de>
15914 PR rtl-optimization/84003
15915 * dse.c (record_store): Only record redundant stores when
15916 the earlier store aliases at least all accesses the later one does.
15918 2018-01-26 Jakub Jelinek <jakub@redhat.com>
15920 PR rtl-optimization/83985
15921 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
15922 REG_CFA_RESTORE insns.
15923 (delete_unmarked_insns): Don't ignore separate shrink wrapping
15924 REG_CFA_RESTORE insns here.
15927 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
15928 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
15930 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
15932 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
15933 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
15934 (arc_init): Likewise.
15935 (arc_override_options): Likewise.
15936 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
15938 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
15940 * config/arc/arc.h (TARGET_DBNZ): Define.
15941 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
15942 properly set the tune attribute.
15943 (dbnz): Use TARGET_DBNZ guard.
15944 * config/arc/arc.opt (mtune): Add core3 option.
15946 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
15948 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
15949 recognize new pic like addresses.
15950 (arc_delegitimize_address): Clean up.
15952 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
15954 * config/arc/arc-arches.def: Option mrf16 valid for all
15956 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
15957 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
15958 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
15959 * config/arc/arc-tables.opt: Regenerate.
15960 * config/arc/arc.c (arc_conditional_register_usage): Handle
15961 reduced register file case.
15962 (arc_file_start): Set must have build attributes.
15963 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
15964 mrf16 option value.
15965 * config/arc/arc.opt (mrf16): Add new option.
15966 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
15967 * config/arc/genmultilib.awk: Handle new mrf16 option.
15968 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
15969 * config/arc/t-multilib: Regenerate.
15970 * doc/invoke.texi (ARC Options): Document mrf16 option.
15972 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
15974 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
15975 * config/arc/arc.c (arc_handle_secure_attribute): New function.
15976 (arc_attribute_table): Add 'secure_call' attribute.
15977 (arc_print_operand): Print secure call operand.
15978 (arc_function_ok_for_sibcall): Don't optimize tail calls when
15980 (arc_is_secure_call_p): New function. * config/arc/arc.md
15981 (call_i): Add support for sjli instruction.
15982 (call_value_i): Likewise.
15983 * config/arc/constraints.md (Csc): New constraint.
15985 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
15986 John Eric Martin <John.Martin@emmicro-us.com>
15988 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
15989 * config/arc/arc.c (_arc_jli_section): New struct.
15990 (arc_jli_section): New type.
15991 (rc_jli_sections): New static variable.
15992 (arc_handle_jli_attribute): New function.
15993 (arc_attribute_table): Add jli_always and jli_fixed attribute.
15994 (arc_file_end): New function.
15995 (TARGET_ASM_FILE_END): Define.
15996 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
15997 (arc_add_jli_section): New function.
15998 (jli_call_scan): Likewise.
15999 (arc_reorg): Call jli_call_scan.
16000 (arc_output_addsi): Remove 'S' from printing asm operand.
16001 (arc_is_jli_call_p): New function.
16002 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
16004 (movhi_insn): Likewise.
16005 (movsi_insn): Likewise.
16006 (movsi_set_cc_insn): Likewise.
16007 (loadqi_update): Likewise.
16008 (load_zeroextendqisi_update): Likewise.
16009 (load_signextendqisi_update): Likewise.
16010 (loadhi_update): Likewise.
16011 (load_zeroextendhisi_update): Likewise.
16012 (load_signextendhisi_update): Likewise.
16013 (loadsi_update): Likewise.
16014 (loadsf_update): Likewise.
16015 (movsicc_insn): Likewise.
16016 (bset_insn): Likewise.
16017 (bxor_insn): Likewise.
16018 (bclr_insn): Likewise.
16019 (bmsk_insn): Likewise.
16020 (bicsi3_insn): Likewise.
16021 (cmpsi_cc_c_insn): Likewise.
16022 (movsi_ne): Likewise.
16023 (movsi_cond_exec): Likewise.
16024 (clrsbsi2): Likewise.
16025 (norm_f): Likewise.
16035 (call_i): Remove 'S' asm letter, add jli instruction.
16036 (call_value_i): Likewise.
16037 * config/arc/arc.op (mjli-always): New option.
16038 * config/arc/constraints.md (Cji): New constraint.
16039 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
16041 (subsf3_fpx): Likewise.
16042 (mulsf3_fpx): Likewise.
16043 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
16045 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
16046 function attrbutes.
16047 * doc/invoke.texi (ARC): Document mjli-always option.
16049 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
16051 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
16052 avoid addition with 0 and use incw and decw where possible.
16054 2018-01-26 Richard Biener <rguenther@suse.de>
16056 PR tree-optimization/81082
16057 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
16058 association if it requires casting to unsigned.
16059 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
16060 from fold_plusminus_mult_expr to catch important cases late when
16061 range info is available.
16063 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16065 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
16066 * configure.ac (hidden_linkonce): New test.
16067 * configure: Regenerate.
16068 * config.in: Regenerate.
16070 2018-01-26 Julia Koval <julia.koval@intel.com>
16072 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
16073 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
16074 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
16075 _mm_mask_bitshuffle_epi64_mask): Fix type.
16076 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
16077 USI_FTYPE_V4DI_V4DI_USI): Remove.
16078 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
16079 __builtin_ia32_vpshufbitqmb256_mask,
16080 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
16081 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
16082 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
16084 2018-01-26 Alan Modra <amodra@gmail.com>
16087 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
16088 UNSPEC_VBPERMQ. Sort other unspecs.
16090 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
16092 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
16094 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
16096 PR middle-end/83055
16097 * predict.c (drop_profile): Do not push/pop cfun; update also
16099 (handle_missing_profiles): Fix logic looking for zero profiles.
16101 2018-01-25 Jakub Jelinek <jakub@redhat.com>
16103 PR middle-end/83977
16104 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
16105 on functions with #pragma omp declare simd or functions with simd
16107 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
16108 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
16109 Remove trailing \n from warning_at calls.
16111 2018-01-25 Tom de Vries <tom@codesourcery.com>
16114 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
16115 for neutered workers.
16117 2018-01-24 Joseph Myers <joseph@codesourcery.com>
16120 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
16121 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
16123 2018-01-24 Jeff Law <law@redhat.com>
16126 * i386.c (get_probe_interval): Move to earlier point.
16127 (ix86_compute_frame_layout): If -fstack-clash-protection and
16128 the frame is larger than the probe interval, then use pushes
16129 to save registers rather than reg->mem moves.
16130 (ix86_expand_prologue): Remove conditional for int_registers_saved
16133 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
16136 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
16137 min/max for never referenced object.
16139 2018-01-24 Jakub Jelinek <jakub@redhat.com>
16141 PR middle-end/83977
16142 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
16144 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
16145 attributes from DECL_ATTRIBUTES (decl) without affecting
16146 DECL_ATTRIBUTES (current_function_decl).
16147 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
16148 functions with non-NULL DECL_ABSTRACT_ORIGIN.
16150 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
16152 PR tree-optimization/83979
16153 * fold-const.c (fold_comparison): Use constant_boolean_node
16154 instead of boolean_{true,false}_node.
16156 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
16158 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
16161 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16163 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
16164 Simplify the clause that sets the length attribute.
16165 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
16166 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
16167 clause that sets the length attribute.
16168 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
16170 2018-01-24 Tom de Vries <tom@codesourcery.com>
16173 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
16174 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
16175 Add strict parameter.
16176 (prevent_branch_around_nothing): Insert dummy insn between branch to
16177 label and label with no ptx insn inbetween.
16178 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
16180 2018-01-24 Tom de Vries <tom@codesourcery.com>
16183 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
16184 for neutered threads in warp.
16185 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
16187 2018-01-24 Richard Biener <rguenther@suse.de>
16189 PR tree-optimization/83176
16190 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
16193 2018-01-24 Richard Biener <rguenther@suse.de>
16195 PR tree-optimization/82819
16196 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
16197 code generating pluses that are no-ops in the target precision.
16199 2018-01-24 Richard Biener <rguenther@suse.de>
16201 PR middle-end/84000
16202 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
16204 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
16206 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
16207 to merge probabilities.
16208 * predict.c (probably_never_executed): Also mark as cold functions
16209 with global 0 profile and guessed local profile.
16210 * profile-count.c (profile_probability::combine_with_count): New
16212 * profile-count.h (profile_probability::operator*,
16213 profile_probability::operator*=, profile_probability::operator/,
16214 profile_probability::operator/=): Reduce precision to adjusted
16215 and set value to guessed on contradictory divisions.
16216 (profile_probability::combine_with_freq): Remove.
16217 (profile_probability::combine_wiht_count): Declare.
16218 (profile_count::force_nonzero):: Set to adjusted.
16219 (profile_count::probability_in):: Set quality to adjusted.
16220 * tree-ssa-tail-merge.c (replace_block_by): Use
16221 combine_with_count.
16223 2018-01-23 Andrew Waterman <andrew@sifive.com>
16224 Jim Wilson <jimw@sifive.com>
16226 * config/riscv/riscv.c (riscv_stack_boundary): New.
16227 (riscv_option_override): Set riscv_stack_boundary. Handle
16228 riscv_preferred_stack_boundary_arg.
16229 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
16230 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
16231 (STACK_BOUNDARY): Set to riscv_stack_boundary.
16232 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
16233 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
16234 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
16236 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
16239 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
16240 of struct ix86_frame.
16241 (ix86_expand_epilogue): Likewise. Add a local variable for
16242 the reg_save_offset field in struct ix86_frame.
16244 2018-01-23 Bin Cheng <bin.cheng@arm.com>
16246 PR tree-optimization/82604
16247 * tree-loop-distribution.c (enum partition_kind): New enum item
16248 PKIND_PARTIAL_MEMSET.
16249 (partition_builtin_p): Support above new enum item.
16250 (generate_code_for_partition): Ditto.
16251 (compute_access_range): Differentiate cases that equality can be
16252 proven at all loops, the innermost loops or no loops.
16253 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
16254 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
16255 (finalize_partitions, distribute_loop): Don't fuse partition of
16256 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
16257 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
16258 parloop is enabled.
16260 2018-01-23 Martin Liska <mliska@suse.cz>
16262 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
16263 order to ignore the predictor.
16264 (PRED_POLYMORPHIC_CALL): Likewise.
16265 (PRED_RECURSIVE_CALL): Likewise.
16267 2018-01-23 Martin Liska <mliska@suse.cz>
16269 * tree-profile.c (tree_profiling): Print function header to
16270 aware reader which function we are working on.
16271 * value-prof.c (gimple_find_values_to_profile): Do not print
16272 not interesting value histograms.
16274 2018-01-23 Martin Liska <mliska@suse.cz>
16276 * profile-count.h (enum profile_quality): Add
16277 profile_uninitialized as the first value. Do not number values
16278 as they are zero based.
16279 (profile_count::verify): Update sanity check.
16280 (profile_probability::verify): Likewise.
16282 2018-01-23 Nathan Sidwell <nathan@acm.org>
16284 * doc/invoke.texi (ffor-scope): Deprecate.
16286 2018-01-23 David Malcolm <dmalcolm@redhat.com>
16288 PR tree-optimization/83510
16289 * domwalk.c (set_all_edges_as_executable): New function.
16290 (dom_walker::dom_walker): Convert bool param
16291 "skip_unreachable_blocks" to enum reachability. Move setup of
16292 edge flags to set_all_edges_as_executable and only do it when
16293 reachability is REACHABLE_BLOCKS.
16294 * domwalk.h (enum dom_walker::reachability): New enum.
16295 (dom_walker::dom_walker): Convert bool param
16296 "skip_unreachable_blocks" to enum reachability.
16297 (set_all_edges_as_executable): New decl.
16298 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
16299 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
16301 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
16302 but converting true to REACHABLE_BLOCKS.
16303 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
16305 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
16306 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
16307 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
16309 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
16310 if check_all_array_refs will be called.
16312 2018-01-23 David Malcolm <dmalcolm@redhat.com>
16314 * tree.c (selftest::test_location_wrappers): Add more test
16317 2018-01-23 David Malcolm <dmalcolm@redhat.com>
16319 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
16320 (selftest::test_bit_in_range): Likewise.
16322 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
16325 * doc/sourcebuild.texi (vect_float): Say that the selector
16326 only describes the situation when -funsafe-math-optimizations is on.
16327 (vect_float_strict): Document.
16329 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
16331 PR tree-optimization/83965
16332 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
16333 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
16334 instead of checking only for a reduction.
16335 (vect_recog_widen_sum_pattern): Likewise.
16337 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
16339 * predict.c (probably_never_executed): Only use precise profile info.
16340 (compute_function_frequency): Skip after inlining hack since we now
16341 have quality checking.
16343 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
16345 * profile-count.h (profile_probability::very_unlikely,
16346 profile_probability::unlikely, profile_probability::even): Set
16347 precision to guessed.
16349 2018-01-23 Richard Biener <rguenther@suse.de>
16351 PR tree-optimization/83963
16352 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
16353 Properly terminate dominator walk when crossing the exit edge not
16354 when visiting its source block.
16356 2018-01-23 Jakub Jelinek <jakub@redhat.com>
16359 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
16360 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
16362 2018-01-22 Jakub Jelinek <jakub@redhat.com>
16364 PR tree-optimization/83957
16365 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
16366 semicolon after for body surrounded by braces.
16368 PR tree-optimization/83081
16369 * profile-count.h (profile_probability::split): New method.
16370 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
16371 Use profile_probability::split.
16372 (do_compare_rtx_and_jump): Fix adjustment of probabilities
16373 when splitting a single conditional jump into 2.
16375 2018-01-22 David Malcolm <dmalcolm@redhat.com>
16377 PR tree-optimization/69452
16378 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
16381 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16383 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
16384 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
16385 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
16387 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16389 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
16391 * config/rl78/rl78.md (movdi): New define_expand.
16392 * config/rl78/rl78.c (rl78_split_movdi): New function.
16394 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
16397 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
16399 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
16400 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
16401 128-bit to produce an UNSPEC move to get the double word with the
16402 signbit and then a shift directly to do signbit.
16403 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
16404 implementation with a new version that just does either a direct
16405 move or a regular move. Move memory interface to separate insns.
16406 Move insns so they are next to the expander.
16407 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
16408 with signbit move. Split big and little endian case.
16409 (signbit<mode>2_dm_mem_le): Likewise.
16410 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
16411 (signbit<mode>2_dm2): Likewise.
16413 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16415 * config/rl78/rl78.md (anddi3): New define_expand.
16417 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16419 * config/rl78/rl78.md (umindi3): New define_expand.
16421 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16423 * config/rl78/rl78.md (smindi3): New define_expand.
16425 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16427 * config/rl78/rl78.md (smaxdi3): New define_expand.
16429 2018-01-22 Carl Love <cel@us.ibm.com>
16431 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
16432 LVX_V1TI): Add macro expansion.
16433 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
16434 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
16435 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
16436 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
16437 Change check to determine if the instruction is a byte reversing
16438 entry. Fix typo in comment.
16439 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
16440 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
16441 Add def_builtin calls for new builtins.
16442 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
16443 Add define_insn expansion.
16445 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16447 * config/rl78/rl78.md (umaxdi3): New define_expand.
16449 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
16451 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
16452 for non-QImode registers.
16454 2018-01-22 Richard Biener <rguenther@suse.de>
16456 PR tree-optimization/83963
16457 * graphite-scop-detection.c (scop_detection::get_sese): Delay
16458 including the loop exit block.
16459 (scop_detection::merge_sese): Likewise.
16460 (scop_detection::add_scop): Do it here instead.
16462 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16464 * doc/sourcebuild.texi (arm_softfloat): Document.
16466 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
16469 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
16470 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
16471 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
16473 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16474 David Edelsohn <dje.gcc@gmail.com>
16477 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
16478 Change "crset eq" to "crset 2".
16479 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
16480 (*call_indirect_aix<mode>_nospec): Likewise.
16481 (*call_value_indirect_aix<mode>_nospec): Likewise.
16482 (*call_indirect_elfv2<mode>_nospec): Likewise.
16483 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
16484 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
16485 change assembly output from . to $.
16486 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
16487 (indirect_jump<mode>_nospec): Change assembly output from . to $.
16488 (*tablejump<mode>_internal1_nospec): Likewise.
16490 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
16493 * config/sh/sh_optimize_sett_clrt.cc:
16494 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
16496 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
16498 PR tree-optimization/83940
16499 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
16500 offset_dt to vect_constant_def rather than vect_unknown_def_type.
16501 (vect_check_load_store_mask): Add a mask_dt_out parameter and
16502 use it to pass back the definition type.
16503 (vect_check_store_rhs): Likewise rhs_dt_out.
16504 (vect_build_gather_load_calls): Add a mask_dt argument and use
16505 it instead of a call to vect_is_simple_use.
16506 (vectorizable_store): Update calls to vect_check_load_store_mask
16507 and vect_check_store_rhs. Use the dt returned by the latter instead
16508 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
16509 instead of calls to vect_is_simple_use. Pass the scalar rather
16510 than the vector operand to vect_is_simple_use when handling
16511 second and subsequent copies of an rhs value.
16512 (vectorizable_load): Update calls to vect_check_load_store_mask
16513 and vect_build_gather_load_calls. Use the cached mask_dt and
16514 gs_info.offset_dt instead of calls to vect_is_simple_use.
16516 2018-01-20 Jakub Jelinek <jakub@redhat.com>
16518 PR middle-end/83945
16519 * tree-emutls.c: Include gimplify.h.
16520 (lower_emutls_2): New function.
16521 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
16522 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
16523 it before further processing.
16526 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
16527 UINTVAL (trueop1) instead of INTVAL (op1).
16529 2018-01-19 Jakub Jelinek <jakub@redhat.com>
16533 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
16534 INCOMING_FRAME_SP_OFFSET if not defined.
16535 (scan_trace): Add ENTRY argument. If true and
16536 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
16537 emit a note to adjust the CFA offset.
16538 (create_cfi_notes): Adjust scan_trace callers.
16539 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
16540 INCOMING_FRAME_SP_OFFSET in the CIE.
16541 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
16542 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
16544 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
16545 * doc/tm.texi: Regenerated.
16547 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
16549 PR rtl-optimization/83147
16550 * lra-constraints.c (remove_inheritance_pseudos): Use
16551 lra_substitute_pseudo_within_insn.
16553 2018-01-19 Tom de Vries <tom@codesourcery.com>
16554 Cesar Philippidis <cesar@codesourcery.com>
16557 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
16559 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
16562 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
16563 spaces for function labels.
16565 2018-01-19 Martin Liska <mliska@suse.cz>
16567 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
16568 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
16569 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
16570 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
16571 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
16572 (PRED_CONST_RETURN): Change from 69 to 65.
16573 (PRED_NULL_RETURN): Change from 91 to 71.
16574 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
16575 (PRED_LOOP_GUARD): Change from 66 to 73.
16577 2018-01-19 Martin Liska <mliska@suse.cz>
16579 * predict.c (predict_insn_def): Add new assert.
16580 (struct branch_predictor): Change type to signed integer.
16581 (test_prediction_value_range): Amend test to cover
16582 PROB_UNINITIALIZED.
16583 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
16584 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
16585 (PRED_LOOP_ITERATIONS_MAX): Likewise.
16586 (PRED_LOOP_IV_COMPARE): Likewise.
16587 * predict.h (PROB_UNINITIALIZED): Define new constant.
16589 2018-01-19 Martin Liska <mliska@suse.cz>
16591 * predict.c (dump_prediction): Add new format for
16592 analyze_brprob.py script which is enabled with -details
16594 * profile-count.h (precise_p): New function.
16596 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
16598 PR tree-optimization/83922
16599 * tree-vect-loop.c (vect_verify_full_masking): Return false if
16600 there are no statements that need masking.
16601 (vect_active_double_reduction_p): New function.
16602 (vect_analyze_loop_operations): Use it when handling phis that
16603 are not in the loop header.
16605 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
16607 PR tree-optimization/83914
16608 * tree-vect-loop.c (vectorizable_induction): Don't convert
16609 init_expr or apply the peeling adjustment for inductions
16610 that are nested within the vectorized loop.
16612 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16614 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
16617 2018-01-18 Jakub Jelinek <jakub@redhat.com>
16621 * function.h (gimplify_parameters): Add gimple_seq * argument.
16622 * function.c: Include gimple.h and options.h.
16623 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
16624 for the added local temporaries if needed.
16625 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
16626 if there are any parameter cleanups, wrap whole body into a
16627 try/finally with the cleanups.
16629 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
16632 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
16633 Use GET_MODE_CLASS for scalar floating point.
16635 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
16639 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
16640 Fix call of call_cgraph_insertion_hooks.
16642 2018-01-18 Martin Sebor <msebor@redhat.com>
16644 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
16646 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
16649 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
16652 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
16655 * common.opt: (-ffile-prefix-map): New option.
16656 * opts.c (common_handle_option): Defer it.
16657 * opts-global.c (handle_common_deferred_options): Handle it.
16658 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
16659 * file-prefix-map.h: New file.
16660 (remap_debug_filename, add_debug_prefix_map): ...here.
16661 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
16662 * final.c (debug_prefix_map, add_debug_prefix_map
16663 remap_debug_filename): Move to...
16664 * file-prefix-map.c: New file.
16665 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
16666 generalize, get rid of alloca(), use strrchr() instead of strchr().
16667 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
16668 Implement in terms of add_prefix_map().
16669 (remap_macro_filename, remap_debug_filename): Implement in term of
16671 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
16672 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
16673 * dbxout.c: Include file-prefix-map.h.
16674 * varasm.c: Likewise.
16675 * vmsdbgout.c: Likewise.
16676 * xcoffout.c: Likewise.
16677 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
16678 * doc/cppopts.texi (-fmacro-prefix-map): Document.
16679 * doc/invoke.texi (-ffile-prefix-map): Document.
16680 (-fdebug-prefix-map): Update description.
16682 2018-01-18 Martin Liska <mliska@suse.cz>
16684 * config/i386/i386.c (indirect_thunk_name): Document that also
16686 (output_indirect_thunk): Document why both instructions
16687 (pause and lfence) are generated.
16689 2018-01-18 Richard Biener <rguenther@suse.de>
16691 PR tree-optimization/83887
16692 * graphite-scop-detection.c
16693 (scop_detection::get_nearest_dom_with_single_entry): Remove.
16694 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
16695 (scop_detection::merge_sese): Re-implement with a flood-fill
16696 algorithm that properly finds a SESE region if it exists.
16698 2018-01-18 Jakub Jelinek <jakub@redhat.com>
16701 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
16702 pointer_diff optimizations use view_convert instead of convert.
16704 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16706 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
16707 Generate different code for -mno-speculate-indirect-jumps.
16708 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
16709 (*call_indirect_aix<mode>): Disable for
16710 -mno-speculate-indirect-jumps.
16711 (*call_indirect_aix<mode>_nospec): New define_insn.
16712 (*call_value_indirect_aix<mode>): Disable for
16713 -mno-speculate-indirect-jumps.
16714 (*call_value_indirect_aix<mode>_nospec): New define_insn.
16715 (*sibcall_nonlocal_sysv<mode>): Generate different code for
16716 -mno-speculate-indirect-jumps.
16717 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
16719 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
16721 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
16722 long double type, set the flags for noting the default long double
16723 type, even if we don't pass or return a long double type.
16725 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
16728 * ipa-inline.c (flatten_function): Do not overwrite final inlining
16731 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
16733 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
16734 support for merge[hl].
16735 (fold_mergehl_helper): New helper function.
16736 (tree-vector-builder.h): New #include for tree_vector_builder usage.
16737 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
16738 (altivec_vmrglw_direct): Add xxmrglw insn.
16740 2018-01-17 Andrew Waterman <andrew@sifive.com>
16742 * config/riscv/riscv.c (riscv_conditional_register_usage): If
16743 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
16745 2018-01-17 David Malcolm <dmalcolm@redhat.com>
16748 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
16749 call the lto_location_cache before reading the
16750 DECL_SOURCE_LOCATION of the types.
16752 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
16753 Richard Sandiford <richard.sandiford@linaro.org>
16755 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
16756 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
16757 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
16758 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
16759 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
16761 * config/aarch64/constraints.md (aarch64_movti_operand):
16763 * config/aarch64/predicates.md (Uti): Add new constraint.
16765 2018-01-17 Carl Love <cel@us.ibm.com>
16767 * config/rs6000/vsx.md (define_expand xl_len_r,
16768 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
16769 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
16771 (define_expand, define_insn): Move the shift left from the
16772 define_insn to the define_expand for lxvl and stxvl instructions.
16773 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
16774 and XL_LEN_R definitions to PURE.
16776 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
16778 * config/i386/i386.c (indirect_thunk_name): Declare regno
16779 as unsigned int. Compare regno with INVALID_REGNUM.
16780 (output_indirect_thunk): Ditto.
16781 (output_indirect_thunk_function): Ditto.
16782 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
16783 in the call to output_indirect_thunk_function.
16785 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
16787 PR middle-end/83884
16788 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
16789 rather than the size of inner_type to determine the stack slot size
16790 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
16792 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
16795 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
16798 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
16800 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
16801 endian Linux systems to optionally enable multilibs for selecting
16802 the long double type if the user configured an explicit type.
16803 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
16804 have no long double multilibs if not defined.
16805 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
16806 warn if the user used -mabi={ieee,ibm}longdouble and we built
16807 multilibs for long double.
16808 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
16809 appropriate multilib option.
16810 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
16812 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
16813 for building long double multilibs.
16814 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
16816 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
16818 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
16821 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
16823 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
16826 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
16829 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
16832 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
16834 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
16835 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
16837 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
16839 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
16840 different rtl trees depending on TARGET_64BIT.
16841 (rs6000_gen_lvx): Likewise.
16843 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
16845 * config/visium/visium.md (nop): Tweak comment.
16846 (hazard_nop): Likewise.
16848 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16850 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
16851 -mspeculate-indirect-jumps.
16852 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
16853 for -mno-speculate-indirect-jumps.
16854 (*call_indirect_elfv2<mode>_nospec): New define_insn.
16855 (*call_value_indirect_elfv2<mode>): Disable for
16856 -mno-speculate-indirect-jumps.
16857 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
16858 (indirect_jump): Emit different RTL for
16859 -mno-speculate-indirect-jumps.
16860 (*indirect_jump<mode>): Disable for
16861 -mno-speculate-indirect-jumps.
16862 (*indirect_jump<mode>_nospec): New define_insn.
16863 (tablejump): Emit different RTL for
16864 -mno-speculate-indirect-jumps.
16865 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
16866 (tablejumpsi_nospec): New define_expand.
16867 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
16868 (tablejumpdi_nospec): New define_expand.
16869 (*tablejump<mode>_internal1): Disable for
16870 -mno-speculate-indirect-jumps.
16871 (*tablejump<mode>_internal1_nospec): New define_insn.
16872 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
16875 2018-01-16 Artyom Skrobov tyomitch@gmail.com
16877 * caller-save.c (insert_save): Drop unnecessary parameter. All
16880 2018-01-16 Jakub Jelinek <jakub@redhat.com>
16881 Richard Biener <rguenth@suse.de>
16884 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
16885 return early, inline manually is_gimple_sizepos. Make sure if we
16886 call gimplify_expr we don't end up with a gimple constant.
16887 * tree.c (variably_modified_type_p): Don't return true for
16888 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
16889 * gimplify.h (is_gimple_sizepos): Remove.
16891 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
16893 PR tree-optimization/83857
16894 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
16895 vectorizable_live_operation for pure SLP statements.
16896 (vectorizable_live_operation): Handle PHIs.
16898 2018-01-16 Richard Biener <rguenther@suse.de>
16900 PR tree-optimization/83867
16901 * tree-vect-stmts.c (vect_transform_stmt): Precompute
16902 nested_in_vect_loop_p since the scalar stmt may get invalidated.
16904 2018-01-16 Jakub Jelinek <jakub@redhat.com>
16907 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
16908 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
16909 If off is not INTEGER_CST, issue a may not be aligned warning
16910 rather than isn't aligned. Use isn%'t rather than isn't.
16911 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
16913 <case MULT_EXPR>: Improve the case when bottom and one of the
16914 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
16915 operand, in that case check if the other operand is multiple of
16916 bottom divided by the INTEGER_CST operand.
16918 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
16921 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
16922 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
16923 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
16924 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
16925 * config/pa/pa.c (pa_function_arg_advance): Likewise.
16926 (pa_function_arg, pa_arg_partial_bytes): Likewise.
16927 (pa_function_arg_size): New function.
16929 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
16931 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
16932 in a separate statement.
16934 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
16936 PR tree-optimization/83847
16937 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
16938 group gathers and scatters.
16940 2018-01-16 Jakub Jelinek <jakub@redhat.com>
16942 PR rtl-optimization/86620
16943 * params.def (max-sched-ready-insns): Bump minimum value to 1.
16945 PR rtl-optimization/83213
16946 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
16947 to last if both are JUMP_INSNs.
16949 PR tree-optimization/83843
16950 * gimple-ssa-store-merging.c
16951 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
16952 store_immediate_info for bswap/nop orig_stores.
16954 2018-01-15 Andrew Waterman <andrew@sifive.com>
16956 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
16958 <UDIV>: Increase cost if !TARGET_DIV.
16960 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
16962 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
16963 (define_attr "cr_logical_3op"): New.
16964 (cceq_ior_compare): Adjust.
16965 (cceq_ior_compare_complement): Adjust.
16966 (*cceq_rev_compare): Adjust.
16967 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
16968 (is_cracked_insn): Adjust.
16969 (insn_must_be_first_in_group): Adjust.
16970 * config/rs6000/40x.md: Adjust.
16971 * config/rs6000/440.md: Adjust.
16972 * config/rs6000/476.md: Adjust.
16973 * config/rs6000/601.md: Adjust.
16974 * config/rs6000/603.md: Adjust.
16975 * config/rs6000/6xx.md: Adjust.
16976 * config/rs6000/7450.md: Adjust.
16977 * config/rs6000/7xx.md: Adjust.
16978 * config/rs6000/8540.md: Adjust.
16979 * config/rs6000/cell.md: Adjust.
16980 * config/rs6000/e300c2c3.md: Adjust.
16981 * config/rs6000/e500mc.md: Adjust.
16982 * config/rs6000/e500mc64.md: Adjust.
16983 * config/rs6000/e5500.md: Adjust.
16984 * config/rs6000/e6500.md: Adjust.
16985 * config/rs6000/mpc.md: Adjust.
16986 * config/rs6000/power4.md: Adjust.
16987 * config/rs6000/power5.md: Adjust.
16988 * config/rs6000/power6.md: Adjust.
16989 * config/rs6000/power7.md: Adjust.
16990 * config/rs6000/power8.md: Adjust.
16991 * config/rs6000/power9.md: Adjust.
16992 * config/rs6000/rs64.md: Adjust.
16993 * config/rs6000/titan.md: Adjust.
16995 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
16997 * config/i386/predicates.md (indirect_branch_operand): Rewrite
16998 ix86_indirect_branch_register logic.
17000 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
17002 * config/i386/constraints.md (Bs): Update
17003 ix86_indirect_branch_register check. Don't check
17004 ix86_indirect_branch_register with GOT_memory_operand.
17006 * config/i386/predicates.md (GOT_memory_operand): Don't check
17007 ix86_indirect_branch_register here.
17008 (GOT32_symbol_operand): Likewise.
17010 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
17012 * config/i386/predicates.md (constant_call_address_operand):
17013 Rewrite ix86_indirect_branch_register logic.
17014 (sibcall_insn_operand): Likewise.
17016 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
17018 * config/i386/constraints.md (Bs): Replace
17019 ix86_indirect_branch_thunk_register with
17020 ix86_indirect_branch_register.
17022 * config/i386/i386.md (indirect_jump): Likewise.
17023 (tablejump): Likewise.
17024 (*sibcall_memory): Likewise.
17025 (*sibcall_value_memory): Likewise.
17026 Peepholes of indirect call and jump via memory: Likewise.
17027 * config/i386/i386.opt: Likewise.
17028 * config/i386/predicates.md (indirect_branch_operand): Likewise.
17029 (GOT_memory_operand): Likewise.
17030 (call_insn_operand): Likewise.
17031 (sibcall_insn_operand): Likewise.
17032 (GOT32_symbol_operand): Likewise.
17034 2018-01-15 Jakub Jelinek <jakub@redhat.com>
17036 PR middle-end/83837
17037 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
17038 type rather than type addr's type points to.
17039 (expand_omp_atomic_mutex): Likewise.
17040 (expand_omp_atomic): Likewise.
17042 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
17045 * config/i386/i386.c (output_indirect_thunk_function): Use
17046 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
17047 for __x86_return_thunk.
17049 2018-01-15 Richard Biener <rguenther@suse.de>
17051 PR middle-end/83850
17052 * expmed.c (extract_bit_field_1): Fix typo.
17054 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17057 * config/arm/iterators.md (VF): New mode iterator.
17058 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
17059 Remove integer-related logic from pattern.
17060 (neon_vabd<mode>_3): Likewise.
17062 2018-01-15 Jakub Jelinek <jakub@redhat.com>
17064 PR middle-end/82694
17065 * common.opt (fstrict-overflow): No longer an alias.
17066 (fwrapv-pointer): New option.
17067 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
17068 also for pointer types based on flag_wrapv_pointer.
17069 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
17070 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
17071 opts->x_flag_wrapv got set.
17072 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
17073 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
17074 POINTER_TYPE_OVERFLOW_UNDEFINED.
17075 * match.pd: Likewise in address comparison pattern.
17076 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
17078 2018-01-15 Richard Biener <rguenther@suse.de>
17081 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
17082 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
17083 Reset type names to their identifier if their TYPE_DECL doesn't
17084 have linkage (and thus is used for ODR and devirt).
17085 (save_debug_info_for_decl): Remove.
17086 (save_debug_info_for_type): Likewise.
17087 (add_tree_to_fld_list): Adjust.
17088 * tree-pretty-print.c (dump_generic_node): Make dumping of
17089 type names more robust.
17091 2018-01-15 Richard Biener <rguenther@suse.de>
17093 * BASE-VER: Bump to 8.0.1.
17095 2018-01-14 Martin Sebor <msebor@redhat.com>
17098 * builtins.c (check_access): Avoid warning when the no-warning bit
17101 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
17103 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
17104 * ira-color (allocno_hard_regs_compare): Likewise.
17106 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
17109 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
17110 Use .pushsection/.popsection.
17112 2018-01-14 Martin Sebor <msebor@redhat.com>
17115 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
17117 2018-01-14 Jakub Jelinek <jakub@redhat.com>
17119 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
17120 entry from extra_headers.
17121 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
17122 extra_headers, make the list bitwise identical to the i?86-*-* one.
17124 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
17126 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
17127 -mcmodel=large with -mindirect-branch=thunk,
17128 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
17129 -mfunction-return=thunk-extern.
17130 * doc/invoke.texi: Document -mcmodel=large is incompatible with
17131 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
17132 -mfunction-return=thunk and -mfunction-return=thunk-extern.
17134 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
17136 * config/i386/i386.c (print_reg): Print the name of the full
17137 integer register without '%'.
17138 (ix86_print_operand): Handle 'V'.
17139 * doc/extend.texi: Document 'V' modifier.
17141 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
17143 * config/i386/constraints.md (Bs): Disallow memory operand for
17144 -mindirect-branch-register.
17146 * config/i386/predicates.md (indirect_branch_operand): Likewise.
17147 (GOT_memory_operand): Likewise.
17148 (call_insn_operand): Likewise.
17149 (sibcall_insn_operand): Likewise.
17150 (GOT32_symbol_operand): Likewise.
17151 * config/i386/i386.md (indirect_jump): Call convert_memory_address
17152 for -mindirect-branch-register.
17153 (tablejump): Likewise.
17154 (*sibcall_memory): Likewise.
17155 (*sibcall_value_memory): Likewise.
17156 Disallow peepholes of indirect call and jump via memory for
17157 -mindirect-branch-register.
17158 (*call_pop): Replace m with Bw.
17159 (*call_value_pop): Likewise.
17160 (*sibcall_pop_memory): Replace m with Bs.
17161 * config/i386/i386.opt (mindirect-branch-register): New option.
17162 * doc/invoke.texi: Document -mindirect-branch-register option.
17164 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
17166 * config/i386/i386-protos.h (ix86_output_function_return): New.
17167 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
17168 set function_return_type.
17169 (indirect_thunk_name): Add ret_p to indicate thunk for function
17171 (output_indirect_thunk_function): Pass false to
17172 indirect_thunk_name.
17173 (ix86_output_indirect_branch_via_reg): Likewise.
17174 (ix86_output_indirect_branch_via_push): Likewise.
17175 (output_indirect_thunk_function): Create alias for function
17176 return thunk if regno < 0.
17177 (ix86_output_function_return): New function.
17178 (ix86_handle_fndecl_attribute): Handle function_return.
17179 (ix86_attribute_table): Add function_return.
17180 * config/i386/i386.h (machine_function): Add
17181 function_return_type.
17182 * config/i386/i386.md (simple_return_internal): Use
17183 ix86_output_function_return.
17184 (simple_return_internal_long): Likewise.
17185 * config/i386/i386.opt (mfunction-return=): New option.
17186 (indirect_branch): Mention -mfunction-return=.
17187 * doc/extend.texi: Document function_return function attribute.
17188 * doc/invoke.texi: Document -mfunction-return= option.
17190 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
17192 * config/i386/i386-opts.h (indirect_branch): New.
17193 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
17194 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
17195 with local indirect jump when converting indirect call and jump.
17196 (ix86_set_indirect_branch_type): New.
17197 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
17198 (indirectlabelno): New.
17199 (indirect_thunk_needed): Likewise.
17200 (indirect_thunk_bnd_needed): Likewise.
17201 (indirect_thunks_used): Likewise.
17202 (indirect_thunks_bnd_used): Likewise.
17203 (INDIRECT_LABEL): Likewise.
17204 (indirect_thunk_name): Likewise.
17205 (output_indirect_thunk): Likewise.
17206 (output_indirect_thunk_function): Likewise.
17207 (ix86_output_indirect_branch_via_reg): Likewise.
17208 (ix86_output_indirect_branch_via_push): Likewise.
17209 (ix86_output_indirect_branch): Likewise.
17210 (ix86_output_indirect_jmp): Likewise.
17211 (ix86_code_end): Call output_indirect_thunk_function if needed.
17212 (ix86_output_call_insn): Call ix86_output_indirect_branch if
17214 (ix86_handle_fndecl_attribute): Handle indirect_branch.
17215 (ix86_attribute_table): Add indirect_branch.
17216 * config/i386/i386.h (machine_function): Add indirect_branch_type
17217 and has_local_indirect_jump.
17218 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
17220 (tablejump): Likewise.
17221 (*indirect_jump): Use ix86_output_indirect_jmp.
17222 (*tablejump_1): Likewise.
17223 (simple_return_indirect_internal): Likewise.
17224 * config/i386/i386.opt (mindirect-branch=): New option.
17225 (indirect_branch): New.
17228 (thunk-inline): Likewise.
17229 (thunk-extern): Likewise.
17230 * doc/extend.texi: Document indirect_branch function attribute.
17231 * doc/invoke.texi: Document -mindirect-branch= option.
17233 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
17236 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
17238 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
17240 * ipa-inline.c (want_inline_small_function_p): Return false if
17241 inlining has already failed with CIF_FINAL_ERROR.
17242 (update_caller_keys): Call want_inline_small_function_p before
17244 (update_callee_keys): Likewise.
17246 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
17248 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
17250 (rs6000_quadword_masked_address_p): Likewise.
17251 (quad_aligned_load_p): Likewise.
17252 (quad_aligned_store_p): Likewise.
17253 (const_load_sequence_p): Add comment to describe the outer-most loop.
17254 (mimic_memory_attributes_and_flags): New function.
17255 (rs6000_gen_stvx): Likewise.
17256 (replace_swapped_aligned_store): Likewise.
17257 (rs6000_gen_lvx): Likewise.
17258 (replace_swapped_aligned_load): Likewise.
17259 (replace_swapped_load_constant): Capitalize argument name in
17260 comment describing this function.
17261 (rs6000_analyze_swaps): Add a third pass to search for vector loads
17262 and stores that access quad-word aligned addresses and replace
17263 with stvx or lvx instructions when appropriate.
17264 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
17265 New function prototype.
17266 (rs6000_quadword_masked_address_p): Likewise.
17267 (rs6000_gen_lvx): Likewise.
17268 (rs6000_gen_stvx): Likewise.
17269 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
17270 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
17271 when memory address is aligned.
17272 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
17273 this split to select lvx instruction when memory address is aligned.
17274 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
17275 instruction when memory address is aligned.
17276 (*vsx_le_perm_load_v16qi): Likewise.
17277 (four unnamed splitters): Modify to select the stvx instruction
17278 when memory is aligned.
17280 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
17282 * predict.c (determine_unlikely_bbs): Handle correctly BBs
17283 which appears in the queue multiple times.
17285 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17286 Alan Hayward <alan.hayward@arm.com>
17287 David Sherwood <david.sherwood@arm.com>
17289 * tree-vectorizer.h (vec_lower_bound): New structure.
17290 (_loop_vec_info): Add check_nonzero and lower_bounds.
17291 (LOOP_VINFO_CHECK_NONZERO): New macro.
17292 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
17293 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
17294 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
17295 fields. Make seg_len the distance travelled, not including the
17297 (dr_direction_indicator): Declare.
17298 (dr_zero_step_indicator): Likewise.
17299 (dr_known_forward_stride_p): Likewise.
17300 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
17302 (runtime_alias_check_p): Allow runtime alias checks with
17304 (operator ==): Compare access_size and align.
17305 (prune_runtime_alias_test_list): Rework for new distinction between
17306 the access_size and seg_len.
17307 (create_intersect_range_checks_index): Likewise. Cope with polynomial
17309 (get_segment_min_max): New function.
17310 (create_intersect_range_checks): Use it.
17311 (dr_step_indicator): New function.
17312 (dr_direction_indicator): Likewise.
17313 (dr_zero_step_indicator): Likewise.
17314 (dr_known_forward_stride_p): Likewise.
17315 * tree-loop-distribution.c (data_ref_segment_size): Return
17316 DR_STEP * (niters - 1).
17317 (compute_alias_check_pairs): Update call to the dr_with_seg_len
17319 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
17320 (vect_preserves_scalar_order_p): New function, split out from...
17321 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
17322 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
17323 (vect_vfa_access_size): New function.
17324 (vect_vfa_align): Likewise.
17325 (vect_compile_time_alias): Take access_size_a and access_b arguments.
17326 (dump_lower_bound): New function.
17327 (vect_check_lower_bound): Likewise.
17328 (vect_small_gap_p): Likewise.
17329 (vectorizable_with_step_bound_p): Likewise.
17330 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
17331 depencies if the vectorization factor is 1. Convert the checks
17332 for nonzero steps into checks on the bounds of DR_STEP. Try using
17333 a bunds check for variable steps if the minimum required step is
17334 relatively small. Update calls to the dr_with_seg_len
17335 constructor and to vect_compile_time_alias.
17336 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
17338 (vect_loop_versioning): Call it.
17339 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
17341 (vect_estimate_min_profitable_iters): Account for any bounds checks.
17343 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17344 Alan Hayward <alan.hayward@arm.com>
17345 David Sherwood <david.sherwood@arm.com>
17347 * doc/sourcebuild.texi (vect_scatter_store): Document.
17348 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
17350 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
17352 * genopinit.c (main): Add supports_vec_scatter_store and
17353 supports_vec_scatter_store_cached to target_optabs.
17354 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
17355 IFN_MASK_SCATTER_STORE.
17356 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
17358 * internal-fn.h (internal_store_fn_p): Declare.
17359 (internal_fn_stored_value_index): Likewise.
17360 * internal-fn.c (scatter_store_direct): New macro.
17361 (expand_scatter_store_optab_fn): New function.
17362 (direct_scatter_store_optab_supported_p): New macro.
17363 (internal_store_fn_p): New function.
17364 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
17365 IFN_MASK_SCATTER_STORE.
17366 (internal_fn_mask_index): Likewise.
17367 (internal_fn_stored_value_index): New function.
17368 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
17369 for scatter stores.
17370 * optabs-query.h (supports_vec_scatter_store_p): Declare.
17371 * optabs-query.c (supports_vec_scatter_store_p): New function.
17372 * tree-vectorizer.h (vect_get_store_rhs): Declare.
17373 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
17374 true for scatter stores.
17375 (vect_gather_scatter_fn_p): Handle scatter stores too.
17376 (vect_check_gather_scatter): Consider using scatter stores if
17377 supports_vec_scatter_store_p.
17378 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
17379 scatter stores too.
17380 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
17381 internal_fn_stored_value_index.
17382 (check_load_store_masking): Handle scatter stores too.
17383 (vect_get_store_rhs): Make public.
17384 (vectorizable_call): Use internal_store_fn_p.
17385 (vectorizable_store): Handle scatter store internal functions.
17386 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
17387 when deciding whether the end of the group has been reached.
17388 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
17389 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
17390 (mask_scatter_store<mode>): New insns.
17392 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17393 Alan Hayward <alan.hayward@arm.com>
17394 David Sherwood <david.sherwood@arm.com>
17396 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
17397 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
17398 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
17400 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
17401 Use vect_truncate_gather_scatter_offset if we can't treat the
17402 operation as a normal gather load or scatter store.
17403 (get_group_load_store_type): Take the gather_scatter_info
17404 as argument. Try using a gather load or scatter store for
17405 single-element groups.
17406 (get_load_store_type): Update calls to get_group_load_store_type
17407 and vect_use_strided_gather_scatters_p.
17409 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17410 Alan Hayward <alan.hayward@arm.com>
17411 David Sherwood <david.sherwood@arm.com>
17413 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
17414 optional tree argument.
17415 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
17417 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
17418 but continue to use the current value as a fallback.
17419 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
17420 to compare the updates.
17421 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
17422 (get_load_store_type): Use it when handling a strided access.
17423 (vect_get_strided_load_store_ops): New function.
17424 (vect_get_data_ptr_increment): Likewise.
17425 (vectorizable_load): Handle strided gather loads. Always pass
17426 a step to vect_create_data_ref_ptr and bump_vector_ptr.
17428 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17429 Alan Hayward <alan.hayward@arm.com>
17430 David Sherwood <david.sherwood@arm.com>
17432 * doc/md.texi (gather_load@var{m}): Document.
17433 (mask_gather_load@var{m}): Likewise.
17434 * genopinit.c (main): Add supports_vec_gather_load and
17435 supports_vec_gather_load_cached to target_optabs.
17436 * optabs-tree.c (init_tree_optimization_optabs): Use
17437 ggc_cleared_alloc to allocate target_optabs.
17438 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
17439 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
17441 * internal-fn.h (internal_load_fn_p): Declare.
17442 (internal_gather_scatter_fn_p): Likewise.
17443 (internal_fn_mask_index): Likewise.
17444 (internal_gather_scatter_fn_supported_p): Likewise.
17445 * internal-fn.c (gather_load_direct): New macro.
17446 (expand_gather_load_optab_fn): New function.
17447 (direct_gather_load_optab_supported_p): New macro.
17448 (direct_internal_fn_optab): New function.
17449 (internal_load_fn_p): Likewise.
17450 (internal_gather_scatter_fn_p): Likewise.
17451 (internal_fn_mask_index): Likewise.
17452 (internal_gather_scatter_fn_supported_p): Likewise.
17453 * optabs-query.c (supports_at_least_one_mode_p): New function.
17454 (supports_vec_gather_load_p): Likewise.
17455 * optabs-query.h (supports_vec_gather_load_p): Declare.
17456 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
17457 and memory_type field.
17458 (NUM_PATTERNS): Bump to 15.
17459 * tree-vect-data-refs.c: Include internal-fn.h.
17460 (vect_gather_scatter_fn_p): New function.
17461 (vect_describe_gather_scatter_call): Likewise.
17462 (vect_check_gather_scatter): Try using internal functions for
17463 gather loads. Recognize existing calls to a gather load function.
17464 (vect_analyze_data_refs): Consider using gather loads if
17465 supports_vec_gather_load_p.
17466 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
17467 (vect_get_gather_scatter_offset_type): Likewise.
17468 (vect_convert_mask_for_vectype): Likewise.
17469 (vect_add_conversion_to_patterm): Likewise.
17470 (vect_try_gather_scatter_pattern): Likewise.
17471 (vect_recog_gather_scatter_pattern): New pattern recognizer.
17472 (vect_vect_recog_func_ptrs): Add it.
17473 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
17474 internal_fn_mask_index and internal_gather_scatter_fn_p.
17475 (check_load_store_masking): Take the gather_scatter_info as an
17476 argument and handle gather loads.
17477 (vect_get_gather_scatter_ops): New function.
17478 (vectorizable_call): Check internal_load_fn_p.
17479 (vectorizable_load): Likewise. Handle gather load internal
17481 (vectorizable_store): Update call to check_load_store_masking.
17482 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
17483 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
17484 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
17485 (aarch64_gather_scale_operand_d): New predicates.
17486 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
17487 (mask_gather_load<mode>): New insns.
17489 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17490 Alan Hayward <alan.hayward@arm.com>
17491 David Sherwood <david.sherwood@arm.com>
17493 * optabs.def (fold_left_plus_optab): New optab.
17494 * doc/md.texi (fold_left_plus_@var{m}): Document.
17495 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
17496 * internal-fn.c (fold_left_direct): Define.
17497 (expand_fold_left_optab_fn): Likewise.
17498 (direct_fold_left_optab_supported_p): Likewise.
17499 * fold-const-call.c (fold_const_fold_left): New function.
17500 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
17501 * tree-parloops.c (valid_reduction_p): New function.
17502 (gather_scalar_reductions): Use it.
17503 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
17504 (vect_finish_replace_stmt): Declare.
17505 * tree-vect-loop.c (fold_left_reduction_fn): New function.
17506 (needs_fold_left_reduction_p): New function, split out from...
17507 (vect_is_simple_reduction): ...here. Accept reductions that
17508 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
17509 (vect_force_simple_reduction): Also store the reduction type in
17510 the assignment's STMT_VINFO_REDUC_TYPE.
17511 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
17512 (merge_with_identity): New function.
17513 (vect_expand_fold_left): Likewise.
17514 (vectorize_fold_left_reduction): Likewise.
17515 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
17516 scalar phi in place for it. Check for target support and reject
17517 cases that would reassociate the operation. Defer the transform
17518 phase to vectorize_fold_left_reduction.
17519 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
17520 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
17521 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
17523 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17525 * tree-if-conv.c (predicate_mem_writes): Remove redundant
17526 call to ifc_temp_var.
17528 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17529 Alan Hayward <alan.hayward@arm.com>
17530 David Sherwood <david.sherwood@arm.com>
17532 * target.def (legitimize_address_displacement): Take the original
17533 offset as a poly_int.
17534 * targhooks.h (default_legitimize_address_displacement): Update
17536 * targhooks.c (default_legitimize_address_displacement): Likewise.
17537 * doc/tm.texi: Regenerate.
17538 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
17539 as an argument, moving assert of ad->disp == ad->disp_term to...
17540 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
17541 Try calling targetm.legitimize_address_displacement before expanding
17542 the address rather than afterwards, and adjust for the new interface.
17543 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
17544 Match the new hook interface. Handle SVE addresses.
17545 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
17546 new hook interface.
17548 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17550 * Makefile.in (OBJS): Add early-remat.o.
17551 * target.def (select_early_remat_modes): New hook.
17552 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
17553 * doc/tm.texi: Regenerate.
17554 * targhooks.h (default_select_early_remat_modes): Declare.
17555 * targhooks.c (default_select_early_remat_modes): New function.
17556 * timevar.def (TV_EARLY_REMAT): New timevar.
17557 * passes.def (pass_early_remat): New pass.
17558 * tree-pass.h (make_pass_early_remat): Declare.
17559 * early-remat.c: New file.
17560 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
17562 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
17564 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17565 Alan Hayward <alan.hayward@arm.com>
17566 David Sherwood <david.sherwood@arm.com>
17568 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
17569 vfm1 with a bound_epilog parameter.
17570 (vect_do_peeling): Update calls accordingly, and move the prologue
17571 call earlier in the function. Treat the base bound_epilog as 0 for
17572 fully-masked loops and retain vf - 1 for other loops. Add 1 to
17573 this base when peeling for gaps.
17574 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
17575 with fully-masked loops.
17576 (vect_estimate_min_profitable_iters): Handle the single peeled
17577 iteration in that case.
17579 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17580 Alan Hayward <alan.hayward@arm.com>
17581 David Sherwood <david.sherwood@arm.com>
17583 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
17584 single-element interleaving even if the size is not a power of 2.
17585 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
17586 accesses for single-element interleaving if the group size is
17589 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17590 Alan Hayward <alan.hayward@arm.com>
17591 David Sherwood <david.sherwood@arm.com>
17593 * doc/md.texi (fold_extract_last_@var{m}): Document.
17594 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
17595 * optabs.def (fold_extract_last_optab): New optab.
17596 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
17597 * internal-fn.c (fold_extract_direct): New macro.
17598 (expand_fold_extract_optab_fn): Likewise.
17599 (direct_fold_extract_optab_supported_p): Likewise.
17600 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
17601 * tree-vect-loop.c (vect_model_reduction_cost): Handle
17602 EXTRACT_LAST_REDUCTION.
17603 (get_initial_def_for_reduction): Do not create an initial vector
17604 for EXTRACT_LAST_REDUCTION reductions.
17605 (vectorizable_reduction): Leave the scalar phi in place for
17606 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
17607 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
17608 epilogue code for EXTRACT_LAST_REDUCTION and defer the
17609 transform phase to vectorizable_condition.
17610 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
17612 (vect_finish_stmt_generation): ...here.
17613 (vect_finish_replace_stmt): New function.
17614 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
17615 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
17617 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
17619 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17620 Alan Hayward <alan.hayward@arm.com>
17621 David Sherwood <david.sherwood@arm.com>
17623 * doc/md.texi (extract_last_@var{m}): Document.
17624 * optabs.def (extract_last_optab): New optab.
17625 * internal-fn.def (EXTRACT_LAST): New internal function.
17626 * internal-fn.c (cond_unary_direct): New macro.
17627 (expand_cond_unary_optab_fn): Likewise.
17628 (direct_cond_unary_optab_supported_p): Likewise.
17629 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
17630 loops using EXTRACT_LAST.
17631 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
17632 (extract_last_<mode>): ...this optab.
17633 (vec_extract<mode><Vel>): Update accordingly.
17635 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17636 Alan Hayward <alan.hayward@arm.com>
17637 David Sherwood <david.sherwood@arm.com>
17639 * target.def (empty_mask_is_expensive): New hook.
17640 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
17641 * doc/tm.texi: Regenerate.
17642 * targhooks.h (default_empty_mask_is_expensive): Declare.
17643 * targhooks.c (default_empty_mask_is_expensive): New function.
17644 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
17645 if the target says that empty masks are expensive.
17646 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
17648 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
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 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
17655 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
17656 (vect_use_loop_mask_for_alignment_p): New function.
17657 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
17658 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
17659 niters_skip argument. Make sure that the first niters_skip elements
17660 of the first iteration are inactive.
17661 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
17662 Update call to vect_set_loop_masks_directly.
17663 (get_misalign_in_elems): New function, split out from...
17664 (vect_gen_prolog_loop_niters): ...here.
17665 (vect_update_init_of_dr): Take a code argument that specifies whether
17666 the adjustment should be added or subtracted.
17667 (vect_update_init_of_drs): Likewise.
17668 (vect_prepare_for_masked_peels): New function.
17669 (vect_do_peeling): Skip prologue peeling if we're using a mask
17670 instead. Update call to vect_update_inits_of_drs.
17671 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
17673 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
17674 alignment. Do not include the number of peeled iterations in
17675 the minimum threshold in that case.
17676 (vectorizable_induction): Adjust the start value down by
17677 LOOP_VINFO_MASK_SKIP_NITERS iterations.
17678 (vect_transform_loop): Call vect_prepare_for_masked_peels.
17679 Take the number of skipped iterations into account when calculating
17681 * tree-vect-stmts.c (vect_gen_while_not): New function.
17683 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17684 Alan Hayward <alan.hayward@arm.com>
17685 David Sherwood <david.sherwood@arm.com>
17687 * doc/sourcebuild.texi (vect_fully_masked): Document.
17688 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
17689 default value to 0.
17690 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
17692 (vect_analyze_loop_2): ...here. Don't check the vectorization
17693 factor against the number of loop iterations if the loop is
17696 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17697 Alan Hayward <alan.hayward@arm.com>
17698 David Sherwood <david.sherwood@arm.com>
17700 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
17701 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
17702 (dump_groups): Update accordingly.
17703 (iv_use::mem_type): New member variable.
17704 (address_p): New function.
17705 (record_use): Add a mem_type argument and initialize the new
17707 (record_group_use): Add a mem_type argument. Use address_p.
17708 Remove obsolete null checks of base_object. Update call to record_use.
17709 (find_interesting_uses_op): Update call to record_group_use.
17710 (find_interesting_uses_cond): Likewise.
17711 (find_interesting_uses_address): Likewise.
17712 (get_mem_type_for_internal_fn): New function.
17713 (find_address_like_use): Likewise.
17714 (find_interesting_uses_stmt): Try find_address_like_use before
17715 calling find_interesting_uses_op.
17716 (addr_offset_valid_p): Use the iv mem_type field as the type
17717 of the addressed memory.
17718 (add_autoinc_candidates): Likewise.
17719 (get_address_cost): Likewise.
17720 (split_small_address_groups_p): Use address_p.
17721 (split_address_groups): Likewise.
17722 (add_iv_candidate_for_use): Likewise.
17723 (autoinc_possible_for_pair): Likewise.
17724 (rewrite_groups): Likewise.
17725 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
17726 (determine_group_iv_cost): Update after split of USE_ADDRESS.
17727 (get_alias_ptr_type_for_ptr_address): New function.
17728 (rewrite_use_address): Rewrite address uses in calls that were
17729 identified by find_address_like_use.
17731 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17732 Alan Hayward <alan.hayward@arm.com>
17733 David Sherwood <david.sherwood@arm.com>
17735 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
17737 * gimple-expr.h (is_gimple_addressable: Likewise.
17738 * gimple-expr.c (is_gimple_address): Likewise.
17739 * internal-fn.c (expand_call_mem_ref): New function.
17740 (expand_mask_load_optab_fn): Use it.
17741 (expand_mask_store_optab_fn): Likewise.
17743 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17744 Alan Hayward <alan.hayward@arm.com>
17745 David Sherwood <david.sherwood@arm.com>
17747 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
17748 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
17749 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
17750 (cond_umax@var{mode}): Document.
17751 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
17752 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
17753 (cond_umin_optab, cond_umax_optab): New optabs.
17754 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
17755 (COND_IOR, COND_XOR): New internal functions.
17756 * internal-fn.h (get_conditional_internal_fn): Declare.
17757 * internal-fn.c (cond_binary_direct): New macro.
17758 (expand_cond_binary_optab_fn): Likewise.
17759 (direct_cond_binary_optab_supported_p): Likewise.
17760 (get_conditional_internal_fn): New function.
17761 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
17762 Cope with reduction statements that are vectorized as calls rather
17764 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
17765 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
17766 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
17767 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
17768 (UNSPEC_COND_EOR): New unspecs.
17769 (optab): Add mappings for them.
17770 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
17771 (sve_int_op, sve_fp_op): New int attributes.
17773 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17774 Alan Hayward <alan.hayward@arm.com>
17775 David Sherwood <david.sherwood@arm.com>
17777 * optabs.def (while_ult_optab): New optab.
17778 * doc/md.texi (while_ult@var{m}@var{n}): Document.
17779 * internal-fn.def (WHILE_ULT): New internal function.
17780 * internal-fn.h (direct_internal_fn_supported_p): New override
17781 that takes two types as argument.
17782 * internal-fn.c (while_direct): New macro.
17783 (expand_while_optab_fn): New function.
17784 (convert_optab_supported_p): Likewise.
17785 (direct_while_optab_supported_p): New macro.
17786 * wide-int.h (wi::udiv_ceil): New function.
17787 * tree-vectorizer.h (rgroup_masks): New structure.
17788 (vec_loop_masks): New typedef.
17789 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
17790 and fully_masked_p.
17791 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
17792 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
17793 (vect_max_vf): New function.
17794 (slpeel_make_loop_iterate_ntimes): Delete.
17795 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
17796 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
17797 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
17798 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
17799 internal-fn.h, stor-layout.h and optabs-query.h.
17800 (vect_set_loop_mask): New function.
17801 (add_preheader_seq): Likewise.
17802 (add_header_seq): Likewise.
17803 (interleave_supported_p): Likewise.
17804 (vect_maybe_permute_loop_masks): Likewise.
17805 (vect_set_loop_masks_directly): Likewise.
17806 (vect_set_loop_condition_masked): Likewise.
17807 (vect_set_loop_condition_unmasked): New function, split out from
17808 slpeel_make_loop_iterate_ntimes.
17809 (slpeel_make_loop_iterate_ntimes): Rename to..
17810 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
17811 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
17812 (vect_do_peeling): Update call accordingly.
17813 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
17815 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
17816 mask_compare_type, can_fully_mask_p and fully_masked_p.
17817 (release_vec_loop_masks): New function.
17818 (_loop_vec_info): Use it to free the loop masks.
17819 (can_produce_all_loop_masks_p): New function.
17820 (vect_get_max_nscalars_per_iter): Likewise.
17821 (vect_verify_full_masking): Likewise.
17822 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
17823 retries, and free the mask rgroups before retrying. Check loop-wide
17824 reasons for disallowing fully-masked loops. Make the final decision
17825 about whether use a fully-masked loop or not.
17826 (vect_estimate_min_profitable_iters): Do not assume that peeling
17827 for the number of iterations will be needed for fully-masked loops.
17828 (vectorizable_reduction): Disable fully-masked loops.
17829 (vectorizable_live_operation): Likewise.
17830 (vect_halve_mask_nunits): New function.
17831 (vect_double_mask_nunits): Likewise.
17832 (vect_record_loop_mask): Likewise.
17833 (vect_get_loop_mask): Likewise.
17834 (vect_transform_loop): Handle the case in which the final loop
17835 iteration might handle a partial vector. Call vect_set_loop_condition
17836 instead of slpeel_make_loop_iterate_ntimes.
17837 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
17838 (check_load_store_masking): New function.
17839 (prepare_load_store_mask): Likewise.
17840 (vectorizable_store): Handle fully-masked loops.
17841 (vectorizable_load): Likewise.
17842 (supportable_widening_operation): Use vect_halve_mask_nunits for
17844 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
17845 (vect_gen_while): New function.
17846 * config/aarch64/aarch64.md (umax<mode>3): New expander.
17847 (aarch64_uqdec<mode>): New insn.
17849 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17850 Alan Hayward <alan.hayward@arm.com>
17851 David Sherwood <david.sherwood@arm.com>
17853 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
17854 (reduc_xor_scal_optab): New optabs.
17855 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
17856 (reduc_xor_scal_@var{m}): Document.
17857 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
17858 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
17859 internal functions.
17860 * fold-const-call.c (fold_const_call): Handle them.
17861 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
17862 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
17863 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
17864 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
17865 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
17866 (UNSPEC_XORV): New unspecs.
17867 (optab): Add entries for them.
17868 (BITWISEV): New int iterator.
17869 (bit_reduc_op): New int attributes.
17871 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17872 Alan Hayward <alan.hayward@arm.com>
17873 David Sherwood <david.sherwood@arm.com>
17875 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
17876 * internal-fn.def (VEC_SHL_INSERT): New internal function.
17877 * optabs.def (vec_shl_insert_optab): New optab.
17878 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
17879 (duplicate_and_interleave): Likewise.
17880 * tree-vect-loop.c: Include internal-fn.h.
17881 (neutral_op_for_slp_reduction): New function, split out from
17882 get_initial_defs_for_reduction.
17883 (get_initial_def_for_reduction): Handle option 2 for variable-length
17884 vectors by loading the neutral value into a vector and then shifting
17885 the initial value into element 0.
17886 (get_initial_defs_for_reduction): Replace the code argument with
17887 the neutral value calculated by neutral_op_for_slp_reduction.
17888 Use gimple_build_vector for constant-length vectors.
17889 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
17890 but the first group_size elements have a neutral value.
17891 Use duplicate_and_interleave otherwise.
17892 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
17893 Update call to get_initial_defs_for_reduction. Handle SLP
17894 reductions for variable-length vectors by creating one vector
17895 result for each scalar result, with the elements associated
17896 with other scalar results stubbed out with the neutral value.
17897 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
17898 Require IFN_VEC_SHL_INSERT for double reductions on
17899 variable-length vectors, or SLP reductions that have
17900 a neutral value. Require can_duplicate_and_interleave_p
17901 support for variable-length unchained SLP reductions if there
17902 is no neutral value, such as for MIN/MAX reductions. Also require
17903 the number of vector elements to be a multiple of the number of
17904 SLP statements when doing variable-length unchained SLP reductions.
17905 Update call to vect_create_epilog_for_reduction.
17906 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
17907 and remove initial values.
17908 (duplicate_and_interleave): Make public.
17909 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
17910 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
17912 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17913 Alan Hayward <alan.hayward@arm.com>
17914 David Sherwood <david.sherwood@arm.com>
17916 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
17917 (can_duplicate_and_interleave_p): New function.
17918 (vect_get_and_check_slp_defs): Take the vector of statements
17919 rather than just the current one. Remove excess parentheses.
17920 Restriction rejectinon of vect_constant_def and vect_external_def
17921 for variable-length vectors to boolean types, or types for which
17922 can_duplicate_and_interleave_p is false.
17923 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
17924 (duplicate_and_interleave): New function.
17925 (vect_get_constant_vectors): Use gimple_build_vector for
17926 constant-length vectors and suitable variable-length constant
17927 vectors. Use duplicate_and_interleave for other variable-length
17928 vectors. Don't defer the update when inserting new statements.
17930 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17931 Alan Hayward <alan.hayward@arm.com>
17932 David Sherwood <david.sherwood@arm.com>
17934 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
17935 min_profitable_iters doesn't go negative.
17937 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17938 Alan Hayward <alan.hayward@arm.com>
17939 David Sherwood <david.sherwood@arm.com>
17941 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
17942 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
17943 * optabs.def (vec_mask_load_lanes_optab): New optab.
17944 (vec_mask_store_lanes_optab): Likewise.
17945 * internal-fn.def (MASK_LOAD_LANES): New internal function.
17946 (MASK_STORE_LANES): Likewise.
17947 * internal-fn.c (mask_load_lanes_direct): New macro.
17948 (mask_store_lanes_direct): Likewise.
17949 (expand_mask_load_optab_fn): Handle masked operations.
17950 (expand_mask_load_lanes_optab_fn): New macro.
17951 (expand_mask_store_optab_fn): Handle masked operations.
17952 (expand_mask_store_lanes_optab_fn): New macro.
17953 (direct_mask_load_lanes_optab_supported_p): Likewise.
17954 (direct_mask_store_lanes_optab_supported_p): Likewise.
17955 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
17957 (vect_load_lanes_supported): Likewise.
17958 * tree-vect-data-refs.c (strip_conversion): New function.
17959 (can_group_stmts_p): Likewise.
17960 (vect_analyze_data_ref_accesses): Use it instead of checking
17961 for a pair of assignments.
17962 (vect_store_lanes_supported): Take a masked_p parameter.
17963 (vect_load_lanes_supported): Likewise.
17964 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
17965 vect_store_lanes_supported and vect_load_lanes_supported.
17966 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
17967 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
17968 parameter. Don't allow gaps for masked accesses.
17969 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
17970 and vect_load_lanes_supported.
17971 (get_load_store_type): Take a masked_p parameter and update
17972 call to get_group_load_store_type.
17973 (vectorizable_store): Update call to get_load_store_type.
17974 Handle IFN_MASK_STORE_LANES.
17975 (vectorizable_load): Update call to get_load_store_type.
17976 Handle IFN_MASK_LOAD_LANES.
17978 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
17979 Alan Hayward <alan.hayward@arm.com>
17980 David Sherwood <david.sherwood@arm.com>
17982 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
17984 * config/aarch64/aarch64-protos.h
17985 (aarch64_sve_struct_memory_operand_p): Declare.
17986 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
17987 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
17988 (VPRED, vpred): Handle SVE structure modes.
17989 * config/aarch64/constraints.md (Utx): New constraint.
17990 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
17991 (aarch64_sve_struct_nonimmediate_operand): New predicates.
17992 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
17993 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
17994 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
17995 structure modes. Split into pieces after RA.
17996 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
17997 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
17999 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
18000 SVE structure modes.
18001 (aarch64_classify_address): Likewise.
18002 (sizetochar): Move earlier in file.
18003 (aarch64_print_operand): Handle SVE register lists.
18004 (aarch64_array_mode): New function.
18005 (aarch64_sve_struct_memory_operand_p): Likewise.
18006 (TARGET_ARRAY_MODE): Redefine.
18008 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18009 Alan Hayward <alan.hayward@arm.com>
18010 David Sherwood <david.sherwood@arm.com>
18012 * target.def (array_mode): New target hook.
18013 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
18014 * doc/tm.texi: Regenerate.
18015 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
18016 * hooks.c (hook_optmode_mode_uhwi_none): New function.
18017 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
18018 targetm.array_mode.
18019 * stor-layout.c (mode_for_array): Likewise. Support polynomial
18022 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18023 Alan Hayward <alan.hayward@arm.com>
18024 David Sherwood <david.sherwood@arm.com>
18026 * fold-const.c (fold_binary_loc): Check the argument types
18027 rather than the result type when testing for a vector operation.
18029 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18031 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
18032 * doc/tm.texi: Regenerate.
18034 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18035 Alan Hayward <alan.hayward@arm.com>
18036 David Sherwood <david.sherwood@arm.com>
18038 * doc/invoke.texi (-msve-vector-bits=): Document new option.
18039 (sve): Document new AArch64 extension.
18040 * doc/md.texi (w): Extend the description of the AArch64
18041 constraint to include SVE vectors.
18042 (Upl, Upa): Document new AArch64 predicate constraints.
18043 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
18045 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
18046 (msve-vector-bits=): New option.
18047 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
18048 SVE when these are disabled.
18049 (sve): New extension.
18050 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
18051 modes. Adjust their number of units based on aarch64_sve_vg.
18052 (MAX_BITSIZE_MODE_ANY_MODE): Define.
18053 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
18054 aarch64_addr_query_type.
18055 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
18056 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
18057 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
18058 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
18059 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
18060 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
18061 (aarch64_simd_imm_zero_p): Delete.
18062 (aarch64_check_zero_based_sve_index_immediate): Declare.
18063 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
18064 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
18065 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
18066 (aarch64_sve_float_mul_immediate_p): Likewise.
18067 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
18068 rather than an rtx.
18069 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
18070 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
18071 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
18072 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
18073 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
18074 (aarch64_regmode_natural_size): Likewise.
18075 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
18076 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
18078 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
18079 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
18080 for VG and the SVE predicate registers.
18081 (V_ALIASES): Add a "z"-prefixed alias.
18082 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
18083 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
18084 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
18085 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
18086 (REG_CLASS_NAMES): Add entries for them.
18087 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
18088 and the predicate registers.
18089 (aarch64_sve_vg): Declare.
18090 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
18091 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
18092 (REGMODE_NATURAL_SIZE): Define.
18093 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
18095 * config/aarch64/aarch64.c: Include cfgrtl.h.
18096 (simd_immediate_info): Add a constructor for series vectors,
18097 and an associated step field.
18098 (aarch64_sve_vg): New variable.
18099 (aarch64_dbx_register_number): Handle VG and the predicate registers.
18100 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
18101 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
18102 (VEC_ANY_DATA, VEC_STRUCT): New constants.
18103 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
18104 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
18105 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
18106 (aarch64_get_mask_mode): New functions.
18107 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
18108 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
18109 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
18110 predicate modes and predicate registers. Explicitly restrict
18111 GPRs to modes of 16 bytes or smaller. Only allow FP registers
18112 to store a vector mode if it is recognized by
18113 aarch64_classify_vector_mode.
18114 (aarch64_regmode_natural_size): New function.
18115 (aarch64_hard_regno_caller_save_mode): Return the original mode
18117 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
18118 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
18119 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
18120 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
18122 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
18123 does not overlap dest if the function is frame-related. Handle
18125 (aarch64_split_add_offset): New function.
18126 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
18127 them aarch64_add_offset.
18128 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
18129 and update call to aarch64_sub_sp.
18130 (aarch64_add_cfa_expression): New function.
18131 (aarch64_expand_prologue): Pass extra temporary registers to the
18132 functions above. Handle the case in which we need to emit new
18133 DW_CFA_expressions for registers that were originally saved
18134 relative to the stack pointer, but now have to be expressed
18135 relative to the frame pointer.
18136 (aarch64_output_mi_thunk): Pass extra temporary registers to the
18138 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
18139 IP0 and IP1 values for SVE frames.
18140 (aarch64_expand_vec_series): New function.
18141 (aarch64_expand_sve_widened_duplicate): Likewise.
18142 (aarch64_expand_sve_const_vector): Likewise.
18143 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
18144 Handle SVE constants. Use emit_move_insn to move a force_const_mem
18145 into the register, rather than emitting a SET directly.
18146 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
18147 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
18148 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
18149 (offset_9bit_signed_scaled_p): New functions.
18150 (aarch64_replicate_bitmask_imm): New function.
18151 (aarch64_bitmask_imm): Use it.
18152 (aarch64_cannot_force_const_mem): Reject expressions involving
18153 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
18154 (aarch64_classify_index): Handle SVE indices, by requiring
18155 a plain register index with a scale that matches the element size.
18156 (aarch64_classify_address): Handle SVE addresses. Assert that
18157 the mode of the address is VOIDmode or an integer mode.
18158 Update call to aarch64_classify_symbol.
18159 (aarch64_classify_symbolic_expression): Update call to
18160 aarch64_classify_symbol.
18161 (aarch64_const_vec_all_in_range_p): New function.
18162 (aarch64_print_vector_float_operand): Likewise.
18163 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
18164 "vN" for FP registers with SVE modes. Handle (const ...) vectors
18165 and the FP immediates 1.0 and 0.5.
18166 (aarch64_print_address_internal): Handle SVE addresses.
18167 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
18168 (aarch64_regno_regclass): Handle predicate registers.
18169 (aarch64_secondary_reload): Handle big-endian reloads of SVE
18171 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
18172 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
18173 (aarch64_convert_sve_vector_bits): New function.
18174 (aarch64_override_options): Use it to handle -msve-vector-bits=.
18175 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
18176 rather than an rtx.
18177 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
18178 Handle SVE vector and predicate modes. Accept VL-based constants
18179 that need only one temporary register, and VL offsets that require
18180 no temporary registers.
18181 (aarch64_conditional_register_usage): Mark the predicate registers
18182 as fixed if SVE isn't available.
18183 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
18184 Return true for SVE vector and predicate modes.
18185 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
18186 rather than an unsigned int. Handle SVE modes.
18187 (aarch64_preferred_simd_mode): Update call accordingly. Handle
18189 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
18191 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
18192 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
18193 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
18194 (aarch64_sve_float_mul_immediate_p): New functions.
18195 (aarch64_sve_valid_immediate): New function.
18196 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
18197 Explicitly reject structure modes. Check for INDEX constants.
18198 Handle PTRUE and PFALSE constants.
18199 (aarch64_check_zero_based_sve_index_immediate): New function.
18200 (aarch64_simd_imm_zero_p): Delete.
18201 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
18202 vector modes. Accept constants in the range of CNT[BHWD].
18203 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
18204 ask for an Advanced SIMD mode.
18205 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
18206 (aarch64_simd_vector_alignment): Handle SVE predicates.
18207 (aarch64_vectorize_preferred_vector_alignment): New function.
18208 (aarch64_simd_vector_alignment_reachable): Use it instead of
18210 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
18211 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
18213 (MAX_VECT_LEN): Delete.
18214 (expand_vec_perm_d): Add a vec_flags field.
18215 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
18216 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
18217 (aarch64_evpc_ext): Don't apply a big-endian lane correction
18219 (aarch64_evpc_rev): Rename to...
18220 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
18221 (aarch64_evpc_rev_global): New function.
18222 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
18223 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
18225 (aarch64_evpc_sve_tbl): New function.
18226 (aarch64_expand_vec_perm_const_1): Update after rename of
18227 aarch64_evpc_rev. Handle SVE permutes too, trying
18228 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
18229 than aarch64_evpc_tbl.
18230 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
18231 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
18232 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
18233 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
18234 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
18235 (aarch64_expand_sve_vcond): New functions.
18236 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
18237 of aarch64_vector_mode_p.
18238 (aarch64_dwarf_poly_indeterminate_value): New function.
18239 (aarch64_compute_pressure_classes): Likewise.
18240 (aarch64_can_change_mode_class): Likewise.
18241 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
18242 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
18243 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
18244 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
18245 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
18246 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
18247 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
18248 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
18250 (Dn, Dl, Dr): Accept const as well as const_vector.
18251 (Dz): Likewise. Compare against CONST0_RTX.
18252 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
18253 of "vector" where appropriate.
18254 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
18255 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
18256 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
18257 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
18258 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
18259 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
18260 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
18261 (v_int_equiv): Extend to SVE modes.
18262 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
18264 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
18265 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
18266 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
18267 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
18268 (SVE_COND_FP_CMP): New int iterators.
18269 (perm_hilo): Handle the new unpack unspecs.
18270 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
18272 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
18273 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
18274 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
18275 (aarch64_equality_operator, aarch64_constant_vector_operand)
18276 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
18277 (aarch64_sve_nonimmediate_operand): Likewise.
18278 (aarch64_sve_general_operand): Likewise.
18279 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
18280 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
18281 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
18282 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
18283 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
18284 (aarch64_sve_float_arith_immediate): Likewise.
18285 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
18286 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
18287 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
18288 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
18289 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
18290 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
18291 (aarch64_sve_float_arith_operand): Likewise.
18292 (aarch64_sve_float_arith_with_sub_operand): Likewise.
18293 (aarch64_sve_float_mul_operand): Likewise.
18294 (aarch64_sve_vec_perm_operand): Likewise.
18295 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
18296 (aarch64_mov_operand): Accept const_poly_int and const_vector.
18297 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
18298 as well as const_vector.
18299 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
18300 in file. Use CONST0_RTX and CONSTM1_RTX.
18301 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
18302 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
18303 Use aarch64_simd_imm_zero.
18304 * config/aarch64/aarch64-sve.md: New file.
18305 * config/aarch64/aarch64.md: Include it.
18306 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
18307 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
18308 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
18309 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
18310 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
18311 (sve): New attribute.
18312 (enabled): Disable instructions with the sve attribute unless
18314 (movqi, movhi): Pass CONST_POLY_INT operaneds through
18315 aarch64_expand_mov_immediate.
18316 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
18317 CNT[BHSD] immediates.
18318 (movti): Split CONST_POLY_INT moves into two halves.
18319 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
18320 Split additions that need a temporary here if the destination
18321 is the stack pointer.
18322 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
18323 (*add<mode>3_poly_1): New instruction.
18324 (set_clobber_cc): New expander.
18326 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18328 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
18329 parameter and use it instead of GET_MODE_SIZE (innermode). Use
18330 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
18331 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
18332 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
18333 Change innermode from fixed_mode_size to machine_mode.
18334 (simplify_subreg): Update call accordingly. Handle a constant-sized
18335 subreg of a variable-length CONST_VECTOR.
18337 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
18338 Alan Hayward <alan.hayward@arm.com>
18339 David Sherwood <david.sherwood@arm.com>
18341 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
18342 (add_offset_to_base): New function, split out from...
18343 (create_mem_ref): ...here. When handling a scale other than 1,
18344 check first whether the address is valid without the offset.
18345 Add it into the base if so, leaving the index and scale as-is.
18347 2018-01-12 Jakub Jelinek <jakub@redhat.com>
18350 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
18351 fold_for_warn before checking if arg2 is INTEGER_CST.
18353 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
18355 * config/rs6000/predicates.md (load_multiple_operation): Delete.
18356 (store_multiple_operation): Delete.
18357 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
18358 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
18359 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
18360 guarded by TARGET_STRING.
18361 (rs6000_output_load_multiple): Delete.
18362 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
18363 OPTION_MASK_STRING / TARGET_STRING handling.
18364 (print_operand) <'N', 'O'>: Add comment that these are unused now.
18365 (const rs6000_opt_masks) <"string">: Change mask to 0.
18366 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
18367 (MASK_STRING): Delete.
18368 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
18370 (load_multiple): Delete.
18377 (store_multiple): Delete.
18384 (movmemsi_8reg): Delete.
18385 (corresponding unnamed define_insn): Delete.
18386 (movmemsi_6reg): Delete.
18387 (corresponding unnamed define_insn): Delete.
18388 (movmemsi_4reg): Delete.
18389 (corresponding unnamed define_insn): Delete.
18390 (movmemsi_2reg): Delete.
18391 (corresponding unnamed define_insn): Delete.
18392 (movmemsi_1reg): Delete.
18393 (corresponding unnamed define_insn): Delete.
18394 * config/rs6000/rs6000.opt (mno-string): New.
18395 (mstring): Replace by deprecation warning stub.
18396 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
18398 2018-01-12 Jakub Jelinek <jakub@redhat.com>
18400 * regrename.c (regrename_do_replace): If replacing the same
18401 reg multiple times, try to reuse last created gen_raw_REG.
18404 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
18405 main to workaround a bug in GDB.
18407 2018-01-12 Tom de Vries <tom@codesourcery.com>
18410 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
18412 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
18414 PR rtl-optimization/80481
18415 * ira-color.c (get_cap_member): New function.
18416 (allocnos_conflict_by_live_ranges_p): Use it.
18417 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
18418 (setup_slot_coalesced_allocno_live_ranges): Ditto.
18420 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
18423 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
18424 (*saddl_se_1): Ditto.
18425 (*ssubsi_1): Ditto.
18426 (*ssubl_se_1): Ditto.
18428 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
18430 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
18431 rather than wi::to_widest for DR_INITs.
18432 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
18433 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
18434 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
18436 (vect_analyze_group_access_1): Note that here.
18438 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
18440 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
18441 polynomial type sizes.
18443 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
18445 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
18446 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
18447 (gimple_add_tmp_var): Likewise.
18449 2018-01-12 Martin Liska <mliska@suse.cz>
18451 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
18452 (gimple_alloc_sizes): Likewise.
18453 (dump_gimple_statistics): Use PRIu64 in printf format.
18454 * gimple.h: Change uint64_t to int.
18456 2018-01-12 Martin Liska <mliska@suse.cz>
18458 * tree-core.h: Use uint64_t instead of int.
18459 * tree.c (tree_node_counts): Likewise.
18460 (tree_node_sizes): Likewise.
18461 (dump_tree_statistics): Use PRIu64 in printf format.
18463 2018-01-12 Martin Liska <mliska@suse.cz>
18465 * Makefile.in: As qsort_chk is implemented in vec.c, add
18466 vec.o to linkage of gencfn-macros.
18467 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
18468 passing the info to record_node_allocation_statistics.
18469 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
18471 * ggc-common.c (struct ggc_usage): Add operator== and use
18472 it in operator< and compare function.
18473 * mem-stats.h (struct mem_usage): Likewise.
18474 * vec.c (struct vec_usage): Remove operator< and compare
18475 function. Can be simply inherited.
18477 2018-01-12 Martin Jambor <mjambor@suse.cz>
18480 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
18481 * tree-ssa-math-opts.c: Include domwalk.h.
18482 (convert_mult_to_fma_1): New function.
18483 (fma_transformation_info): New type.
18484 (fma_deferring_state): Likewise.
18485 (cancel_fma_deferring): New function.
18486 (result_of_phi): Likewise.
18487 (last_fma_candidate_feeds_initial_phi): Likewise.
18488 (convert_mult_to_fma): Added deferring logic, split actual
18489 transformation to convert_mult_to_fma_1.
18490 (math_opts_dom_walker): New type.
18491 (math_opts_dom_walker::after_dom_children): New method, body moved
18492 here from pass_optimize_widening_mul::execute, added deferring logic
18494 (pass_optimize_widening_mul::execute): Moved most of code to
18495 math_opts_dom_walker::after_dom_children.
18496 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
18497 * config/i386/i386.c (ix86_option_override_internal): Added
18498 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
18500 2018-01-12 Richard Biener <rguenther@suse.de>
18503 * dwarf2out.c (gen_variable_die): Do not reset old_die for
18504 inline instance vars.
18506 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
18509 * config/rx/rx.c (rx_is_restricted_memory_address):
18510 Handle SUBREG case.
18512 2018-01-12 Richard Biener <rguenther@suse.de>
18514 PR tree-optimization/80846
18515 * target.def (split_reduction): New target hook.
18516 * targhooks.c (default_split_reduction): New function.
18517 * targhooks.h (default_split_reduction): Declare.
18518 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
18519 target requests first reduce vectors by combining low and high
18521 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
18522 (get_vectype_for_scalar_type_and_size): Export.
18523 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
18524 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
18525 * doc/tm.texi: Regenerate.
18526 * config/i386/i386.c (ix86_split_reduction): Implement
18527 TARGET_VECTORIZE_SPLIT_REDUCTION.
18529 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
18532 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
18533 in PIC mode except for TARGET_VXWORKS_RTP.
18534 * config/sparc/sparc.c: Include cfgrtl.h.
18535 (TARGET_INIT_PIC_REG): Define.
18536 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
18537 (sparc_pic_register_p): New predicate.
18538 (sparc_legitimate_address_p): Use it.
18539 (sparc_legitimize_pic_address): Likewise.
18540 (sparc_delegitimize_address): Likewise.
18541 (sparc_mode_dependent_address_p): Likewise.
18542 (gen_load_pcrel_sym): Remove 4th parameter.
18543 (load_got_register): Adjust call to above. Remove obsolete stuff.
18544 (sparc_expand_prologue): Do not call load_got_register here.
18545 (sparc_flat_expand_prologue): Likewise.
18546 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
18547 (sparc_use_pseudo_pic_reg): New function.
18548 (sparc_init_pic_reg): Likewise.
18549 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
18550 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
18552 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
18554 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
18555 Add item for branch_cost.
18557 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
18559 PR rtl-optimization/83565
18560 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
18561 not extend the result to a larger mode for rotate operations.
18562 (num_sign_bit_copies1): Likewise.
18564 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
18567 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
18569 Use values-Xc.o for -pedantic.
18570 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
18572 2018-01-12 Martin Liska <mliska@suse.cz>
18575 * ipa-devirt.c (final_warning_record::grow_type_warnings):
18577 (possible_polymorphic_call_targets): Use it.
18578 (ipa_devirt): Likewise.
18580 2018-01-12 Martin Liska <mliska@suse.cz>
18582 * profile-count.h (enum profile_quality): Use 0 as invalid
18583 enum value of profile_quality.
18585 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
18587 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
18588 -mext-string options.
18590 2018-01-12 Richard Biener <rguenther@suse.de>
18592 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
18593 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
18594 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
18596 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
18598 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
18600 * configure.ac (--with-long-double-format): Add support for the
18601 configuration option to change the default long double format on
18603 * config.gcc (powerpc*-linux*-*): Likewise.
18604 * configure: Regenerate.
18605 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
18606 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
18607 used without modification.
18609 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18611 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
18612 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
18613 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
18614 MISC_BUILTIN_SPEC_BARRIER.
18615 (rs6000_init_builtins): Likewise.
18616 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
18618 (speculation_barrier): New define_insn.
18619 * doc/extend.texi: Document __builtin_speculation_barrier.
18621 2018-01-11 Jakub Jelinek <jakub@redhat.com>
18624 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
18625 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
18626 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
18628 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
18629 integral modes instead of "ss" and "sd".
18630 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
18631 vectors with 32-bit and 64-bit elements.
18632 (vecdupssescalarmodesuffix): New mode attribute.
18633 (vec_dup<mode>): Use it.
18635 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
18638 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
18639 frame if argument is passed on stack.
18641 2018-01-11 Jakub Jelinek <jakub@redhat.com>
18644 * ree.c (combine_reaching_defs): Optimize also
18645 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
18646 reg2=any_extend(exp); reg1=reg2;, formatting fix.
18648 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
18650 PR middle-end/83189
18651 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
18653 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
18655 PR middle-end/83718
18656 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
18657 after they are computed.
18659 2018-01-11 Bin Cheng <bin.cheng@arm.com>
18661 PR tree-optimization/83695
18662 * gimple-loop-linterchange.cc
18663 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
18664 reset cached scev information after interchange.
18665 (pass_linterchange::execute): Remove call to scev_reset_htab.
18667 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18669 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
18670 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
18671 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
18672 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
18673 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
18674 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
18675 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
18676 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
18677 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
18678 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
18679 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
18680 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
18681 (V_lane_reg): Likewise.
18682 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
18684 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
18685 (vfmal_lane_low<mode>_intrinsic,
18686 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
18687 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
18688 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
18689 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
18690 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
18691 vfmsl_lane_high<mode>_intrinsic): New define_insns.
18693 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18695 * config/arm/arm-cpus.in (fp16fml): New feature.
18696 (ALL_SIMD): Add fp16fml.
18697 (armv8.2-a): Add fp16fml as an option.
18698 (armv8.3-a): Likewise.
18699 (armv8.4-a): Add fp16fml as part of fp16.
18700 * config/arm/arm.h (TARGET_FP16FML): Define.
18701 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
18703 * config/arm/arm-modes.def (V2HF): Define.
18704 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
18705 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
18706 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
18707 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
18708 vfmsl_low, vfmsl_high): New set of builtins.
18709 * config/arm/iterators.md (PLUSMINUS): New code iterator.
18710 (vfml_op): New code attribute.
18711 (VFMLHALVES): New int iterator.
18712 (VFML, VFMLSEL): New mode attributes.
18713 (V_reg): Define mapping for V2HF.
18714 (V_hi, V_lo): New mode attributes.
18715 (VF_constraint): Likewise.
18716 (vfml_half, vfml_half_selector): New int attributes.
18717 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
18719 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
18720 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
18722 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
18723 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
18724 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
18725 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
18727 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
18728 Document new effective target and option set.
18730 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18732 * config/arm/arm-cpus.in (armv8_4): New feature.
18733 (ARMv8_4a): New fgroup.
18734 (armv8.4-a): New arch.
18735 * config/arm/arm-tables.opt: Regenerate.
18736 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
18737 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
18738 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
18739 Add matching rules for -march=armv8.4-a and extensions.
18740 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
18742 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
18745 * config/rx/rx.md (BW): New mode attribute.
18746 (sync_lock_test_and_setsi): Add mode suffix to insn output.
18748 2018-01-11 Richard Biener <rguenther@suse.de>
18750 PR tree-optimization/83435
18751 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
18752 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
18753 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
18755 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
18756 Alan Hayward <alan.hayward@arm.com>
18757 David Sherwood <david.sherwood@arm.com>
18759 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
18761 (aarch64_classify_address): Initialize it. Track polynomial offsets.
18762 (aarch64_print_address_internal): Use it to check for a zero offset.
18764 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
18765 Alan Hayward <alan.hayward@arm.com>
18766 David Sherwood <david.sherwood@arm.com>
18768 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
18769 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
18770 Return a poly_int64 rather than a HOST_WIDE_INT.
18771 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
18772 rather than a HOST_WIDE_INT.
18773 * config/aarch64/aarch64.h (aarch64_frame): Protect with
18774 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
18775 hard_fp_offset, frame_size, initial_adjust, callee_offset and
18776 final_offset from HOST_WIDE_INT to poly_int64.
18777 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
18778 to_constant when getting the number of units in an Advanced SIMD
18780 (aarch64_builtin_vectorized_function): Check for a constant number
18782 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
18784 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
18785 attribute instead of GET_MODE_NUNITS.
18786 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
18787 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
18788 GET_MODE_SIZE for fixed-size registers.
18789 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
18790 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
18791 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
18792 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
18793 (aarch64_print_operand, aarch64_print_address_internal)
18794 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
18795 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
18796 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
18797 Handle polynomial GET_MODE_SIZE.
18798 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
18799 wider than SImode without modification.
18800 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
18801 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
18802 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
18803 passing and returning SVE modes.
18804 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
18805 rather than GEN_INT.
18806 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
18807 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
18808 (aarch64_allocate_and_probe_stack_space): Likewise.
18809 (aarch64_layout_frame): Cope with polynomial offsets.
18810 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
18811 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
18812 polynomial offsets.
18813 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
18814 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
18815 poly_int64 rather than a HOST_WIDE_INT.
18816 (aarch64_get_separate_components, aarch64_process_components)
18817 (aarch64_expand_prologue, aarch64_expand_epilogue)
18818 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
18819 (aarch64_anchor_offset): New function, split out from...
18820 (aarch64_legitimize_address): ...here.
18821 (aarch64_builtin_vectorization_cost): Handle polynomial
18822 TYPE_VECTOR_SUBPARTS.
18823 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
18825 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
18826 number of elements from the PARALLEL rather than the mode.
18827 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
18828 rather than GET_MODE_BITSIZE.
18829 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
18830 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
18831 (aarch64_expand_vec_perm_const_1): Handle polynomial
18832 d->perm.length () and d->perm elements.
18833 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
18834 Apply to_constant to d->perm elements.
18835 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
18836 polynomial CONST_VECTOR_NUNITS.
18837 (aarch64_move_pointer): Take amount as a poly_int64 rather
18839 (aarch64_progress_pointer): Avoid temporary variable.
18840 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
18841 the mode attribute instead of GET_MODE.
18843 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
18844 Alan Hayward <alan.hayward@arm.com>
18845 David Sherwood <david.sherwood@arm.com>
18847 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
18848 x exists before using it.
18849 (aarch64_add_constant_internal): Rename to...
18850 (aarch64_add_offset_1): ...this. Replace regnum with separate
18851 src and dest rtxes. Handle the case in which they're different,
18852 including when the offset is zero. Replace scratchreg with an rtx.
18853 Use 2 additions if there is no spare register into which we can
18854 move a 16-bit constant.
18855 (aarch64_add_constant): Delete.
18856 (aarch64_add_offset): Replace reg with separate src and dest
18857 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
18858 Use aarch64_add_offset_1.
18859 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
18860 an rtx rather than an int. Take the delta as a poly_int64
18861 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
18862 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
18863 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
18864 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
18865 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
18866 and aarch64_add_sp.
18867 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
18868 aarch64_add_constant.
18870 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
18872 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
18873 Use scalar_float_mode.
18875 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
18877 * config/aarch64/aarch64-simd.md
18878 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
18879 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
18880 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
18881 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
18882 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
18883 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
18884 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
18885 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
18886 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
18887 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
18889 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
18892 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
18893 targ_options->x_arm_arch_string is non NULL.
18895 2018-01-11 Tamar Christina <tamar.christina@arm.com>
18897 * config/aarch64/aarch64.h
18898 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
18900 2018-01-11 Sudakshina Das <sudi.das@arm.com>
18903 * expmed.c (emit_store_flag_force): Swap if const op0
18904 and change VOIDmode to mode of op0.
18906 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
18908 PR rtl-optimization/83761
18909 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
18910 than bytes to mode_for_size.
18912 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
18914 PR middle-end/83189
18915 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
18916 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
18919 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
18921 PR middle-end/83575
18922 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
18923 when in layout mode.
18924 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
18925 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
18928 2018-01-10 Michael Collison <michael.collison@arm.com>
18930 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
18931 * config/aarch64/aarch64-option-extension.def: Add
18932 AARCH64_OPT_EXTENSION of 'fp16fml'.
18933 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
18934 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
18935 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
18936 * config/aarch64/constraints.md (Ui7): New constraint.
18937 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
18938 (VFMLA_SEL_W): Ditto.
18941 (VFMLA16_LOW): New int iterator.
18942 (VFMLA16_HIGH): Ditto.
18943 (UNSPEC_FMLAL): New unspec.
18944 (UNSPEC_FMLSL): Ditto.
18945 (UNSPEC_FMLAL2): Ditto.
18946 (UNSPEC_FMLSL2): Ditto.
18947 (f16mac): New code attribute.
18948 * config/aarch64/aarch64-simd-builtins.def
18949 (aarch64_fmlal_lowv2sf): Ditto.
18950 (aarch64_fmlsl_lowv2sf): Ditto.
18951 (aarch64_fmlalq_lowv4sf): Ditto.
18952 (aarch64_fmlslq_lowv4sf): Ditto.
18953 (aarch64_fmlal_highv2sf): Ditto.
18954 (aarch64_fmlsl_highv2sf): Ditto.
18955 (aarch64_fmlalq_highv4sf): Ditto.
18956 (aarch64_fmlslq_highv4sf): Ditto.
18957 (aarch64_fmlal_lane_lowv2sf): Ditto.
18958 (aarch64_fmlsl_lane_lowv2sf): Ditto.
18959 (aarch64_fmlal_laneq_lowv2sf): Ditto.
18960 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
18961 (aarch64_fmlalq_lane_lowv4sf): Ditto.
18962 (aarch64_fmlsl_lane_lowv4sf): Ditto.
18963 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
18964 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
18965 (aarch64_fmlal_lane_highv2sf): Ditto.
18966 (aarch64_fmlsl_lane_highv2sf): Ditto.
18967 (aarch64_fmlal_laneq_highv2sf): Ditto.
18968 (aarch64_fmlsl_laneq_highv2sf): Ditto.
18969 (aarch64_fmlalq_lane_highv4sf): Ditto.
18970 (aarch64_fmlsl_lane_highv4sf): Ditto.
18971 (aarch64_fmlalq_laneq_highv4sf): Ditto.
18972 (aarch64_fmlsl_laneq_highv4sf): Ditto.
18973 * config/aarch64/aarch64-simd.md:
18974 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
18975 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
18976 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
18977 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
18978 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
18979 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
18980 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
18981 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
18982 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
18983 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
18984 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
18985 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
18986 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
18987 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
18988 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
18989 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
18990 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
18991 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
18992 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
18993 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
18994 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
18995 (vfmlsl_low_u32): Ditto.
18996 (vfmlalq_low_u32): Ditto.
18997 (vfmlslq_low_u32): Ditto.
18998 (vfmlal_high_u32): Ditto.
18999 (vfmlsl_high_u32): Ditto.
19000 (vfmlalq_high_u32): Ditto.
19001 (vfmlslq_high_u32): Ditto.
19002 (vfmlal_lane_low_u32): Ditto.
19003 (vfmlsl_lane_low_u32): Ditto.
19004 (vfmlal_laneq_low_u32): Ditto.
19005 (vfmlsl_laneq_low_u32): Ditto.
19006 (vfmlalq_lane_low_u32): Ditto.
19007 (vfmlslq_lane_low_u32): Ditto.
19008 (vfmlalq_laneq_low_u32): Ditto.
19009 (vfmlslq_laneq_low_u32): Ditto.
19010 (vfmlal_lane_high_u32): Ditto.
19011 (vfmlsl_lane_high_u32): Ditto.
19012 (vfmlal_laneq_high_u32): Ditto.
19013 (vfmlsl_laneq_high_u32): Ditto.
19014 (vfmlalq_lane_high_u32): Ditto.
19015 (vfmlslq_lane_high_u32): Ditto.
19016 (vfmlalq_laneq_high_u32): Ditto.
19017 (vfmlslq_laneq_high_u32): Ditto.
19018 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
19019 (AARCH64_FL_FOR_ARCH8_4): New.
19020 (AARCH64_ISA_F16FML): New ISA flag.
19021 (TARGET_F16FML): New feature flag for fp16fml.
19022 (doc/invoke.texi): Document new fp16fml option.
19024 2018-01-10 Michael Collison <michael.collison@arm.com>
19026 * config/aarch64/aarch64-builtins.c:
19027 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
19028 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
19029 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
19030 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
19031 (AARCH64_ISA_SHA3): New ISA flag.
19032 (TARGET_SHA3): New feature flag for sha3.
19033 * config/aarch64/iterators.md (sha512_op): New int attribute.
19034 (CRYPTO_SHA512): New int iterator.
19035 (UNSPEC_SHA512H): New unspec.
19036 (UNSPEC_SHA512H2): Ditto.
19037 (UNSPEC_SHA512SU0): Ditto.
19038 (UNSPEC_SHA512SU1): Ditto.
19039 * config/aarch64/aarch64-simd-builtins.def
19040 (aarch64_crypto_sha512hqv2di): New builtin.
19041 (aarch64_crypto_sha512h2qv2di): Ditto.
19042 (aarch64_crypto_sha512su0qv2di): Ditto.
19043 (aarch64_crypto_sha512su1qv2di): Ditto.
19044 (aarch64_eor3qv8hi): Ditto.
19045 (aarch64_rax1qv2di): Ditto.
19046 (aarch64_xarqv2di): Ditto.
19047 (aarch64_bcaxqv8hi): Ditto.
19048 * config/aarch64/aarch64-simd.md:
19049 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
19050 (aarch64_crypto_sha512su0qv2di): Ditto.
19051 (aarch64_crypto_sha512su1qv2di): Ditto.
19052 (aarch64_eor3qv8hi): Ditto.
19053 (aarch64_rax1qv2di): Ditto.
19054 (aarch64_xarqv2di): Ditto.
19055 (aarch64_bcaxqv8hi): Ditto.
19056 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
19057 (vsha512h2q_u64): Ditto.
19058 (vsha512su0q_u64): Ditto.
19059 (vsha512su1q_u64): Ditto.
19060 (veor3q_u16): Ditto.
19061 (vrax1q_u64): Ditto.
19062 (vxarq_u64): Ditto.
19063 (vbcaxq_u16): Ditto.
19064 * config/arm/types.md (crypto_sha512): New type attribute.
19065 (crypto_sha3): Ditto.
19066 (doc/invoke.texi): Document new sha3 option.
19068 2018-01-10 Michael Collison <michael.collison@arm.com>
19070 * config/aarch64/aarch64-builtins.c:
19071 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
19072 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
19073 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
19074 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
19075 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
19076 (AARCH64_ISA_SM4): New ISA flag.
19077 (TARGET_SM4): New feature flag for sm4.
19078 * config/aarch64/aarch64-simd-builtins.def
19079 (aarch64_sm3ss1qv4si): Ditto.
19080 (aarch64_sm3tt1aq4si): Ditto.
19081 (aarch64_sm3tt1bq4si): Ditto.
19082 (aarch64_sm3tt2aq4si): Ditto.
19083 (aarch64_sm3tt2bq4si): Ditto.
19084 (aarch64_sm3partw1qv4si): Ditto.
19085 (aarch64_sm3partw2qv4si): Ditto.
19086 (aarch64_sm4eqv4si): Ditto.
19087 (aarch64_sm4ekeyqv4si): Ditto.
19088 * config/aarch64/aarch64-simd.md:
19089 (aarch64_sm3ss1qv4si): Ditto.
19090 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
19091 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
19092 (aarch64_sm4eqv4si): Ditto.
19093 (aarch64_sm4ekeyqv4si): Ditto.
19094 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
19095 (sm3part_op): Ditto.
19096 (CRYPTO_SM3TT): Ditto.
19097 (CRYPTO_SM3PART): Ditto.
19098 (UNSPEC_SM3SS1): New unspec.
19099 (UNSPEC_SM3TT1A): Ditto.
19100 (UNSPEC_SM3TT1B): Ditto.
19101 (UNSPEC_SM3TT2A): Ditto.
19102 (UNSPEC_SM3TT2B): Ditto.
19103 (UNSPEC_SM3PARTW1): Ditto.
19104 (UNSPEC_SM3PARTW2): Ditto.
19105 (UNSPEC_SM4E): Ditto.
19106 (UNSPEC_SM4EKEY): Ditto.
19107 * config/aarch64/constraints.md (Ui2): New constraint.
19108 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
19109 * config/arm/types.md (crypto_sm3): New type attribute.
19110 (crypto_sm4): Ditto.
19111 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
19112 (vsm3tt1aq_u32): Ditto.
19113 (vsm3tt1bq_u32): Ditto.
19114 (vsm3tt2aq_u32): Ditto.
19115 (vsm3tt2bq_u32): Ditto.
19116 (vsm3partw1q_u32): Ditto.
19117 (vsm3partw2q_u32): Ditto.
19118 (vsm4eq_u32): Ditto.
19119 (vsm4ekeyq_u32): Ditto.
19120 (doc/invoke.texi): Document new sm4 option.
19122 2018-01-10 Michael Collison <michael.collison@arm.com>
19124 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
19125 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
19126 (AARCH64_FL_FOR_ARCH8_4): New.
19127 (AARCH64_FL_V8_4): New flag.
19128 (doc/invoke.texi): Document new armv8.4-a option.
19130 2018-01-10 Michael Collison <michael.collison@arm.com>
19132 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
19133 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
19134 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
19135 * config/aarch64/aarch64-option-extension.def: Add
19136 AARCH64_OPT_EXTENSION of 'sha2'.
19137 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
19138 (crypto): Disable sha2 and aes if crypto disabled.
19139 (crypto): Enable aes and sha2 if enabled.
19140 (simd): Disable sha2 and aes if simd disabled.
19141 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
19143 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
19144 (TARGET_SHA2): New feature flag for sha2.
19145 (TARGET_AES): New feature flag for aes.
19146 * config/aarch64/aarch64-simd.md:
19147 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
19148 conditional on TARGET_AES.
19149 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
19150 (aarch64_crypto_sha1hsi): Make pattern conditional
19152 (aarch64_crypto_sha1hv4si): Ditto.
19153 (aarch64_be_crypto_sha1hv4si): Ditto.
19154 (aarch64_crypto_sha1su1v4si): Ditto.
19155 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
19156 (aarch64_crypto_sha1su0v4si): Ditto.
19157 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
19158 (aarch64_crypto_sha256su0v4si): Ditto.
19159 (aarch64_crypto_sha256su1v4si): Ditto.
19160 (doc/invoke.texi): Document new aes and sha2 options.
19162 2018-01-10 Martin Sebor <msebor@redhat.com>
19164 PR tree-optimization/83781
19165 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
19168 2018-01-11 Martin Sebor <msebor@gmail.com>
19169 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
19171 PR tree-optimization/83501
19172 PR tree-optimization/81703
19174 * tree-ssa-strlen.c (get_string_cst): Rename...
19175 (get_string_len): ...to this. Handle global constants.
19176 (handle_char_store): Adjust.
19178 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
19179 Jim Wilson <jimw@sifive.com>
19181 * config/riscv/riscv-protos.h (riscv_output_return): New.
19182 * config/riscv/riscv.c (struct machine_function): New naked_p field.
19183 (riscv_attribute_table, riscv_output_return),
19184 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
19185 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
19186 (riscv_compute_frame_info): Only compute frame->mask if not a naked
19188 (riscv_expand_prologue): Add early return for naked function.
19189 (riscv_expand_epilogue): Likewise.
19190 (riscv_function_ok_for_sibcall): Return false for naked function.
19191 (riscv_set_current_function): New.
19192 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
19193 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
19194 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
19195 * doc/extend.texi (RISC-V Function Attributes): New.
19197 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
19199 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
19200 check for 128-bit long double before checking TCmode.
19201 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
19202 128-bit long doubles before checking TFmode or TCmode.
19203 (FLOAT128_IBM_P): Likewise.
19205 2018-01-10 Martin Sebor <msebor@redhat.com>
19207 PR tree-optimization/83671
19208 * builtins.c (c_strlen): Unconditionally return zero for the empty
19210 Use -Warray-bounds for warnings.
19211 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
19212 for non-constant array indices with COMPONENT_REF, arrays of
19213 arrays, and pointers to arrays.
19214 (gimple_fold_builtin_strlen): Determine and set length range for
19215 non-constant character arrays.
19217 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
19219 PR middle-end/81897
19220 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
19223 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
19225 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
19227 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
19230 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
19231 VECTOR_MEM_ALTIVEC_OR_VSX_P.
19232 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
19233 indexed_or_indirect_operand predicate.
19234 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
19235 (*vsx_le_perm_load_v8hi): Likewise.
19236 (*vsx_le_perm_load_v16qi): Likewise.
19237 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
19238 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
19239 (*vsx_le_perm_store_v8hi): Likewise.
19240 (*vsx_le_perm_store_v16qi): Likewise.
19241 (eight unnamed splitters): Likewise.
19243 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
19245 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
19246 * config/rs6000/emmintrin.h: Likewise.
19247 * config/rs6000/mmintrin.h: Likewise.
19248 * config/rs6000/xmmintrin.h: Likewise.
19250 2018-01-10 David Malcolm <dmalcolm@redhat.com>
19253 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
19255 * tree.c (tree_nop_conversion): Return true for location wrapper
19257 (maybe_wrap_with_location): New function.
19258 (selftest::check_strip_nops): New function.
19259 (selftest::test_location_wrappers): New function.
19260 (selftest::tree_c_tests): Call it.
19261 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
19262 (maybe_wrap_with_location): New decl.
19263 (EXPR_LOCATION_WRAPPER_P): New macro.
19264 (location_wrapper_p): New inline function.
19265 (tree_strip_any_location_wrapper): New inline function.
19267 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
19270 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
19271 stack_realign_offset for the largest alignment of stack slot
19273 (ix86_find_max_used_stack_alignment): New function.
19274 (ix86_finalize_stack_frame_flags): Use it. Set
19275 max_used_stack_alignment if we don't realign stack.
19276 * config/i386/i386.h (machine_function): Add
19277 max_used_stack_alignment.
19279 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
19281 * config/arm/arm.opt (-mbranch-cost): New option.
19282 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
19285 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
19288 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
19289 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
19291 2018-01-10 Richard Biener <rguenther@suse.de>
19294 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
19295 early out so it also covers the case where we have a non-NULL
19298 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
19300 PR tree-optimization/83753
19301 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
19302 for non-strided grouped accesses if the number of elements is 1.
19304 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
19307 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
19308 * i386.h (TARGET_USE_GATHER): Define.
19309 * x86-tune.def (X86_TUNE_USE_GATHER): New.
19311 2018-01-10 Martin Liska <mliska@suse.cz>
19314 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
19315 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
19317 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
19318 CLEANUP_NO_PARTITIONING is not set.
19320 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
19322 * doc/rtl.texi: Remove documentation of (const ...) wrappers
19323 for vectors, as a partial revert of r254296.
19324 * rtl.h (const_vec_p): Delete.
19325 (const_vec_duplicate_p): Don't test for vector CONSTs.
19326 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
19327 * expmed.c (make_tree): Likewise.
19330 * common.md (E, F): Use CONSTANT_P instead of checking for
19332 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
19333 checking for CONST_VECTOR.
19335 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
19337 PR middle-end/83575
19338 * predict.c (force_edge_cold): Handle in more sane way edges
19339 with no prediction.
19341 2018-01-09 Carl Love <cel@us.ibm.com>
19343 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
19345 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
19346 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
19347 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
19348 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
19349 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
19350 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
19351 * config/rs6000/rs6000-protos.h: Add extern defition for
19352 rs6000_generate_float2_double_code.
19353 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
19355 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
19356 (float2_v2df): Add define_expand.
19358 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
19361 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
19362 op_mode in the force_to_mode call.
19364 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
19366 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
19367 instead of checking each element individually.
19368 (aarch64_evpc_uzp): Likewise.
19369 (aarch64_evpc_zip): Likewise.
19370 (aarch64_evpc_ext): Likewise.
19371 (aarch64_evpc_rev): Likewise.
19372 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
19373 instead of checking each element individually. Return true without
19375 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
19376 whether all selected elements come from the same input, instead of
19377 checking each element individually. Remove calls to gen_rtx_REG,
19378 start_sequence and end_sequence and instead assert that no rtl is
19381 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
19383 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
19384 order of HIGH and CONST checks.
19386 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
19388 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
19389 if the destination isn't an SSA_NAME.
19391 2018-01-09 Richard Biener <rguenther@suse.de>
19393 PR tree-optimization/83668
19394 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
19396 (canonicalize_loop_form): ... here, renamed from ...
19397 (canonicalize_loop_closed_ssa_form): ... this and amended to
19398 swap successor edges for loop exit blocks to make us use
19399 the RPO order we need for initial schedule generation.
19401 2018-01-09 Joseph Myers <joseph@codesourcery.com>
19403 PR tree-optimization/64811
19404 * match.pd: When optimizing comparisons with Inf, avoid
19405 introducing or losing exceptions from comparisons with NaN.
19407 2018-01-09 Martin Liska <mliska@suse.cz>
19410 * asan.c (shadow_mem_size): Add gcc_assert.
19412 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
19414 Don't save registers in main().
19417 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
19418 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
19419 * config/avr/avr.c (avr_set_current_function): Don't error if
19420 naked, OS_task or OS_main are specified at the same time.
19421 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
19423 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
19425 * common/config/avr/avr-common.c (avr_option_optimization_table):
19426 Switch on -mmain-is-OS_task for optimizing compilations.
19428 2018-01-09 Richard Biener <rguenther@suse.de>
19430 PR tree-optimization/83572
19431 * graphite.c: Include cfganal.h.
19432 (graphite_transform_loops): Connect infinite loops to exit
19433 and remove fake edges at the end.
19435 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
19437 * ipa-inline.c (edge_badness): Revert accidental checkin.
19439 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
19442 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
19443 symbols; not inline clones.
19445 2018-01-09 Jakub Jelinek <jakub@redhat.com>
19448 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
19449 hard registers. Formatting fixes.
19451 PR preprocessor/83722
19452 * gcc.c (try_generate_repro): Pass
19453 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
19454 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
19457 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
19458 Kito Cheng <kito.cheng@gmail.com>
19460 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
19461 (riscv_leaf_function_p): Delete.
19462 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
19464 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19466 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
19468 (do_ifelse): New function.
19469 (do_isel): New function.
19470 (do_sub3): New function.
19471 (do_add3): New function.
19472 (do_load_mask_compare): New function.
19473 (do_overlap_load_compare): New function.
19474 (expand_compare_loop): New function.
19475 (expand_block_compare): Call expand_compare_loop() when appropriate.
19476 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
19477 option description.
19478 (-mblock-compare-inline-loop-limit): New option.
19480 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19483 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
19484 Reverse order of second and third operands in first alternative.
19485 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
19486 of first and second elements in UNSPEC_VPERMR vector.
19487 (altivec_expand_vec_perm_le): Likewise.
19489 2018-01-08 Jeff Law <law@redhat.com>
19491 PR rtl-optimizatin/81308
19492 * tree-switch-conversion.c (cfg_altered): New file scoped static.
19493 (process_switch): If group_case_labels makes a change, then set
19495 (pass_convert_switch::execute): If a switch is converted, then
19496 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
19498 PR rtl-optimization/81308
19499 * recog.c (split_all_insns): Conditionally cleanup the CFG after
19502 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
19504 PR target/83663 - Revert r255946
19505 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
19506 generation for cases where splatting a value is not useful.
19507 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
19508 across a vec_duplicate and a paradoxical subreg forming a vector
19509 mode to a vec_concat.
19511 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19513 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
19514 -march=armv8.3-a variants.
19515 * config/arm/t-multilib: Likewise.
19516 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
19518 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19520 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
19522 (cceq_ior_compare_complement): Give it a name so I can use it, and
19523 change boolean_or_operator predicate to boolean_operator so it can
19524 be used to generate a crand.
19525 (eqne): New code iterator.
19526 (bd/bd_neg): New code_attrs.
19527 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
19528 a single define_insn.
19529 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
19530 decrement (bdnzt/bdnzf/bdzt/bdzf).
19531 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
19532 with the new names of the branch decrement patterns, and added the
19533 names of the branch decrement conditional patterns.
19535 2018-01-08 Richard Biener <rguenther@suse.de>
19537 PR tree-optimization/83563
19538 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
19541 2018-01-08 Richard Biener <rguenther@suse.de>
19543 PR middle-end/83713
19544 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
19546 2018-01-08 Richard Biener <rguenther@suse.de>
19548 PR tree-optimization/83685
19549 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
19550 references to abnormals.
19552 2018-01-08 Richard Biener <rguenther@suse.de>
19555 * dwarf2out.c (output_indirect_strings): Handle empty
19556 skeleton_debug_str_hash.
19557 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
19559 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
19561 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
19562 (emit_store_direct): Likewise.
19563 (arc_trampoline_adjust_address): Likewise.
19564 (arc_asm_trampoline_template): New function.
19565 (arc_initialize_trampoline): Use asm_trampoline_template.
19566 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
19567 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
19568 * config/arc/arc.md (flush_icache): Delete pattern.
19570 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
19572 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
19573 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
19576 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
19579 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
19580 by not USED_FOR_TARGET.
19581 (make_pass_resolve_sw_modes): Likewise.
19583 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
19585 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
19588 2018-01-08 Richard Biener <rguenther@suse.de>
19590 PR middle-end/83580
19591 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
19593 2018-01-08 Richard Biener <rguenther@suse.de>
19595 PR middle-end/83517
19596 * match.pd ((t * 2) / 2) -> t): Add missing :c.
19598 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
19600 PR middle-end/81897
19601 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
19602 basic blocks with a small number of successors.
19603 (convert_control_dep_chain_into_preds): Improve handling of
19605 (dump_predicates): Split apart into...
19606 (dump_pred_chain): ...here...
19607 (dump_pred_info): ...and here.
19608 (can_one_predicate_be_invalidated_p): Add debugging printfs.
19609 (can_chain_union_be_invalidated_p): Improve check for invalidation
19611 (uninit_uses_cannot_happen): Avoid unnecessary if
19612 convert_control_dep_chain_into_preds yielded nothing.
19614 2018-01-06 Martin Sebor <msebor@redhat.com>
19616 PR tree-optimization/83640
19617 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
19618 subtracting negative offset from size.
19619 (builtin_access::overlap): Adjust offset bounds of the access to fall
19620 within the size of the object if possible.
19622 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
19624 PR rtl-optimization/83699
19625 * expmed.c (extract_bit_field_1): Restrict the vector usage of
19626 extract_bit_field_as_subreg to cases in which the extracted
19627 value is also a vector.
19629 * lra-constraints.c (process_alt_operands): Test for the equivalence
19630 substitutions when detecting a possible reload cycle.
19632 2018-01-06 Jakub Jelinek <jakub@redhat.com>
19635 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
19636 by default if flag_selective_schedling{,2}. Formatting fixes.
19638 PR rtl-optimization/83682
19639 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
19640 if it has non-VECTOR_MODE element mode.
19641 (vec_duplicate_p): Likewise.
19643 PR middle-end/83694
19644 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
19645 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
19647 2018-01-05 Jakub Jelinek <jakub@redhat.com>
19650 * config/i386/i386-builtin.def
19651 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
19652 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
19653 Require also OPTION_MASK_ISA_AVX512F in addition to
19654 OPTION_MASK_ISA_GFNI.
19655 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
19656 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
19657 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
19658 to OPTION_MASK_ISA_GFNI.
19659 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
19660 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
19661 OPTION_MASK_ISA_AVX512BW.
19662 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
19663 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
19664 addition to OPTION_MASK_ISA_GFNI.
19665 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
19666 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
19667 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
19668 to OPTION_MASK_ISA_GFNI.
19669 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
19670 a requirement for all ISAs rather than any of them with a few
19672 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
19674 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
19675 bitmasks to be enabled with 3 exceptions, instead of requiring any
19676 enabled ISA with lots of exceptions.
19677 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
19678 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
19679 Change avx512bw in isa attribute to avx512f.
19680 * config/i386/sgxintrin.h: Add license boilerplate.
19681 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
19682 to __AVX512F__ and __AVX512VL to __AVX512VL__.
19683 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
19684 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
19686 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
19687 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
19688 temporarily sse2 rather than sse if not enabled already.
19691 * config/i386/sse.md (VI248_VLBW): Rename to ...
19692 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
19693 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
19694 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
19695 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
19696 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
19697 mode iterator instead of VI248_VLBW.
19699 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
19701 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
19702 (record_modified): Skip clobbers; add debug output.
19703 (param_change_prob): Use sreal frequencies.
19705 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
19707 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
19708 punt for user-aligned variables.
19710 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
19712 * tree-chrec.c (chrec_contains_symbols): Return true for
19715 2018-01-05 Sudakshina Das <sudi.das@arm.com>
19718 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
19719 of (x|y) == x for BICS pattern.
19721 2018-01-05 Jakub Jelinek <jakub@redhat.com>
19723 PR tree-optimization/83605
19724 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
19725 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
19728 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
19730 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
19731 * config/epiphany/rtems.h: New file.
19733 2018-01-04 Jakub Jelinek <jakub@redhat.com>
19734 Uros Bizjak <ubizjak@gmail.com>
19737 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
19738 QIreg_operand instead of register_operand predicate.
19739 * config/i386/i386.c (ix86_rop_should_change_byte_p,
19740 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
19741 comments instead of -fmitigate[-_]rop.
19743 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
19746 * cgraphunit.c (symbol_table::compile): Switch to text_section
19747 before calling assembly_start debug hook.
19748 * run-rtl-passes.c (run_rtl_passes): Likewise.
19751 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
19753 * tree-vrp.c (extract_range_from_binary_expr_1): Check
19754 range_int_cst_p rather than !symbolic_range_p before calling
19755 extract_range_from_multiplicative_op_1.
19757 2018-01-04 Jeff Law <law@redhat.com>
19759 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
19760 redundant test in assertion.
19762 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
19764 * doc/rtl.texi: Document machine_mode wrapper classes.
19766 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
19768 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
19769 using tree_to_uhwi.
19771 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
19773 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
19774 the VEC_PERM_EXPR fold to fail.
19776 2018-01-04 Jakub Jelinek <jakub@redhat.com>
19779 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
19780 to switched_sections.
19782 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
19785 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
19786 test for d.testing.
19788 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
19791 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
19792 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
19794 2018-01-04 Jakub Jelinek <jakub@redhat.com>
19797 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
19798 is BLKmode and bitpos not zero or mode change is needed.
19800 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
19803 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
19806 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
19809 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
19810 instead of MULT rtx. Update all corresponding splitters.
19811 (*saddl_se): Ditto.
19812 (*ssub<modesuffix>): Ditto.
19813 (*ssubl_se): Ditto.
19814 (*cmp_sadd_di): Update split patterns.
19815 (*cmp_sadd_si): Ditto.
19816 (*cmp_sadd_sidi): Ditto.
19817 (*cmp_ssub_di): Ditto.
19818 (*cmp_ssub_si): Ditto.
19819 (*cmp_ssub_sidi): Ditto.
19820 * config/alpha/predicates.md (const23_operand): New predicate.
19821 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
19822 Look for ASHIFT, not MULT inner operand.
19823 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
19825 2018-01-04 Martin Liska <mliska@suse.cz>
19827 PR gcov-profile/83669
19828 * gcov.c (output_intermediate_file): Add version to intermediate
19830 * doc/gcov.texi: Document new field 'version' in intermediate
19831 file format. Fix location of '-k' option of gcov command.
19833 2018-01-04 Martin Liska <mliska@suse.cz>
19836 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
19838 2018-01-04 Jakub Jelinek <jakub@redhat.com>
19840 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
19842 2018-01-03 Martin Sebor <msebor@redhat.com>
19844 PR tree-optimization/83655
19845 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
19846 checking calls with invalid arguments.
19848 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19850 * tree-vect-stmts.c (vect_get_store_rhs): New function.
19851 (vectorizable_mask_load_store): Delete.
19852 (vectorizable_call): Return false for masked loads and stores.
19853 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
19854 instead of gimple_assign_rhs1.
19855 (vectorizable_load): Handle IFN_MASK_LOAD.
19856 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
19858 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19860 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
19862 (vectorizable_mask_load_store): ...here.
19863 (vectorizable_load): ...and here.
19865 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19867 * tree-vect-stmts.c (vect_build_all_ones_mask)
19868 (vect_build_zero_merge_argument): New functions, split out from...
19869 (vectorizable_load): ...here.
19871 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19873 * tree-vect-stmts.c (vect_check_store_rhs): New function,
19875 (vectorizable_mask_load_store): ...here.
19876 (vectorizable_store): ...and here.
19878 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19880 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
19882 (vectorizable_mask_load_store): ...here.
19884 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19886 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
19887 (vect_model_store_cost): Take a vec_load_store_type instead of a
19889 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
19890 (vect_model_store_cost): Take a vec_load_store_type instead of a
19892 (vectorizable_mask_load_store): Update accordingly.
19893 (vectorizable_store): Likewise.
19894 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
19896 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19898 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
19899 IFN_MASK_LOAD calls here rather than...
19900 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
19902 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19903 Alan Hayward <alan.hayward@arm.com>
19904 David Sherwood <david.sherwood@arm.com>
19906 * expmed.c (extract_bit_field_1): For vector extracts,
19907 fall back to extract_bit_field_as_subreg if vec_extract
19910 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19911 Alan Hayward <alan.hayward@arm.com>
19912 David Sherwood <david.sherwood@arm.com>
19914 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
19915 they are variable or constant sized.
19916 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
19917 slots for constant-sized data.
19919 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19920 Alan Hayward <alan.hayward@arm.com>
19921 David Sherwood <david.sherwood@arm.com>
19923 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
19924 handling COND_EXPRs with boolean comparisons, try to find a better
19925 basis for the mask type than the boolean itself.
19927 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19929 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
19930 is calculated and how it can be overridden.
19931 * genmodes.c (max_bitsize_mode_any_mode): New variable.
19932 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
19934 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
19937 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19938 Alan Hayward <alan.hayward@arm.com>
19939 David Sherwood <david.sherwood@arm.com>
19941 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
19942 Remove the mode argument.
19943 (aarch64_simd_valid_immediate): Remove the mode and inverse
19945 * config/aarch64/iterators.md (bitsize): New iterator.
19946 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
19947 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
19948 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
19949 aarch64_simd_valid_immediate.
19950 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
19951 (aarch64_reg_or_bic_imm): Likewise.
19952 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
19953 with an insn_type enum and msl with a modifier_type enum.
19954 Replace element_width with a scalar_mode. Change the shift
19955 to unsigned int. Add constructors for scalar_float_mode and
19956 scalar_int_mode elements.
19957 (aarch64_vect_float_const_representable_p): Delete.
19958 (aarch64_can_const_movi_rtx_p)
19959 (aarch64_simd_scalar_immediate_valid_for_move)
19960 (aarch64_simd_make_constant): Update call to
19961 aarch64_simd_valid_immediate.
19962 (aarch64_advsimd_valid_immediate_hs): New function.
19963 (aarch64_advsimd_valid_immediate): Likewise.
19964 (aarch64_simd_valid_immediate): Remove mode and inverse
19965 arguments. Rewrite to use the above. Use const_vec_duplicate_p
19966 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
19967 and aarch64_float_const_representable_p on the result.
19968 (aarch64_output_simd_mov_immediate): Remove mode argument.
19969 Update call to aarch64_simd_valid_immediate and use of
19970 simd_immediate_info.
19971 (aarch64_output_scalar_simd_mov_immediate): Update call
19974 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19975 Alan Hayward <alan.hayward@arm.com>
19976 David Sherwood <david.sherwood@arm.com>
19978 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
19979 (mode_nunits): Likewise CONST_MODE_NUNITS.
19980 * machmode.def (ADJUST_NUNITS): Document.
19981 * genmodes.c (mode_data::need_nunits_adj): New field.
19982 (blank_mode): Update accordingly.
19983 (adj_nunits): New variable.
19984 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
19986 (emit_mode_size_inline): Set need_bytesize_adj for all modes
19987 listed in adj_nunits.
19988 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
19989 listed in adj_nunits. Don't emit case statements for such modes.
19990 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
19991 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
19992 nothing if adj_nunits is nonnull.
19993 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
19994 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
19995 (emit_mode_fbit): Update use of print_maybe_const_decl.
19996 (emit_move_size): Likewise. Treat the array as non-const
19998 (emit_mode_adjustments): Handle adj_nunits.
20000 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20002 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
20003 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
20004 (VECTOR_MODES): Use it.
20005 (make_vector_modes): Take the prefix as an argument.
20007 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20008 Alan Hayward <alan.hayward@arm.com>
20009 David Sherwood <david.sherwood@arm.com>
20011 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
20012 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
20013 for MODE_VECTOR_BOOL.
20014 * machmode.def (VECTOR_BOOL_MODE): Document.
20015 * genmodes.c (VECTOR_BOOL_MODE): New macro.
20016 (make_vector_bool_mode): New function.
20017 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
20019 * lto-streamer-in.c (lto_input_mode_table): Likewise.
20020 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
20022 * stor-layout.c (int_mode_for_mode): Likewise.
20023 * tree.c (build_vector_type_for_mode): Likewise.
20024 * varasm.c (output_constant_pool_2): Likewise.
20025 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
20026 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
20027 for MODE_VECTOR_BOOL.
20028 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
20029 of mode class checks.
20030 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
20031 instead of a list of mode class checks.
20032 (expand_vector_scalar_condition): Likewise.
20033 (type_for_widest_vector_mode): Handle BImode as an inner mode.
20035 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20036 Alan Hayward <alan.hayward@arm.com>
20037 David Sherwood <david.sherwood@arm.com>
20039 * machmode.h (mode_size): Change from unsigned short to
20041 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
20042 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
20043 or if measurement_type is not polynomial.
20044 (fixed_size_mode::includes_p): Check for constant-sized modes.
20045 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
20046 return a poly_uint16 rather than an unsigned short.
20047 (emit_mode_size): Change the type of mode_size from unsigned short
20048 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
20049 (emit_mode_adjustments): Cope with polynomial vector sizes.
20050 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
20052 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
20054 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
20055 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
20056 * caller-save.c (setup_save_areas): Likewise.
20057 (replace_reg_with_saved_mem): Likewise.
20058 * calls.c (emit_library_call_value_1): Likewise.
20059 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
20060 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
20061 (gen_lowpart_for_combine): Likewise.
20062 * convert.c (convert_to_integer_1): Likewise.
20063 * cse.c (equiv_constant, cse_insn): Likewise.
20064 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
20065 (cselib_subst_to_values): Likewise.
20066 * dce.c (word_dce_process_block): Likewise.
20067 * df-problems.c (df_word_lr_mark_ref): Likewise.
20068 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
20069 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
20070 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
20071 (rtl_for_decl_location): Likewise.
20072 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
20073 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
20074 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
20075 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
20076 (expand_expr_real_1): Likewise.
20077 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
20078 (pad_below): Likewise.
20079 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
20080 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
20081 * ira.c (get_subreg_tracking_sizes): Likewise.
20082 * ira-build.c (ira_create_allocno_objects): Likewise.
20083 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
20084 (ira_sort_regnos_for_alter_reg): Likewise.
20085 * ira-costs.c (record_operand_costs): Likewise.
20086 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
20087 (resolve_simple_move): Likewise.
20088 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
20089 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
20090 (lra_constraints): Likewise.
20091 (CONST_POOL_OK_P): Reject variable-sized modes.
20092 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
20093 (add_pseudo_to_slot, lra_spill): Likewise.
20094 * omp-low.c (omp_clause_aligned_alignment): Likewise.
20095 * optabs-query.c (get_best_extraction_insn): Likewise.
20096 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
20097 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
20098 (expand_mult_highpart, valid_multiword_target_p): Likewise.
20099 * recog.c (offsettable_address_addr_space_p): Likewise.
20100 * regcprop.c (maybe_mode_change): Likewise.
20101 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
20102 * regrename.c (build_def_use): Likewise.
20103 * regstat.c (dump_reg_info): Likewise.
20104 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
20105 (find_reloads, find_reloads_subreg_address): Likewise.
20106 * reload1.c (eliminate_regs_1): Likewise.
20107 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
20108 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
20109 (simplify_binary_operation_1, simplify_subreg): Likewise.
20110 * targhooks.c (default_function_arg_padding): Likewise.
20111 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
20112 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
20113 (verify_gimple_assign_ternary): Likewise.
20114 * tree-inline.c (estimate_move_cost): Likewise.
20115 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
20116 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
20117 (get_address_cost_ainc): Likewise.
20118 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
20119 (vect_supportable_dr_alignment): Likewise.
20120 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
20121 (vectorizable_reduction): Likewise.
20122 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
20123 (vectorizable_operation, vectorizable_load): Likewise.
20124 * tree.c (build_same_sized_truth_vector_type): Likewise.
20125 * valtrack.c (cleanup_auto_inc_dec): Likewise.
20126 * var-tracking.c (emit_note_insn_var_location): Likewise.
20127 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
20128 (ADDR_VEC_ALIGN): Likewise.
20130 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20131 Alan Hayward <alan.hayward@arm.com>
20132 David Sherwood <david.sherwood@arm.com>
20134 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
20136 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
20137 or if measurement_type is polynomial.
20138 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
20139 * combine.c (make_extraction): Likewise.
20140 * dse.c (find_shift_sequence): Likewise.
20141 * dwarf2out.c (mem_loc_descriptor): Likewise.
20142 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
20143 (extract_bit_field, extract_low_bits): Likewise.
20144 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
20145 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
20146 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
20147 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
20148 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
20149 * reload.c (find_reloads): Likewise.
20150 * reload1.c (alter_reg): Likewise.
20151 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
20152 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
20153 * tree-if-conv.c (predicate_mem_writes): Likewise.
20154 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
20155 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
20156 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
20157 * valtrack.c (dead_debug_insert_temp): Likewise.
20158 * varasm.c (mergeable_constant_section): Likewise.
20159 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
20161 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20162 Alan Hayward <alan.hayward@arm.com>
20163 David Sherwood <david.sherwood@arm.com>
20165 * expr.c (expand_assignment): Cope with polynomial mode sizes
20166 when assigning to a CONCAT.
20168 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20169 Alan Hayward <alan.hayward@arm.com>
20170 David Sherwood <david.sherwood@arm.com>
20172 * machmode.h (mode_precision): Change from unsigned short to
20174 (mode_to_precision): Return a poly_uint16 rather than an unsigned
20176 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
20177 or if measurement_type is not polynomial.
20178 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
20179 in which the mode is already known to be a scalar_int_mode.
20180 * genmodes.c (emit_mode_precision): Change the type of mode_precision
20181 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
20183 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
20184 for GET_MODE_PRECISION.
20185 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
20186 for GET_MODE_PRECISION.
20187 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
20189 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
20190 (expand_field_assignment, make_extraction): Likewise.
20191 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
20192 (get_last_value): Likewise.
20193 * convert.c (convert_to_integer_1): Likewise.
20194 * cse.c (cse_insn): Likewise.
20195 * expr.c (expand_expr_real_1): Likewise.
20196 * lra-constraints.c (simplify_operand_subreg): Likewise.
20197 * optabs-query.c (can_atomic_load_p): Likewise.
20198 * optabs.c (expand_atomic_load): Likewise.
20199 (expand_atomic_store): Likewise.
20200 * ree.c (combine_reaching_defs): Likewise.
20201 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
20202 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
20203 * tree.h (type_has_mode_precision_p): Likewise.
20204 * ubsan.c (instrument_si_overflow): Likewise.
20206 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20207 Alan Hayward <alan.hayward@arm.com>
20208 David Sherwood <david.sherwood@arm.com>
20210 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
20211 polynomial numbers of units.
20212 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
20213 (valid_vector_subparts_p): New function.
20214 (build_vector_type): Remove temporary shim and take the number
20215 of units as a poly_uint64 rather than an int.
20216 (build_opaque_vector_type): Take the number of units as a
20217 poly_uint64 rather than an int.
20218 * tree.c (build_vector_from_ctor): Handle polynomial
20219 TYPE_VECTOR_SUBPARTS.
20220 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
20221 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
20222 (build_vector_from_val): If the number of units is variable,
20223 use build_vec_duplicate_cst for constant operands and
20224 VEC_DUPLICATE_EXPR otherwise.
20225 (make_vector_type): Remove temporary is_constant ().
20226 (build_vector_type, build_opaque_vector_type): Take the number of
20227 units as a poly_uint64 rather than an int.
20228 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
20230 * cfgexpand.c (expand_debug_expr): Likewise.
20231 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
20232 (store_constructor, expand_expr_real_1): Likewise.
20233 (const_scalar_mask_from_tree): Likewise.
20234 * fold-const-call.c (fold_const_reduction): Likewise.
20235 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
20236 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
20237 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
20238 (fold_relational_const): Likewise.
20239 (native_interpret_vector): Likewise. Change the size from an
20240 int to an unsigned int.
20241 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
20242 TYPE_VECTOR_SUBPARTS.
20243 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
20244 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
20245 duplicating a non-constant operand into a variable-length vector.
20246 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
20247 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
20248 * ipa-icf.c (sem_variable::equals): Likewise.
20249 * match.pd: Likewise.
20250 * omp-simd-clone.c (simd_clone_subparts): Likewise.
20251 * print-tree.c (print_node): Likewise.
20252 * stor-layout.c (layout_type): Likewise.
20253 * targhooks.c (default_builtin_vectorization_cost): Likewise.
20254 * tree-cfg.c (verify_gimple_comparison): Likewise.
20255 (verify_gimple_assign_binary): Likewise.
20256 (verify_gimple_assign_ternary): Likewise.
20257 (verify_gimple_assign_single): Likewise.
20258 * tree-pretty-print.c (dump_generic_node): Likewise.
20259 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
20260 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
20261 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
20262 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
20263 (vect_shift_permute_load_chain): Likewise.
20264 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
20265 (expand_vector_condition, optimize_vector_constructor): Likewise.
20266 (lower_vec_perm, get_compute_type): Likewise.
20267 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
20268 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
20269 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
20270 (vect_recog_mask_conversion_pattern): Likewise.
20271 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
20272 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
20273 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
20274 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
20275 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
20276 (vectorizable_shift, vectorizable_operation, vectorizable_store)
20277 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
20278 (supportable_widening_operation): Likewise.
20279 (supportable_narrowing_operation): Likewise.
20280 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
20282 * varasm.c (output_constant): Likewise.
20284 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20285 Alan Hayward <alan.hayward@arm.com>
20286 David Sherwood <david.sherwood@arm.com>
20288 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
20289 so that both the length == 3 and length != 3 cases set up their
20290 own permute vectors. Add comments explaining why we know the
20291 number of elements is constant.
20292 (vect_permute_load_chain): Likewise.
20294 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20295 Alan Hayward <alan.hayward@arm.com>
20296 David Sherwood <david.sherwood@arm.com>
20298 * machmode.h (mode_nunits): Change from unsigned char to
20300 (ONLY_FIXED_SIZE_MODES): New macro.
20301 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
20302 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
20303 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
20305 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
20306 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
20307 or if measurement_type is not polynomial.
20308 * genmodes.c (ZERO_COEFFS): New macro.
20309 (emit_mode_nunits_inline): Make mode_nunits_inline return a
20311 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
20312 Use ZERO_COEFFS when emitting initializers.
20313 * data-streamer.h (bp_pack_poly_value): New function.
20314 (bp_unpack_poly_value): Likewise.
20315 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
20316 for GET_MODE_NUNITS.
20317 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
20318 for GET_MODE_NUNITS.
20319 * tree.c (make_vector_type): Remove temporary shim and make
20320 the real function take the number of units as a poly_uint64
20321 rather than an int.
20322 (build_vector_type_for_mode): Handle polynomial nunits.
20323 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
20324 * emit-rtl.c (const_vec_series_p_1): Likewise.
20325 (gen_rtx_CONST_VECTOR): Likewise.
20326 * fold-const.c (test_vec_duplicate_folding): Likewise.
20327 * genrecog.c (validate_pattern): Likewise.
20328 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
20329 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
20330 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
20331 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
20332 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
20333 * rtlanal.c (subreg_get_info): Likewise.
20334 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
20335 (vect_grouped_load_supported): Likewise.
20336 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
20337 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
20338 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
20339 (simplify_const_unary_operation, simplify_binary_operation_1)
20340 (simplify_const_binary_operation, simplify_ternary_operation)
20341 (test_vector_ops_duplicate, test_vector_ops): Likewise.
20342 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
20343 instead of CONST_VECTOR_NUNITS.
20344 * varasm.c (output_constant_pool_2): Likewise.
20345 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
20346 explicit-encoded elements in the XVEC for variable-length vectors.
20348 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20350 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
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 * coretypes.h (fixed_size_mode): Declare.
20357 (fixed_size_mode_pod): New typedef.
20358 * builtins.h (target_builtins::x_apply_args_mode)
20359 (target_builtins::x_apply_result_mode): Change type to
20360 fixed_size_mode_pod.
20361 * builtins.c (apply_args_size, apply_result_size, result_vector)
20362 (expand_builtin_apply_args_1, expand_builtin_apply)
20363 (expand_builtin_return): Update accordingly.
20365 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20367 * cse.c (hash_rtx_cb): Hash only the encoded elements.
20368 * cselib.c (cselib_hash_rtx): Likewise.
20369 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
20370 CONST_VECTOR encoding.
20372 2018-01-03 Jakub Jelinek <jakub@redhat.com>
20373 Jeff Law <law@redhat.com>
20376 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
20377 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
20378 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
20379 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
20382 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
20383 explicitly probe *sp in a noreturn function if there were any callee
20384 register saves or frame pointer is needed.
20386 2018-01-03 Jakub Jelinek <jakub@redhat.com>
20389 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
20390 BLKmode for ternary, binary or unary expressions.
20393 * var-tracking.c (delete_vta_debug_insn): New inline function.
20394 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
20395 insns from get_insns () to NULL instead of each bb separately.
20396 Use delete_vta_debug_insn. No longer static.
20397 (vt_debug_insns_local, variable_tracking_main_1): Adjust
20398 delete_vta_debug_insns callers.
20399 * rtl.h (delete_vta_debug_insns): Declare.
20400 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
20401 instead of variable_tracking_main.
20403 2018-01-03 Martin Sebor <msebor@redhat.com>
20405 PR tree-optimization/83603
20406 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
20407 arguments past the endof the argument list in functions declared
20408 without a prototype.
20409 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
20410 Avoid checking when arguments are null.
20412 2018-01-03 Martin Sebor <msebor@redhat.com>
20415 * doc/extend.texi (attribute const): Fix a typo.
20416 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
20417 issuing -Wsuggest-attribute for void functions.
20419 2018-01-03 Martin Sebor <msebor@redhat.com>
20421 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
20422 offset_int::from instead of wide_int::to_shwi.
20423 (maybe_diag_overlap): Remove assertion.
20424 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
20425 * gimple-ssa-sprintf.c (format_directive): Same.
20426 (parse_directive): Same.
20427 (sprintf_dom_walker::compute_format_length): Same.
20428 (try_substitute_return_value): Same.
20430 2018-01-03 Jeff Law <law@redhat.com>
20432 PR middle-end/83654
20433 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
20434 non-constant residual for zero at runtime and avoid probing in
20435 that case. Reorganize code for trailing problem to mirror handling
20438 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
20440 PR tree-optimization/83501
20441 * tree-ssa-strlen.c (get_string_cst): New.
20442 (handle_char_store): Call get_string_cst.
20444 2018-01-03 Martin Liska <mliska@suse.cz>
20446 PR tree-optimization/83593
20447 * tree-ssa-strlen.c: Include tree-cfg.h.
20448 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
20449 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
20450 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
20452 (strlen_dom_walker::before_dom_children): Call
20453 gimple_purge_dead_eh_edges. Dump tranformation with details
20455 (strlen_dom_walker::before_dom_children): Update call by adding
20456 new argument cleanup_eh.
20457 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
20459 2018-01-03 Martin Liska <mliska@suse.cz>
20462 * cif-code.def (VARIADIC_THUNK): New enum value.
20463 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
20466 2018-01-03 Jan Beulich <jbeulich@suse.com>
20468 * sse.md (mov<mode>_internal): Tighten condition for when to use
20469 vmovdqu<ssescalarsize> for TI and OI modes.
20471 2018-01-03 Jakub Jelinek <jakub@redhat.com>
20473 Update copyright years.
20475 2018-01-03 Martin Liska <mliska@suse.cz>
20478 * ipa-visibility.c (function_and_variable_visibility): Skip
20479 functions with noipa attribure.
20481 2018-01-03 Jakub Jelinek <jakub@redhat.com>
20483 * gcc.c (process_command): Update copyright notice dates.
20484 * gcov-dump.c (print_version): Ditto.
20485 * gcov.c (print_version): Ditto.
20486 * gcov-tool.c (print_version): Ditto.
20487 * gengtype.c (create_file): Ditto.
20488 * doc/cpp.texi: Bump @copying's copyright year.
20489 * doc/cppinternals.texi: Ditto.
20490 * doc/gcc.texi: Ditto.
20491 * doc/gccint.texi: Ditto.
20492 * doc/gcov.texi: Ditto.
20493 * doc/install.texi: Ditto.
20494 * doc/invoke.texi: Ditto.
20496 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20498 * vector-builder.h (vector_builder::m_full_nelts): Change from
20499 unsigned int to poly_uint64.
20500 (vector_builder::full_nelts): Update prototype accordingly.
20501 (vector_builder::new_vector): Likewise.
20502 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
20503 (vector_builder::operator ==): Likewise.
20504 (vector_builder::finalize): Likewise.
20505 * int-vector-builder.h (int_vector_builder::int_vector_builder):
20506 Take the number of elements as a poly_uint64 rather than an
20508 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
20509 from unsigned int to poly_uint64.
20510 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
20511 (vec_perm_indices::new_vector): Likewise.
20512 (vec_perm_indices::length): Likewise.
20513 (vec_perm_indices::nelts_per_input): Likewise.
20514 (vec_perm_indices::input_nelts): Likewise.
20515 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
20516 number of elements per input as a poly_uint64 rather than an
20517 unsigned int. Use the original encoding for variable-length
20518 vectors, rather than clamping each individual element.
20519 For the second and subsequent elements in each pattern,
20520 clamp the step and base before clamping their sum.
20521 (vec_perm_indices::series_p): Handle polynomial element counts.
20522 (vec_perm_indices::all_in_range_p): Likewise.
20523 (vec_perm_indices_to_tree): Likewise.
20524 (vec_perm_indices_to_rtx): Likewise.
20525 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
20526 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
20527 (tree_vector_builder::new_binary_operation): Handle polynomial
20528 element counts. Return false if we need to know the number
20529 of elements at compile time.
20530 * fold-const.c (fold_vec_perm): Punt if the number of elements
20531 isn't known at compile time.
20533 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20535 * vec-perm-indices.h (vec_perm_builder): Change element type
20536 from HOST_WIDE_INT to poly_int64.
20537 (vec_perm_indices::element_type): Update accordingly.
20538 (vec_perm_indices::clamp): Handle polynomial element_types.
20539 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
20540 (vec_perm_indices::all_in_range_p): Likewise.
20541 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
20543 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
20544 polynomial vec_perm_indices element types.
20545 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
20546 * fold-const.c (fold_vec_perm): Likewise.
20547 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
20548 * tree-vect-generic.c (lower_vec_perm): Likewise.
20549 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
20550 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
20551 element type to HOST_WIDE_INT.
20553 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20554 Alan Hayward <alan.hayward@arm.com>
20555 David Sherwood <david.sherwood@arm.com>
20557 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
20558 rather than an int. Use plus_constant.
20559 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
20560 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
20562 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20563 Alan Hayward <alan.hayward@arm.com>
20564 David Sherwood <david.sherwood@arm.com>
20566 * calls.c (emit_call_1, expand_call): Change struct_value_size from
20567 a HOST_WIDE_INT to a poly_int64.
20569 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20570 Alan Hayward <alan.hayward@arm.com>
20571 David Sherwood <david.sherwood@arm.com>
20573 * calls.c (load_register_parameters): Cope with polynomial
20574 mode sizes. Require a constant size for BLKmode parameters
20575 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
20576 forces a parameter to be padded at the lsb end in order to
20577 fill a complete number of words, require the parameter size
20578 to be ordered wrt UNITS_PER_WORD.
20580 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20581 Alan Hayward <alan.hayward@arm.com>
20582 David Sherwood <david.sherwood@arm.com>
20584 * reload1.c (spill_stack_slot_width): Change element type
20585 from unsigned int to poly_uint64_pod.
20586 (alter_reg): Treat mode sizes as polynomial.
20588 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20589 Alan Hayward <alan.hayward@arm.com>
20590 David Sherwood <david.sherwood@arm.com>
20592 * reload.c (complex_word_subreg_p): New function.
20593 (reload_inner_reg_of_subreg, push_reload): Use it.
20595 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20596 Alan Hayward <alan.hayward@arm.com>
20597 David Sherwood <david.sherwood@arm.com>
20599 * lra-constraints.c (process_alt_operands): Reject matched
20600 operands whose sizes aren't ordered.
20601 (match_reload): Refer to this check here.
20603 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20604 Alan Hayward <alan.hayward@arm.com>
20605 David Sherwood <david.sherwood@arm.com>
20607 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
20608 that the mode size is in the set {1, 2, 4, 8, 16}.
20610 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20611 Alan Hayward <alan.hayward@arm.com>
20612 David Sherwood <david.sherwood@arm.com>
20614 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
20615 Use plus_constant instead of gen_rtx_PLUS.
20617 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20618 Alan Hayward <alan.hayward@arm.com>
20619 David Sherwood <david.sherwood@arm.com>
20621 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
20622 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
20623 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
20624 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
20625 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
20626 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
20627 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
20628 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
20629 * config/i386/i386.c (ix86_push_rounding): ...this new function.
20630 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
20632 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
20633 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
20634 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
20635 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
20636 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
20637 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
20638 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
20639 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
20640 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
20641 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
20643 * expr.c (emit_move_resolve_push): Treat the input and result
20644 of PUSH_ROUNDING as a poly_int64.
20645 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
20646 (emit_push_insn): Likewise.
20647 * lra-eliminations.c (mark_not_eliminable): Likewise.
20648 * recog.c (push_operand): Likewise.
20649 * reload1.c (elimination_effects): Likewise.
20650 * rtlanal.c (nonzero_bits1): Likewise.
20651 * calls.c (store_one_arg): Likewise. Require the padding to be
20652 known at compile time.
20654 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20655 Alan Hayward <alan.hayward@arm.com>
20656 David Sherwood <david.sherwood@arm.com>
20658 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
20659 Use plus_constant instead of gen_rtx_PLUS.
20661 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20662 Alan Hayward <alan.hayward@arm.com>
20663 David Sherwood <david.sherwood@arm.com>
20665 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
20666 rather than an int.
20668 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20669 Alan Hayward <alan.hayward@arm.com>
20670 David Sherwood <david.sherwood@arm.com>
20672 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
20673 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
20674 via stack temporaries. Treat the mode size as polynomial too.
20676 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20677 Alan Hayward <alan.hayward@arm.com>
20678 David Sherwood <david.sherwood@arm.com>
20680 * expr.c (expand_expr_real_2): When handling conversions involving
20681 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
20682 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
20683 as a poly_uint64 too.
20685 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20686 Alan Hayward <alan.hayward@arm.com>
20687 David Sherwood <david.sherwood@arm.com>
20689 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
20691 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20692 Alan Hayward <alan.hayward@arm.com>
20693 David Sherwood <david.sherwood@arm.com>
20695 * combine.c (can_change_dest_mode): Handle polynomial
20696 REGMODE_NATURAL_SIZE.
20697 * expmed.c (store_bit_field_1): Likewise.
20698 * expr.c (store_constructor): Likewise.
20699 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
20700 and polynomial REGMODE_NATURAL_SIZE.
20701 (gen_lowpart_common): Likewise.
20702 * reginfo.c (record_subregs_of_mode): Likewise.
20703 * rtlanal.c (read_modify_subreg_p): Likewise.
20705 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20706 Alan Hayward <alan.hayward@arm.com>
20707 David Sherwood <david.sherwood@arm.com>
20709 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
20710 numbers of elements.
20712 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20713 Alan Hayward <alan.hayward@arm.com>
20714 David Sherwood <david.sherwood@arm.com>
20716 * match.pd: Cope with polynomial numbers of vector elements.
20718 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20719 Alan Hayward <alan.hayward@arm.com>
20720 David Sherwood <david.sherwood@arm.com>
20722 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
20723 in a POINTER_PLUS_EXPR.
20725 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20726 Alan Hayward <alan.hayward@arm.com>
20727 David Sherwood <david.sherwood@arm.com>
20729 * omp-simd-clone.c (simd_clone_subparts): New function.
20730 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
20731 (ipa_simd_modify_function_body): Likewise.
20733 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20734 Alan Hayward <alan.hayward@arm.com>
20735 David Sherwood <david.sherwood@arm.com>
20737 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
20738 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
20739 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
20740 (expand_vector_condition, vector_element): Likewise.
20741 (subparts_gt): New function.
20742 (get_compute_type): Use subparts_gt.
20743 (count_type_subparts): Delete.
20744 (expand_vector_operations_1): Use subparts_gt instead of
20745 count_type_subparts.
20747 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20748 Alan Hayward <alan.hayward@arm.com>
20749 David Sherwood <david.sherwood@arm.com>
20751 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
20752 (vect_compile_time_alias): ...this new function. Do the calculation
20753 on poly_ints rather than trees.
20754 (vect_prune_runtime_alias_test_list): Update call accordingly.
20756 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20757 Alan Hayward <alan.hayward@arm.com>
20758 David Sherwood <david.sherwood@arm.com>
20760 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
20762 (vect_schedule_slp_instance): Likewise.
20764 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20765 Alan Hayward <alan.hayward@arm.com>
20766 David Sherwood <david.sherwood@arm.com>
20768 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
20769 constant and extern definitions for variable-length vectors.
20770 (vect_get_constant_vectors): Note that the number of units
20771 is known to be constant.
20773 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20774 Alan Hayward <alan.hayward@arm.com>
20775 David Sherwood <david.sherwood@arm.com>
20777 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
20778 of units as polynomial. Choose between WIDE and NARROW based
20781 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20782 Alan Hayward <alan.hayward@arm.com>
20783 David Sherwood <david.sherwood@arm.com>
20785 * tree-vect-stmts.c (simd_clone_subparts): New function.
20786 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
20788 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20789 Alan Hayward <alan.hayward@arm.com>
20790 David Sherwood <david.sherwood@arm.com>
20792 * tree-vect-stmts.c (vectorizable_call): Treat the number of
20793 vectors as polynomial. Use build_index_vector for
20794 IFN_GOMP_SIMD_LANE.
20796 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20797 Alan Hayward <alan.hayward@arm.com>
20798 David Sherwood <david.sherwood@arm.com>
20800 * tree-vect-stmts.c (get_load_store_type): Treat the number of
20801 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
20802 for variable-length vectors.
20803 (vectorizable_mask_load_store): Treat the number of units as
20804 polynomial, asserting that it is constant if the condition has
20805 already been enforced.
20806 (vectorizable_store, vectorizable_load): Likewise.
20808 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20809 Alan Hayward <alan.hayward@arm.com>
20810 David Sherwood <david.sherwood@arm.com>
20812 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
20813 of units as polynomial. Punt if we can't tell at compile time
20814 which vector contains the final result.
20816 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20817 Alan Hayward <alan.hayward@arm.com>
20818 David Sherwood <david.sherwood@arm.com>
20820 * tree-vect-loop.c (vectorizable_induction): Treat the number
20821 of units as polynomial. Punt on SLP inductions. Use an integer
20822 VEC_SERIES_EXPR for variable-length integer reductions. Use a
20823 cast of such a series for variable-length floating-point
20826 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20827 Alan Hayward <alan.hayward@arm.com>
20828 David Sherwood <david.sherwood@arm.com>
20830 * tree.h (build_index_vector): Declare.
20831 * tree.c (build_index_vector): New function.
20832 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
20833 of units as polynomial, forcibly converting it to a constant if
20834 vectorizable_reduction has already enforced the condition.
20835 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
20836 to create a {1,2,3,...} vector.
20837 (vectorizable_reduction): Treat the number of units as polynomial.
20838 Choose vectype_in based on the largest scalar element size rather
20839 than the smallest number of units. Enforce the restrictions
20842 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20843 Alan Hayward <alan.hayward@arm.com>
20844 David Sherwood <david.sherwood@arm.com>
20846 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
20847 number of units as polynomial.
20849 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20850 Alan Hayward <alan.hayward@arm.com>
20851 David Sherwood <david.sherwood@arm.com>
20853 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
20854 * target.def (autovectorize_vector_sizes): Return the vector sizes
20855 by pointer, using vector_sizes rather than a bitmask.
20856 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
20857 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
20858 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
20860 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
20861 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
20862 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
20863 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
20864 * omp-general.c (omp_max_vf): Likewise.
20865 * omp-low.c (omp_clause_aligned_alignment): Likewise.
20866 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
20867 * tree-vect-loop.c (vect_analyze_loop): Likewise.
20868 * tree-vect-slp.c (vect_slp_bb): Likewise.
20869 * doc/tm.texi: Regenerate.
20870 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
20872 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
20873 the vector size as a poly_uint64 rather than an unsigned int.
20874 (current_vector_size): Change from an unsigned int to a poly_uint64.
20875 (get_vectype_for_scalar_type): Update accordingly.
20876 * tree.h (build_truth_vector_type): Take the size and number of
20877 units as a poly_uint64 rather than an unsigned int.
20878 (build_vector_type): Add a temporary overload that takes
20879 the number of units as a poly_uint64 rather than an unsigned int.
20880 * tree.c (make_vector_type): Likewise.
20881 (build_truth_vector_type): Take the number of units as a poly_uint64
20882 rather than an unsigned int.
20884 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20885 Alan Hayward <alan.hayward@arm.com>
20886 David Sherwood <david.sherwood@arm.com>
20888 * target.def (get_mask_mode): Take the number of units and length
20889 as poly_uint64s rather than unsigned ints.
20890 * targhooks.h (default_get_mask_mode): Update accordingly.
20891 * targhooks.c (default_get_mask_mode): Likewise.
20892 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
20893 * doc/tm.texi: Regenerate.
20895 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20896 Alan Hayward <alan.hayward@arm.com>
20897 David Sherwood <david.sherwood@arm.com>
20899 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
20900 * omp-general.c (omp_max_vf): Likewise.
20901 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
20902 (expand_omp_simd): Handle polynomial safelen.
20903 * omp-low.c (omplow_simd_context): Add a default constructor.
20904 (omplow_simd_context::max_vf): Change from int to poly_uint64.
20905 (lower_rec_simd_input_clauses): Update accordingly.
20906 (lower_rec_input_clauses): Likewise.
20908 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20909 Alan Hayward <alan.hayward@arm.com>
20910 David Sherwood <david.sherwood@arm.com>
20912 * tree-vectorizer.h (vect_nunits_for_cost): New function.
20913 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
20914 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
20915 (vect_analyze_slp_cost): Likewise.
20916 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
20917 (vect_model_load_cost): Likewise.
20919 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20920 Alan Hayward <alan.hayward@arm.com>
20921 David Sherwood <david.sherwood@arm.com>
20923 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
20924 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
20925 from an unsigned int * to a poly_uint64_pod *.
20926 (calculate_unrolling_factor): New function.
20927 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
20929 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
20930 Alan Hayward <alan.hayward@arm.com>
20931 David Sherwood <david.sherwood@arm.com>
20933 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
20934 from an unsigned int to a poly_uint64.
20935 (_loop_vec_info::slp_unrolling_factor): Likewise.
20936 (_loop_vec_info::vectorization_factor): Change from an int
20938 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
20939 (vect_get_num_vectors): New function.
20940 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
20941 (vect_get_num_copies): Use vect_get_num_vectors.
20942 (vect_analyze_data_ref_dependences): Change max_vf from an int *
20943 to an unsigned int *.
20944 (vect_analyze_data_refs): Change min_vf from an int * to a
20946 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
20947 than an unsigned HOST_WIDE_INT.
20948 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
20949 (vect_analyze_data_ref_dependence): Change max_vf from an int *
20950 to an unsigned int *.
20951 (vect_analyze_data_ref_dependences): Likewise.
20952 (vect_compute_data_ref_alignment): Handle polynomial vf.
20953 (vect_enhance_data_refs_alignment): Likewise.
20954 (vect_prune_runtime_alias_test_list): Likewise.
20955 (vect_shift_permute_load_chain): Likewise.
20956 (vect_supportable_dr_alignment): Likewise.
20957 (dependence_distance_ge_vf): Take the vectorization factor as a
20958 poly_uint64 rather than an unsigned HOST_WIDE_INT.
20959 (vect_analyze_data_refs): Change min_vf from an int * to a
20961 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
20962 vfm1 as a poly_uint64 rather than an int. Make the same change
20963 for the returned bound_scalar.
20964 (vect_gen_vector_loop_niters): Handle polynomial vf.
20965 (vect_do_peeling): Likewise. Update call to
20966 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
20967 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
20969 * tree-vect-loop.c (vect_determine_vectorization_factor)
20970 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
20971 (vect_get_known_peeling_cost): Likewise.
20972 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
20973 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
20974 (vect_transform_loop): Likewise. Use the lowest possible VF when
20975 updating the upper bounds of the loop.
20976 (vect_min_worthwhile_factor): Make static. Return an unsigned int
20977 rather than an int.
20978 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
20979 polynomial unroll factors.
20980 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
20981 (vect_make_slp_decision): Likewise.
20982 (vect_supported_load_permutation_p): Likewise, and polynomial
20984 (vect_analyze_slp_cost): Handle polynomial vf.
20985 (vect_slp_analyze_node_operations): Likewise.
20986 (vect_slp_analyze_bb_1): Likewise.
20987 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
20988 than an unsigned HOST_WIDE_INT.
20989 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
20990 (vectorizable_load): Handle polynomial vf.
20991 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
20993 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
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 * match.pd: Handle bit operations involving three constants
21000 and try to fold one pair.
21002 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
21004 * tree-vect-loop-manip.c: Include gimple-fold.h.
21005 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
21006 niters_maybe_zero parameters. Handle other cases besides a step of 1.
21007 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
21008 Add a path that uses a step of VF instead of 1, but disable it
21010 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
21011 and niters_no_overflow parameters. Update calls to
21012 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
21013 Create a new SSA name if the latter choses to use a ste other
21014 than zero, and return it via niters_vector_mult_vf_var.
21015 * tree-vect-loop.c (vect_transform_loop): Update calls to
21016 vect_do_peeling, vect_gen_vector_loop_niters and
21017 slpeel_make_loop_iterate_ntimes.
21018 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
21019 (vect_gen_vector_loop_niters): Update declarations after above changes.
21021 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
21023 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
21024 128-bit round to integer instructions.
21025 (ceil<mode>2): Likewise.
21026 (btrunc<mode>2): Likewise.
21027 (round<mode>2): Likewise.
21029 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
21031 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
21032 unaligned VSX load/store on P8/P9.
21033 (expand_block_clear): Allow the use of unaligned VSX
21034 load/store on P8/P9.
21036 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21038 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
21040 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
21041 swap associated with both a load and a store.
21043 2018-01-02 Andrew Waterman <andrew@sifive.com>
21045 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
21046 * config/riscv/riscv.md (clear_cache): Use it.
21048 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
21050 * web.c: Remove out-of-date comment.
21052 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21054 * expr.c (fixup_args_size_notes): Check that any existing
21055 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
21056 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
21057 (emit_single_push_insn): ...here.
21059 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21061 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
21062 (const_vector_encoded_nelts): New function.
21063 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
21064 (const_vector_int_elt, const_vector_elt): Declare.
21065 * emit-rtl.c (const_vector_int_elt_1): New function.
21066 (const_vector_elt): Likewise.
21067 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
21068 of CONST_VECTOR_ELT.
21070 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21072 * expr.c: Include rtx-vector-builder.h.
21073 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
21074 directly on the tree encoding.
21075 (const_vector_from_tree): Likewise.
21076 * optabs.c: Include rtx-vector-builder.h.
21077 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
21078 sequence of "u" values.
21079 * vec-perm-indices.c: Include rtx-vector-builder.h.
21080 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
21081 directly on the vec_perm_indices encoding.
21083 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21085 * doc/rtl.texi (const_vector): Describe new encoding scheme.
21086 * Makefile.in (OBJS): Add rtx-vector-builder.o.
21087 * rtx-vector-builder.h: New file.
21088 * rtx-vector-builder.c: Likewise.
21089 * rtl.h (rtx_def::u2): Add a const_vector field.
21090 (CONST_VECTOR_NPATTERNS): New macro.
21091 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
21092 (CONST_VECTOR_DUPLICATE_P): Likewise.
21093 (CONST_VECTOR_STEPPED_P): Likewise.
21094 (CONST_VECTOR_ENCODED_ELT): Likewise.
21095 (const_vec_duplicate_p): Check for a duplicated vector encoding.
21096 (unwrap_const_vec_duplicate): Likewise.
21097 (const_vec_series_p): Check for a non-duplicated vector encoding.
21098 Say that the function only returns true for integer vectors.
21099 * emit-rtl.c: Include rtx-vector-builder.h.
21100 (gen_const_vec_duplicate_1): Delete.
21101 (gen_const_vector): Call gen_const_vec_duplicate instead of
21102 gen_const_vec_duplicate_1.
21103 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
21104 (gen_const_vec_duplicate): Use rtx_vector_builder.
21105 (gen_const_vec_series): Likewise.
21106 (gen_rtx_CONST_VECTOR): Likewise.
21107 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
21108 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
21109 Build a new vector rather than modifying a CONST_VECTOR in-place.
21110 (handle_special_swappables): Update call accordingly.
21111 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
21112 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
21113 Build a new vector rather than modifying a CONST_VECTOR in-place.
21114 (handle_special_swappables): Update call accordingly.
21116 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21118 * simplify-rtx.c (simplify_const_binary_operation): Use
21119 CONST_VECTOR_ELT instead of XVECEXP.
21121 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21123 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
21124 the selector elements to be different from the data elements
21125 if the selector is a VECTOR_CST.
21126 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
21127 ssizetype for the selector.
21129 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21131 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
21132 before testing each element individually.
21133 * tree-vect-generic.c (lower_vec_perm): Likewise.
21135 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21137 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
21138 * selftest-run-tests.c (selftest::run_tests): Call it.
21139 * vector-builder.h (vector_builder::operator ==): New function.
21140 (vector_builder::operator !=): Likewise.
21141 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
21142 (vec_perm_indices::all_from_input_p): New function.
21143 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
21144 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
21145 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
21146 instead of reading the VECTOR_CST directly. Detect whether both
21147 vector inputs are the same before constructing the vec_perm_indices,
21148 and update the number of inputs argument accordingly. Use the
21149 utility functions added above. Only construct sel2 if we need to.
21151 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21153 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
21154 the broadcast of the low byte.
21155 (expand_mult_highpart): Use an explicit encoding for the permutes.
21156 * optabs-query.c (can_mult_highpart_p): Likewise.
21157 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
21158 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
21159 (vectorizable_bswap): Likewise.
21160 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
21161 explicit encoding for the power-of-2 permutes.
21162 (vect_permute_store_chain): Likewise.
21163 (vect_grouped_load_supported): Likewise.
21164 (vect_permute_load_chain): Likewise.
21166 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21168 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
21169 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
21170 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
21171 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
21172 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
21173 (vect_gen_perm_mask_any): Likewise.
21175 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21177 * int-vector-builder.h: New file.
21178 * vec-perm-indices.h: Include int-vector-builder.h.
21179 (vec_perm_indices): Redefine as an int_vector_builder.
21180 (auto_vec_perm_indices): Delete.
21181 (vec_perm_builder): Redefine as a stand-alone class.
21182 (vec_perm_indices::vec_perm_indices): New function.
21183 (vec_perm_indices::clamp): Likewise.
21184 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
21185 (vec_perm_indices::new_vector): New function.
21186 (vec_perm_indices::new_expanded_vector): Update for new
21187 vec_perm_indices class.
21188 (vec_perm_indices::rotate_inputs): New function.
21189 (vec_perm_indices::all_in_range_p): Operate directly on the
21190 encoded form, without computing elided elements.
21191 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
21192 encoding. Update for new vec_perm_indices class.
21193 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
21194 the given vec_perm_builder.
21195 (expand_vec_perm_var): Update vec_perm_builder constructor.
21196 (expand_mult_highpart): Use vec_perm_builder instead of
21197 auto_vec_perm_indices.
21198 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
21199 vec_perm_indices instead of auto_vec_perm_indices. Use a single
21200 or double series encoding as appropriate.
21201 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
21202 vec_perm_indices instead of auto_vec_perm_indices.
21203 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
21204 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
21205 (vect_permute_store_chain): Likewise.
21206 (vect_grouped_load_supported): Likewise.
21207 (vect_permute_load_chain): Likewise.
21208 (vect_shift_permute_load_chain): Likewise.
21209 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
21210 (vect_transform_slp_perm_load): Likewise.
21211 (vect_schedule_slp_instance): Likewise.
21212 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
21213 (vectorizable_mask_load_store): Likewise.
21214 (vectorizable_bswap): Likewise.
21215 (vectorizable_store): Likewise.
21216 (vectorizable_load): Likewise.
21217 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
21218 vec_perm_indices instead of auto_vec_perm_indices. Use
21219 tree_to_vec_perm_builder to read the vector from a tree.
21220 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
21221 vec_perm_builder instead of a vec_perm_indices.
21222 (have_whole_vector_shift): Use vec_perm_builder and
21223 vec_perm_indices instead of auto_vec_perm_indices. Leave the
21224 truncation to calc_vec_perm_mask_for_shift.
21225 (vect_create_epilog_for_reduction): Likewise.
21226 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
21227 from auto_vec_perm_indices to vec_perm_indices.
21228 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
21229 instead of changing individual elements.
21230 (aarch64_vectorize_vec_perm_const): Use new_vector to install
21231 the vector in d.perm.
21232 * config/arm/arm.c (expand_vec_perm_d::perm): Change
21233 from auto_vec_perm_indices to vec_perm_indices.
21234 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
21235 instead of changing individual elements.
21236 (arm_vectorize_vec_perm_const): Use new_vector to install
21237 the vector in d.perm.
21238 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
21239 Update vec_perm_builder constructor.
21240 (rs6000_expand_interleave): Likewise.
21241 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
21242 (rs6000_expand_interleave): Likewise.
21244 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21246 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
21247 to qimode could truncate the indices.
21248 * optabs.c (expand_vec_perm_var): Likewise.
21250 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21252 * Makefile.in (OBJS): Add vec-perm-indices.o.
21253 * vec-perm-indices.h: New file.
21254 * vec-perm-indices.c: Likewise.
21255 * target.h (vec_perm_indices): Replace with a forward class
21257 (auto_vec_perm_indices): Move to vec-perm-indices.h.
21258 * optabs.h: Include vec-perm-indices.h.
21259 (expand_vec_perm): Delete.
21260 (selector_fits_mode_p, expand_vec_perm_var): Declare.
21261 (expand_vec_perm_const): Declare.
21262 * target.def (vec_perm_const_ok): Replace with...
21263 (vec_perm_const): ...this new hook.
21264 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
21265 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
21266 * doc/tm.texi: Regenerate.
21267 * optabs.def (vec_perm_const): Delete.
21268 * doc/md.texi (vec_perm_const): Likewise.
21269 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
21270 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
21271 expand_vec_perm for constant permutation vectors. Assert that
21272 the mode of variable permutation vectors is the integer equivalent
21273 of the mode that is being permuted.
21274 * optabs-query.h (selector_fits_mode_p): Declare.
21275 * optabs-query.c: Include vec-perm-indices.h.
21276 (selector_fits_mode_p): New function.
21277 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
21278 is defined, instead of checking whether the vec_perm_const_optab
21279 exists. Use targetm.vectorize.vec_perm_const instead of
21280 targetm.vectorize.vec_perm_const_ok. Check whether the indices
21281 fit in the vector mode before using a variable permute.
21282 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
21283 vec_perm_indices instead of an rtx.
21284 (expand_vec_perm): Replace with...
21285 (expand_vec_perm_const): ...this new function. Take the selector
21286 as a vec_perm_indices rather than an rtx. Also take the mode of
21287 the selector. Update call to shift_amt_for_vec_perm_mask.
21288 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
21289 Use vec_perm_indices::new_expanded_vector to expand the original
21290 selector into bytes. Check whether the indices fit in the vector
21291 mode before using a variable permute.
21292 (expand_vec_perm_var): Make global.
21293 (expand_mult_highpart): Use expand_vec_perm_const.
21294 * fold-const.c: Includes vec-perm-indices.h.
21295 * tree-ssa-forwprop.c: Likewise.
21296 * tree-vect-data-refs.c: Likewise.
21297 * tree-vect-generic.c: Likewise.
21298 * tree-vect-loop.c: Likewise.
21299 * tree-vect-slp.c: Likewise.
21300 * tree-vect-stmts.c: Likewise.
21301 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
21303 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
21304 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
21305 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
21306 (aarch64_vectorize_vec_perm_const): ...this new function.
21307 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
21308 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
21309 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
21310 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
21311 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
21312 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
21313 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
21315 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
21316 check for NEON modes.
21317 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
21318 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
21319 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
21320 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
21322 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
21323 the old VEC_PERM_CONST conditions.
21324 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
21325 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
21326 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
21327 (ia64_vectorize_vec_perm_const_ok): Merge into...
21328 (ia64_vectorize_vec_perm_const): ...this new function.
21329 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
21330 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
21331 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
21332 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
21333 * config/mips/mips.c (mips_expand_vec_perm_const)
21334 (mips_vectorize_vec_perm_const_ok): Merge into...
21335 (mips_vectorize_vec_perm_const): ...this new function.
21336 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
21337 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
21338 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
21339 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
21340 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
21341 (rs6000_expand_vec_perm_const): Delete.
21342 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
21344 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
21345 (altivec_expand_vec_perm_const_le): Take each operand individually.
21346 Operate on constant selectors rather than rtxes.
21347 (altivec_expand_vec_perm_const): Likewise. Update call to
21348 altivec_expand_vec_perm_const_le.
21349 (rs6000_expand_vec_perm_const): Delete.
21350 (rs6000_vectorize_vec_perm_const_ok): Delete.
21351 (rs6000_vectorize_vec_perm_const): New function.
21352 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
21353 an element count and rtx array.
21354 (rs6000_expand_extract_even): Update call accordingly.
21355 (rs6000_expand_interleave): Likewise.
21356 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
21357 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
21358 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
21359 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
21360 (rs6000_expand_vec_perm_const): Delete.
21361 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
21362 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
21363 (altivec_expand_vec_perm_const_le): Take each operand individually.
21364 Operate on constant selectors rather than rtxes.
21365 (altivec_expand_vec_perm_const): Likewise. Update call to
21366 altivec_expand_vec_perm_const_le.
21367 (rs6000_expand_vec_perm_const): Delete.
21368 (rs6000_vectorize_vec_perm_const_ok): Delete.
21369 (rs6000_vectorize_vec_perm_const): New function. Remove stray
21370 reference to the SPE evmerge intructions.
21371 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
21372 an element count and rtx array.
21373 (rs6000_expand_extract_even): Update call accordingly.
21374 (rs6000_expand_interleave): Likewise.
21375 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
21376 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
21378 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
21380 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21382 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
21383 vector mode and that that mode matches the mode of the data
21385 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
21386 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
21387 directly using expand_vec_perm_1 when forcing selectors into
21389 (expand_vec_perm_var): New function, split out from expand_vec_perm.
21391 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21393 * optabs-query.h (can_vec_perm_p): Delete.
21394 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
21395 * optabs-query.c (can_vec_perm_p): Split into...
21396 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
21397 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
21398 particular selector is valid.
21399 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
21400 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
21401 (vect_grouped_load_supported): Likewise.
21402 (vect_shift_permute_load_chain): Likewise.
21403 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
21404 (vect_transform_slp_perm_load): Likewise.
21405 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
21406 (vectorizable_bswap): Likewise.
21407 (vect_gen_perm_mask_checked): Likewise.
21408 * fold-const.c (fold_ternary_loc): Likewise. Don't take
21409 implementations of variable permutation vectors into account
21410 when deciding which selector to use.
21411 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
21412 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
21413 with a false third argument.
21414 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
21415 to test whether the constant selector is valid and can_vec_perm_var_p
21416 to test whether a variable selector is valid.
21418 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21420 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
21421 * optabs-query.c (can_vec_perm_p): Likewise.
21422 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
21423 instead of vec_perm_indices.
21424 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
21425 (vect_gen_perm_mask_checked): Likewise,
21426 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
21427 (vect_gen_perm_mask_checked): Likewise,
21429 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
21431 * optabs-query.h (qimode_for_vec_perm): Declare.
21432 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
21433 (qimode_for_vec_perm): ...this new function.
21434 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
21436 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
21438 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
21439 does not have a conditional at the top.
21441 2018-01-02 Richard Biener <rguenther@suse.de>
21443 * ipa-inline.c (big_speedup_p): Fix expression.
21445 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
21448 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
21451 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
21455 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
21456 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
21457 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
21458 cond_taken_branch_cost 3->4.
21460 2018-01-01 Jakub Jelinek <jakub@redhat.com>
21462 PR tree-optimization/83581
21463 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
21464 TODO_cleanup_cfg if any changes have been made.
21466 PR middle-end/83608
21467 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
21468 convert_modes if target mode has the right side, but different mode
21471 PR middle-end/83609
21472 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
21473 last argument when extracting from CONCAT. If either from_real or
21474 from_imag is NULL, use expansion through memory. If result is not
21475 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
21476 the parts directly to inner mode, if even that fails, use expansion
21479 PR middle-end/83623
21480 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
21481 check for bswap in mode rather than HImode and use that in expand_unop
21484 Copyright (C) 2018 Free Software Foundation, Inc.
21486 Copying and distribution of this file, with or without modification,
21487 are permitted in any medium without royalty provided the copyright
21488 notice and this notice are preserved.